Published on 16/11/2025
Building Eligibility and Enrichment Strategies that are Ethical, Scientific, and Inspection-Ready
From Target Treatment Population to Eligibility Architecture
Eligibility criteria are the gatekeepers of your evidence. They determine who can enter the study, what the resulting population represents, and how credibly results generalize to real-world use. Well-designed criteria balance participant protection, internal validity, and external validity—and they do so transparently. This balance is grounded in Good Clinical Practice under the ICH (E6[R3] quality by design; E8[R1] fit-for-purpose quality), supported by regulators including the U.S. Start with a precise definition of the target treatment population. Describe the condition, disease stage, line of therapy, concomitant standard of care (SoC), and clinically meaningful subgroups (age bands, organ function, comorbidities). Map disease epidemiology to identify groups that must be represented to support labeling and payer decisions. This map informs which exclusions are necessary for safety or interpretability—and, critically, which “convenience exclusions” to remove. Write inclusions that identify true need. Inclusion criteria should capture diagnostic certainty, severity thresholds, and baseline stability needed to interpret endpoints, with clear source data and timing (e.g., “laboratory confirmation within 14 days prior to randomization”). For multi-region programs, anchor definitions to internationally recognized frameworks or specific test kits/assays and provide equivalency tables to avoid regional drift. Challenge every exclusion with three tests: (1) Safety—is there evidence of unacceptable risk at screening or during intervention? (2) Scientific integrity—does the exclusion avoid unmanageable confounding or endpoint interference? (3) Feasibility—could we instead mitigate risk with monitoring or rescue rules rather than exclude? If an exclusion only makes logistics easier or enriches responders without a principled rationale, it threatens justice and external validity. Document this analysis in a short “eligibility rationale memo” filed in the Trial Master File (TMF). Operational specificity prevents screening chaos. Name acceptable tests (with units and reference ranges), specify allowable windows, and point to adjudication rules for borderline values. Provide decision trees for complex cases (e.g., controlled atrial fibrillation, prior malignancy in remission) so sites make consistent calls. Align windows with the Schedule of Assessments to avoid re-testing churn and screen failures that disproportionately affect participants with limited resources. Link eligibility to your estimand strategy. Per ICH E9(R1), the estimand defines how intercurrent events (ICEs) are handled. Avoid exclusions that merely remove expected ICEs (e.g., excluding anyone likely to need rescue therapy) if your primary estimand is treatment-policy and intends to reflect rescue in the effect definition. Instead, capture ICEs rigorously and analyze as designed. This maintains interpretability and ethical access. Ethics and independence. IRBs/IECs expect that eligibility supports fair selection and respect for persons. Keep consent language consistent with eligibility burdens (e.g., contraception), risks of screen failure, and any run-in steps. Ensure vulnerable populations are not excluded reflexively; tailor safeguards instead. These principles are recognized across FDA/EMA/PMDA/TGA/WHO oversight cultures. Safety-protective, not overprotective. Exclusions should address specific, demonstrated risks. Replace blanket categories with nuanced thresholds and mitigation. Examples: Diagnostic certainty and misclassification control. Define reference standards (e.g., imaging sequences, assay platforms) and centralized confirmation where needed. For biomarkers, specify cut-points, sample handling, and retest rules. Provide contingency for indeterminate results to prevent inadvertent exclusion due to pre-analytical error. Run-in periods—use sparingly and transparently. Placebo or adherence run-ins can reduce variability, but they also select for adherent, side-effect–tolerant participants, potentially inflating effect sizes and undermining generalizability. If used, justify scientifically, keep them short, and disclose attrition in public summaries and the clinical study report. Consider alternative variance controls (training, centralized reading) before choosing run-ins. Age, sex, and reproductive status. Explicitly include older adults unless clear risk or interpretability issues exist; adjust visit burdens instead of excluding. Include women of childbearing potential with proportionate safeguards and respect gender-diverse participants with clear, inclusive language for reproductive risk and contraception. Special populations. For pediatrics, align with ICH E11/E11A and ensure age-appropriate assent/permission and weight- or surface area–based dosing. For hepatic/renal impairment, plan dedicated cohorts or sub-studies if excluded from the main protocol to avoid a post-approval evidence gap. Device and combination-product nuance. For device trials, eligibility must cover anatomical suitability and prior device exposure; for drug–device combinations, include usability or training criteria that protect safety without filtering out people who can be trained effectively. Maintain human factors documentation to support criteria rationale. Operational equity matters. Eligibility that requires frequent, narrow-window testing or specialty procedures risks excluding participants with limited access. Consider decentralized options (home phlebotomy, local labs) where valid, and offer support (transport, childcare) to avoid disproportionate screen failures. This is consistent with public-health ethics endorsed by the WHO and scrutiny recognizable to FDA/EMA. Prescreening and algorithmic fairness. If electronic health record (EHR) queries or algorithms identify candidates, validate that they do not systematically miss language groups, older adults, or those with less frequent visits. Keep specifications, fairness checks, and corrective actions on file. Record “eligible but not approached” with reasons to detect selection bias. Document consistency. Synchronize eligibility wording across protocol, IRT checks, eConsent, site job aids, and registries. Mismatches are common inspection findings. Provide example scenarios in the site manual (e.g., borderline QTc, prior therapies) with escalation paths to medical monitors. Enrichment improves the chance of detecting a true effect by selecting participants who are more likely to have events (prognostic enrichment), more likely to respond (predictive enrichment), or more consistently measured (practical enrichment). The strategy should match your clinical question and estimand, be justified in the protocol, and be evaluated for its impact on labeling and real-world applicability. Prognostic enrichment (risk-based). Enroll individuals at higher baseline risk of the primary outcome to increase event rates (e.g., specific risk scores, biomarkers of disease activity). Define risk algorithms prospectively and ensure they are feasible in routine care; otherwise, post-approval use may not reflect the trial population. Consider stratification by risk bands to protect balance and enable interpretable subgroup summaries. Predictive enrichment (biomarker-defined responders). Restrict to or weight enrollment toward biomarker-positive populations where mechanism supports differential response (e.g., genetic variants, receptor expression). Pre-specify assay methods, cut-points, turnaround time, and specimen logistics. Clarify whether the biomarker is integral (required for enrollment) or integrated (used for prospective stratification/analysis). Discuss how negative or indeterminate biomarker results are handled. Align justification with scientific plausibility and regulatory expectations recognizable to FDA and EMA. Practical enrichment (variance reduction). Improve measurement precision without narrowing who can participate: central read for imaging, training/certifying raters for ClinROs, standardized device calibration, and stable background therapy periods where clinically indicated. These steps can increase power without excluding populations. Adaptive enrichment. Some designs allow preplanned restriction to a responsive subgroup at interim based on predictive markers. To remain credible, define decision criteria, maintain Type I error control (e.g., combination tests/closed testing), and update the primary estimand if the target population changes. Keep an Adaptation Specifications Document cross-referenced in the Statistical Analysis Plan and TMF. Multiplicity and labeling implications. If you plan confirmatory inference in multiple populations (overall and biomarker-positive), pre-specify a testing hierarchy or alpha-sharing scheme with strong familywise control. Closed testing or graphical alpha approaches are common. Explicitly state how labeling and clinical guidance will reference results (e.g., indicated only for biomarker-positive patients). Assay and logistics readiness. Predictive enrichment hinges on reliable testing. Qualify labs, validate assays across regions, and ensure sample transport times support enrollment. Consider a reflex testing strategy at prescreening to avoid delays. Maintain chain-of-custody and proficiency testing records—routine inspection pulls for programs spanning the PMDA and TGA jurisdictions. Equity lens on enrichment. Prognostic or predictive filters can unintentionally exclude underserved groups (e.g., biomarkers discovered in homogenous cohorts). Test performance across demographic strata and languages; mitigate with inclusive cut-points or parallel cohorts if appropriate. Transparently report screen-failure reasons and demographics in the CSR and lay summaries, consistent with the WHO transparency ethos. Power, sample size, and event rates. Enrichment changes assumptions. Recalculate power under enriched prevalence and event models, account for screen-failure inflation on timelines/budget, and ensure drug supply aligns with revised accrual patterns. Document scenarios in a simulation appendix; regulators value evidence that operating characteristics remain robust under realistic variability. External generalizability safeguards. If enrichment narrows the population, plan complementary evidence: pragmatic or RWE studies, post-approval registries, or bridging cohorts. In your benefit–risk narrative, explain how the enriched evidence informs real-world decisions and where uncertainty remains. Screening discipline is your inspection story. Maintain a unified screening log covering prescreened, approached, consented, screen-failed, randomized, and non-randomized eligibles—with standardized reasons (medical vs. logistics vs. consent declines vs. biomarker ineligible). Record interpreter use, accommodations provided, and algorithmic prescreening flags. Inspectors frequently ask for this log first. Misclassification and overrides. Define how to adjudicate borderline cases and when medical monitors may override strict thresholds (with written justification). Keep an audit trail and harmonize with statistical analysis (e.g., covariate adjustment using actual baseline values even if the stratification category entered in IRT was wrong). Amendments with control. As SoC evolves or feasibility signals emerge, eligibility may need adjustment. Use a version-controlled decision log: rationale, safety/scientific basis, alternatives considered, impact on estimand and power, IRB/IEC and authority submissions, re-consent decisions, translations, site retraining, and IRT updates. Synchronize registries and public summaries to maintain transparency. Training and job aids. Provide role-specific training: coordinators on source verification for key criteria; investigators on adjudication of complex comorbidities; pharmacists on prohibited med checks; lab staff on biomarker handling; raters on ClinRO criteria; and call centers on neutral scripts. Keep attendance logs and competency checks in the TMF. Centralized monitoring and QTLs. Trend screen-failure rates, top failure reasons, time from prescreen to randomization, and approach rates by subgroup/site. Define Quality Tolerance Limits (e.g., ≤10% logistics-driven screen failures after accommodations; ≥90% approach of eligibles; biomarker turnaround within X days; mis-stratification ≤1%). Breaches trigger CAPA and targeted site support. Privacy and data governance. Eligibility often requires sensitive data and specimens. Align privacy notices and HIPAA/GDPR/UK-GDPR artifacts with what you collect for screening. Use coded IDs and honest-broker models for biomarker testing where possible. Contracts with labs should fix hosting regions and incident response SLAs—expectations recognizable to FDA, EMA, PMDA, and TGA. Public transparency and community trust. Register trials before enrollment; update records when eligibility or enrichment changes materially; and provide lay summaries that clearly explain who was included and why. Where enrichment limits broad applicability, articulate plans for further evidence. This supports public trust in line with the WHO transparency principles. What to file—fast retrieval index. Ready-to-use checklist (actionable excerpt). Bottom line. Eligibility criteria and enrichment are powerful levers. When designed with ethical intent, scientific rigor, and operational realism—and proven by impeccable documentation—they produce evidence that regulators trust, clinicians can apply, and participants can access fairly across the U.S., EU/UK, Japan, and Australia.Writing Eligibility That is Ethical, Scientific, and Feasible
Enrichment Strategies that Increase Power without Sacrificing Credibility
Execution Evidence: Logs, Governance, and a Practical Compliance Checklist