Who Designs the Trial Is Now a Regulatory Variable

A 15-year audit in the Journal of Clinical Epidemiology finds that across four high-impact journals from 2008 to 2023, authorship from low- and middle-income countries did not keep pace with LMIC participant enrollment in the same trials (JCE S0895-4356(26)00211-8). A companion paper in the same journal audits SARS-CoV-2 vaccine trials on ClinicalTrials.gov and shows that older adults — 76% of COVID-19 deaths — are screened out by eligibility criteria the authors call “exclusion by design”. Read together, these are not EDI think-pieces. They are evidence that the target population of regulatory interest and the population actually studied are routinely different populations, and that the people deciding which is which often sit nowhere near the participants.

What the two papers actually measure

The authorship study is a straightforward proportionality test: LMIC participants in, LMIC authors out, across NEJM, Lancet, JAMA, and BMJ. The gap persists across the 15-year window, with no clear trend toward closure. The implicit benchmark is the ICMJE authorship criteria — substantial contribution to design, analysis, or interpretation — which means the authorship gap is also a plausible proxy for who writes the SAP and owns the analytic decisions. That is the version biometrics leaders should care about. It also lines up uncomfortably with Derek Lowe’s reporting that NIH and NASA grantees are now being directed to strip foreign-collaboration language from active awards — a mechanism for the gap to widen, not close.

The “exclusion by design” paper is more operationally direct. The authors extract eligibility criteria from hundreds of registered SARS-CoV-2 vaccine trials and catalogue the restrictions — comorbidity caps, concomitant medication exclusions, frailty and functional thresholds — that disproportionately remove the ≥65 cohort. The methodological consequence is an ICH E9(R1) problem stated in the vocabulary regulators already use: the population attribute of the estimand does not match the population in whom benefit-risk will be inferred. Bridging via external controls or RWE only relocates the problem, because eligibility-criteria misalignment is exactly what makes synthetic-control transportability assumptions fragile.

Why this lands now

The regulatory frame is already in motion. FDA’s June 2024 draft guidance on Diversity Action Plans, issued under FDORA §3604, requires sponsors to submit a DAP before Phase 3 or other pivotal trials, and ICH E17 puts multi-regional design squarely in scope. A third JCE paper this cycle — a focused mapping review — concludes that despite the rhetoric, “practical guidance for applying inclusion principles across all clinical trial stages remains limited.” The gap between guidance posture and operational specifics is precisely where sponsors get questions from review divisions.

The geopolitical backdrop sharpens the operational stakes. WHO declared the Bundibugyo ebolavirus outbreak in DRC and Uganda a PHEIC in mid-May; CEPI has put $62M behind three candidate vaccines, with trials still months away and no licensed Bundibugyo vaccine in existence (Ervebo covers Zaire only). At the same time, STAT reports that terminated NIH grants to a 10-center emerging-infections network — $82M committed, ~$14.9M unspent at termination — severed the international relationships that would normally seed local statistical and investigator capacity. The trials that get designed under these conditions will be the textbook case the authorship paper is measuring, in real time.

The operational read

None of this requires a SAP rewrite this quarter. It does require sponsors to expect that DAP reviewers will read the eligibility table the way the “exclusion by design” authors did — as a list of structural exclusions to defend, not a list of safety thresholds to wave through. And it requires biometrics leaders to take a position on who is named on the analysis. Authorship is not a vanity question when the FDA is asking who designed the trial for the population it is meant to serve. The methodological community is already responding on the design side — see Novel Statistical Designs and Considerations to Support Diversity and Inclusion in Clinical Trials — but the audit data say the conduct side has not caught up.

Protocol read: Diversity in trials is migrating from ethics framing to estimand framing, and sponsors who cannot explain the gap between their enrolled population and their target population will be explaining it to a reviewer instead.

What to do now:

• Audit Phase 3 eligibility criteria against target-population demographics before DAP submission; flag age, comorbidity, and concomitant-med exclusions that lack a safety rationale.
• Pre-specify estimand population attributes that name the geriatric or LMIC subgroup of regulatory interest, not just the ITT population as enrolled.
• For trials with LMIC sites, document analytic ownership — who writes the SAP, who signs the CSR — and treat persistent HIC concentration as a finding to address, not a default.