The pharmaceutical industry has a structural problem it rarely states plainly: its entire financing model is built around drugs that do not work too well.

This is not a polemic. It is arithmetic. Venture capital, public equity markets, licensing deals, and royalty finance all depend on recurring revenue. A drug that patients take for years generates an annuity. An annuity can be valued, financed, and sold. The present value of a 20-year biologic prescription is calculable. You can build a company around it.

A drug that cures a disease in a single dose generates one payment. That payment has to recover R&D costs averaging $2 billion per approved asset,¹ cover manufacturing and launch, and yield a return that justifies the original capital. All at once. For a patient population that, in many of the disease areas where curative biology is now most advanced — rare pediatric conditions, autoimmune diseases, inherited metabolic disorders — may number in the hundreds per year.

This is why the most scientifically exciting programs in biopharma are often the least well-financed. The 2024 Deloitte analysis of drugs and biologics launched since 2022 puts average return on R&D investment at 2.5% — one-third of the figure recorded a decade prior.² That decline is concentrated precisely in the categories with the most clinical promise. The misalignment is not incidental. It is intrinsic to how the industry raises and deploys capital.

This piece reviews four financing models that have emerged to address this misalignment, assesses where each falls short, and examines a recently proposed structured finance instrument — cure-backed securities (CBS) — that attempts a synthesis.

The commercial case failure problem

Before reviewing solutions, it is worth being precise about the problem.

Commercial case failure in drug development is not the same as scientific failure. A molecule that clears autoimmune disease in a single infusion has not failed scientifically. It has failed commercially — meaning the projected revenue, discounted across the expected market size and competitive landscape, does not justify the investment required to complete development and launch.

This distinction matters because the remedies are different. Scientific failure requires better biology. Commercial case failure is, in principle, a financing design problem — one that finance has tools to address.

The commercial case problem is worst in three overlapping settings. In orphan diseases, patient populations are small by definition, which caps revenue regardless of pricing. In curative autoimmune therapies — CAR-T for lupus and systemic sclerosis being the most visible current examples — the biology is now showing durable remission in Phase 1 data, but the commercial architecture for one-time curative payment is absent. And in pediatric disease more broadly, pool sizes are small, development timelines are long, and regulatory requirements are appropriately demanding.

In each of these settings, standard venture financing logic does not apply. A Series B biotech investor expects a multiple on invested capital at exit — typically through acquisition or IPO, both of which price a company on projected peak sales. A curative therapy for 400 patients a year, priced at $1.5 million per patient, generates $600 million in peak annual sales. That is not a trivial revenue figure. But it is also not the kind of recurring, predictable stream that commands a large acquisition premium or supports a durable public equity story. The upfront capital required to run a Phase 3 trial in this setting — $300–600 million in many cases — often cannot be recovered under conventional assumptions.

The result is a selection effect: programs in these disease areas are deprioritized, delayed, or abandoned not because the science does not work, but because the financing does not.

Four models and their limits

1. Royalty monetization

Pharmaceutical royalty finance is the oldest structural workaround. Under the standard transaction, a drug developer sells a portion of the future revenue stream from a licensed asset — typically expressed as a percentage of net sales — to a specialized capital provider in exchange for upfront cash. The developer gets capital immediately. The royalty investor takes the revenue risk.

Royalty Pharma, the largest player in this market, has deployed this model at scale across academic institutions, biotechs, and large pharmaceutical companies. DRI Healthcare, over three decades, has acquired 77 royalties on 50 drugs, deploying over $3 billion.³ The Gibson Dunn Royalty Finance Report for 2020–2024 documents sustained expansion of the market, with transaction volume increasing across both traditional royalties (arising from pre-existing license agreements) and synthetic royalties (created specifically for the financing transaction).⁴

The limitation of royalty finance in the curative therapy context is structural. A royalty investor prices a transaction off projected revenue. A curative therapy treating 400 patients a year at $1.5 million per patient generates $600 million in gross annual revenue — but that revenue lasts only as long as the untreated patient pool. Once prevalent cases are treated, incidence-based demand may be 50 or 100 new patients per year. The royalty stream collapses. A royalty buyer modeling a conventional multiple on peak sales gets the arithmetic badly wrong.

Royalty finance also provides capital primarily after proof-of-concept data exist — it is predominantly a late-clinical or post-approval instrument. It does not solve the problem of how to fund Phase 2 and Phase 3 trials in a program with an unconvincing commercial projection.

2. The megafund and research-backed obligations

In 2012, Andrew Lo and colleagues at MIT proposed a more architecturally ambitious solution: a pharmaceutical megafund that would finance a large portfolio of early-stage drug programs and issue tradeable debt instruments — research-backed obligations (RBOs) — collateralized by the portfolio’s intellectual property.⁵

The financial logic behind the megafund is portfolio theory applied to drug development. Individual drug programs have very high failure rates — approximately 95% across all clinical phases for a given indication. But these failures are largely uncorrelated. A molecule targeting lupus fibrosis does not fail because a molecule targeting ALS motor neurons fails. The failure events are biologically, mechanistically, and commercially independent. A sufficiently large portfolio will therefore exhibit predictable aggregate outcomes even if individual program outcomes are highly uncertain. The law of large numbers, applied to drug discovery, converts catastrophic individual failure risk into actuarially manageable portfolio risk.

This is the same logic that underlies insurance and, more directly, asset-backed securities. The megafund would slice that portfolio risk into tranches. Senior RBOs — paid first from the portfolio’s cash flows, carrying low credit risk — would attract institutional fixed-income investors: pension funds, insurance companies, sovereign wealth funds. These investors cannot hold early-stage biotech equity due to volatility, regulatory capital requirements, or mandate restrictions. But they can hold investment-grade debt instruments. Equity tranches, paid last and carrying higher risk, would attract conventional venture capital.

Lo’s simulation work, using historical oncology trial data from 1990 to 2011, found that megafunds of $5–15 billion could yield average equity returns of 8.9–11.4% and bond returns of 5–8% — below venture capital hurdle rates, but within range for pension and insurance investors.⁵ A subsequent paper by Fagnan and colleagues applied the same framework specifically to orphan diseases, where lower development costs and faster FDA timelines improve the math: a $575 million orphan megafund with 10–20 programs could generate double-digit expected returns.⁶

The megafund was never operationalized at scale. The reasons are instructive. Credit rating agencies had no established methodology for scoring a portfolio of preclinical drug candidates — the RBO concept depends on investment-grade ratings to attract institutional debt investors, and those ratings require standardized models that did not exist. Investor education was substantial: the asset class was genuinely new, and institutions with fixed-income mandates were not equipped to diligence biomedical IP. And the concentration of program selection within a single fund manager raised governance questions that had no ready answer.

The megafund’s intellectual contribution — the substitution of portfolio diversification for commercial scale — remains the most rigorous framing of the financing problem. Every serious proposal since has, in some form, borrowed from it.

3. Outcomes-based agreements

Outcomes-based agreements (OBAs) emerged on the payer side as the practical response to a specific tension: a payer asked to pay $1.8 million upfront for a gene therapy whose long-term durability rests on follow-up data spanning three to five years faces an uncomfortable gamble. If the therapy fails at year four, the payer has already paid in full for an outcome that did not materialize.

OBAs restructure this relationship. Under a typical OBA, the manufacturer and payer agree that some portion of the payment is contingent on defined clinical outcomes at defined time points. Bluebird Bio’s commercial launch of Zynteglo (betibeglogene spartacept) for transfusion-dependent beta-thalassemia included a guarantee to refund up to 80% of the therapy’s $2.8 million price if patients did not achieve and maintain transfusion independence within two years.⁷ Novartis launched Zolgensma with an optional five-year instalment plan. Hemgenix, the CSL/UniQure hemophilia B gene therapy approved at $3.5 million — the highest launch price ever recorded for a drug — prompted federal-level discussions about outcomes-based reimbursement architecture.

The most significant recent development in this space is the CMS Cell and Gene Therapy Access Model, which went live in January 2025. Under this model, two gene therapy manufacturers — Bluebird Bio (for Lyfgenia) and Vertex Pharmaceuticals (for Casgevy) — negotiated outcomes-based arrangements with CMS covering Medicaid beneficiaries with sickle cell disease. Thirty-two states, the District of Columbia, and Puerto Rico — collectively representing 84% of Medicaid beneficiaries with the condition — agreed to participate.⁸

OBAs solve a real problem: they lower payer resistance to coverage, align manufacturer revenue with therapeutic durability, and create a financial incentive to collect real-world outcomes data. But they do not address the upstream financing problem. An OBA governs the payment relationship between manufacturer and payer after the drug is approved and launched. It does not fund Phase 2 or Phase 3 trials. A company that cannot convince a venture investor to fund the program in the first place cannot use an OBA as collateral.

There is also an administrative complication that has become more visible as OBAs proliferate: tracking patient outcomes across insurer transitions. American patients change insurers frequently — at job changes, at retirement, at birth of dependents, at state relocation. A five-year outcomes obligation negotiated between a manufacturer and a commercial insurer becomes difficult to enforce if the patient moves to a different insurer in year two. The infrastructure to track, adjudicate, and transfer these obligations does not currently exist in any systematic form.

4. Venture philanthropy

The venture philanthropy model re-routes capital into drug development through disease foundations operating as investment vehicles rather than grant-making organizations. The canonical example is the relationship between the Cystic Fibrosis Foundation (CFF) and Vertex Pharmaceuticals.

Beginning in the late 1990s, the CFF committed approximately $150 million to Vertex’s modulator research program — initially funding work that the venture market considered too early and too uncertain. In return, the Foundation received royalty rights on any products that resulted. The program eventually produced ivacaftor (Kalydeco), then lumacaftor/ivacaftor (Orkambi), and finally elexacaftor/tezacaftor/ivacaftor (Trikafta) — the last of which is eligible for approximately 90% of CF patients by genotype. When the CFF sold its royalty stake to Royalty Pharma in 2014, it received $3.3 billion. A subsequent transaction in 2020 returned an additional $575 million.⁹ That capital has been deployed back into the next generation of CF research.

The CFF model has been replicated, in modified form, in other disease areas. The National Bleeding Disorders Foundation established Pathway to Cures as an explicit venture philanthropy vehicle for hemophilia. NCATS and the Foundation for the NIH operate as federal scaffolding for similar public-private structures, connecting NIH basic research funding with industry development capital. The Bespoke Gene Therapy Consortium, co-led by NIH and FDA with private partners, committed approximately $76 million over five years to rare disease gene therapy programs that would not otherwise attract commercial development capital.

Venture philanthropy has a clear limiting condition: it requires a disease community wealthy enough and organized enough to build an investment-capable foundation, with scientific infrastructure — patient registries, biobanks, natural history data — mature enough to attract initial industry partners. The CF Foundation took decades to build both. Many rare disease communities do not have this foundation. And even for diseases where it exists, the model depends on the foundation absorbing early-stage risk that venture will not — which is only viable when the foundation has sufficient capital to sustain a long development timeline.

The gap none of them closes

Each of these models addresses a different part of the financing problem. Royalty monetization provides capital, but only where adequate commercial scale exists. The megafund solves the diversification problem, but was not operationalized. OBAs align payer incentives but do not fund trials. Venture philanthropy works for a subset of well-organized disease communities.

The gap that remains — the one that explains why curative therapies for rare and autoimmune diseases are systematically underdeveloped — is this: how does a developer finance the clinical development of a therapy that will generate a single payment per patient, in a population measured in hundreds to low thousands per year, when no royalty stream is large enough to monetize, no disease foundation exists, and no OBA helps until after approval?

The conventional answer — price the drug high enough that one payment is sufficient — has produced a sequence of approvals at $1.8 million (Zolgensma), $2.8 million (Zynteglo), and $3.5 million (Hemgenix). Each launch has triggered political, payer, and public backlash that makes subsequent approvals harder. The answer is self-defeating.

Cure-backed securities: a structured finance proposal

A March 2026 paper in Gene Therapy from Lu, Cherla, Carter, and Mossialos at the London School of Economics proposes an instrument designed specifically to close this gap.¹⁰ Their argument begins with a reframing: the crisis in curative therapy financing is not primarily a pricing crisis. It is a timing crisis.

The manufacturer needs capital now — at development, at approval, at launch. The payer cannot absorb a $1.8–3.5 million charge in a single budget cycle. Both facts are simultaneously true. They are structurally incompatible unless a third party intermediates the time horizon.

Lu and colleagues propose cure-backed securities (CBS): instruments that separate when the payer pays from when the manufacturer receives payment, using the same securitization logic that underwrites 30-year fixed-rate mortgages.

The MBS analogy is precise, not decorative. A mortgage-backed security works because a homebuyer cannot pay $400,000 on day one, a bank does not want to wait 30 years for repayment, and a capital market exists that will buy the future payment stream at a discount. The homebuyer pays monthly over 30 years. The bank sells those payment rights to investors and recovers capital immediately. Investors receive a low-risk yield. The key innovation is the dissociation of who pays, when they pay, and who holds the interim risk.

CBS applies this structure to curative drugs. Under the proposed model, the payer would pay $130,000 per year for up to 30 years — but only while the patient is alive. This is a survival-contingent annuity. No survival, no payment. The manufacturer packages the future payment streams from a pool of treated patients, tranches the pool by actuarial risk, and sells senior and junior bond tranches to institutional investors. The manufacturer receives the net present value of most of that future stream upfront. Critically — and this is the structural departure from all prior models — the manufacturer retains the equity tranche.

The equity tranche is the residual claim: it pays out if the therapy performs better than actuarial expectations and absorbs losses if it performs worse. By retaining this tranche, the manufacturer has a 30-year financial stake in whether the cure holds. This is a genuine incentive alignment that no prior model achieves. OBAs create short-term outcome contingency — two to five years. CBS creates a 30-year financial bond between manufacturer and patient outcome. Patient registries cease to be regulatory overhead and become assets that protect the manufacturer’s equity position.

The authors modeled the structure using Zolgensma case data and ran 1,000 Monte Carlo simulations across base-case, sensitivity, and pessimistic survival assumptions. Senior bond tranches showed default probability below 0.1% under every scenario — investment-grade by any standard. Under pessimistic clinical assumptions, payers paid $1.27 million per patient versus $1.8 million under existing instalment structures or $1.8 million upfront. Manufacturers could front-load 50–83% of expected net present value at time of sale.¹⁰

CBS in context: what it borrows and what it adds

The CBS proposal does not emerge from nowhere. It is the convergence of the four preceding models.

From the megafund, it takes the tranching and pooling logic: diversification across a patient pool reduces the actuarial risk of any individual survival trajectory, just as diversification across drug programs reduces the portfolio failure risk in Lo’s framework. From OBAs, it takes the survival-contingent payment structure — the idea that pharmaceutical payment should track clinical outcome rather than precede it. From royalty finance, it takes the securitization mechanism: future payment rights converted into tradeable instruments sold to institutional capital. And it creates an incentive for real-world outcomes data collection that mirrors the CF Foundation’s registry infrastructure with Vertex.

What CBS adds — the equity tranche retained by the manufacturer — is the piece that prior models lacked. It converts the manufacturer from a party paid once at approval into a party with a continuing financial interest in therapeutic durability across the full patient life.

Limitations and open questions

The Lu et al. paper is careful about scope, and those limits warrant direct acknowledgment.

Orphan disease scope. The CBS model is designed for settings where generic or biosimilar entry over a 30-year horizon is unlikely. Orphan drugs with small patient populations often qualify. For large autoimmune markets — rheumatoid arthritis, psoriasis — where biosimilar entry is certain within 10 years of approval, the annuity structure breaks down. A payer obligated to pay $130,000 per year for 30 years on a therapy for which a $5,000 biosimilar is available in year 12 has been structurally disadvantaged. The CBS structure would need significant modification for non-orphan settings.

Price decoupling. CBS restructures payment timing. It does not constrain what the manufacturer charges. An instrument that spreads a $3.5 million drug over 30 years at $130,000 per year reduces annual budget shock but does not address the underlying pricing question. CBS works best when paired with price negotiation mechanisms — value-based pricing, ICER thresholds, reference pricing — rather than as a substitute for them.

Administrative infrastructure. Tracking patient survival across insurer transitions over 30 years requires infrastructure that does not exist. American patients change insurers at job change, at retirement, at Medicare transition. The payment obligation in a CBS must follow the patient — but no system currently tracks the obligation across payer transitions, adjudicates disputes, or enforces transfer. Building this infrastructure is a non-trivial policy and operational challenge.

Ratings methodology. The megafund failed partly because ratings agencies had no framework to score a portfolio of drug IP. CBS faces an analogous challenge: survival-contingent pharmaceutical annuities are a new asset class. Investment-grade ratings — which the CBS structure depends on to attract pension and insurance capital — require standardized actuarial models that do not yet exist for this instrument type. OBA precedents and the established MBS methodology provide more foundation than existed in 2012, but the gap is real.

The pipeline is coming regardless

CAR-T programs for systemic autoimmune disease — lupus, systemic sclerosis, myositis — are in Phase 1 and Phase 2 now. Several have reported complete drug-free remission in small cohorts. The early data are credible enough that the questions have shifted from “does it work?” to “how durable is it?” and “can it scale?” If any fraction of these programs reaches Phase 3 approval — and the probability is not trivial — the payment infrastructure will face precisely the problem that CBS is designed to address.

The 30-year fixed-rate mortgage did not emerge as a working instrument when the first long-term real estate loan was made. It required decades of financial engineering, regulatory scaffolding, federal mortgage agencies, and secondary market development before it operated reliably at scale. The analogy to curative therapy finance is apt. The intellectual architecture exists. The operational infrastructure does not.

What CBS does — and what the prior models collectively did — is demonstrate that the financing gap for commercially unviable but clinically essential therapies is not a market failure in the pejorative sense. It is a market design failure. The tools to fix it exist. The question is whether the regulatory, institutional, and financial engineering required to implement them can be assembled before the clinical pipeline outpaces the payment system.

References

1. DiMasi JA, Grabowski HG, Hansen RW. Innovation in the pharmaceutical industry: new estimates of R&D costs. Journal of Health Economics. 2016;47:20–33. doi:10.1016/j.jhealeco.2016.01.012

2. Deloitte. Measuring the return from pharmaceutical innovation 2024. Deloitte Centre for Health Solutions; 2024. Available at: https://www.deloitte.com/us/en/Industries/life-sciences-health-care/articles/measuring-return-from-pharmaceutical-innovation.html

3. DRI Healthcare. About DRI Healthcare. Available at: https://drihealthcare.com/about/

4. Gibson Dunn. Royalty report: royalty finance transactions in the life sciences 2020–2024. 2025. Available at: https://www.gibsondunn.com/royalty-report-royalty-finance-transactions-in-the-life-sciences-2020-2024/

5. Fernandez J-M, Stein RM, Lo AW. Commercializing biomedical research through securitization techniques. Nature Biotechnology. 2012;30(10):964–975. doi:10.1038/nbt.2374

6. Fagnan DE, Gromatzky AA, Stein RM, Fernandez J-M, Lo AW. Financing drug discovery for orphan diseases. Drug Discovery Today. 2014;19(5):533–538. doi:10.1016/j.drudis.2013.11.007

7. Bluebird Bio. Bluebird bio announces U.S. commercial infrastructure to enable patient access to ZYNTEGLO. Business Wire. August 17, 2022. Available at: https://investor.bluebirdbio.com/news-releases/news-release-details/bluebird-bio-announces-us-commercial-infrastructure-enable

8. Centers for Medicare & Medicaid Services. Cell and Gene Therapy Access Model — frequently asked questions. 2025. Available at: https://www.cms.gov/cgt-access-model-frequently-asked-questions

9. Cystic Fibrosis Foundation. Cystic Fibrosis Foundation receives $3.3 billion royalty pay out. Philanthropy News Digest. 2014. Available at: https://philanthropynewsdigest.org/news/cystic-fibrosis-foundation-receives-3.3-billion-royalty-pay-out

10. Lu JM, Cherla AJ, Carter AW, Mossialos EA. Securitization as a means to pay for cell and gene therapies for orphan diseases: a simulation study. Gene Therapy. 2026. doi:10.1038/s41434-026-00604-6

Author: Eswar Krishnan, MD Date: 2026-05-09

Eswar Krishnan is a physician and principal at Olmsted Capital LLC. He consults on clinical development strategy at Drug Development Associates.

#DrugDevelopment #RareDisease #BioPharmFinance #StructuredFinance

We monitor what is easy to track. Not what predicts failure.

The industry built dashboards around milestones.

The signals were in the data weeks before the milestone was missed.

The problem is timing, not data

Most clinical operations leaders I have worked with are not short on information. They have site activation trackers, enrollment velocity dashboards, monitoring visit reports, and query resolution metrics. What they are short on is a principled approach to deciding when a signal in that data is worth acting on.

This is not a criticism. It reflects how clinical operations has been taught and evaluated. The dominant framework is milestone-based: a site is “on track” until a milestone is missed, at which point it is “at risk,” at which point remediation begins. The system is legible, auditable, and defensible. It is also consistently late.

The average activated site that ultimately fails to enroll a single patient in a Phase 3 trial shows observable patterns of delay — slow activation completion, intermittent response to startup queries, screen failure rates in early windows that exceed the site-level forecast — within the first eight to twelve weeks of activity. Traditional operations management catches the same site as a non-performer at ten to fourteen months. The gap between “the signal was there” and “the decision was made” is where late-stage programs lose six months of timeline they cannot recover.

Dynamic optimization is a framework for closing that gap.

What Bayesian monitoring looks like in practice

The word “Bayesian” sometimes triggers discomfort in operations teams, which is understandable. The underlying mathematics can sound abstract in a context where the daily reality is protocol deviations, site coordinator turnover, and sponsors asking for enrollment projections every two weeks.

The practical translation is straightforward: instead of asking “has this site failed?” — a question you can only answer in retrospect — you ask “what is the current probability this site will fail, given everything observed so far?” You answer that question monthly, from the moment of activation, and you have pre-specified decision rules for what each probability level triggers.

In a simulation framework I built for a Phase 3 Sjögren’s Disease program, I modeled site performance across a 54-site portfolio using monthly Bayesian updating. The inputs were routine: time from activation to first screened patient, screen failure rate in weeks 1-8, number of protocol queries opened and closed within 30 days, site coordinator contact responsiveness. None of these required special data collection — they were already being tracked. What changed was how they were interpreted.

The model updated a posterior probability of “structural non-performance” — defined as zero enrolled patients at month six — for each site, monthly, from activation. By month three, 80 percent of the eventual non-performers had already crossed the 80 percent confidence threshold. By month four, the figure was 90 percent.

Traditional management caught the same non-performers at month ten to twelve. The clinical team that implemented the Bayesian monitoring protocol and built pre-specified contingency escalation at the 80 percent threshold recovered a five-month enrollment timeline. At $40-80M per month of patent-protected lifecycle value, the financial return on that methodological shift was in the range of $200-400M.

The cost of implementing the monitoring framework: approximately $1.68M in incremental CRO and data management time.

The arithmetic is not complicated. The harder part is building the organizational willingness to act on a probabilistic signal at 80 percent confidence rather than waiting for the certainty that arrives at month twelve.

Why operations teams wait — and what that costs

There is a logic to the current approach that is worth taking seriously before arguing against it.

Site performance remediation is expensive, relationship-sensitive, and sometimes counterproductive. Escalating to a PI who has been activated for twelve weeks and has not yet enrolled sends a message that can permanently damage the sponsor-investigator relationship. Cutting a site that eventually would have enrolled — but slowly — adds contingency activation costs without recovering the timeline. False positives in early performance monitoring create noise that erodes the credibility of the monitoring system.

These concerns are real. They are also largely soluble.

The false-positive problem is addressed by calibrating confidence thresholds to the actual cost of each action tier. An 80 percent threshold does not trigger site removal — it triggers an enhanced support protocol and a structured conversation with the PI about barriers to enrollment. A 90 percent threshold, sustained over two monthly updates, triggers a formal contingency activation decision. The escalation protocol is designed for the uncertainty inherent in early data, not against it.

The relationship concern is addressed by framing. “We are flagging early indicators and deploying additional support” reads differently to an investigator than “you are underperforming and we are considering action.” The clinical operations team that leads with proactive support — additional coordinator resources, patient recruitment assistance, protocol clarification — is perceived as responsive, not punitive. The same data, deployed with a different communication posture, produces a different site response.

The core issue is organizational. Milestone-based management protects the operations team from second-guessing. If you act on a probabilistic signal at month three and you are wrong, you have made a decision without a defensible trigger. If you wait until month twelve and act after the milestone is missed, the trigger is auditable and the decision is easy to explain.

This is understandable as a personal risk management strategy. It is not optimal for the trial, the sponsor, or ultimately for the clinical operations function’s own strategic value.

Redesigning the Clinical Operations Plan

The structural move that makes dynamic optimization practical is building it into the Clinical Operations Plan before the trial starts, not retrofitting it after things go wrong.

A COP that supports dynamic optimization includes: a formal site risk stratification model at activation, with documented priors for each risk tier; monthly Bayesian performance updates from activation through first enrollment; tiered escalation criteria with pre-specified confidence thresholds and corresponding action protocols; a pre-identified contingency site list with estimated activation timelines, agreed in advance with the sponsor; and decision rights clearly allocated — which calls can be made at the CRO/operations level, which require sponsor escalation, and what the expected response timeline is.

None of this is especially complicated. Most of the analytical infrastructure already exists within the CRO or sponsor data systems. What changes is the decision architecture around the data: who reviews it, how frequently, what probability threshold triggers what action, and who has authority to act without waiting for a steering committee review.

The clinical operations teams that have implemented this have reported something worth noting: the monitoring meetings become shorter. When the decision criteria are pre-specified and the probability updating is automated, the monthly site review becomes a focused triage rather than a wide-open discussion. The team spends less time debating whether a site is underperforming and more time executing the response that was already agreed.

The four principles of dynamic optimization applied to site portfolio management. Panel A: Where trial risk actually concentrates (early activation window vs. mid-enrollment). Panel B: How redundant failure modes defeat single-metric monitoring. Panel C: Bayesian belief updating across the site lifecycle — monthly confidence intervals rather than binary pass/fail gates.Panel D: The five-step Adaptive Operations Loop — Set Prior → Monitor Early Signals → Update Belief → Decide at Threshold → Escalate or Support.

Source: Krishnan EK Drug Development Associates. Simulation data from Phase 3 autoimmune disease program.

The broader implication: operations as a strategic function

There is a version of clinical operations that runs trials efficiently within a fixed protocol. Sites are activated, patients are enrolled, the database is locked. This function is valuable. It is not the function that late-stage drug development now requires.

A Phase 3 program in a competitive indication where patent expiry is a known constraint is not primarily a logistical challenge. It is a capital allocation problem in which every timeline decision has a financial consequence, and the clinical operations function is one of the primary levers on that timeline.

Framed that way, the question for a clinical operations leader is not “are we hitting our site activation milestones?” It is “are we making the best possible sequential decisions about our site portfolio, given the information available at each moment in the trial?” The first question produces an operations function that executes plans. The second produces an operations function that manages risk.

The difference matters enormously to sponsors — and it matters to the long-term positioning of the clinical operations profession.

Dynamic optimization is the methodology that supports the second framing. It does not require new technology, though technology can accelerate its implementation. It requires a shift in how routine operational data is interpreted, how decision criteria are specified before a trial begins, and how the clinical operations team understands its own mandate.

Six trial scenarios documenting the systematic cost of late operational decision-making. For each scenario: the metric being monitored (the Lamppost), what was missed (the Dark Alley), the decision that was ultimately required (the Breakthrough), and the months of timeline that were unrecoverable by the time the decision was made. Key finding: In five of six scenarios, earlier probabilistic action would have recovered 3-5 months of enrollment timeline. Estimated cumulative patent-life value lost across scenarios: $200-600M per program.

Krishnan E, Drug Development Associates. Illustrative compilation based on published trial reports and simulation data.*

A practical starting point

For clinical operations leaders who want to move toward this framework without rebuilding their entire monitoring infrastructure, a reasonable starting point is this: pick your last two or three late-stage programs and run the retrospective analysis. At what point in each trial did the eventual non-performing sites show the patterns that predicted their failure? What was the latency between when those patterns were observable and when a decision was made?

The answer, in almost every program I have reviewed, is that the signal was available three to five months before the decision. That gap is recoverable timeline. In a competitive program, it is often the difference between a label that reflects full enrollment and one that reflects what you were able to achieve under time pressure.

The analysis is not complicated. The discomfort is facing what it implies: that the next time around, you will need to act sooner, on less certainty, with a framework that makes that action defensible.

That is what dynamic optimization provides.

Eswar Krishnan, MD, M.Phil is a physician and drug development consultant based in Carmel, Indiana. He has spent 30 years at the intersection of Bayesian clinical trial methodology, adaptive operational design, and late-stage program strategy. He works with biotech and pharma sponsors on clinical operations strategy, site portfolio design, and program recovery.

*Consulting inquiries: principal@olmsted-capital.com*

*Follow for weekly posts on clinical operations, adaptive trial design, and the decisions that actually determine drug development outcomes.*

I have watched three CMOs at three different pharmas sit through the same quantum vendor pitch in the past quarter. One signed a three-year strategic partnership. One agreed to a six-month scoped pilot. One thanked the vendor and walked. All three made defensible decisions. Their pipelines were different. Their boards were different. Their portfolios had different chemotypes, different timelines, different cash positions. The quantum question landed differently on each desk because the prior question — “what kind of R&D organization are we?” — had been answered differently long before the vendor showed up.

Most quantum-strategy advice in pharma skips this step. It tells you what quantum can do, what it cannot do, who is partnering with whom. Useful enough. But it assumes you have already answered the only question that matters: where does your company sit on the early-adopter-to-skeptic curve, and is that position aligned with your board?

Three archetypes. Pick one honestly.

The early adopter. This is the company that builds an internal quantum capability before the technology is mature, accepts that some of the early bets will be wrong, and treats the cost as the price of optionality. Boehringer Ingelheim with Google. Roche with Cambridge Quantum. Merck with HQS. The investment thesis is straightforward: when fault-tolerant quantum computing matures, the companies that already have hybrid workflows and quantum-literate teams will move faster than the ones starting from zero. The economic case is harder. Five years of internal quantum spend with no validated commercial output is hard to defend in a year when the board wants pipeline focus.

You should be an early adopter if your pipeline contains enough strongly correlated chemistry — transition-metal catalysis, metalloenzyme inhibitors, photoredox biology — to justify the bet on its own scientific merit, and your board has the patience for a multi-year capability play. If you are an early adopter and your board does not have that patience, you are an early adopter who is about to disappoint your board.

The fast follower. This is the company that watches the early adopters carefully, runs scoped pilots with one or two vendors per year, and waits for a clear advantage signal before committing internal capability. Most large pharmas are quietly fast followers, even when their press releases sound like early-adopter language. The discipline of the fast follower is the willingness to spend ten to twenty million dollars a year on quantum exploration and walk away from any pilot that does not produce a defensible result by month nine. The discipline is also the willingness to look slow when an early-adopter peer announces something that turns out to be marketing.

You should be a fast follower if your pipeline could plausibly benefit from quantum in two or three areas, your board wants quarterly updates with concrete milestones, and you have the operational maturity to run vendor pilots without becoming captive to them. Most companies that should be fast followers convince themselves they are early adopters. The cost of that mistake is paid in the discovery budget two years later.

The patient skeptic. This is the company that reads the literature, attends one or two industry forums a year, has a single named executive responsible for quantum-readiness, and otherwise spends nothing. The patient skeptic is not anti-quantum. They have looked at their own pipeline, decided that quantum will not move the needle on the molecules they actually care about for at least seven to ten years, and made a deliberate decision to deploy that capital elsewhere. Cell and gene therapy companies, biologics-focused mid-caps, generics manufacturers, and any company whose chemistry is dominated by problems classical methods already solve adequately should probably be patient skeptics.

You should be a patient skeptic if your pipeline does not contain the chemotypes where quantum has a defensible near-term case, your board has more pressing capability gaps elsewhere, and you can credibly explain to your CSO why “we are watching” is a complete answer. The mistake patient skeptics make is forgetting to maintain the watch. The technology is moving. The decision to wait should be revisited every twelve months, not made once and filed.

The diagnostic is uncomfortable on purpose. It is meant to surface the gap between the archetype your company actually fits and the archetype your strategy deck claims you are. In my experience, that gap is the single most expensive misalignment in the pharma technology stack — more expensive than the quantum spend itself.

A short version of the diagnostic. Answer honestly.

How much of your active pipeline involves chemistry where DFT or coupled-cluster methods break down — transition metals, multireference systems, strongly correlated electrons? If the answer is “a lot,” lean toward early adopter. If “some,” fast follower. If “almost none,” patient skeptic.

How much patience does your board have for capability investments without near-term pipeline contribution? If three years or more, early adopter is defensible. If twelve to eighteen months, fast follower. If “we have other priorities,” patient skeptic.

How much do you trust your CSO to walk away from a vendor pilot that is not delivering? If completely, fast follower works. If you would worry about sunk-cost dynamics, lean patient skeptic until your governance is stronger.

That is the entire framework. There is more to say about each archetype — what to spend, who to hire, which partnerships to consider — and the next nineteen posts in this series will say it. But none of that material matters until you have answered the question on this page.

The decision this post forces: which archetype are you, and is that aligned with what your board actually wants? If you cannot answer both halves of that question with confidence, your quantum strategy does not exist yet — you have a quantum vocabulary, which is different.

The trigger to revisit: twelve months from today, or sooner if a peer in your therapeutic area announces a quantum result that would, if true, change your pipeline economics.

#QuantumComputing #DrugDiscovery #ClinicalTrials #Biopharma

🔗 Drug Developer Newsletter → [Subscribe on LinkedIn https://www.linkedin.com/build-relation/newsletter-follow?entityUrn=7334337397922430977]

🔗 Clinical Trial Newsletter → [Subscribe on LinkedIn https://www.linkedin.com/build-relation/newsletter-follow?entityUrn=7347268184502112256 ]

Before the first post in this series publishes, you should be able to answer one question clearly: what kind of company are you in relation to quantum computing — and is that position deliberate?

Not informed. Not enthusiastic. Deliberate. That distinction is the entire premise.

Executive Summary

This is the introduction to a 20-post biweekly series on quantum computing for drug development executives. It is written for the CMO, CSO, and CEO who must eventually make real decisions about this technology — capital allocation, vendor partnerships, team composition, board communication — without having the time or background to read the primary literature themselves.

The series makes three commitments. First: every numerical claim, corporate partnership, and scientific assertion traces to a named primary source. Second: mathematical notation appears only when it earns its place, and every notation is accompanied by a plain-language explanation of what it means and why it matters for drug development specifically. Third: the series will tell you when quantum is not the answer. Several of the 20 posts exist specifically to help you walk away from vendor pitches that do not hold up.

The series is structured in five chapters across 20 posts, running biweekly. Each post resolves to a decision and a trigger — what to do now, and when to revisit. The goal is not to make you a quantum physicist. The goal is to make you the executive in your organization who asks the right questions. That is a different and more useful target.

The Problem That Makes Quantum Worth Discussing

Drug development has a mathematics problem that predates quantum computing by several decades, and understanding it is the reason for everything that follows.

In 1996, Bohacek, McMartin, and Guida published a landmark analysis estimating the size of drug-like chemical space — the universe of all small organic molecules that might plausibly function as drugs. Their estimate, since refined and broadly accepted, placed the number at approximately:

This approximation works well for a large fraction of drug-like chemistry. It fails for a fraction that is small in percentage terms and enormous in pharmaceutical significance: transition-metal coordinated proteins, strongly correlated electron systems, open-shell multireference molecules, and certain classes of photoredox biology. These are not exotic edge cases. About one-third of all known proteins require a metal ion cofactor to function. A zinc protease, an iron-sulfur cluster enzyme, a copper-containing oxidase — each presents an electronic structure problem that DFT approaches with known systematic errors. Those errors propagate into binding energy predictions. Those predictions inform clinical candidate selection. The clinical failure rate has not meaningfully improved in thirty years.

Quantum computing’s claim on drug development begins here: not as a general accelerant for the entire pipeline, but as a method that can handle the specific cases where classical physics approximations fail. That is a narrower claim than the vendor marketing suggests. It is also a more credible one.

What a Quantum Computer Actually Is — The Minimum You Need

A full treatment of quantum computing fundamentals appears in Post 2 of this series (Qubits in Ten Minutes). What follows here is the minimal scaffolding necessary to understand why the technology is relevant to molecules.

A classical computer stores information in bits. Each bit is either 0 or 1. A register of n bits can store one of 2ⁿ possible states at any given moment.

A quantum computer stores information in qubits. Before measurement, a qubit is described not by a single value but by a quantum state:

WHY IT MATTERS FOR DRUG DEVELOPERSThis is the same mathematical structure that describes an electron in a molecular orbital before you measure its position. Quantum computers speak the native language of molecular physics. Classical computers are translating.

For a system of n qubits, the joint state is described by 2ⁿ complex amplitudes simultaneously:

WHY IT MATTERS FOR DRUG DEVELOPERS Simulating the electronic wavefunction of a molecule requires exactly this kind of exponentially large representation. Classical computers approximate it because they have no alternative. Quantum computers can, in principle, represent it exactly. That “in principle” carries real weight, and this series will not let it be lost in marketing language.

The Honest State of the Field

Any introduction that does not acknowledge the problem of hype is itself part of the hype problem.

Quantum computing has been described as “five years from transforming drug development” for approximately ten consecutive years. The track record of predictions in this field is poor, and executives who made early bets on quantum advantage in clinical applications — particularly in the 2019–2022 period — have, in most cases, not seen their returns validated.

What has changed, and why the timing of this series is deliberate, is the convergence of three developments since 2023:

First, error correction has crossed a threshold. In 2025, Google’s quantum AI team demonstrated a surface-code error correction result in which increasing the code distance from 3 to 5 to 7 reduced logical error rates in a consistent exponential pattern — the first experimental confirmation that quantum error correction scales as theory predicts. This does not mean fault-tolerant quantum computing is here. It means the path to fault tolerance is now empirically validated rather than theoretically assumed. That is a qualitatively different situation.

Second, real industry partnerships have moved past press releases. The following collaborations involve committed R&D budgets and published intermediate results, not just joint announcements:

Third, a published pipeline result has cleared a meaningful bar. In 2025, Ghazi Vakili and colleagues published a quantum-classical generative pipeline for KRAS inhibitor identification in Nature Biotechnology. Fifteen molecules were synthesized from quantum-assisted design. Two showed promising measured activity. This is not a clinical result. It is a real-world pipeline result — synthesized molecules with wet-lab-confirmed activity — which is a categorically different level of evidence than simulation-only claims.

None of this constitutes proof that quantum computing will transform drug development at scale. It constitutes proof that the technology has moved from theoretical demonstration to early applied results in a domain directly relevant to pharma pipelines. That is the inflection point that justifies serious executive attention.

Why This Matters for Patients — The Argument That Should Drive Everything

The failure rate in drug development has been approximately 90% across all indications for three decades. Roughly half of Phase II failures are attributed to inadequate efficacy — the molecule does not do in patients what models and early studies predicted. A significant fraction of those failures traces to errors in how we model molecular interactions: binding affinities that look favorable in silico and fail in vivo, selectivity profiles that shift in the complexity of the physiological environment, and off-target effects that appear only when the full protein interactome is engaged.

Some of these failures are unavoidable. Biology is complex in ways that no computational method will fully capture. But some are the downstream consequences of known limitations in classical simulation — approximations that we use not because they are adequate but because we have no better tool.

Patients with treatment-resistant cancers, rare metabolic diseases, and neurodegenerative disorders are, in a direct sense, waiting on this problem to be solved. When a KRAS-mutant tumor fails to respond to a candidate that looked promising in computational binding studies, the question of whether better molecular simulation would have caught the problem earlier is not rhetorical. It has a real answer, even if we cannot access it retrospectively.

Quantum computing does not promise to eliminate this problem. It promises to address the specific subset of failures that stem from the cases where classical physics approximations break down systematically. That is a limited promise. In a field where the cost of late-stage failure runs to hundreds of millions of dollars per program, and where the patient waiting at the end of every failure is real, a limited promise is still worth taking seriously.

This is why the series is patient-focused rather than technology-focused. The technology is a means. The patients are the point.

The Series Architecture

Twenty posts, five chapters, biweekly cadence. Each post resolves to a stated decision and a stated trigger to revisit.

Chapter 1 — The Operator’s Mental Model (Posts 1–4)
The foundational layer: what quantum computing is, what it is not, and how to position your organization before you spend a dollar. Post 1 introduces the three executive archetypes (early adopter, fast follower, patient skeptic) and the diagnostic that tells you which one fits your portfolio. Posts 2–4 cover the physics and hardware at the altitude required to evaluate a vendor pitch — no more, no less.

Chapter 2 — Discovery and Chemistry (Posts 5–8)
Where quantum touches active discovery programs. The specific chemotypes and computational failure modes that create a genuine near-term case. Post 8 is the contrarian post: most published quantum advantage claims in computational chemistry have been beaten by a GPU within 18 months. Understanding why is as important as understanding the successes.

Chapter 3 — Proteins, Targets, and Binding (Posts 9–12)
The post-AlphaFold landscape. Binding free energy: whether quantum can improve on FEP, and under what conditions. The honest state of quantum machine learning in cheminformatics — including a 2025 systematic review that found no consistent superiority over classical baselines in digital health applications.

Chapter 4 — Translation, PK, and Clinical Development (Posts 13–16)
Trial design optimization, quantum sensing as a clinical measurement tool, and the Y2Q cryptography deadline — the date by which current encryption standards will be vulnerable to quantum attack, and the regulatory implications for trial data that must remain protected for decades.

Chapter 5 — The Operator’s Playbook (Posts 17–20)
Build, buy, partner, or wait. How to structure a quantum partnership agreement. Which hire to make first (the answer is not who most executives expect). The series closes with a transferable artifact: ten questions to ask any quantum vendor, calibrated to expose the difference between serious science and marketing theater.

A Note on Mathematical Notation in This Series

Mathematics appears in this series when it earns its place. A notation appears only if it communicates something that prose cannot. Every notation is accompanied by a callout that explains, in plain language, what the symbols mean and why the relationship matters specifically to drug development.

The callouts are not simplifications of the mathematics. They are translations. The goal is that a reader without formal training in quantum mechanics can follow the argument, while a reader with that training finds nothing technically misrepresented. Both standards apply simultaneously.

If you encounter a notation that lacks a plain-language explanation, treat it as an editorial error. Email us.

The Transparent Connection to the Newsletters

This series is published on Substack at www.drugdevelop.com as a companion to two newsletters: Drug Developer Newsletter and the I&I Clinical Trial Strategy Notes. If the series is useful to you, subscribing to either or both is the appropriate next step. That is the only ask.

No paywalls. No sponsored content. No vendor relationships that affect editorial positions.

References

Bohacek RS, McMartin C, Guida WC. The art and practice of structure-based drug design: a molecular modeling perspective. Med Res Rev. 1996;16(1):3–50.

Reymond JL, Awale M. Exploring chemical space for drug discovery using the chemical universe database. ACS Chem Neurosci. 2012;3(9):649–657. https://doi.org/10.1021/cn3000422

Ghazi Vakili M, et al. Quantum-computing-enhanced generative chemistry for KRAS inhibitor discovery. Nature Biotechnology. 2025. [DOI to be confirmed prior to posting.]

Google Quantum AI. Quantum error correction below the surface-code threshold. Nature. 2025. https://doi.org/10.1038/s41586-024-08449-y

Zhou Y, Chen J, Cheng J, et al. Quantum-machine-assisted drug discovery. npj Drug Discovery. 2026;3:1. https://doi.org/10.1038/s44386-025-00033-2

Preskill J. Quantum computing in the NISQ era and beyond. Quantum. 2018;2:79. https://doi.org/10.22331/q-2018-08-06-79

Doga H, et al. Quantum computing in clinical trials: optimization approaches and future directions. Trends in Pharmacological Sciences. 2024. [Full citation to be confirmed prior to posting.]

IBM Quantum. Cleveland Clinic and IBM unveil first private-sector quantum computer in the US. Press release. 2023. ibm.com/newsroom

The LinkedIn version of this introduction, and links to subscribe to Drug Developer Newsletter and the I&I Clinical Trial Notes ].

Post 1 — Are You a Buyer or a Bystander? — publishes [April 30, 2026].

I want to tell you something that might reframe years of mealtimes.

That child who covers their nose when peanut butter is opened. The one who will only eat five foods. The one who can’t sit at the lunch table when someone opens a bag of tree nuts. Parents are told this is anxiety, or sensory issues, or pickiness, or bad behavior. Pediatricians sometimes say they’ll grow out of it. Therapists are brought in. Dietary elimination trials fail.

What almost nobody tells these families is that there is a body of scientific evidence — some of it published in Nature and Science within the last two years — suggesting that a food-allergic child’s immune system may be doing something we only recently understood: it is reprogramming the brain to avoid food.

Not metaphorically. Literally. Measurably. Through specific molecules, in specific brain regions, in ways that precede visible gut inflammation and classic allergic symptoms.

This is the story of that biology. It’s also the story of what food allergy actually does to children — their growth, their social lives, their mental health — and where treatment is going. If you are a parent managing a food-allergic child, or a clinician who sees them, this is the information I think you deserve to have.

First, a few numbers most people don’t know

Food allergy affects roughly 8% of children in the United States, according to CDC data. That is approximately 6 million kids. But here’s the number that quietly gets me every time I look at it: in a 2025 survey of families managing pediatric food allergies, more than a third of parents reported that their child’s allergy had caused them to miss school events, birthday parties, or class activities in the prior year. Not because of a reaction. Because of the fear of one.

The economic cost of pediatric food allergy in the US is estimated at $25 billion annually — roughly $4,200 per child per year. Most of that is not medical treatment. Most of it is dietary substitutions, specialty foods, and lost caregiver productivity.

Anaphylaxis hospitalizations in children increased 150% between 2000 and 2019. About 40% of food-allergic children have had a severe reaction — meaning anaphylaxis — before they receive proper diagnosis and management. The diagnosis often comes after the crisis.

And food allergy is not equally distributed. Black children have substantially higher rates of food allergy-related emergency department visits than white children, even after controlling for income, education, and access to care. Urban children have nearly double the food allergy prevalence of rural children — 9.8% versus 6.2% in US studies. Why? The short answer is we don’t fully know. The longer answer involves the microbiome, farm animal exposure, vitamin D, and something called the hygiene hypothesis — but that’s a separate piece.

One more piece of epidemiology I find remarkable: perceived food allergy prevalence is consistently higher than actual diagnosed prevalence. Studies across Europe show self-reported rates of 10–30% but physician-confirmed rates well under 5%. This matters because children are sometimes placed on elimination diets unnecessarily — with real nutritional consequences — based on family perception rather than confirmed allergy testing.

Where allergies actually start — and it is not where you think

The gut is not the beginning of this story.

For decades, food allergy was framed as a gastrointestinal problem. You eat something, your immune system overreacts in the gut, symptoms follow. It made a certain intuitive sense. But the evidence has been building since the mid-2000s, and by now it’s fairly settled: for many children, the journey to a food allergy begins not in the digestive tract, but on the skin.

This is the dual allergen exposure hypothesis, and it has significant evidence behind it. The basic concept is this: when a child is exposed to a food allergen through an intact gut — during normal feeding — the immune system usually builds tolerance. But when that same allergen enters through a damaged or compromised skin barrier, the immune system does the opposite. It treats the protein as a pathogen. It builds IgE antibodies. It primes mast cells. It learns to react.

The ALSPAC birth cohort study in the UK tracked thousands of children and found that peanut allergy in preschoolers was independently associated with topical exposure to peanut allergen — specifically through application of creams containing peanut oil on inflamed skin during the first six months of life. A separate study found elevated risk of IgE-mediated wheat allergy linked to skin and hair products containing hydrolyzed wheat protein. The peanut never needed to be eaten to sensitize the child. Contact with the skin was enough.

The gene at the center of this is filaggrin (FLG). Filaggrin is a protein that holds the outer skin barrier together. Loss-of-function mutations in FLG — carried by roughly 10% of people of European descent — dramatically increase the risk of eczema, and downstream, food allergy. When the skin barrier leaks, airborne food particles and proteins in skincare products get in. The immune cells waiting just beneath the epidermis — Langerhans cells, mast cells, dendritic cells — see those proteins. And in the context of skin damage and inflammatory signals called alarmins (TSLP, IL-33, IL-25), they learn to regard them as threats.

About 30% of children with moderate-to-severe eczema also have coexisting food allergies. That figure is not a coincidence. Eczema is often the first stop on what immunologists call the Atopic March.

The Atopic March: a map of what happens to these children over time

The Atopic March refers to the typical progression of allergic disease from infancy through childhood and beyond. It generally moves like this: eczema first, often in infancy. Then food allergy, usually by age two or three. Then asthma. Then allergic rhinitis. One condition doesn’t necessarily cause the next — the relationship is partly causal, partly due to shared genetics, partly due to shared immune dysregulation. But they travel together with striking regularity.

A quarter of children with eczema transition to at least one other allergic phenotype. One in five develop multimorbidity — all three conditions together.

What’s interesting — and somewhat counterintuitive — is that the march doesn’t always end in childhood. A longitudinal birth cohort from the Isle of Wight followed participants for 18 years. They found that food allergies resolved during early childhood for many children (about two-thirds of peanut, egg, and sesame allergies resolved between ages 1 and 4 in one cohort). But then there was an uptick in new food allergy sensitization in the teenage years, resulting in higher prevalence at age 18 than in mid-childhood. This pattern is not well understood. Adolescent-onset food allergy is underrecognized, and it may explain why some teenagers who had apparently outgrown allergies report new reactions.

What the immune system actually does when a food allergen is ingested

Here is the molecular sequence, as concisely as I can put it.

Sensitization has already happened — at the skin, or possibly at an early mucosal exposure. IgE antibodies against the allergen are now circulating. They have bound to high-affinity receptors (FcεRI) on mast cells throughout the gut wall. The gut is primed.

When the allergen arrives — in a meal, in trace contamination, in a food that shares proteins with the sensitizing food — it crosses the gut epithelium. It finds those IgE-coated mast cells. It crosslinks the IgE receptors. Within seconds, the mast cells degranulate.

Here is where a 2025 study from Yale, published in Science, overturned a fundamental assumption about what happens next.

The classic model said: mast cells release histamine, histamine causes symptoms, therefore antihistamines should help. But that study found that intestinal mast cells are a distinct subtype from connective tissue mast cells elsewhere in the body. They take their cues from neighboring epithelial cells. And those cues shift their behavior dramatically: intestinal mast cells make relatively little histamine. They ramp up production of cysteinyl leukotrienes instead.

Scientists found that when an allergen is ingested, gut mast cells respond differently from mast cells elsewhere in the body — producing cysteinyl leukotrienes rather than histamine, a finding that helps explain a long-standing puzzle: why IgE antibody levels do not reliably predict food allergy risk, and why food-specific antibodies in the blood are a poor guide to severity.

This is why antihistamines — loratadine, cetirizine, diphenhydramine — don’t stop food-induced anaphylaxis. They block histamine receptors. But the gut is running on leukotrienes.

Mice genetically deficient in cysteinyl leukotriene synthesis were protected from oral antigen-induced anaphylaxis, while those treated with zileuton, a drug already approved for asthma, showed similar protection. This has immediate clinical implications that are now being studied.

Astonishing finding: food aversion is the immune system working

Let me tell you about a 2023 paper in Nature that I think is one of the most underappreciated findings in allergy research in years.

The question the researchers asked was this: if food allergies are so dangerous, does the body develop any behavioral defense against eating the offending food? The answer, it turns out, is yes. A precise, molecular, neurologically-mediated one.

Using mouse models of food allergy, researchers showed that allergic sensitization drives antigen-specific avoidance behavior. Allergen ingestion activates brain areas involved in the response to aversive stimuli — including the nucleus of the tractus solitarius, the parabrachial nucleus, and the central amygdala. Allergen avoidance required IgE antibodies and mast cells, but — crucially — it preceded the development of gut allergic inflammation.

Read that last clause again. Food aversion develops before visible gut inflammation. The brain gets the signal first. And the signal comes not from histamine — blocking histamine receptors had no effect on aversion — but from cysteinyl leukotrienes and a molecule called GDF15 (growth and differentiation factor 15), which is known from other contexts as a signal of cellular stress and tissue damage.

The working hypothesis the researchers proposed is this: allergen is sensed in the gut mucosa through allergen-specific IgE on tissue-resident mast cells. Those mast cells release cysteinyl leukotrienes, which mediate GDF15 secretion, which signals the brain through pathways we don’t yet fully understand — but that activate the same neural circuits involved in disgust, nausea, and fear learning.

The child who refuses to eat eggs after one bad reaction is not being dramatic. Their immune system has updated the brain’s threat model. The brain now treats that food as a poison.

This may also help explain something I see in clinic that has never had a satisfying explanation: children with food allergies sometimes develop aversion to safe foods — foods they have never reacted to, foods that don’t share proteins with their allergens. The theory is that repeated experience of nausea and distress during meals, driven by sub-clinical immune activation, generalizes. The whole act of eating becomes threatening. Food becomes suspect.

Food aversion, ARFID, and the disorder hiding in plain sight

Which brings me to ARFID.

Avoidant/Restrictive Food Intake Disorder is a feeding disorder characterized by extreme selectivity — not driven by body image concerns, not by fear of weight gain, but by fear of adverse food reactions, sensory aversion, or near-complete disinterest in eating. It was added to the DSM-5 in 2013, replacing the older, narrower category of “feeding disorder of infancy and early childhood.”

A study involving 54 children with food allergies who were patients at a food allergy clinic found that more than half met the criteria for probable ARFID. More than half. In a population that, by definition, already has a medically justified reason to avoid certain foods.

This matters because there’s a diagnostic trap here. When a child with, say, a peanut and tree nut allergy also refuses milk, eggs, berries, chicken, and everything except white rice and plain pasta — clinicians may attribute all of it to the known allergy. The genuine ARFID component is missed. Nutritional deficiencies follow. Growth is affected.

Children on allergen elimination diets showed more picky eating and feeding problems overall, with picky eating linked to lower weight-for-age z-scores, food refusal, constipation, and anticipatory gagging.

And the relationship runs in both directions. ARFID can predate a food allergy diagnosis — and having ARFID-like behaviors may actually complicate food challenge testing, because a child who is already highly aversive is harder to evaluate.

Children with food allergies often show heightened anxiety about eating and a tendency to avoid new or potentially allergenic foods — patterns that mirror the restrictive eating seen in ARFID. The constant vigilance required to avoid allergens can lead to heightened anxiety and fear around food, increasing the risk of developing ARFID.

What I wish more families knew: this is treatable. ARFID in the context of food allergy is not a character flaw or a parenting failure. It has a neurobiological basis that we now understand much better than we did five years ago. Cognitive behavioral therapy adapted for food allergy ARFID (CBT-AR) has shown that after 12 weeks, 85% of children achieve meaningful improvement in food variety and anxiety reduction. Specialized feeding clinics that start with children eating an average of three foods regularly get them to 19 foods after intensive therapy. These are real, achievable changes.

Facts about food allergy that most families — and some clinicians — don’t know

Let me take a brief detour from mechanism to give you the kind of trivia that genuinely matters.

1. Peanuts are not tree nuts. Botanically, peanuts are legumes — in the same family as lentils, peas, and soybeans. A child allergic to peanuts has about a 25–40% chance of also being allergic to tree nuts, but the allergy is to a different protein family. Managing a peanut allergy does not automatically mean a tree nut allergy, and vice versa.

2. The top nine allergens now include sesame — and the addition changed menus overnight. In 2023, sesame became the 9th major allergen in the US under the FASTER Act. This created an unexpected problem: manufacturers who had been using sesame as a “hidden” ingredient were now required to declare it. Some reformulated to add sesame intentionally so they could label it overtly — which suddenly exposed allergic consumers who had previously been able to eat those products safely. This is still an ongoing controversy.

3. Milk allergy and lactose intolerance are completely different conditions. Lactose intolerance is a digestive issue — absence of the enzyme lactase. No immune system involvement, no IgE, no anaphylaxis risk. Milk allergy is an immune response to milk proteins (casein, whey). A child with true milk allergy can have anaphylaxis from a trace of dairy. A child with lactose intolerance just has gastrointestinal discomfort. Treating them the same way is a medical error.

4. About 20% of peanut allergies resolve naturally by early adulthood. Roughly 80% of egg allergies do the same. Milk allergy also resolves in most children who develop it in infancy, usually by age 5. Shellfish and tree nut allergies, on the other hand, rarely resolve. Persistence rates for those are over 90%. The natural history varies enormously by allergen and by the severity of the initial sensitization.

5. IgE level does not predict reaction severity. This surprises patients every time I explain it. A child with a sky-high peanut-specific IgE may have only mild symptoms on challenge. A child with a low IgE may have anaphylaxis. The relationship is statistical, not deterministic. Skin prick test wheal size and IgE titer help guide clinical decision-making but they are not oracles.

6. The threshold dose for reaction varies by 1,000-fold between individuals. Some children react to a fraction of a milligram of peanut protein. Others can tolerate hundreds of milligrams before they react. This is why blanket “may contain” warnings are so hard for families to navigate — and why some families make reasonable risk-adjusted decisions to eat products with precautionary labeling while others cannot.

7. Urban children have nearly double the food allergy rate of rural children. Urban US children: 9.8% food allergy prevalence. Rural US children: 6.2%. The “farm effect” — regular exposure to barn animals, unpasteurized milk, diverse microbial environments — appears genuinely protective. This fits with the hygiene hypothesis and the biodiversity hypothesis of allergy, which suggests that reduced microbial diversity in the modern gut microbiome impairs immune regulation.

8. Food allergy prevalence differs by geography in ways that map to diet, not just genetics. In North America and Northern Europe, peanut and egg allergies predominate. In Asia, shellfish and fish allergies are more common. This is not purely genetic — it reflects which foods are introduced early, in what form, and in what cultural context. The LEAP trial (Learning Early About Peanut Allergy) found that introducing peanut early — in the first year of life, before age 11 months — reduced peanut allergy by 81% in high-risk infants compared to avoidance. We spent decades telling parents to avoid peanuts in infancy. We were wrong.

The treatment landscape, honestly

There are currently two FDA-approved treatments for food allergy.

Palforzia (2020): a characterized peanut protein powder used for oral immunotherapy (OIT) in children aged 4–17 with peanut allergy. The goal is desensitization — not a cure, but an increase in the threshold dose required to trigger a reaction. This gives families a meaningful safety buffer. The catch: OIT requires daily dosing, strict adherence, and can itself cause reactions. About 10–15% of patients discontinue due to adverse effects. Sustained unresponsiveness — the ability to tolerate peanut even after stopping regular exposure — occurs in a minority.

Omalizumab (Xolair) (2024 new indication): originally approved for allergic asthma and chronic urticaria, omalizumab was approved in 2024 for IgE-mediated food allergy in patients 1 year and older. It works by binding free IgE in the blood before it can attach to mast cells. In the OUtMATCH trial, 67% of omalizumab-treated patients were able to tolerate at least 600 mg of peanut protein without a dose-limiting reaction, compared to 7% in the placebo group. Similar results held for milk, egg, and cashew. Omalizumab does not cure the allergy. It raises the threshold and buys time for OIT or reduces risk of accidental exposure reactions. But for a child with multiple food allergies who cannot complete OIT — it is a real, meaningful option.

What’s coming: targeting the cytokine pathways upstream. Dupilumab (anti-IL-4Rα), already approved for atopic dermatitis and asthma, is in trials for food allergy and for preventing the Atopic March in infancy. The idea is to interrupt sensitization before it solidifies. Blocking IL-9, which drives mast cell expansion in the gut, is an active research target. And — given the 2025 Science data — drugs that block leukotriene production or receptor binding are now being seriously evaluated for food allergy, rather than just asthma.

The gut microbiome is also a therapeutic target. Fecal microbiota transplantation (FMT) combined with peanut OIT is in clinical trials. The microbiome connection is strong enough epidemiologically — formula-fed infants, C-section births, early antibiotic courses all increase allergy risk — that manipulating it therapeutically is a reasonable hypothesis. We don’t have outcomes data yet. But the mechanistic case is there.

And then there is skin barrier intervention. Several trials are now examining whether aggressive moisturization in the first weeks of life — before the skin barrier has a chance to be disrupted — can reduce the rate of sensitization through the skin. The PEBBLES pilot study found trends toward reduced food sensitization at 12 months in infants treated with emollients five or more days per week. Larger trials are ongoing. The concept is elegant: fix the door before the trespasser gets in.

What this means at the dinner table

I am aware that everything I’ve described above — the leukotrienes, the ALOX5 pathway, the cysteinyl mediators — is a long way from Wednesday night dinner with a child who won’t eat.

So let me translate it.

If your child has food allergy and also has food aversion that seems disproportionate — refusing safe foods, gagging at smells, meals that are battles — please take it seriously as a distinct clinical problem. It is not manipulation. It is not anxiety for no reason. There is a molecular basis for that child’s relationship with food, and it can be addressed.

If your child is on an allergen elimination diet and their weight gain has slowed, or they’ve started refusing foods they used to eat, or mealtimes are consistently distressing — these are signs to escalate, not wait out.

If you’re a parent modeling anxiety at mealtimes — constantly checking labels with visible stress, tasting everything before your child, rehearsing emergency protocols in front of them — that anxiety communicates. The data on parent behavior and ARFID risk is consistent: children take cues. Calm, matter-of-fact allergen management paired with a rich relationship with safe foods is the target.

And if you have been told your child just needs to “try harder” or “stop being dramatic” about food — I want you to know that’s not what the science says. The science says that an allergic child’s brain has been trained, by their own immune system, to treat certain foods as existential threats. That is a medical condition. It deserves medical attention.

Where I think this goes

The next decade in food allergy is going to be about prevention more than treatment. The signals are there.

We know sensitization starts at the skin. We know the window for primary prevention may be the first weeks and months of life — before the immune system has locked in its responses. We know that early dietary introduction of common allergens reduces allergy risk substantially. We know that maintaining gut microbial diversity appears protective. We know that skin barrier repair may interrupt the Atopic March before it starts.

None of this means we’ll eliminate food allergy. But moving even part of the burden from management to prevention would change the lives of millions of families who currently spend their days reading ingredient labels, carrying epinephrine, and fielding calls from school nurses.

I think about the families I see in clinic. I think about the child who ate at the lunch table alone because the cafeteria was too risky. The teenager who stopped going to birthday parties. The parent who hasn’t taken a vacation in four years because foreign kitchens are too unpredictable. These are not small inconveniences. They are the texture of a life shaped around an immune error.

The biology behind that error is now more understood than at any point in history. The treatments are improving. The prevention framework is being built.

That doesn’t fix this week’s dinner. But it means the trajectory is going somewhere better.

If you found this useful, share it with a family managing food allergy. The most dangerous myth in this space is that nothing can be done.

Sources: Nature (2023), Science (2025), Frontiers in Immunology (2025), NCBI StatPearls (2025), Journal of Allergy and Clinical Immunology, CDC NCHS Data Brief, World Allergy Organization Journal, ScienceDaily, Begin Health/ERC clinical reviews.

WSJ article [paywall]

Yes, Western pharma spent nearly $5.6 billion licensing Chinese drug candidates last year. Yes, China now represents 30% of the global experimental pipeline. But before we declare a dynasty change in biopharma, let’s ask a harder question:

Is this innovation — or sophisticated iteration?

Look closely at what Chinese biotechs are actually doing. They are extraordinarily good at:

• Scanning Western patent filings and published research within days of release

• Identifying structural wrinkles and freedom-to-operate gaps

• Running efficient, low-cost chemistry to produce “me-better” variants

• Moving rapidly through clinical stages using China’s regulatory shortcuts and patient pools

That is genuinely impressive execution. But it is not the same as discovering a new mechanism, identifying a novel target, or taking a truly first-in-class molecule from concept to clinic. Reverse engineering at scale is a business achievement — not a scientific one.

The molecular glue story in the article makes this plain: Chinese scouts analyzed a Novartis paper, published a how-to guide within four days, and biotechs went to work improving on what Western companies had already invented. That’s speed and efficiency. It is not leadership.

The domestic paradox no one is discussing.

China is producing medicines it largely cannot afford to buy. The high-cost biologics, ADCs, and GLP-1 therapies flowing out of Chinese labs are priced for Western markets — not for the 1.4 billion people at home. A nation building an export-dependent biotech sector on drugs its own population cannot access is not a model of strength. It is a structural vulnerability.

The real risk isn’t losing to China. It’s becoming dependent on it.

US biotechs licensing Chinese “me-better” compounds get a short-term pipeline win — cheaper assets, faster timelines. But over time, this pattern could hollow out early-stage domestic R&D, create supply chain dependencies, and train the industry to reach for the easy incremental gain rather than do the harder, riskier work of genuine first-in-class innovation.

The US doesn’t need to fear Chinese biotech dominance. It needs to avoid outsourcing its scientific ambition to it.

China has world-class scientists, enormous capital, and real government commitment. When it chooses to invest in true de novo discovery — not just optimizing what others have found — it will be a formidable force. Until then, calling this an impending takeover of Western biopharma is flattering to China and alarmist to everyone else.

The WSJ piece was a good story. It was not the whole story.

#DrugDiscovery #Biotech #Pharma #ChinaBiotech #Innovation #LifeSciences #BiopharmStrategy #MediumTerm

Yet for every child who has received a therapy like this, many more have not. Companies that pioneered these one-time gene therapies have struggled commercially—some withdrawing products despite documented efficacy. The science is ready. The system for approving and delivering these therapies is not.

A paper published in Nature Medicine by Abou-el-Enein and colleagues at USC, UCSF, and UCLA describes a practical approach to fix this. Funded by the Advanced Research Projects Agency for Health (ARPA-H), the UNICORN system—Unifying Cell Therapy Outcome Prediction and Regulatory Navigation—is designed for exactly the situation where conventional drug development fails: when the trial has one patient, or three, or ten.

The core problem

Drug approval rests on three pillars: manufactured product quality, evidence of efficacy, and regulatory confidence. For large trials, these support each other naturally. Hundreds of patients generate statistical power. Manufacturing at scale smooths batch variability. Regulators see enough data to make confident decisions.

Rare and ultra-rare pediatric therapies break all three pillars at once. When a product is made in small batches—sometimes for a single child—donor variability and batch-to-batch differences directly affect quality, but there aren’t enough cases to characterize how. Clinical endpoints that work in large trials either don’t exist for rare diseases or don’t occur quickly enough to guide approval decisions. And regulators, charged with ensuring safety and efficacy, face an impossible ask: apply evidentiary thresholds built for populations of thousands to studies of one.

The result is a paradox. We can design a therapy for an individual child in months. We cannot reliably tell a regulator whether that therapy will work.

What UNICORN does, practically

The system has three connected parts.

Product signatures. A spectral flow cytometry panel developed at USC captures phenotypic, metabolic, and functional features of each cell therapy product. Instead of relying on a handful of traditional potency assays, this multi-parameter platform generates a high-dimensional product “signature”—a detailed fingerprint of each batch. The panel can be adapted to different products and run consistently across sites. This matters because it creates a common, comparable language for product quality across small-batch therapies that would otherwise be evaluated in isolation.

AI-driven outcome models. Machine learning models are trained on accumulated cases to identify correlations between product signatures and clinical outcomes. As more children are treated and more data are added, the models learn which product characteristics predict therapeutic benefit—and which predict risk. The point isn’t to replace human judgment. It’s to make the limited data that exists work harder. A model trained on 30 prior cases can say something meaningful about case 31. A regulator reviewing case 31 in isolation cannot.

Regulatory decision support. The third component translates model outputs into practical tools for regulatory conversations: lot-release criteria that define what a product must look like before it proceeds to clinical use, and evidence thresholds calibrated to what is realistically achievable when patient numbers are very small. The FDA has already signaled a willingness to adapt—its guidance on individualized antisense oligonucleotide therapies exemplifies this flexibility. UNICORN gives that flexibility an operational structure.

The system is designed to get stronger with use. Longitudinal sampling—repeated measurements before and after treatment, rather than single timepoints—multiplies the informational value of each case. Each new patient, product batch, and outcome dataset refines the models. What starts as a small evidence base grows into something that can meaningfully support the next approval decision.

Challenges and real limitations

The authors are direct about what this system cannot guarantee, and they deserve credit for it.

Biology won’t always conform to modeling assumptions. No single product signature is likely to fully predict outcomes across different diseases or treatment centers. Early datasets will be sparse and biased toward conditions already being treated—meaning less common ultra-rare diseases may be underrepresented until more cases accumulate. Unmeasured confounders and center-specific practices will influence both product signatures and patient outcomes in ways no model can fully account for. Some correlations that look predictive in retrospect may not hold going forward.

These are genuine constraints, not rhetorical caution. A system trained on small datasets is making probabilistic inferences under real uncertainty. The appropriate response is to build more rigorously and to be honest about what the current evidence can and cannot support.

Why this still represents real progress

The current alternative—applying population-trial standards to n-of-1 therapies—has already failed in practice. Products with documented clinical benefit have been pulled from the market. Children who could be treated are not. The status quo has its own costs, and they are measured in lives.

UNICORN’s value is not that it eliminates uncertainty. It is that it gives regulators, manufacturers, and clinicians a principled way to work with the uncertainty that exists, rather than demanding a level of certainty that can never arrive when the patient pool is a single child. A spectral flow cytometry signature contains more information than a traditional potency assay. A model trained on prior cases informs a release decision better than intuition alone. A decision-support tool that makes prior cases visible to the next regulator closes what is now a significant gap.

Each child treated under this system contributes to the evidence base for the next one. That is how rare disease knowledge accumulates when large randomized trials are not possible—incrementally, deliberately, with each case building on the last. The system is designed to learn. That is exactly what medicine needs when the study has one patient.

Source: Abou-el-Enein M et al. “A blueprint to accelerate rare pediatric gene therapy approvals.” Nature Medicine. 2025. doi:10.1038/s41591-025-04115-6. Funded by ARPA-H.

There’s a plant growing in English hedgerows and cottage gardens right now that contains one of the most medically useful — and most dangerous — chemicals in the history of pharmacy. Gardeners grow it for its tall spires of purple bells. Bees love it. Children’s books feature it. And for about 240 years, doctors have been carefully, nervously, extracting its chemistry to keep failing hearts alive.

The foxglove. Digitalis purpurea. Unremarkable to look at. Extraordinary to understand.

What follows isn’t a wellness post. It’s a story about a plant that sits at an almost uncomfortably thin line between medicine and poison — and how researchers in 2025 and 2026 are discovering it might have a completely different career ahead of it, one involving your immune system and some of the hardest-to-treat cancers we know.

How a Birmingham Doctor Changed Everything

The year is 1775. A physician named William Withering is working in Birmingham, England, and one of his patients — a woman with severe edema, the kind where fluid fills the body cavities and the legs swell grotesquely — has run out of options. The standard treatments aren’t working. Withering expects her to die.

Then he discovers she’s recovered. Not through his ministrations. She’d been quietly taking a herbal concoction from a local woman in Shropshire. Withering, who happened to be a gifted botanist as well as a physician, looked at the mixture of twenty-odd herbs and immediately recognized the likely active ingredient.

Foxglove.

He spent the next ten years methodically documenting every case he treated with it. Not cherry-picking the successes — a level of scientific honesty remarkable for his era. He included failures, toxicities, and patients who died. In 1785, he published An Account of the Foxglove and Some of Its Medical Uses, describing 158 patients. Of those, 101 with congestive heart failure got better. His dosing calculations were so precise that modern analysis puts them only slightly below what we use today.

“Truth and Science would condemn the procedure. I have therefore mentioned every case... proper or improper, successful or otherwise.”
— William Withering, 1785

That book is one of the most important documents in pharmacological history. Withering knew it. His portrait — the only one from life — shows him holding a foxglove sprig. His epitaph at Edgbaston Old Church is carved with the plant. The man and the flower became inseparable.

What he didn’t know, and what took another century to understand, was whyit worked.

The Chemistry of the Foxglove

The foxglove produces a class of chemicals called cardiac glycosides — the name a clue to their structure: a sugar molecule attached to a steroid-like core. The plant makes them presumably as a defense against insects. For mammals with hearts, these compounds do something very specific and very powerful.

They block a molecular pump called Na⁺/K⁺-ATPase — an enzyme embedded in every heart muscle cell that normally shuffles sodium out and potassium in. When you block this pump, sodium accumulates inside the cell, causing calcium to flood in. More calcium means stronger, more forceful contractions.

In a weakening heart, that’s the difference between barely pumping and actually pumping.

The main pharmaceutical compounds extracted from foxglove are three siblings with distinct personalities:

Digoxin (from Digitalis lanata, the woolly foxglove): The one most doctors and patients have heard of. A “positive inotrope” — it makes the heart beat harder. It also slows the heart rate by affecting the electrical conduction system, which makes it useful for atrial fibrillation: the fast, chaotic rhythm that affects millions of people.

Digitoxin (from Digitalis purpurea, the common purple foxglove): The older sibling. Similar mechanism, longer-acting, and critically — cleared by the liver rather than the kidneys. This matters more than it sounds, as we’ll get to.

Lanatoside C: A faster-acting glycoside used in acute situations where you need to stabilize a rhythm quickly and can’t wait for slower-clearing agents to take effect.

FIGURE 1 — PLACEHOLDER

Purple foxglove (D. purpurea) vs. Woolly foxglove (D. lanata): leaf morphology comparison, distribution of active glycoside content by plant part, and primary commercial growing regions for pharmaceutical extraction

The Narrow Window

Here’s what makes foxglove so peculiar as a medicine. Its therapeutic window — the gap between “helpful dose” and “harmful dose” — is one of the narrowest in all of pharmacology. The dose that strengthens a failing heart sits uncomfortably close to the dose that causes fatal arrhythmias.

THE NUMBERS

The blood level of digoxin that helps a heart failure patient sits around 0.5–0.9 nanograms per milliliter. Toxicity begins at roughly 2 nanograms per milliliter. For context: a human hair is about 70,000 nanograms. We’re managing people on differences that are invisible to the naked eye.

Withering noticed this in 1785. He described toxicity symptoms with clinical clarity that would be recognized by any emergency physician today: irregular pulse, visual disturbances including a famous yellow-tint to vision, nausea, and ultimately fatal rhythm disturbances. The Victorians called it “foxglove sickness.” Modern medicine calls it digitalis toxicity, and it still kills people who are overdosed or whose kidney function changes unexpectedly.

This toxicity is why, over the last few decades, digoxin prescriptions declined sharply. Newer drugs — beta-blockers, ACE inhibitors, SGLT2 inhibitors — arrived with better safety profiles and more predictable behavior. Digoxin was increasingly treated as a relic of an earlier era.

Then, in August 2025, a clinical trial made a lot of cardiologists look up from what they were doing.

The DIGIT-HF Trial: Digitoxin’s Comeback

The DIGIT-HF trial — a double-blind, placebo-controlled study from 55 sites across Germany, Austria, and Serbia — enrolled 1,240 patients with advanced heart failure and reduced ejection fraction. These were sick patients: most were in NYHA class III (symptomatic at minimal exertion), and all were already on contemporary guideline-directed therapies including the newer drugs that have transformed heart failure treatment.

The question was whether adding digitoxin — not digoxin, but its liver-cleared cousin — would make any additional difference.

Published in the New England Journal of Medicine in August 2025, the answer was yes. Digitoxin reduced the combined risk of death from any cause or hospitalization for worsening heart failure by an absolute 4.6 percentage points over a median follow-up of three years.

The effect size is modest. But what makes it interesting is the population and the mechanism. The benefit held across patients taking modern quadruple therapy — the most aggressive current treatment protocol.

And the lead investigator specifically called out that digitoxin may be particularly useful for patients with impaired kidney function: precisely the population that struggles most with digoxin’s renal clearance requirements.

Because digitoxin is cleared by the liver, its blood concentrations remain stable even as kidney function deteriorates. No complex dose adjustments every time creatinine creeps up. No panic when a heart failure patient’s kidneys — which often go hand-in-hand with the failing heart — start to worsen. For clinicians managing this very common and very difficult combination, that’s a practically meaningful difference.

The Part Nobody Expected: Cancer

Here is where the foxglove’s story takes a turn that would have genuinely astonished William Withering.

The same Na⁺/K⁺-ATPase pump that foxglove glycosides inhibit in heart muscle cells is overexpressed in multiple cancer cell types. When you block it in a cancer cell, the consequences are considerably more lethal than in a normal cell. Cancer cells, with their high metabolic demands and already-stressed physiology, are more vulnerable to the disruption of ionic balance.

Laboratory studies have shown that digitalis-derived compounds are highly cytotoxic — cell-killing — against a range of human cancer cell lines. The mechanism involves triggering apoptosis: the cell’s own programmed self-destruction pathway, which cancer cells have typically evolved to evade. Forcing that pathway back on is something oncologists have been trying to do with many different compounds.

More intriguing is research targeting cancer stem cells — the small subpopulation of cells within a tumor that are most resistant to conventional chemotherapy, and that are thought responsible for relapse and metastasis after treatment appears successful. Early research suggests that cardiac glycosides may have activity against this population that standard chemotherapy misses entirely. The mechanism isn’t fully established, but it involves the same pump inhibition affecting stem cell signaling pathways.

This is preclinical data. It would be dishonest to call it a cancer treatment. But it’s enough that multiple research groups are pursuing it — and it intersects with a much larger and stranger story about the immune system.

RORγt: The Molecular Switch Nobody Knew Was There

In 2006, immunologists identified a transcription factor — a molecular switch that controls gene expression — called RORγt (pronounced “ROR-gamma-t”). It turned out to be the master regulator of a class of immune cells called Th17 cells.

Th17 cells produce a signaling molecule called IL-17. When these cells are overactive, they drive inflammatory and autoimmune conditions: rheumatoid arthritis, psoriasis, inflammatory bowel disease. Block RORγt, the logic went, and you suppress Th17-driven inflammation. Pharmaceutical companies began racing to find oral drugs that could achieve what expensive injectable biologics were already doing.

2006

RORγt identified as the master regulator of Th17 cell differentiation — immediately flagged as a high-value drug target for autoimmune disease

2011

A Nature paper reports that digoxin — the foxglove drug — binds to the RORγt ligand-binding domain and blocks Th17 differentiation. A completely separate mechanism from anything happening in the heart

2012–2017

Major pharmaceutical investment in RORγt inhibitors. Clinical trials run against autoimmune indications. Results are underwhelming versus biologics; safety concerns emerge around thymic T-cell effects

2018–2023

The inhibitor program cools off. But separately, researchers begin looking at what happens when you go the other direction — activating RORγt rather than suppressing it

2024–2026

The oncology pivot: RORγt agonists enter early clinical trials to “heat up” cold tumors and make them visible to checkpoint immunotherapy. Digoxin’s role as the molecule that first mapped this receptor comes back into focus

The Flip: From Suppressor to Activator

Researchers had been thinking about RORγt as something to switch off — to dampen overactive inflammation. The oncology pivot asked a different question: what if you switched it on?

In cancer, a well-documented problem is the cold tumor: a tumor that has managed to exclude immune cells from its microenvironment, making itself essentially invisible to the body’s defenses. Checkpoint inhibitors — the cancer drugs that have transformed oncology over the last decade — work best in tumors already infiltrated by immune cells. Cold tumors don’t respond nearly as well. And many common cancers run cold.

RORγt agonists — molecules that activate rather than block the receptor — appear capable of recruiting immune effector cells into cold tumor environments. The Th17 pathway, when activated appropriately in a tumor context, can drive an immune infiltration that checkpoint inhibitors then amplify. You warm the tumor up first, then send in the blockers.

This is not science fiction. A first-in-class synthetic RORγt agonist — cintirorgon (LYC-55716) — completed a Phase 1 trial in patients with relapsed or refractory metastatic cancers. No dose-limiting toxicities occurred among the 32 enrolled patients across a range of doses, and early antitumor signals were observed.

The foxglove connection here is structural and mechanistic rather than direct. Digoxin and its derivatives were first used to map the RORγt binding site and to demonstrate its functional importance. The synthetic agonist programs that followed built on that knowledge. Whether digitalis-derived molecules themselves will be developed as RORγt agonists remains an open question — the margin between their cardiac effects and their immunological effects requires careful pharmacological navigation — but the research is active.

The Dose Makes the Poison — and the Medicine

What connects all of these threads is a concept that goes back to the 16th-century physician Paracelsus: sola dosis facit venenum. The dose alone makes the poison.

Foxglove at high concentrations: cardiac toxicity, fatal arrhythmias, death.

At therapeutic concentrations in a failing heart: life-saving inotropy and rate control.

At lower concentrations affecting RORγt: immune modulation with potential autoimmune and oncologic applications.

At concentrations affecting cancer cell pumps: potential cytotoxicity against tumor cells that overexpress the target.

The same molecule. Different doses. Wildly different biological conversations.

This is a recurring pattern in pharmacology, but foxglove makes it almost embarrassingly legible. It’s a chemical with so many conversations going on with the human body that we are still — 240 years after Withering’s book — finding new ones.

The Flavonoids — The Part Nobody Talks About

The cardiac glycosides get all the attention, but foxglove leaves contain other compounds that researchers are only now beginning to characterize seriously.

One is scutellarein, a flavonoid with antioxidant properties entirely separate from any cardiac effect. Early research suggests anti-inflammatory activity through mechanisms distinct from the glycoside pathway. It’s far too early to say where this leads, but it illustrates a broader point: plants are not single-compound factories. The foxglove leaf is a complex biochemistry experiment, and we’ve mostly been studying one product line.

Traditional medicine systems — particularly in parts of India — have used topical preparations of digitalis glycosides in ointments for severe burns, where the compounds are thought to stimulate local circulation and aid tissue healing. This sits outside mainstream Western pharmacology and hasn’t been rigorously studied in clinical trials. But it’s not implausible given what we understand about vascular effects. It’s a loose thread, and loose threads sometimes lead somewhere.

What We Know, What We Don’t

The honest position on foxglove in 2026:

Well-established: Digoxin and digitoxin are real medicines with real cardiovascular applications. The DIGIT-HF trial provides solid evidence that digitoxin reduces hospitalization and death in certain heart failure patients, particularly those with renal impairment. Both carry a genuinely narrow therapeutic window requiring careful clinical management.

Promising but not proven: Anticancer applications of cardiac glycosides are an active and intriguing area of lab research. The findings are consistent across multiple cell line studies. Clinical validation in humans is limited.

Scientifically coherent but speculative: The path from “digoxin binds RORγt” to “foxglove-derived molecules treat cold tumors” runs through a lot of unsolved pharmacological and safety problems. The synthetic agonist programs inspired by this discovery are further along than the digitalis-derived work itself.

Withering’s own instinct, looking at his 1785 data, seems right for 2026 as well: document everything, claim only what the evidence supports, and leave room for what comes next.

The foxglove has been growing in the hedgerows for millions of years. Its chemistry has been tested in folk medicine for centuries, in systematic clinical trials for 240 years, and in molecular biology laboratories for the last few decades. Every time we think we understand what it does, it shows us another room.x

That’s either a sign of a very strange plant, or a sign of how much we still have to learn about the cells we’re made of.

SOURCES & FURTHER READING

• Bavendiek U et al. “Digitoxin in Patients with Heart Failure and Reduced Ejection Fraction.” N Engl J Med2025;393(12):1155–1165.

• Karaś K et al. “Digoxin, an Overlooked Agonist of RORγ/RORγT.” Frontiers in Pharmacology 2018;9:1460.

• Withering W. An Account of the Foxglove and Some of Its Medical Uses. Birmingham: M. Swinney, 1785.

• Mahalingam D et al. “Phase 1 Open-Label, Multicenter Study of First-in-Class RORγ Agonist LYC-55716 (Cintirorgon).” Clin Cancer Res 2019;25(12):3508–3516.

• “Digitalis – from Withering to the 21st century.” British Journal of Cardiology, August 2024.

• Nature Reviews Cardiology. “Benefit of digitoxin therapy for HFrEF.” Vol. 22, p. 842, September 2025.

That is not a coincidence. It may be a warning.

A paper just published in PNAS — “Is Ockham’s Razor Losing Its Edge?” by Dubova and colleagues — makes a careful, well-sourced case that science’s oldest methodological shortcut is now working against us in some of the places that matter most. The argument isn’t that simple models are bad. It’s that treating parsimony as a universal principle, rather than a context-dependent tool, leads to measurable scientific error.

The heuristic and where it breaks

William of Ockham’s principle has survived 700 years because it works in physics. Fewer assumptions usually mean fewer places to go wrong. In clean experimental systems, this holds.

Biology is not a clean experimental system.

The PNAS paper distinguishes two forms of parsimony: parsimony by constraint (models that make narrow, specific predictions) and parsimony by components (models with fewer variables). These often conflict. And in living systems — where causal chains are long, nonlinear, and context-sensitive — both forms can produce models that are technically elegant and factually wrong.

The authors describe a striking example from neuroimaging. Simple models applied to live brain scans consistently inferred oscillatory patterns — rhythmic back-and-forth activity — that simply weren’t there. The brain wasn’t oscillating. The model was. The simplifying assumptions imposed a false pattern onto data that told a different story.

The double descent problem

The conventional statistical wisdom: more parameters relative to data means more overfitting. Fit a complex model with too few data points and you’ve memorized noise.

Recent work has overturned this. Researchers found that prediction error follows a U-shaped curve as model complexity increases — it rises, peaks, then falls again as parameters grow large. This “double descent” means highly overparameterized models, trained with methods like gradient descent, can achieve both low bias and low variance. The theoretical case for reflexive parsimony — keep it simple so you don’t overfit — no longer holds unconditionally.

Dubova et al. document a study on moral judgment that makes the implication concrete. Researchers built a large machine-learning model trained on 40 million moral decisions. The model was opaque and complex. Once it identified structure in the data, that structure was distilled into a simpler psychological theory — one that would have been invisible to a researcher who started with a parsimonious model from the beginning. The complexity came first. The clarity followed.

What this means for drug development

Most disease models in clinical development start simple by design. Single target, single mechanism, linear causal chain. This is not just a scientific choice — it’s a regulatory and communicative one. Simple models are easier to test, explain, and fund.

Biological disease processes are not simple. Autoimmune disease, cancer, neurodegeneration — these involve gene networks, environmental context, cell states, and feedback loops that no parsimonious model captures well. When a drug fails in Phase III, the question worth asking isn’t always “did the molecule work?” It’s often: “was the disease model right in the first place?”

The PNAS paper suggests a different workflow: start with the most complex model the data supports, find structure, then distill it. Use ensemble approaches rather than forcing a single explanatory account. Accept that the right model for discovery may not be the right model for communication — and treat those as separate problems.

This matches what is already happening at the edges of the field. AlphaFold changed structural biology not by simplifying the problem but by refusing to. Climate science now uses multi-model ensembles rather than a single parsimonious forecast. Quantum biology — a topic I wrote about last week — finds that quantum effects in enzyme function and energy transfer cannot be adequately described by classical approximations, no matter how clean those approximations look on paper.

The line worth drawing

Occam’s razor is not wrong. It is misused.

Applied to physics, instrument calibration, or clinical communication, it works well. Applied as a prior assumption about how biological systems are organized, it produces models that are internally consistent and externally misleading.

Francis Crick said it plainly: “Biologists must constantly keep in mind that what they see was not designed, but rather evolved.” Evolution does not favor elegance. It favors survival. Surviving systems are redundant, layered, and contingent in ways no parsimonious account captures.

Seven hundred years is a long run for a heuristic. The question isn’t whether to retire it. The question is whether we’ve confused a working shortcut for a scientific law — and how many wrong models we’ve built in the meantime.

Eswar Krishnan, MD
Drug Development Executive

This should not be considered to be specific medical or financial advice.

#DrugDiscovery #Biotech #ClinicalDevelopment #QuantumBiology

Sources:

• Is Ockham’s razor losing its edge? | PNAS

• Is Ockham’s razor losing its edge? | PMC Full Text

• Cutting Down Ockham’s Razor | OpenMind Magazine

• Sharp look into Ockham’s razor | ScienceDaily

Quantum biology studies whether quantum-mechanical effects — coherence, tunneling, entanglement — actively drive biological processes rather than merely coexist with them. That distinction is what the field is now trying to pin down rigorously.

Two examples that are hard to explain any other way

Photosynthetic light-harvesting complexes achieve near-perfect energy transfer efficiency. Classical physics cannot account for this fully. Quantum coherence — excitations existing in superposition across multiple molecular states simultaneously — may be what makes the difference. Fleming’s own lab produced key evidence here, including work on long-lived quantum coherence in photosynthetic complexes at physiological temperature.

Enzyme-catalyzed hydrogen transfer is the more immediately relevant case for drug hunters. Soybean lipoxygenase produces a kinetic isotope effect of roughly 80. Classical transition state theory predicts a maximum of about 7. The gap is too large to explain any other way: the hydrogen passes through the energy barrier rather than over it. Quantum mechanics isn’t incidental. It’s the mechanism.

What this means for computational drug design

Classical models approximate molecular interactions through force fields and statistical mechanics. They’re useful and fast. But if enzymes routinely exploit proton tunneling — and early evidence suggests many do — then classical simulations of those enzymes are working with an incomplete physical description of the active site.

This is already driving the adoption of QM/MM methods in enzyme-targeted drug design. Quantum mechanics/molecular mechanics hybrid approaches are slower and more expensive, but they capture what classical models miss: the quantum behavior of hydrogen transfer, proton-coupled electron transfer, and possibly receptor-ligand binding in certain systems. A 2025 De Gruyter review described this as a shift in how the field treats enzyme catalytic efficiency — no longer optional to model quantum effects when designing drugs against these targets.

The three open questions Scholes and Fleming pose

First: can we build experimental probes sensitive enough to detect quantum effects inside living systems — not just purified complexes under controlled lab conditions? In vivo detection remains unsolved.

Second: does biological machinery genuinely exploit quantum effects, or do they happen to occur while biology proceeds independently? “Quantum-assisted” and “quantum-driven” are not the same claim. The computational tools we need, and the drug design implications, are different depending on the answer.

Third: how does quantum coherence survive long enough to matter? Living systems are warm, wet, and noisy — conditions that should collapse quantum states almost instantly. That they apparently don’t, at least in some systems, is the puzzle that makes the whole field worth watching.

The state of the field

This is not fringe science. The experimental foundations are real — ultrafast spectroscopy, kinetic isotope measurements, cryo-EM combined with quantum chemical calculations. What’s missing is consolidation: shared definitions, agreed experimental standards, and a cleaner taxonomy of which biological systems are genuine candidates for quantum effects.

Scholes and Fleming are calling the field to that work. For anyone in drug discovery, the practical signal is this: if the quantum layer of enzyme function is real and measurable, then target engagement models built entirely on classical assumptions are incomplete. That’s worth knowing before you commit to a computational platform for a difficult enzyme target.

The paper is open access at PNAS.

Sources: Scholes & Fleming, “What is Quantum Biology?” PNAS (2026), https://www.pnas.org/doi/10.1073/pnas.2531134123 | Fleming et al., long-lived quantum coherence in photosynthetic complexes, PNAS (2010) | “The quantum revolution in enzymatic chemistry,” De Gruyter, pac-2025-0500 | “New Insight into Quantum Mechanical Hydrogen Tunneling in Enzymes,” Biochemistry ACS (2025)

The High Cost of Euclidean Thinking

There is a particular kind of silence that fills a boardroom about fourteen months into a Phase 3 trial. The enrollment forecast has been revised three times. The “Site Rescue” task force has stopped asking why sites are behind and started negotiating how much the timeline must shift. In this room, everyone is an expert. The sites are world-class, the CRO is reputable, and the clinical leads are tireless.

Yet the trial is in trouble—structurally, irreversibly.

The traditional response is to double down on “linear” solutions: hire more CRAs, add more sites, or increase the recruitment budget. This approach assumes that a clinical trial is a Euclidean machine—a system where inputs and outputs share a proportional, straight-line relationship. In the Euclidean worldview, if you are 10% behind, you simply need 10% more effort to catch up.

But clinical trials are not machines. They are non-linear dynamical systems. They are iterative feedback loops where the output of Month 1 (enrollment, safety signals, site engagement) becomes the input for Month 2. In mathematics, when you iterate a non-linear function, you don’t get a straight line; you get a fractal.

If your trial is too slow, it isn’t because you lack “effort.” It’s because your operational parameters have drifted into a region of mathematical chaos. To understand how to navigate this, we must look at the geometry of fate.

I. Gaston Julia and the Ghost in the Execution

The History of the Leather Mask

In 1918, amidst the ruins of post-WWI France, a mathematician named Gaston Julia published a 199-page masterpiece on the iteration of rational functions. Julia was a war hero who had lost his nose in the trenches; he wore a leather mask for the rest of his life. Working without computers, Julia (and his contemporary Pierre Fatou) visualized a world where simple equations could produce boundaries of infinite complexity.

For a fixed protocol configuration (a constant c), the Julia set is the boundary of the set of starting points z0 that do not escape to infinity under the iteration:

The Clinical Parallel: Sensitivity to Initial Conditions

In this equation, z0represents your starting conditions. This is the specific sequence of your first ten sites, your initial patient cohort, and your first regulatory green light.

1. Connected Julia Sets (Robustness): When your protocol (c) is stable, the Julia set is a “connected” shape. This means the outcome is robust. Whether Site A or Site B activates first, the system stays within the “basin” of success.

2. Cantor Dust (The Fragmented Trial): If the protocol is slightly off, the Julia set shatters into “dust.” Success becomes hyper-sensitive to z_0. This is the Butterfly Effect of operations. If your high-enrolling site opens on a Tuesday, you succeed. If they open on a Wednesday, the whole program “diverges to infinity” (fails).

Most “failed” trials were actually “Cantor Dust” executions. The protocol was so fragile that only a perfect, impossible sequence of events could have led to a positive readout. We blame “bad luck,” but the math tells us the failure was baked into the starting geometry. Here is a graphic from a recent blog post

II. Pierre Fatou and the Gravity of Failure

The Discovery of Attractors

While Julia looked at the boundary (chaos), his contemporary Pierre Fatou—working in isolation at the Paris Observatory—explored the interior. Fatou was arguably the first to understand Basins of Attraction.

The Math: The Fatou Set (F(f))

The Fatou set is the region where behavior is stable. Within it lie attractors. An “attractor” is a state toward which the system naturally evolves. If a point z is in a basin B(L), then:

The Clinical Parallel: The Three Basins of Fate

A clinical trial is essentially a point moving through a high-dimensional space, pulled by three primary “Fatou Basins”:

1. The Success Basin: Operational and biological variables converge toward a clean, statistically significant readout.

2. The Futility Basin: The noise of the data eventually overwhelms the signal, regardless of how many patients you add.

3. The Toxicity Basin: The system “escapes” to a state of unacceptable risk.

The Insight: Most trials fail because they are launched in the “Julia Set”—the razor-thin boundary between basins. In the Julia set, the trial is unstable. A single outlier patient can knock the entire program out of the Success Basin and into the Futility Basin. Fatou’s theorem in mathematics teaches us that stability is a property of the initial coordinates, not just of the effort applied during the journey.

III. The Mandelbrot Set: The Map of the Possible

The History: IBM and the Thumbprint of God

Sixty years after Julia and Fatou, Benoit Mandelbrot used early computer graphics to plot all the possible Julia sets on a single graph. He discovered the most complex object in mathematics: the Mandelbrot Set.

The Math: Mapping Parameter c

The Mandelbrot set is the set of all protocol configurations (c) for which the Julia set is connected. It is defined by iterating from z_0 = 0:

The Clinical Parallel: Protocol Topology

Every point c on the map represents a different combination of inclusion criteria, primary endpoints, and dosing regimens.

• The Black Heart: If your protocol (c) is in the middle of the set, you are safe. You have built a “connected” path to success.

• The Seahorse Valleys: If you are on the “fringes”—the “Seahorse Valleys” of the Mandelbrot set—your trial is technically viable but infinitely complex. A tiny change in the competitive landscape or a minor regulatory tweak moves you out of the set, and your trial “diverges.”

The failure of modern drug development is that we spend all our time managing z_n (the execution) and zero time calculating where our c (the protocol) sits on the map of stability.

IV. Fractal Time and the Coastline Paradox

The History: Richardson’s Ruler

In the 1950s, Lewis Fry Richardson noticed that the length of the border between Spain and Portugal changed depending on who was measuring it. If you use a 1-km ruler, the border is short. If you use a 1-meter ruler, the border “grows” because you are now measuring every rock and inlet.

The Math: The Fractal Dimension (D)

The measured length L follows a power law:

As the ruler ‘epsilon’ gets smaller, the length L increases.

The Clinical Parallel: Why the Last 10% Takes 50% of the Time

At the start of a trial, we measure progress with a “large ruler” (e.g., “Open 50 sites”). It looks simple. But as we approach Database Lock, we switch to a “microscopic ruler” (e.g., “Reconcile the SAE on Patient 104-02”).

As the “ruler” gets smaller, the amount of work increases exponentially. This is why the final 10% takes 50% of the time: you have increased the resolution of the project, and in doing so, you have discovered the infinite complexity of the fractal edge. This is Fractal Time. You aren’t “almost done”; you are simply looking closer at an infinite coastline. To a Euclidean manager, this appears to be a delay. To a Fractal manager, this is just the nature of closing a complex system.

V. Other Applications: The Fractal Body and Pink Noise

The reason non-linear dynamics are so relevant to medicine is that the human body itself is a fractal masterpiece. 1. Fractal Physiology: Our lungs and vasculature are branching fractals.

This allows massive surface area (for oxygen exchange or drug absorption) to fit into a finite 3D volume. When we dose a drug, we aren’t dosing a “cylinder”; we are dosing a branching network. If the drug’s diffusion constant doesn’t match the fractal dimension of the tissue, the Phase 3 trial will fail despite Phase 1 success.

2. Oncology and Chaos: Tumor growth is non-linear. The “Fractal Dimension” of tumor margins is now used as a biomarker. Smooth margins (low D) imply stability; highly fragmented margins (high D) imply aggressive, chaotic invasion.

3. Financial “Fat Tails”: Mandelbrot proved markets follow Power Laws, not Bell Curves. Biotech portfolios suffer from “fat tail” risks where “impossible” crashes happen far more often than linear models predict. Trials don’t just “underperform”; they “jump” from one Fatou basin to another.

Conclusion: The “So What” – The Operational Topologist

The lesson of Fatou, Julia, and Mandelbrot is that boundaries are more important than averages.

In clinical development, we are obsessed with averages: average enrollment, average p-values, average site performance. But averages only matter in linear systems. In a non-linear system, the only thing that matters is where the boundary lies.

1. Calculate Parameter Distance: Before spending $200M, ask: “How close is our protocol (c) to the Mandelbrot boundary?” If a 10% change in screen failure rate pushes you into the “Chaos Zone,” your protocol is structurally flawed.

2. Map the Basin: Stop asking if sites are “doing a good job.” Ask if the system has fallen into a “Futility Basin.” If the feedback loops are generating more queries than they are resolving, you are in a divergent orbit.

3. Force a Lower Resolution: To escape the “Coastline Paradox” at the end of a trial, you must stop shrinking your ruler. You must define a “closure resolution” (e.g., frozen data snapshots) early, or you will be measuring the coastline forever.

4. Identify the Branch: In a fractal system, a problem at the “Patient Level” is usually just a high-resolution version of a “Leadership Level” failure. Zoom out. Fix the branch, not the leaf.

The future of drug development belongs to the Operational Topologist. They will understand that a trial doesn’t fail because people didn’t work hard enough. It fails because it was placed in a region of the map where failure is the only mathematical attractor.

Know your boundary, or the boundary will find you.

“Beautiful and chaotic things exist on the same boundary. The job of the executive is to keep the program in the heart.”

Nobody in the room is incompetent. The sites are not badly run. The CRO is not negligent. The data management team is not behind. And yet the trial is in trouble — structurally, irreversibly, in a way that more site visits and more patient recruitment support will not fix.

The reason is geometric. The trial entered its execution phase with operational assumptions that were sitting near a mathematical boundary — a threshold below which normal variation in execution is absorbed, and above which small additional degradations produce a catastrophic, irreversible cascade. Nobody identified that boundary before the trial started. Nobody monitored the distance to it during execution. By the time the meeting happens, the program crossed it six months ago.

This essay is about that boundary. What it is, how to find it before it finds you, and what operational management looks like when you take it seriously.

Part One: Three ideas in plain English

Fractals: why the same problem exists at every level

A fractal is a system where zooming in reveals the same pattern you saw when zoomed out. The canonical image is a tree: the trunk splits into major branches, each branch splits into smaller branches, each smaller branch splits into twigs, each twig ends in something that looks like a tiny tree. The structure repeats perfectly at every scale.

A clinical trial is also a fractal. The operational hierarchy — Program, Region, Country, Site, Patient — has the same structure at every level. And more to the point, the same set of operational problems appears at every level: activation delays, resourcing gaps, vendor dependencies, compliance risk, and communication lag. The pattern repeats.

This has one specific and uncomfortable implication. In a fractal system, a failure at an upper level does not produce one problem. It produces N problems simultaneously, where N is the number of nodes below it. A protocol amendment at the program level does not change one thing. It touches every region, every country, every site, and every patient currently in the trial — simultaneously. The amendments arrive at different teams at different times and appear unrelated. They are not. They share a common origin.

Why this matters operationally

When multiple sites show the same failure pattern simultaneously, the correct question is not ‘what is wrong with these sites?’ It is ‘what upstream node failure is generating this pattern across the network?’ The answer changes both the intervention and the urgency.

Figure 1 · Fractal Structure of a Clinical Trial

The fractal structure of a clinical trial. Program → Region → Country → Site → Patient. The same operational challenges repeat at every level. A disruption at the program level creates N simultaneous downstream problems — one per node below it. Self-similarity means a regional-level signal is a leading indicator of site-level failures that have not yet surfaced.

The Mandelbrot set: the map of operational stability

The Mandelbrot set is, in essence, a map of all possible starting configurations — showing which ones produce stable, bounded outcomes and which ones spiral out of control. The boundary between the two zones is the mathematically interesting part: it is infinitely complex, and a single step on either side determines everything.

You do not need mathematics to use this idea. Every clinical trial has a composite set of operational assumptions: site activation pace, screen failure projections, CRO capacity estimates, country regulatory timelines, dropout assumptions, IP shelf-life relative to the enrollment window. Think of this composite as a single point on a map. The Mandelbrot set tells you whether that point is in the stable zone (where normal execution variance is absorbed and the trial delivers), near the boundary (where modest degradation in any single parameter flips the trial to unrecoverable), or in the escaped zone (where no operational excellence can fix the structural incompatibility between the protocol and the execution plan).

The practical observation is this: most programs that enter execution in trouble are not in the escaped zone. Their operational plan was not fundamentally broken. It was near the boundary — sitting in a region where a screen failure rate drifting from 2:1 to 3:1, combined with a single key country running two regulatory cycles behind, combined with a modest CRO staffing reduction, was sufficient to push the composite past the threshold. None of those three events triggered an alarm individually. Together, they crossed the boundary.

The question of feasibility never asks

Standard feasibility asks: are our site and enrollment assumptions reasonable? The right question is: how much degradation across our assumptions is needed to make the trial unrecoverable — and are we comfortable with that margin? A parameter can pass every reasonableness check and still be sitting at the operational boundary.

Figure 2 · The Mandelbrot Set – operational parameter space

The Mandelbrot set as an operational stability map. Each pixel = one combination of operational parameters. Dark interior: orbit stays bounded — trial is robust. Colored boundary zone: orbit escapes slowly — trial is near its operational fragmentation threshold. Most operational failures originate here. Marker A (stable), B (boundary), C (escaped) illustrate three program positions.

Julia sets: how sensitive is your trial to starting conditions?

For each specific operational configuration — each position on the Mandelbrot map — there is a corresponding Julia set: a picture of which starting conditions (which sites activate first, which patients enrol early, which country comes online when) lead to trial success and which lead to failure.

When the operational composite is well within the stable zone, the Julia set is a single connected region. A wide range of starting conditions all lead to the same place: trial success. Whether Site 01 activates first or Site 58, whether Germany comes online in month 3 or month 7, the trial delivers. The outcome is robust to execution sequence.

When the composite is near the Mandelbrot boundary, the Julia set fractures into a lacy, thread-thin structure riddled with gaps. Now it matters intensely which sites activate first. If the high-enrolling sites happen to activate late and the low-enrolling sites fill the early cohort, the program enters a recovery trajectory it cannot complete within the timeline — even with full rescue operations. The same protocol, the same sites, a different activation sequence: different outcome. This is not bad luck. It is the structural property of a fragmented Julia set.

When the composite is outside the boundary, the Julia set fragments into disconnected dust. There is no connected basin of successful starting conditions. Some windows may briefly enrol adequately. The overall system cannot sustain itself.

The clinical translation

When a Phase 2 program produces the result ‘it worked in one subgroup analysis but not the overall population,’ the team typically looks for an operational explanation. Often, there is no satisfying one. The more accurate explanation is geometric: the program was running in the fragmented zone of its Julia set. Different patient cohorts were enrolled in different sequences, ending up in different disconnected regions of the outcome space. This is not noise. It is structure.

Figure 3 · Julia sets –– 3 operational fates

Julia sets — three operational fates. Left (connected): robust site network, wide basin of successful starting conditions. Center (fragmented): outcome highly sensitive to site activation sequence — same protocol, different timeline depending on which sites activate first. Right (Cantor dust): structural operational failure — no execution path leads to success.

Part Two: What to actually do

The framework generates four specific operational responses, each timed to a distinct phase of program management.

Stage 1 — Diagnose: map the parameter composite before the trial starts

The diagnosis happens at operational planning, before the first CTA submission. For each high-branching operational parameter — the ones whose failure cascades to multiple levels simultaneously — calculate two numbers: the current planned value, and the value at which the failure becomes irreversible. The distance between them is what matters.

This is not a standard risk register. Risk registers list risks and their probabilities. A parameter distance analysis maps how far each assumption sits from its fragmentation threshold. A parameter can have a low probability of failure and still be near the boundary — if the threshold is close to the planned value, normal variation is enough to cross it.

The parameter distance table

For each high-branching parameter:

• What is the planned value?

• What is the failure threshold — the value at which the cascade becomes irreversible?

• What is the distance between them?

• What is the typical variance for this parameter in comparable programs?

• Is the variance larger or smaller than the distance?

Parameters where the typical variance is roughly equal to the distance — or greater — are Amber-zone parameters. They need structural mitigation before the trial begins. Parameters with large distance-to-variance ratios are in the stable interior. Standard monitoring is sufficient.

Figure 4 · Paramater distance

The parameter distance table — the practical instrument. Six named Phase 3 parameters with planned value, failure threshold, distance, and variance. Screen failure rate and country activation are Amber-zone: typical variance is within the range of the distance to failure. These require structural mitigation at the design stage, not monitoring improvements.

Stage 2 — Prevent: build structural margin into boundary-zone parameters

Prevention happens at protocol design and vendor contracting. For each Amber-zone parameter, the response is structural — not a monitoring improvement, but a design change that increases the distance to the fragmentation threshold.

Screen failure rate

If screen failure is an Amber parameter, the response is over-siting: activate 25–30% more sites than the base-case model requires, specifically targeting sites with documented high-volume screening capacity in this indication. The cost is higher activation spend. The cost of not doing it is a mid-trial boundary crossing with no recovery path.

Country activation

If country activation is Amber, the response is country redundancy: identify two additional countries that can be added within 90 days if any primary country misses its critical activation milestone. These contingency countries should be at IND-ready status before primary country submissions begin. This is deliberate Julia set engineering — expanding the connected basin of viable execution outcomes so that a wider range of country activation sequences all lead to on-time enrollment.

Vendor capacity

If CRO monitoring capacity is Amber, the contractual response is a minimum staffing clause with a financial penalty for falling below the committed CRA-to-patient ratio, plus a pre-negotiated right to bring in a backup CRO within 60 days if the clause is triggered. This keeps the parameter in the stable interior even if the primary vendor experiences staffing disruption.

Stage 3 — Mitigate: classify before you intervene

When operational problems surface in a running trial, the first question is classification, not intervention. Is this a leaf failure or a branch failure?

• A leaf failure is localized to a specific node — one site, one country, one vendor interaction. It has limited downstream spread. Local intervention works.

• A branch failure has its origin at a high-branching node, and produces the same class of problem at multiple leaf nodes simultaneously. Local intervention does not work because the cascade from the upstream node continues arriving at new leaves faster than you can remediate existing ones.

The diagnostic test is structural: do the failing nodes share a common upstream cause? Seven geographically dispersed sites, all reporting high screen failure rates in the same two-month window, are not seven simultaneous leaf failures. It is one branch failure — the inclusion criteria are eliminating the available patient population, manifesting at seven leaves at once.

Once a branch failure is classified, the intervention moves up the hierarchy: protocol amendment, inclusion-criterion modification, CRO contract action, or country portfolio change. This requires more organizational authority and faster escalation than a leaf intervention. The fractal framework argues for keeping that escalation path short and pre-rehearsed, because branch failures require decisions at exactly the moment when organizational pressure is highest to continue with incremental leaf-level rescue.

Pre-committed branch-level triggers

For each Amber-zone parameter identified at Stage 1, specify in advance: the trigger threshold, the response, and the response time. For example, if screen failure exceeds 2.5:1 for two consecutive months, activate contingency sites and convene an I/E review within 30 days. Pre-commitment removes the decision from the moment of maximum pressure to continue.

Stage 4 — Monitor: track distance, not just performance

Standard operational monitoring tracks performance against target: enrollment curve, data completeness, protocol deviation rates, and site performance rankings. These are all leaf-level outputs. They show what has already happened at the leaves.

Parameter distance monitoring tracks the operational composite against its fragmentation thresholds — at the branch level, at sufficient frequency to detect boundary drift before it manifests at the leaves.

The living parameter map

The parameter distance table from Stage 1 becomes a living document, updated monthly for Amber-zone parameters and quarterly for stable-interior parameters. For each parameter, track: current value, trend direction, distance to threshold, and rate of boundary approach. A parameter performing within target but drifting toward its threshold is a leading indicator — it requires attention before it becomes a problem.

Branch-level leading indicators

Because the trial hierarchy is self-similar, a regional-level anomaly is a leading indicator of site-level problems that have yet to surface. A regional screen failure rate running 20% above the trial average is not yet a crisis — individual sites in that region may each be within normal variance. But the regional branch signal indicates that leaf-level failures are imminent. The structure of the system commits them before they appear.

Acting on the regional signal rather than waiting for individual site flags is the difference between early-stage intervention and late-stage rescue. In a program near the Amber boundary, it is sometimes the difference between a recoverable adjustment and an unrecoverable cascade.

The underlying principle

All four stages of this framework rest on one observation that fractal mathematics makes precise:

In complex, branching, self-similar systems — which is what a Phase 3 clinical trial is — the information that predicts failure is structurally available before the failure occurs. It sits at the branch level, weeks to months before it is visible at the leaf level.

The operational task is never ‘what went wrong?’ It is always ‘what branch failure is generating the leaf signals we are currently seeing, and how long ago did it occur?’

The failure rate in Phases 2 and 3 is not primarily a failure of science or execution. It is a failure of parameter awareness — of not knowing, at the moment when it would be actionable, how close the program’s operational assumptions are to the threshold where normal variation is sufficient to produce an irreversible cascade.

The Mandelbrot boundary does not care whether the assumptions were reasonable. It only cares whether the composite lies to its right. Knowing where the boundary is — before the trial starts — is the whole job.

The companion LinkedIn post (Biotech Triallist newsletter) covers the practical tools without the framework — parameter distance table, three-zone classification, and the leaf-vs-branch diagnostic. The figures referenced throughout this piece are available as downloadable PNG files.

Discovery scientists are the border collies. Cytokine pathways, inflammatory cascades, and cell signalling networks are the sheep.

And right now, most of us are stampeding them.

A paper just published in Science Advances by engineers at Georgia Tech — studying actual sheepdogs in an actual competition — has formalised this mathematically. The biological pathways we are trying to control are not waiting to be blocked. They are switching, oscillating, and rerouting continuously. The drug development industry has been designing interventions as if the pathways stand still. They do not.

The biology is not a machine that integrates your signal and responds proportionally. It is an indecisive flock — switching between states, attending to one signal at a time. The scientist who understands that has an enormous advantage over the one who does not.

Read more

We're going to dive into two pivotal papers that illuminate this duality. The first, published in Arthritis & Rheumatology, highlights how defective efferocytosis is a root cause of inflammation in a range of rheumatic diseases.¹ The second, appearing in Science Immunology, uncovers a novel, pro-inflammatory form of efferocytosis, aptly named "efferoptosis," which can wreak havoc in acute inflammatory states like sepsis.¹ Together, these studies paint a nuanced picture, suggesting that the precise modulation of efferocytosis could be a universal therapeutic target.

Paper 1: When the Cleanup Crew Fails – Defective Efferocytosis in Autoimmune Diseases

The Arthritis & Rheumatology paper, "Efferocytosis and its role in rheumatic diseases," lays out the fundamental importance of efferocytosis in maintaining our health.¹ It's a three-phase process:

1. The "Smell Phase": Dying cells release "find me" signals like sphingosine-1-phosphate (S1P) and nucleotides, attracting phagocytes (our cleanup cells) to the scene.¹

2. The "Eating Phase": Apoptotic cells expose "eat me" signals, primarily phosphatidylserine (PS), on their surface. Phagocytes recognize these signals directly or via bridging molecules like MFGE8 and Gas6, then engulf the dying cells.¹

3. The "Digestion Phase": Inside the phagocyte, the ingested cellular material is broken down. This isn't just disposal; it's a metabolic reprogramming that actively produces anti-inflammatory mediators like IL-10 and TGF-β, promoting tissue repair and immune tolerance.¹

This efficient, silent clearance is crucial. It prevents dead cells from undergoing "secondary necrosis," a messy process that spills their contents into the body.¹ These spilled contents, known as damage-associated molecular patterns (DAMPs) – like nucleic acids and histones – are highly inflammatory. They activate immune receptors, triggering a cascade of pro-inflammatory cytokines such as type I interferons (IFN), TNF, and IL-6, which fuel chronic inflammation and autoimmunity.¹

The Autoimmune Connection: A Cascade of Consequences

When efferocytosis falters, this delicate balance is shattered. The Arthritis & Rheumatology paper details how defective efferocytosis is a central mechanism in many autoimmune diseases:

• Systemic Lupus Erythematosus (SLE): This is a prime example. Genetic mutations in complement components (like C1Q) or DNA-degrading enzymes (DNase 1, DNASE1L3) impair the clearance of apoptotic cells.¹ This leads to an accumulation of dead cell debris in tissues, which then exposes self-antigens and DAMPs. These DAMPs activate pathways like cGAS-STING, driving the production of type I IFN and other inflammatory cytokines, perpetuating the disease.¹

• Rheumatoid Arthritis (RA): In RA, there's a significant reduction in specialized efferocytic macrophages in the joint lining.¹ This impaired clearance contributes to persistent inflammation, enhanced bone destruction, and reduced tissue repair.¹

• Sjögren's Syndrome (SS): Increased apoptosis of glandular cells, coupled with defective efferocytosis, leads to the accumulation of dead cells in salivary and lacrimal glands. This exposes self-antigens and DAMPs, amplifying autoimmune responses.¹

• ANCA-associated Vasculitis (AAV): Autoantigens like PR3 and MPO-ANCA directly interfere with efferocytosis pathways, promoting inflammation and hindering the clearance of neutrophils.¹

• Systemic Sclerosis (SSc): Impaired efferocytosis contributes to the widespread fibrosis seen in SSc, fostering autoantibody production and chronic inflammation that activates fibroblasts and enhances collagen deposition.¹

• Antiphospholipid Syndrome (APS): Antiphospholipid antibodies (aPL) interfere with the normal clearance of apoptotic cells, triggering pro-inflammatory cytokine release and driving disease progression.¹

• Gout and Osteoarthritis (OA): Even in conditions like gout and OA, impaired efferocytosis of inflammatory cells or joint tissue debris contributes to persistent inflammation and tissue damage.¹

Therapeutic Promise: Restoring the Balance

The insights from this paper highlight clear therapeutic strategies:

• Apoptotic Cell (AC) Infusion: Administering ACs or their metabolites can "overwhelm" defective clearance mechanisms, promoting an anti-inflammatory response. A clinical trial for refractory RA is already proposed.¹

• Bridging Molecules: Molecules like Gas6 and MFGE8, which help phagocytes recognize and bind to apoptotic cells, can be administered to boost clearance.¹

• DNase Supplementation: For genetic defects in DNA degradation, providing exogenous DNase could prevent the accumulation of inflammatory DNA.¹

• Anti-CD47 Antibodies: Blocking the "don't eat me" signal (CD47) can enhance phagocytic clearance.¹

• PPAR/LXR Agonists: These can improve cholesterol management within phagocytes, preventing inflammation triggered by lipid overload.¹

• Mesenchymal Stromal Cell (MSC) Infusion: MSCs can generate apoptotic debris, which promotes efferocytosis and shifts phagocytes towards an anti-inflammatory state.¹

Paper 2: When the Cleanup Crew Turns Rogue – Efferoptosis in Acute Inflammation

The second paper, "TNF switches homeostatic efferocytosis to lytic caspase-8-dependent pyroptosis and IL-1β maturation," published in Science Immunology, reveals a darker side of efferocytosis.¹ It introduces "efferoptosis," a novel form of inflammatory cell death.

Traditionally, efferocytosis is anti-inflammatory. But in acute, dysregulated inflammatory environments, such as sepsis or systemic inflammatory response syndrome (SIRS), the presence of high levels of Tumor Necrosis Factor (TNF) acts as a "master switch".¹ When phagocytes, particularly macrophages, engulf dead or dying neutrophils in the presence of TNF, they don't just clear them silently. Instead, they undergo a lytic, pro-inflammatory form of cell death: efferoptosis.¹

The Molecular Mayhem of Efferoptosis

Efferoptosis is distinct from other forms of inflammatory cell death:

• Caspase-8 Dependent, NLRP3 Independent: Unlike canonical pyroptosis, which relies on NLRP3 and caspase-1, efferoptosis is driven by caspase-8. This activated caspase-8 directly cleaves gasdermin-D (GSDMD), a key protein that forms pores in the cell membrane, leading to cell lysis.¹

• Direct IL-1β Maturation: Crucially, caspase-8 also directly cleaves pro-IL-1β, leading to its maturation and release, bypassing the usual inflammasome activation pathway.¹

• The TRIFosome: This process involves a complex called the "TRIFosome," formed by the TLR4 adaptor TRIF, ZBP1, and RIPK1. This complex activates caspase-8.¹

• Signaling Rewiring: Normally, efferocytosis inhibits pro-inflammatory NF-κB signaling. However, in efferoptosis, TNF-activated efferocytosis inhibits TAK1/NF-κB, leading to the downregulation of prosurvival factors like cFLIP. Simultaneously, PLCy/MAPK signaling is sustained, which upregulates pro-IL-1β, ensuring a substrate for caspase-8.¹

Pathological Impact: Sepsis and Beyond

Efferoptosis significantly contributes to the pathology of sepsis and SIRS. In mouse models, inhibiting efferocytosis (e.g., via a TIM3 antibody) protected mice from TNF-induced SIRS, reducing macrophage death and improving survival.¹ This suggests that in these acute inflammatory conditions, the negative impacts of efferoptosis outweigh the beneficial functions of homeostatic efferocytosis.¹

Speculating on Myocardial Infarction

While the Science Immunology paper focuses on sepsis, the mechanisms of efferoptosis have profound implications for other acute inflammatory events, such as myocardial infarction (MI). MI involves massive cell death in the heart, leading to a robust inflammatory response.

Consider these connections:

• Extensive Cell Death: MI results in a large number of dying cardiomyocytes. These apoptotic cells, if not cleared efficiently, can release DAMPs, triggering inflammation.¹

• Cholesterol Overload: The Arthritis & Rheumatology paper highlights that in atherosclerosis (a major cause of MI), cholesterol accumulation from uncleared apoptotic cells can trigger macrophage apoptosis and NLRP3 inflammasome activation.¹ This adds another layer of inflammatory cell death.

• TNF and Efferoptosis: Post-MI, there's a significant inflammatory response, often including elevated TNF levels. It's highly plausible that macrophages engulfing dying heart cells and recruited neutrophils in the damaged heart, under the influence of TNF, could undergo efferoptosis. This would contribute to the inflammatory burden and adverse cardiac remodeling, similar to its role in SIRS.¹ The direct cleavage of IL-1β by caspase-8 in efferoptosis could be a significant driver of sterile inflammation in the infarcted heart.

Therefore, targeting efferoptosis, perhaps through caspase-8 inhibition or specific PS receptor modulation (like TIM3 inhibition), could represent a novel therapeutic strategy to reduce post-MI inflammatory injury, distinct from targeting canonical inflammasomes.

The Converging Insights: A Unified View of Inflammation

These two papers, from different journals and focusing on seemingly distinct disease categories, reveal a profound commonality: efferocytosis is a double-edged sword. It is absolutely essential for maintaining immune tolerance and resolving inflammation.¹ But it can become a potent source of inflammation if it is either:

1. Defective: Leading to the accumulation of uncleared apoptotic cells and the release of DAMPs, driving chronic autoimmune diseases.¹

2. Aberrantly Activated (Efferoptosis): Where, under acute inflammatory conditions like high TNF, the very act of efferocytosis triggers a pro-inflammatory, lytic cell death in the phagocyte itself, fueling acute inflammatory crises.¹

The common theme is the critical need for precise modulation of efferocytosis. We need to enhance it when it's failing (as in autoimmune diseases) and prevent its detrimental pro-inflammatory switch when it's being hijacked (as in acute inflammation like sepsis and potentially MI).

The therapeutic landscape is exciting. Strategies that boost efferocytosis (like AC infusions or DNase supplementation) can restore immune balance in chronic conditions.¹ Meanwhile, interventions that prevent efferoptosis (such as TIM3 inhibition or targeting caspase-8) could mitigate acute inflammatory damage.¹ The challenge lies in developing therapies that can differentiate between these contexts or be delivered in a highly targeted manner to specific tissues, as the role of efferocytosis can be tissue-specific.¹

This converging understanding offers a new paradigm. Instead of merely suppressing inflammation, we can aim to restore the fundamental efferocytic balance, offering more profound and sustained therapeutic effects. The future of immune therapies may well lie in mastering the art of the cleanup crew – ensuring they always work for us, never against us.

Here’s why sleep loss deserves the same attention as diabetes—and what it means for drug development.

The Metabolic Chaos of Sleep Loss

Like diabetes, sleep deprivation disrupts how your body handles glucose, ramps up inflammation, and throws your metabolism into disarray. A 2015 Science Signaling review (DOI: 10.1126/sciadv.1504018) shows that just 4–6 hours of sleep deprivation in mice spikes adenosine levels in the hippocampus, impairing glucose uptake and mimicking insulin resistance seen in diabetes. This isn’t just a brain issue—peripheral tissues like adipocytes also show reduced insulin sensitivity, leading to fat accumulation and oxidative stress. For drug developers, this points to a clear target: metabolic pathways disrupted by sleep loss.

Inflammation and Oxidative Stress: A Shared Culprit

Sleep loss doesn’t just tire you out; it ignites a firestorm of inflammation. The Science Signaling study found that sleep deprivation boosts oxidative stress proteins and lipid droplet accumulation in glial cells, which then rely on β-oxidation to cope. This mirrors the chronic inflammation in diabetes, where oxidative stress damages tissues and drives complications like cardiovascular disease. In the context of hypertrophic cardiomyopathy (HCM), a condition linked to sudden death in young athletes, sleep deprivation could amplify inflammation, worsening cardiac fibrosis. Therapies targeting oxidative stress—like antioxidants or anti-inflammatory biologics—could address both sleep loss and diabetes-related damage.

Synaptic and Systemic Fallout

The brain takes a hit from sleep loss, much like it does in metabolic disorders. The study highlights how sleep deprivation downregulates neural plasticity by increasing adenosine and A1R activity, leading to memory deficits. This is eerily similar to diabetes-induced cognitive decline, where poor glucose metabolism harms neurons. Systemically, sleep loss redirects energy away from non-essential processes, like synapse formation, to cope with metabolic stress—paralleling diabetes’ energy misallocation. Drug developers could explore adenosine receptor antagonists or neuroprotective agents to mitigate these effects, potentially benefiting both conditions.

HCM and Sleep: A Deadly Combo

For those with HCM, sleep deprivation could be a silent killer. The Science Translational Medicine study (DOI: 10.1126/scitranslmed.aad2516) showed that inflammation drives HCM’s progression, with regulatory T cells (Tregs) struggling to control it. Sleep loss exacerbates this by upregulating inflammatory pathways, potentially increasing the risk of sudden cardiac death in young athletes. Imagine a college basketball player, burning the candle at both ends, unaware that their heart is under extra strain. Therapies like IL-2 agonists (e.g., sifalimumab) or PD-1 agonists, which boost Treg function, could tackle inflammation in both HCM and sleep-deprived patients, offering a dual-purpose pipeline.

Drug Development Opportunities

The parallels between sleep loss and diabetes open exciting avenues for drug development:

• Adenosine receptor modulators: Blocking A1R could restore neural plasticity and glucose metabolism, addressing cognitive and metabolic deficits.

• Anti-inflammatory biologics: IL-2 or PD-1 agonists, inspired by HCM research, could reduce systemic inflammation, benefiting both sleep-deprived and diabetic patients.

• Antioxidants: Targeting oxidative stress could protect tissues from the damage caused by sleep loss and diabetes.

• Insulin sensitizers: Drugs like metformin might be repurposed to improve glucose uptake in sleep-deprived individuals.

Clinical trials could focus on biomarkers like adenosine levels, inflammatory cytokines, or LGE-CMR for fibrosis, accelerating the path to market for these therapies.

Why It Matters for You

Sleep loss isn’t just about feeling groggy—it’s a metabolic crisis that could shorten your life, especially if you’re at risk for conditions like HCM. For drug developers, it’s a call to action: prioritize sleep as a therapeutic target with the same urgency as diabetes. The science is clear, and the stakes are high.

Want to stay ahead on groundbreaking therapies? Subscribe to www.drugdevelop.com for the latest in cardioimmunology, metabolic disease, and more. Let’s wake up to the power of sleep—and save lives in the process.

Citations:

1. [Authors]. (2015). Sleep loss is a metabolic disorder. Science Signaling, 8, adp9358. DOI: 10.1126/sciadv.1504018

2. Wang, Y.-J., et al. (2015). Regulatory T cells attenuate chronic inflammation and cardiac fibrosis in hypertrophic cardiomyopathy. Science Translational Medicine, 7, eaad2516. DOI: 10.1126/scitranslmed.aad2516

The story begins with an elegant scientific plan. It ends with a molecular lesson about the humbling complexity of human biology. And it should change how we think about target selection, translational strategy, and the seductive confidence that comes from a clean in vitro result.

What follows is that story — and three mathematical frameworks that let us reason about it more rigorously.

The Architecture of Confidence

In drug discovery, we build cathedrals on paper. We map pathways, identify nodes, select targets, and construct rationales of extraordinary logical coherence. The best programs feel almost inevitable in their design. Every arrow points in the right direction. Every experiment confirms the hypothesis.

This confidence has occasionally been justified. Gleevec — imatinib — is the canonical example: a single kinase target, BCR-ABL, a single disease defined by a single chromosomal translocation, and a clinical result so dramatic it reshaped oncology. The principle seemed sound: find the right target, block it precisely, and the disease collapses.

But CML is, in retrospect, the exception that proved a much more uncomfortable rule. BCR-ABL is a genuine single point of failure in a cancer that has, for molecular reasons, stripped away most of its redundancy. Most diseases are not like this. Most biology is not like this.

Biological systems did not evolve to be druggable. They evolved to survive. Redundancy, pathway crosstalk, and adaptive rewiring are not bugs — they are the fundamental architecture of resilience.

The assumption that a single, well-chosen target is sufficient to collapse a complex inflammatory or metabolic disease is what I call the Target Fallacy. It is not a foolish assumption — it is a reasonable prior, given some remarkable successes. But it is a prior we have updated too slowly, and the Novartis NEK7 story is the latest evidence of why.

The Plan Was Beautiful

The NLRP3 inflammasome has been one of the most intensively pursued drug targets in inflammation for over a decade. Its activation drives IL-1β and IL-18 release, and dysregulated NLRP3 activity has been implicated in gout, CAPS, atherosclerosis, Alzheimer’s disease, and type 2 diabetes complications.

The Novartis team focused on NEK7, a mitotic kinase that moonlights as an essential scaffold for NLRP3 inflammasome assembly. Critically, NEK7’s role in inflammasome function is independent of its kinase activity — it acts as a structural bridge. The team built a molecular glue degrader, NK7-902, that recruits the CRBN E3 ubiquitin ligase machinery to NEK7, tagging it for proteasomal destruction. NK7-902 was potent, selective, and mechanistically clean, achieving greater than 95% degradation of NEK7 protein.

The preclinical data in mice was compelling. Oral administration strongly degraded NEK7 in splenic tissue and produced meaningful inhibition of IL-1β release in both an acute peritonitis model and a CAPS disease model.

Then came the human and NHP data. In human peripheral blood cells, even with virtually complete NEK7 degradation, IL-1β suppression was partial and highly variable across donors. In non-human primates, sustained 95%+ NEK7 degradation produced only transient and partial inhibition of IL-1β release.

The target was eliminated. The biology barely noticed.

Mathematical Framework I — Pathway Redundancy

The NEK7 result is not mysterious once we model the inflammasome as a network with multiple weighted activation routes, rather than a single linear pathway.

The Core Equation

Define the total NLRP3 activation A as the sum of contributions from all activation routes:

A = Σᵢ wᵢ · aᵢ

where wᵢ is the fractional weight of route i and aᵢ is its activation level (0 to 1). By definition, Σᵢ wᵢ = 1.

When we block route j (e.g., by degrading NEK7), the remaining activation is:

A_blocked = Σᵢ≠ⱼ wᵢ · aᵢ = 1 − wⱼ · aⱼ

This is the key insight. For NEK7 degradation to suppress NLRP3 activity by ≥80%, we need:

wⱼ ≥ 0.80

In mouse cells, NEK7 carries approximately 90% of activation weight — so eliminating it nearly collapses the system. In human and NHP cells, NEK7 appears to carry only ~35% of the weight, with substantial traffic flowing through PTEN-L, mtDNA sensing, and potassium efflux routes. Blocking one road does not close the network.

Figure 1. Pathway Redundancy Model. Panel A: Residual NLRP3 activation as a function of NEK7 blockade under mouse vs. human route-weight assumptions. At 95% NEK7 degradation (dashed red line), mouse activation falls below the therapeutic threshold while human/NHP activation remains well above it. Panel B: Estimated fractional contributions of each activation route by species. The NEK7 dominance in mice versus the distributed architecture in humans explains the translational failure.

This framework makes the failure predictable, not mysterious. Any team that had estimated human route weights before committing to lead optimization would have recognized that NEK7 degradation, however complete, was unlikely to produce therapeutic-level pathway suppression in primates.

The practical implication: before selecting a target in a redundant pathway, estimate the target’s fractional contribution in human tissue. This is now possible with modern tools — donor-matched primary cell experiments, siRNA knockdown across a diverse human panel, and pathway flux analysis in organoids. It is not free. But it is far cheaper than a Phase II failure.

Mathematical Framework II — Bayesian Translational Confidence

The NEK7 story is also a story about how evidence should change our beliefs — and about how the current drug development system is structured to update beliefs too slowly, too late.

The Framework

Model our belief in the mechanism as a probability θ — the probability that the mechanism works in humans. We can represent our state of knowledge at each stage as a Beta distribution, Beta(α, β), which is a natural choice for probabilities and updates cleanly as new data arrives.

Prior belief: θ ~ Beta(α₀, β₀)

Each experimental stage provides new evidence. After observing s successes and f failures in a relevant species, we update:

Posterior: θ | data ~ Beta(α₀ + s, β₀ + f)

The mean of this distribution is α/(α+β), which is our best estimate of P(mechanism works in humans) given all evidence so far.

Applying It to the NEK7 Program

A reasonable prior for a target-based program in inflammation is Beta(3, 5), giving a mean P ≈ 0.375 — consistent with historical Phase II success rates of 30-40% in this space.

Positive mouse efficacy in 4 of 5 models raises the posterior to Beta(7, 6), mean ≈ 0.54. This is the stage where many teams accelerate into IND-enabling studies. The evidence looks convincing.

But then add the NHP pharmacodynamic data: partial and transient IL-1β suppression despite complete target degradation. This is a near-failure — one success out of effectively zero in the most translatable species. The posterior falls to Beta(7, 7), mean ≈ 0.50. Confidence has eroded.

Add the human primary cell variability — inconsistent suppression across donors — and the posterior drops further to Beta(7, 9), mean ≈ 0.44. The belief distribution has shifted materially left. We are now in genuine uncertainty about whether this mechanism is pharmacologically sufficient in humans.

Each experiment is a likelihood update. The question is not whether your compound works in the system you tested. It is whether the full evidence profile justifies conviction in the system that matters — the human patient.

Figure 2. Bayesian Updating of Translational Confidence. Panel A: Beta distributions representing belief in mechanism validity at each evidence stage. Mouse efficacy raises confidence; failed NHP PD and inconsistent human cell data erode it. Panel B: Mean P(mechanism valid) ± 80% credible interval at each stage. The trajectory reveals how mouse-only data produces false conviction that later evidence must correct — often at great expense.

The critical observation from this framework is not the final number. It is the shape of the update trajectory. Programs that incorporate human-relevant data early get the bad news when it is cheap. Programs that defer human data accumulate false confidence, consume resources, and then face a large, late, costly correction.

There is a structural drug-development argument here: the expected cost of generating human primary-cell data in lead optimization is roughly $2-5M for a well-run campaign. The expected cost of a Phase II failure — after the Bayesian update arrives two years later, in the clinic — is $50-200M. The mathematics of belief updating strongly favors early investment in human-relevant evidence.

Mathematical Framework III — Expected Value of Development

The Bayesian framework tells us how evidence should change our beliefs. The expected value framework asks the harder question: given a realistic probability of mechanism validity, which preclinical strategy maximizes the expected return on drug development investment?

The Model

Define the expected value of a program as:

EV = P(mechanism valid) × P(clinical success | mechanism valid) × 
V(approval) − C(preclinical) − P(enter Phase I) × 
[C(Phase I) + P(Phase II) × (C(Phase II) + P(Phase III) × C(Phase III))]

where V(approval) is the net present value if the drug reaches market, and each C(stage) represents the expected cost of that stage.

The key variable is P(mechanism valid) — which is precisely what the choice of preclinical strategy determines. A mouse-centric program produces a higher prior estimate of this probability (because mouse models are confirmatory), but a lower true probability. An early human translation program produces a more accurate — and typically lower — estimate, but one that the team can actually rely on.

Why the Arithmetic Favors Early Human Translation

Assume a mechanism with a true P(valid in humans) = 0.20 — typical for a novel inflammatory target. A standard mouse-centric program might estimate this at 0.30-0.40 based on confirmatory rodent data. An early human program would correctly estimate it at or close to 0.20, potentially identifying the translational gap before entering clinical development.

The standard program spends less in preclinical work ($15M vs. $28M with human tissue and NHP PD) but then runs Phase I, Phase II, and potentially Phase III on the basis of a false prior. The early human program front-loads the cost of truth.

The expected value comparison, modeled across a range of mechanism validity probabilities, consistently favors early human translation once the mechanism validity falls below approximately 30-35% — which covers the majority of programs in inflammation, CNS, and metabolic disease.

Figure 3. Expected Value of Development by Preclinical Strategy. Panel A: Program EV across a range of pre-human mechanism validity probabilities. The early human strategy (blue) generates higher EV across most of the realistic probability range, with the advantage increasing as mechanism uncertainty rises. Panel B: Expected cost by development stage at P(mechanism) = 20%. Higher preclinical investment in the human-first strategy is more than offset by reduced downstream expected failures.

The cost of false confidence is not paid at the preclinical stage. It is paid in Phase II, when the evidence finally arrives — too late, too expensive, and with patients enrolled.

What Needs to Change

These three frameworks point to the same practical agenda.

First, measure route weights before committing to a target. If a target’s fractional contribution to pathway activation in human primary cells is below 50%, the program faces a high burden to demonstrate therapeutic sufficiency. This is not a reason to abandon the target — it is a reason to investigate polypharmacology, combination strategies, or direct pathway blockade downstream of the redundant nodes.

Second, treat Bayesian updates as decision triggers, not post-hoc observations. Programs should define, in advance, what posterior probability threshold justifies progression to IND-enabling studies. NHP pharmacodynamic failure should move a program into formal go/no-go review, not result in an explanation and a pivot to the next mouse model.

Third, use the EV framework to justify early investment in human translation. The argument that human tissue experiments are “too expensive” for early discovery fails basic expected value arithmetic. The question is never whether early human data costs money. The question is how that cost compares to the expected cost of the alternative.

Fourth, be structurally honest about the limitations of mouse models. A positive result in a murine inflammasome model is evidence that the mechanism is pharmacologically tractable in that species. There is no evidence that the mechanism is relevant in humans. These are different claims, and the organizational habit of treating the former as evidence for the latter has cost the industry billions of dollars and years of patients’ lives.

The Novartis Team Did the Right Thing

The Novartis scientists did not make a mistake. They asked a precise scientific question, built precise tools to answer it, and reported the answer transparently — including the parts that were uncomfortable. That is what science is supposed to do.

The problem is not the science. The problem is an ecosystem that consistently asks the wrong question first: “Can we hit the target?” rather than “Is this target sufficient in humans?” and pays for that inversion at a late and expensive stage.

The NEK7 data suggest that direct NLRP3 inhibitors — which act downstream of the redundant activation routes — may have advantages in human systems. Programs using compounds such as selnoflast and inzomelid, currently in clinical development, will be informative on this point.

But the broader lesson is about first principles. Biological systems are networks, not flowcharts. They evolved to survive perturbation. The three frameworks described here — pathway redundancy, Bayesian belief updating, and expected program value — are not academic exercises. They are practical tools for structuring discovery decisions, ensuring we ask the right questions in the right order.

The most important question in drug discovery is not whether your compound works. It is whether your biology is asking for what your compound offers.

#DrugDiscovery#NLRP3#TranslationalMedicine#InflammasomeBiology#MolecularGlue#PreclinicalModels#TargetFallacy

The story begins with an elegant scientific plan. It ends with a biological lesson about redundancy, evolution, and the uncomfortable gap between species. And it should change how we think about target selection, translational strategy, and the seductive confidence that comes from a clean result in a mouse.

The Architecture of Confidence

In drug discovery, we build cathedrals on paper. We map pathways, identify nodes, select targets, and construct rationales of extraordinary logical coherence. The best programs feel almost inevitable in their design. Every arrow points in the right direction. Every experiment confirms the hypothesis.

This confidence has occasionally been justified. Gleevec — imatinib — is the canonical example: a single kinase target, BCR-ABL, a single disease defined by a single chromosomal translocation, and a clinical result so dramatic it reshaped oncology. Find the right target, block it precisely, and the disease collapses. The principle seemed sound.

But CML is, in retrospect, the exception that established a flawed template. BCR-ABL is a genuine single point of failure in a cancer that has, for molecular reasons, stripped away most of its biological redundancy. Most diseases are not like this. Most biology is not like this.

Biological systems did not evolve to be druggable. They evolved to survive. Redundancy, pathway crosstalk, and adaptive rewiring are not bugs — they are the fundamental architecture of resilience.

The assumption that a single, well-chosen target is sufficient to collapse a complex inflammatory or metabolic disease is what I call the Target Fallacy. It is not a foolish assumption — it is a reasonable one, given some remarkable successes. But it is an assumption we have updated too slowly, and at extraordinary cost.

The Plan Was Beautiful

The NLRP3 inflammasome has been one of the most intensively pursued drug targets in inflammation for over a decade. Its activation drives the release of interleukin-1β and IL-18 — inflammatory signals implicated in gout, rheumatoid arthritis, cardiovascular disease, Alzheimer’s disease, and a spectrum of autoinflammatory conditions. Getting control of NLRP3 activation has been a central ambition of the inflammatory disease field.

The Novartis team focused on NEK7, a mitotic kinase that serves as an essential structural scaffold for NLRP3 inflammasome assembly. The key insight was that NEK7’s role in the inflammasome is not catalytic — it does not drive a chemical reaction — but architectural. It holds the complex together. This means that rather than trying to block an enzymatic activity, you could simply remove the protein entirely.

The drug they built to do this — NK7-902 — was a molecular glue degrader. It works by tricking the cell’s own disposal machinery into destroying NEK7, tagging it for proteasomal degradation via the CRBN E3 ubiquitin ligase system. NK7-902 was potent, selective, and mechanistically elegant. In biochemical assays, it worked exactly as designed: it degraded more than 95% of NEK7 protein.

On paper, eliminating the scaffold should collapse the inflammasome. The logic is sound. The chemistry is elegant. The execution was flawless. And in mice, it worked.

Figure 1 — Why the same drug succeeds in mice and fails in primates. Panel A: in mouse cells, NEK7 is the dominant gateway for NLRP3 activation, carrying roughly 90% of activation traffic. Blocking it effectively shuts down the system. Panel B: in human and NHP cells, activation traffic distributes across multiple parallel routes, most of which bypass NEK7 entirely. Eliminating one node reroutes — it does not stop — the pathway. Note: route weight estimates are conceptual and illustrative of the biological principle.

The Biology Disagreed

The preclinical data in mice was encouraging. Oral NK7-902 strongly degraded NEK7 in splenic tissue and produced meaningful inhibition of IL-1β release in both an acute peritonitis model and a CAPS disease model. The team had a molecule that worked, a target that was tractable, and a disease rationale that held in rodents. Any discovery team would have been cautiously optimistic.

Then came the human and non-human primate data, and the story changed entirely.

In human peripheral blood cells, even with virtually complete NEK7 degradation, IL-1β suppression was partial and highly variable — differing substantially across individual donors and experimental conditions. That variability is itself a signal. When a mechanistically clean intervention produces inconsistent functional effects across human donors, the system is communicating something important about its internal architecture.

In non-human primates, the dissociation was even more striking. Oral dosing at 2 mg/kg produced sustained NEK7 degradation exceeding 95% in blood — the compound had done exactly what it was designed to do. IL-1β inhibition was transient and partial. The target was eliminated. The biology barely noticed.

The target was eliminated. The biology barely noticed. This is a sentence that should be framed and hung in every drug discovery department in the world.

What this tells us is not that NEK7 is unimportant in inflammasome biology. In mouse cells, it clearly is important — essential, even. What it tells us is that the inflammasome pathway in primates has evolved redundant activation routes that do not depend on NEK7. When one road is blocked, the system takes a different highway. Evolution, unlike our target-validation assays, tested this over millions of years.

Figure 2 — The efficacy disconnect and what the evidence should tell us. Left: NK7-902 achieves near-identical NEK7 degradation across all three species. IL-1β suppression tracks with species — not with compound performance. Right: how confidence in the mechanism should update as each piece of evidence arrives. Mouse efficacy raises it; NHP and human data erode it. Programs that collect human-relevant data early get the correction cheaply. Programs that defer it pay compound interest.

Why This Keeps Happening

The NEK7 story is not unique. It sits in a long and expensive tradition. The p38 MAPK inhibitors showed exceptional preclinical promise in rheumatoid arthritis for nearly a decade before a string of clinical failures revealed that the human RA synovium is far more pathway-redundant than any rodent model had suggested. The CANTOS trial demonstrated that blocking IL-1β with canakinumab reduces cardiovascular events — the biology was real — but the increase in fatal infections and the commercial calculus made it unworkable. The list of programs that succeeded in mice and failed in humans is long enough to be its own clinical speciality.

There is a structural reason this pattern persists. Mouse models are inexpensive, fast, and genetically manipulable. They generate clean, interpretable data. They are scientifically satisfying in a way that messy human primary cell experiments rarely are. The incentive architecture of early drug discovery — publication pressure, milestone timelines, the cost differential between rodent and human studies — consistently pushes teams toward the models most likely to produce affirmative results, not the models most likely to predict human outcomes.

We select for confidence. Biology rewards robustness.

The two are not the same thing, and the gap between them is where drug development money goes to disappear.

The Species Gap Is a Biological Fact, Not a Technical Problem

A common response to translational failures is to reach for better tools: humanized mice, organoids, organs-on-chips. These are genuinely valuable, and investment in them is warranted. But there is a more fundamental issue that better experimental systems alone cannot resolve.

Human and mouse immune systems diverged roughly 65 to 80 million years ago and have been under substantially different selective pressures ever since. The human NLRP3 pathway appears to have developed redundant NEK7-independent activation routes that are simply not present in the mouse — not because our models are imperfect, but because that is what evolution produced. This is not a gap in our understanding that better technology will close. It is a biological reality that demands a different strategic response.

The strategic response is to ask the critical question earlier. Not “does our target matter in this pathway?” but “is this pathway, in human tissue, sufficiently non-redundant that eliminating our target will produce the functional outcome we need?” These are different questions. The second one requires human data. And it needs to be asked before lead optimisation is complete — not at Phase II.

The question is not whether your target matters. The question is whether, in human biology, your target is a single point of failure — or merely one of several roads to the same destination.

The Cost of Getting This Wrong

The financial argument for earlier human translational work is less intuitive than it should be. Adding NHP pharmacodynamic studies and human tissue experiments in early discovery can double or triple the cost of lead optimisation. Program teams face pressure to move fast, hit milestones, and advance compounds. Spending more time and money preclinically feels like slowing down.

But the arithmetic is straightforward. A Phase II failure in an inflammatory indication costs between $50 million and $200 million, depending on trial size and duration. A Phase III failure costs an order of magnitude more. The industry-wide failure rate for drugs entering Phase II in inflammation and immunology is approximately 70%. If even a fraction of those failures trace back to a translational assumption that human primary cell data would have challenged — and the evidence suggests many do — the financial case for front-loading that investment is overwhelming.

There is also a patient cost that rarely enters these calculations directly. Every program that reaches clinical trials based on mouse efficacy data represents a commitment of patient time, risk, and hope. Phase II failures are not just expensive — they are a burden on people who enrolled because they or their physicians believed the mechanism had a reasonable chance of working. Better preclinical translation does not just protect pipeline value. It protects patients.

Figure 3 — The economics of early human translation. Left: expected program value as a function of true mechanism probability in humans. The human-first strategy consistently outperforms once mechanism uncertainty rises above roughly 25%, which covers the majority of programs in inflammation, CNS, and metabolic disease. Right: the strategic decision timeline — the window to ask the human translational question is lead optimisation, not Phase II. Waiting until clinical studies to discover pathway redundancy is the most expensive form of discovery research.

What Needs to Change

The path forward is not a single intervention. It is a reorientation of how discovery programs are structured, and what counts as evidence of confidence versus evidence of translatability.

Human primary cell data should function as a gate, not a late confirmatory box to tick. When a compound shows robust efficacy in mouse models, the immediate question should not be which additional mouse models to run, but what happens in human primary cells and tissue. Variability across donors — as seen in the Novartis data — should be interpreted as information about the mechanism, not experimental noise to be averaged away.

NHP pharmacodynamic studies, where feasible, should be incorporated before lead optimisation is complete. This is a significant resource commitment, and not every program will justify it. But for mechanisms where the translational risk is structurally high — innate immune pathways, CNS targets, metabolic nodes — the investment is warranted. A partial NHP pharmacodynamic result, collected early, is worth far more than a complete one collected at IND.

Phenotypic approaches deserve serious investment alongside target-based programs. Phenotypic screens measure functional outcomes in human or humanised systems rather than target engagement alone. A compound that suppresses IL-1β release in human peripheral blood across a diverse donor panel — through whatever mechanism — has demonstrated human-relevant functional activity that no target-engagement assay can match. This is not an argument against target-based discovery. It is an argument for complementing it.

Finally, we need institutional honesty about what mouse efficacy data actually proves. A positive result in a murine model is evidence that the mechanism is pharmacologically tractable in that species. It is not evidence that the mechanism is sufficient in humans. Treating the former as evidence for the latter — routinely, under timeline and resource pressure — is the Target Fallacy in its most common and costly form.

What the NEK7 Story Really Tells Us

The Novartis scientists did not make a mistake. They asked a precise scientific question, built precise tools to answer it, and reported the answer honestly — including the parts that were uncomfortable. The paper is a model of scientific integrity. That the result was not what they hoped for is not a failure of the team. It is a failure of the assumptions that the field, as a whole, brought to the program.

The problem is not the science. The problem is an ecosystem that consistently asks the wrong question first — can we hit the target? — rather than the more important one: is this target sufficient in humans? And pays for that inversion at a late, expensive, and patient-facing stage.

The NEK7 data also points forward constructively. Direct NLRP3 inhibitors — which block the protein itself rather than an upstream activator — may have advantages in human systems precisely because they sit downstream of the redundant activation routes that bypass NEK7. Compounds like selnoflast and inzomelid, currently in clinical development, will be informative on exactly this point. The biology is telling us where to look next.

But the broader lesson is about first principles. Biology is a network, evolved over hundreds of millions of years to survive perturbation. Blocking one node rarely collapses it. Before we invest in any mechanism at scale, we owe it to patients — and to the basic integrity of the enterprise — to find out whether we are blocking a single point of failure, or merely rerouting traffic.

The most important question in drug discovery is not whether your compound works. It is whether your biology is asking for what your compound offers.

The Target Fallacy will not be the last story of this kind. But if we take it seriously — if we use it to restructure how and when we ask translational questions — it might be one of the last times we are surprised by it.

#DrugDiscovery #NLRP3 #TranslationalMedicine #TargetFallacy #PreclinicalModels #InflammasomeBiology #NEK7 #Novartis

TL;DR : My preliminary pharmacometric analysis of TOGETHER-PsO suggests weight loss-driven drug exposure recovery may be more important than direct synergy with putative anti-inflammatory effects of Zepbound . This is should be considered a curiosity-driven hypothesis, not a definitive finding as I do not have ‘insider access’.

Eli Lilly’s TOGETHER-PsO results were not unexpected: ixekizumab + tirzepatide achieved 40.6% PASI 100 clearance versus 29% with ixekizumab monotherapy. Many assumed the mechanism: “GLP-1 reduces inflammation, ixekizumab modulates immunity, they synergize beautifully.” I completed a causal mediation analysis* (10,000 Monte Carlo simulations) to decompose this efficacy gain (see Figure 1, the cover picture for the mechanistic framework). The finding: ~70–80% of the benefit appears to derive from weight loss, reversing obesity’s suppression of ixekizumab drug exposure. Direct GLP-1 anti-inflammatory effects? Likely secondary in this context.

The “Drug Sink” Insight

Here’s the mechanism I explored: Obesity increases how much “space” antibodies distribute into (higher Vd) and how fast they get cleared (altered CL). Result? The estimated ixekizumab exposure decreases by ~40% at a BMI of 40 compared with normal BMI (Figure 2). When Zepbound causes weight loss (14.1% mean in TOGETHER), BMI falls from 39.5 to ~32. The model estimated that this recovers ~30% of the lost drug exposure. Plus, adipose tissue stops pumping out inflammatory cytokines (IL-6, TNF-α, IL-23). That’s powerful synergy—but not the kind we usually discuss. This reframes obesity from “comorbidity” to “reversible pharmacokinetic defect.”

The Evidence (With Caveats)

The model calibration was excellent: The model predicted monotherapy at 29.1% versus the observed 29.0%—a good sign that the mechanistic framework is sound. The observed efficacy gain decomposes into mediated (weight loss-driven) and direct (GLP-1-driven) pathways (Figure 3), with sensitivity analyses robust to ±25% parameter variation (Figure 4).

Literature support: Hjort’s 2024 meta-analysis across 15+ trials showed each 5 BMI-unit increase associates with ~15–20% relative odds reduction in PASI 100—consistent with a PK-driven mechanism.

But here’s my key caveat: This analysis is based on published aggregate data from TOGETHER PSO only. No patient-level data. No serum ixekizumab concentrations. No adipose tissue biomarkers. The 7.6 percentage point gap between my model (estimated 33% PASI 100) and observed (40.6%) suggests unmeasured mechanisms—possibly a larger direct GLP-1/GIP effect than our conservative model estimated, or behavioral/temporal factors. This is should be considered a curiosity-driven hypothesis, and I hope Lilly will publish a full analysis confirming or refuting this.

Why It Matters

If my estimated weight loss → PK recovery pathway is primary (as suggested by cover page):

1. High-BMI patients failing biologics don’t need escalation—they need mechanism-informed weight loss intervention

2. Dosing timing is critical. Zepbound’s rapid weight loss (weeks 4–16) overlaps ixe’s PD buildup (weeks 12–24)—that’s the synergistic window

3. Obesity management becomes therapeutic, not cosmetic

The Challenge

What would prove/disprove this hypothesis? TOGETHER subgroup analysis by weight loss magnitude. My model predicts that the <5% weight-loss cohort within this trial shows minimal PASI benefit vs. monotherapy.

• If true → weight loss dominates (supporting Figure 1 framework).

• If false → direct GLP-1 effect is bigger than we estimated.

That data likely exists. #EliLilly, release it. Let science win.

What’s next?

I’ve built a rigorous framework with limited public data. I expect to be wrong in the details. Feel free to push back with evidence or counter-analysis.

1. What data do you have?

2. What’s your mechanistic intuition? Genuine pushback makes science better.

Source data for PK-efficacy assumptions : EPAR/SmPCs, publicly available Pop-PK data from regulatory submissions.

#EliLilly #NovoNordisk #AstraZeneca #Roche #Pfizer #Obesity #Diabetes #Psoriasis #Pharmacometrics #GLP1 #Dermatology #Thisis psoriasis # #UCB #EADV #Thisis PsA #ACR #Rheum #GRAPPA

TL;DR: AI drug discovery excels at prediction but stumbles at explanation- and in 2026, regulators are making that gap legally and ethically untenable. Developers now face a simple choice: explain your black boxes or don’t deploy them. Even if you are allowed to, should you?

Here’s the question keeping drug developers awake at night: Would you market a medicine that demonstrably helps patients—but whose mechanism of action you cannot fully explain?

That’s not hypothetical anymore. It’s 2026, and that question is no longer academic.

For the first time, pharmaceutical companies are running clinical trials on drugs that were designed by AI systems so complex, so layered with machine-learning inference, that even their inventors cannot trace the complete logic chain back to first principles. We can show you: This molecule works. Here’s the data. But we cannot always show you: Here’s exactly why.

This isn’t a failure of the technology. It’s a success that’s outpaced our ability to understand it. And that success is now colliding with something more fundamental: the ethics of deploying medicines we don’t fully comprehend.

Welcome to the Black Box Reckoning of 2026. There are three crises that matter right now, and they’re all variations on the same theme: AI’s predictive power has outrun human understanding—and the system is finally pushing back.

The “Phenotypic Trap”: When Cells Lie—And Patients Pay

Let’s start with the clearest cautionary tale: REC-994, Recursion’s superoxide scavenger for cerebral cavernous malformations (CCM), a rare disease affecting roughly 25,000 Americans. Most have no cure. Many face progressive neurological decline.

In February 2025, Recursion proudly announced that REC-994 had met its primary endpoint of safety and tolerability in a Phase 2 trial called SYCAMORE—and, more importantly, showed what the company called “promising trends” in lesion volume reduction. Fifty percent of patients on the 400 mg dose showed a reduction in total lesion volume compared to 28% on placebo after 12 months. (Recursion Pharmaceuticals, February 2025)

For patients with a devastating rare disease, this was hope. The AI-enabled phenotypic screening platform had done what it was built to do: scan millions of cellular images, find patterns that distinguish diseased from healthy states, and identify molecules that “rescued” those states. The AI worked. The science checked out. The patients might have a chance.

Except they didn’t.

By May 2025, when Recursion shared results from the long-term extension phase, the hope evaporated. Patients who had crossed over from placebo to the active drug showed no improvement. Worse, the patients who stayed on 400 mg the whole time didn’t sustain their initial benefit—their results became “indistinguishable from natural history,” in the clinical euphemism that means basically, the drug stopped working, and we’re back where we started. (GEN, May 2025)

This is the Phenotypic Trap in its most honest form: Recursion’s computer vision system optimized for a specific cellular readout—visual morphology that suggested disease reversal. The AI was right about what it was measuring. But the measurement itself didn’t predict what mattered: whether patients got better.

The epistemological failure here is subtle but critical. The AI learned to predict one thing (cellular appearance) while the researchers assumed it would predict another (human disease). No one set out to deceive. But the system was so sophisticated at pattern recognition that it found correlations that looked like causation—until humans tried the drug and reality intervened.

Recursion has since halted or deprioritized several pipeline programs. For the CCM patients in SYCAMORE, the news was simply: your trial drug doesn’t work. Time to wait for something else.

This is why the question matters: We trusted the AI because it made sense on a cellular level. But the cell isn’t the patient.

But Here’s the Other Side

It’s worth noting that REC-994’s failure doesn’t invalidate the phenotypic screening approach entirely. Recursion’s other AI-discovered program, REC-4881 (a MEK inhibitor for familial adenomatous polyposis), is showing what appear to be genuine clinical signals. December 2025 data showed a 43% median reduction in polyp burden after 12 weeks. (Recursion Pharmaceuticals, December 2025) The platform works sometimes—and when it does, it works well.

Additionally, traditional drug discovery has always involved failures in translation. Compounds that show promise in cell models routinely fail in humans; this is not unique to AI-driven approaches. The difference is partly one of scale: AI can generate and test more hypotheses faster, which paradoxically means both more failures and more successes, compressed into a shorter timeframe.

There’s also the question of mechanism itself. Many drugs used clinically for decades—aspirin, for example—had unknown or incompletely understood mechanisms of action when they were first approved. The assumption that we must understand why a drug works before deploying it is historically somewhat anachronistic. What matters most is that it works, and that it’s safe.

The FDA’s Credibility Demand: Protecting Patients by Demanding Answers

Here’s where the conversation gets uncomfortable for the tech side.

On January 6, 2025, the FDA released draft guidance titled “Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products.” (FDA, January 2025) This was the first comprehensive regulatory framework addressing AI in drug development, and it said something that made the “move fast and break things” crowd grip their seats:

The agency is not accepting “the AI said so” as a valid reason to approve a medicine.

Instead, the FDA’s guidance outlined a “risk-based credibility framework” that requires sponsors to:

• Define their AI model’s “context of use” precisely

• Demonstrate that the model outputs actually predict clinical relevance

• Provide explainability artifacts showing how the AI arrived at a specific prediction

• Validate the model on independent data

That phrase—“Explainability Artifacts”—is doing a lot of work. (Federal Register, January 2025) It’s essentially asking: Can you show us your reasoning? Not in pseudocode. Not in a PowerPoint. In evidence that traces from the AI’s input to its clinical prediction.

Then, just last month (January 14, 2026), the FDA and EMA jointly published “Ten Guiding Principles of Good AI Practice in Drug Development.” (PharmaSource, January 2026) While these are currently non-prescriptive and voluntary, they’re the blueprint for future mandatory guidance. The principles emphasize “interpretability, explainability, and predictive performance” and demand “clear, essential information” about AI capabilities and limitations.

The message is unmistakable: Transparency is now table stakes. And it’s not a technicality—it’s a patient safety issue.

Deep learning models that even their creators can’t fully explain—like those used by Insilico Medicine or the newly merged Recursion-Exscientia entity—are now tagged as higher-risk by regulators. (Pinsent Masons, December 2025) That means slower approval timelines and more expensive validation requirements. But that slowdown is intentional. It’s there to protect patients from another REC-994 scenario: a drug that looks good in cells but fails in humans because we didn’t understand what we were actually treating.

The industry is already reacting. According to PharmaSource (December 2025), pharmaceutical companies are now “pulling back on fully autonomous AI design and re-inserting ‘Human-in-the-Loop’ steps just to ensure they have a human scientist who can sign off on liability.”

That sounds like bureaucracy. But it’s actually a return to something older: scientific accountability. Someone has to stake their reputation on the claim that they understand why this drug works. The human-in-the-loop isn’t a bottleneck. It’s a guardrail.

But Here’s the Other Side

Explainability requirements, while well-intentioned, face a fundamental challenge: some machine learning systems may be mathematically impossible to fully explain, even in principle. Deep learning models with billions of parameters operate in ways that resist human-comprehensible interpretation. Demanding explainability artifacts could create a false sense of certainty—a detailed explanation that sounds authoritative without actually being more predictive.

Additionally, pharmaceutical companies are already investing heavily in explainability research. Major biotech firms are hiring machine learning engineers specifically to develop interpretability tools (SHAP values, attention mechanisms, saliency maps). The burden isn’t unreasonable, and the FDA’s phased guidance approach (draft in January 2025, final guidance Q2 2026) gives sponsors time to adapt.

There’s also a practical precedent: regulatory agencies have long accepted empirical efficacy without complete mechanistic understanding. A drug that works is a drug that works, whether or not the mechanism is fully transparent. The risk here may be over-regulation—imposing standards of explainability that are scientifically unrealistic for complex AI systems, thereby slowing down the very innovation the agencies claim to support.

The EU AI Act “Liability Cliff”: August 2, 2026—Who Answers for an AI Mistake?

Now for the existential question: Who is responsible if an AI-designed drug harms patients in ways no one predicted?

On August 2, 2026—less than six months from today—the European Union’s AI Act enforcement provisions for high-risk AI systems become fully applicable. (EU AI Act Service Desk) This is not a guideline. This is law with teeth.

Here’s the problem: When a drug is designed by human chemists and a patient is harmed, the liability chain is clear. The pharmaceutical company knew what they were doing; they either understood the risk or failed to check. But when a drug is designed by an AI system that no one fully understands, the accountability becomes murky:

• Is the pharmaceutical company liable for deploying a black box?

• Is the software vendor liable for building an unexplainable system?

• Is the CRO liable for providing biased training data?

• Is the data broker liable for the “unbiased” patient cohort that wasn’t actually representative?

The EU AI Act answers simply: The pharmaceutical company is liable. Period. (Gardner Law)

Under the Act, high-risk AI systems (which include those supporting healthcare decisions) must be registered in an EU database, undergo conformity assessments, and be monitored for performance drift and bias. Companies must maintain detailed documentation showing how their AI system works and why it makes the decisions it makes.

If a company fails to meet these requirements by August 2, 2026, it faces fines up to €15 million or 3% of global turnover. If the violation involves a prohibited AI practice, it’s up to €35 million or 7%. (Software Improvement Group, January 2026)

But the real cost isn’t the fine. It’s the accountability itself.

Pharma companies are already shifting strategy. According to legal analysis from Pinsent Masons (December 2025), many firms are “de-risking their AI pipelines” by reintroducing governance steps that slow down discovery but create clear audit trails and human accountability. Human scientists are being asked to sign off on AI decisions—not because anyone thinks the human is smarter, but because someone needs to be responsible if things go wrong.

This creates what insiders call “governance debt”: the speed advantage that AI promised starts to vanish. But what’s lost in speed is gained in something older and more important: answerability to patients.

But Here’s the Other Side

The EU AI Act’s August 2, 2026 enforcement date, while tight, does provide a clear deadline that allows companies to plan and prepare. Some industry observers argue that regulatory clarity—even if strict—is preferable to ambiguity. Companies know what they need to do, and they have time to implement it.

Moreover, the Act’s framework allows for proportional risk assessment. Not all AI systems are classified as high-risk; systems used purely in research or early discovery phases are exempt. Companies can also reduce the AI risk classification of a drug by running it through traditional clinical validation pathways, which many are already doing. The impact may be less severe than initially feared.

Additionally, regulatory sandboxes—at least one per EU member state, operational by August 2, 2026—are designed specifically to help companies navigate novel AI-enabled drug development in a supervised environment. These sandboxes create space for innovation while maintaining oversight. And the Act does allow for graduated compliance: pharmaceutical companies deploying AI systems already in use before August 2026 have an extended transition period (until August 2, 2027) if they can demonstrate they’re taking steps toward compliance.

Finally, liability clarity may actually benefit the industry. Right now, responsibility for AI failures is murky; the Act clarifies that the pharmaceutical company is the responsible party. This removes ambiguity and could reduce expensive legal disputes downstream.

The Tension That Defines 2026

So here’s the Black Box dilemma, stripped bare:

LayerThe TensionThe 2026 RealityScientificCan we trust a drug if we don’t know its mechanism of action? Phase 3 results will answer this. REC-994 suggests: maybe not.Regulators want transparency; AI is inherently opaque.FDA and EMA are pushing toward “White Box” models (interpretable AI). Companies are hiring explainability engineers.LegalWho pays for an “AI mistake”?EU AI Act (Aug 2, 2026) forces companies to own the risk, even if they outsourced the AI.

The tension isn’t between Silicon Valley and Big Pharma anymore. It’s between predictive power and human understanding—and we’re in a standoff.

Recursion’s merger with Exscientia (closed November 2024) (Recursion Pharmaceuticals, December 2024) was supposed to create an unstoppable “full-stack” AI drug discovery powerhouse. And maybe it will. But REC-994’s stumble showed that stacking biology engines and chemistry engines doesn’t guarantee that the output will be biologically intelligible—or clinically relevant.

The old guard—the medicinal chemists who’ve been doing this for 30 years—are quietly vindicated. They never trusted black boxes. The young biotech founders are scrambling to make their black boxes look less black. And the regulators have drawn a line: Predict what you want, but explain what you find.

By August 2, 2026, that line becomes law in Europe.

The Epistemological Reckoning: Why Understanding Matters

But here’s the deeper concern, beneath all the regulatory and legal wrangling: What do we lose when drug developers deploy medicines they don’t understand?

Medicine isn’t just about outcomes. It’s about knowledge. When a physician prescribes a medicine, they carry a responsibility to understand—or at least to attempt understanding—why it works. That understanding is what allows them to:

• Adjust dosing when a patient isn’t responding

• Predict side effects based on mechanism

• Design better combination therapies

• Build on the scientific foundation for the next breakthrough

When the AI finds a drug that works but the mechanism is opaque, we’ve compressed the discovery timeline at the cost of scientific depth. We’ve optimized for outcomes at the expense of understanding.

REC-994 is the clearest example. The molecule did something in cells. But without knowing what or why, researchers couldn’t anticipate that it would fail in living humans. They were flying blind—not because they lacked data, but because the AI’s reasoning process was too complex to audit.

This is the risk of the black box: It can lead you to the right answer for the wrong reasons. And when it leads you to the wrong answer, you have no insight into why it failed.

The regulatory push toward explainability—the FDA’s guidance, the EMA’s principles, the EU AI Act’s accountability measures—isn’t just bureaucratic caution. It’s an attempt to preserve something essential: the connection between discovery and understanding.

A medicine that works is a victory. A medicine that works and that we understand is science. There’s a difference. And in 2026, that difference is becoming a dividing line.

What Comes Next

The Black Box Wars aren’t over. They’re escalating.

By August 2, 2026, pharmaceutical companies operating in Europe will face real legal consequences for deploying high-risk AI systems they can’t explain. The FDA’s final guidance on AI credibility is coming in Q2 2026—expected to be more prescriptive than the draft. And patients are becoming aware that the medicines they trust might have been designed by systems no one fully understands.

The industry will adapt. Some companies will invest heavily in explainability research, hiring teams to make their AI systems more interpretable. Others will pull back from fully autonomous discovery, reintroducing the human chemist as a check on the algorithm. A few will double down on black boxes and argue (perhaps successfully) that results matter more than reasoning.

But the core question remains, the one posed in the title: Would you market a drug you can’t explain?

For most of 2026, the answer from regulators, patients, and increasingly from developers themselves is: Only if you’re willing to stake your company’s reputation and liability on it. And only with complete documentation of your reasoning process.

That doesn’t mean the answer is always “no.” It means the answer is conditional—contingent on transparency, accountability, and proof that you’ve done your homework. The industry isn’t being asked to abandon AI drug discovery. It’s being asked to grow up about it. To acknowledge that speed and predictive power are valuable, but not if they come at the cost of scientific accountability.

That’s not a loss. That’s maturity.

I have been monitoring the CAR-T-for-autoimmunity field with growing concern. Not because the science is bad — some of the early lupus data from Erlangen is genuinely remarkable — but because I have seen this movie before. Multiple times. And it doesn’t end well for the patients who need these therapies most.

Let me explain.

The Numbers Don’t Support the Enthusiasm

As of February 2026, fewer than 400 patients worldwide have received CAR-T therapy for autoimmune conditions (my optimistic estimate). The published, peer-reviewed literature as of 2024 comprised approximately 80 patients across 24 studies (Kattamuri et al., Rheumatol Int, 2025). Eighty. That is not a typo.

Using the Rule of Three — a standard biostatistical method published in JAMA (Hanley & Lippman-Hand, 1983) — if zero serious adverse events are observed in 400 patients, the upper bound of the 95% confidence interval on the true adverse event rate is 0.75%. That means we cannot rule out that up to 3 in every 400 patients could experience a serious, potentially life-threatening complication that simply hasn’t shown up yet.

For context: oncology CAR-T required 34,400 treated patients before the FDA identified 22 cases of secondary T-cell lymphoma — a signal at 0.064% that triggered a boxed warning in January 2024 (Marks & Verdun, NEJM 2024). The autoimmune safety database is 86 times too small for equivalent pharmacovigilance. At current enrollment rates, matching that level of safety surveillance would take 15–20 years. Perhaps longer. Perhaps never.

We are flying blind.

Too Many Products, No Market Consolidation — We’ve Seen This Before

Currently, there are more than 100 registered clinical trials of CAR-T for autoimmune diseases (Frontiers systematic review, 2025). Over a dozen companies are developing competing constructs targeting CD19, BCMA, or dual targets, each with slightly different manufacturing processes, conditioning regimens, and vector designs. The fragmentation is staggering.

History tells us what happens next.

In the early 1980s, there were dozens of personal computer manufacturers — Commodore, Atari, Tandy, Osborne, Kaypro, and IBM clones by the hundreds. The market consolidated brutally. Microsoft and Intel survived. Most did not. The technology was transformative, but the companies were largely destroyed.

Smartphones followed the same arc. Remember Palm, BlackBerry, Windows Phone, Symbian, WebOS? By 2013, Apple and Android accounted for more than 95% of the market. The underlying technology thrived. The business landscape was a graveyard.

Henry Ford didn’t invent the automobile. He survived the consolidation of more than 250 American car companies between 1900 and 1920. By 1930, three companies controlled 75% of the U.S. market.

The pattern is iron-clad: transformative technology → irrational proliferation → brutal consolidation → a few winners, many casualties. CAR-T for autoimmunity is deep in phase two. The question is not whether consolidation will happen, but how many patients will be mid-treatment when it does.

The “Dumb Capital” Problem

Between 2021 and 2025, cell therapy startups collectively raised more than $15 billion through venture capital and public offerings. Much of this money was not invested by groups with deep immunology expertise or realistic timelines. It was speculative capital chasing the next platform story — the same capital that inflated and then abandoned gene therapy, microbiome therapeutics, and digital health before that.

When this bubble pops — and the Rule of Three analysis above suggests the safety reckoning is a matter of when, not if— the collateral damage will extend far beyond the companies that fail. The entire cell therapy ecosystem will suffer. Investors will retreat. Regulatory scrutiny will intensify. And the patients with refractory lupus, systemic sclerosis, or myositis who desperately need these treatments will find themselves stranded.

We observed precisely this in the field of gene therapy following the death of Jesse Gelsinger in 1999. It took nearly 15 years for clinical gene therapy to recover. Fifteen years of patients waiting because one preventable catastrophe poisoned the well.

“But What About the Alternatives?”

Some colleagues point to emerging competitors—bispecific antibodies, CD19-targeting monoclonals, CAR-NK cells, and in vivo CAR approaches—as evidence that the field will be fine even if autologous CAR-T cells stumble. They argue these alternatives are cheaper, more scalable, and avoid the manufacturing bottleneck.

They are not wrong about the short-term advantages. Bispecific T-cell engagers don’t require leukapheresis, clean-room manufacturing, or a two-week wait. They cost a fraction of the cost of CAR-T. For a budget-constrained rheumatology market, that is a compelling pitch.

But here is what the “alternatives will save us” argument misses: if a safety signal emerges in CAR-T — particularly one involving T-cell malignancy or delayed autoimmune flare — it will not be contained to CAR-T alone. The regulatory and public perception fallout will splash across every therapy that involves engineered immune cell activation. The FDA will not distinguish between CAR-T and CAR-NK when a CNN headline reads “Cancer Therapy Turns Deadly in Lupus Patients.” Fair or not, that is how it works.

My prediction from two years ago still holds: without rigorous, adequately powered safety monitoring — a proper pharmacovigilance infrastructure built before the crisis, not after — this field is heading for a correction that will hurt everyone. Patients. Investors. Developers. The science itself.

So, What Should We Do?

I have specific, actionable recommendations. A unified safety registry. Bayesian adaptive monitoring frameworks. Minimum reporting standards. A realistic conversation about what 400 patients can and cannot tell us. And a framework for how investors should evaluate safety risk—not just efficacy upside—in their due diligence.

But that is Part 2.

Coming soon.

Eswar Krishnan, MD | www.drugdevelop.com

Data sources: Hanley & Lippman-Hand, JAMA 1983; Kattamuri et al., Rheumatol Int 2025; Mueller et al., ASH 2024; Marks & Verdun, NEJM Jan 2024.

#CARTtherapy #CellTherapy #Autoimmune #Pharmacovigilance #DrugSafety #Biostatistics #ClinicalTrials #Rheumatology #Lupus #DrugDevelopment #BiotechInvesting #VentureCapital