ICH E21 "Inclusion of Pregnant and Breastfeeding Individuals in Clinical Trials"

US FDA has just endorsed the draft guideline of ICH E21 "Inclusion of Pregnant and Breastfeeding Individuals in Clinical Trials". E21 Training Material can also be found at ich.org. 

The ICH E21 guideline aims to provide recommendations for the appropriate inclusion and retention of pregnant and breastfeeding individuals in clinical trials. This is crucial because historically, these populations have been largely excluded, leading to a significant lack of data on the safety and efficacy of medicinal products during pregnancy and lactation.

For clinical trialists, the guideline emphasizes:

• Proactive Planning: Sponsors should consider strategies to generate data on medicinal product use in pregnancy and breastfeeding from early stages of development, including nonclinical studies through post-approval.

• Risk-Benefit Evaluation: A thorough assessment of potential risks and benefits for both the pregnant or breastfeeding individual and the fetus/breastfed infant is essential.

• Ethical Considerations and Informed Consent: The guideline stresses the importance of obtaining truly informed consent, ensuring clear communication of potential risks and benefits, and avoiding any undue influence or coercion. Efforts should be made to reduce the burden of study procedures on these participants.

• Study Design and Implementation: It provides practical guidance for designing clinical trial protocols, including considerations for data collection, safety monitoring, and follow-up for both mother and child. This also includes pharmacokinetic (PK) and dosing considerations, as physiological changes during pregnancy can impact drug pharmacokinetics.

• Addressing Unmet Medical Need: The inclusion of pregnant and breastfeeding individuals is particularly important for conditions where there is a high unmet medical need for treatment in pregnancy or while breastfeeding, though the guideline's scope is not limited to these scenarios.

• Data Generation: The overarching objective is to facilitate the generation of robust clinical data that allows for evidence-based decision-making on the safe and effective use of medicinal products by these individuals and their healthcare providers.

The ICH E21 guideline is a significant step towards improving the evidence base for prescribing medications during pregnancy and lactation, ultimately aiming for better patient outcomes for these underserved populations.

Just several weeks ago, I wrote a blog post on this same topic "Managing Pregnancies in Clinical Trials: Regulatory Guidance and Best Practices", which offered a comprehensive overview of the historical exclusion of pregnant women from clinical trials, the reasons behind this, and the evolving regulatory landscape advocating for their inclusion. The article emphasized a balanced approach, highlighting FDA's shift towards diversity in trials and the ethical imperative to gather data on drug use in pregnancy to ensure evidence-based care. It delved into practical aspects like informed consent, pre-screening procedures, managing pregnancies discovered during trials (including stopping drug, unblinding, and re-consenting), and the importance of detailed data collection and specialized safety monitoring. The blog post also strongly argued for the ethical benefits of inclusion, asserting that systematic exclusion leads to greater harm by forcing healthcare providers and patients to make decisions without adequate evidence. 

The newly endorsed ICH E21 guideline directly aligns with various FDA guidance described in the blog post and significantly reinforces the "balanced approach" and the ethical arguments. 

• ICH E21 provides a more harmonized, international standard for how this inclusion should be achieved proactively throughout drug development. 
• ICH E21 offers detailed recommendations on study design, ethical considerations, pharmacokinetic and dosing adjustments, safety monitoring, and data collection specifics for pregnant and breastfeeding individuals, directly addressing the "when and how" question you pose at the end of your article. 
• ICH E21 reiterates the need for IRBs with maternal-fetal expertise, suggesting consultation with specialists for study design and safety monitoring. 
• ICH E21 highlights the critical importance of robust informed consent and ongoing safety monitoring for both the mother and the child. 
• ICH E21 essentially provides the detailed, internationally agreed-upon framework, moving the industry further towards systematic and responsible inclusion rather than historical exclusion.

Rank Analysis of Covariance (RANK ANCOVA) versus Aligned Rank Stratified Wilcoxon Test

In clinical trials where outcome measures violate the assumptions required for parametric statistical methods such as analysis of covariance (ANCOVA), non-parametric approaches are often employed. The classical two-sample Wilcoxon rank-sum test (also known as the Mann–Whitney U test) can be used in such cases; however, this method does not allow for adjustment of covariates.

Randomized clinical trials often adjust for baseline covariates (FDA 2021 guidance "Adjusting for Covariates in Randomized Clinical Trials for Drugs and Biological Products") to improve precision. When the outcome is non‐normal or ordinal, rank‐based methods offer alternatives to parametric ANCOVA. Two popular approaches are: (1) Rank ANCOVA – an ANCOVA on rank‐transformed data (the “rank‐transformation ANCOVA” of Conover and Iman), and (2) Aligned‐Rank Stratified Wilcoxon – a Wilcoxon rank‐sum test applied to responses “aligned” by covariate effects (e.g. Hodges‐Lehmann or van Elteren style). Below we compare their theory, practical use, pros/cons, and SAS implementations.

Theoretical Properties

• Rank ANCOVA (Rank‐Transform ANCOVA): 

Replace the outcome Y by its overall ranks and fit a usual linear ANCOVA model (treatment + covariates). Under the null it tests whether the treatment coefficient in the rank‐scale model is zero. In effect, the null is “no difference in the adjusted rank‐distributions” between groups. This approach assumes a linear additive model for covariate effects (on the rank‐scale) with homogeneous slopes across treatments. Conover and Iman showed the rank‐transform inherits the robustness and power properties of rank tests in regression. However it is not fully model‐free: violations of slope homogeneity or covariate–treatment interactions can invalidate the test. The test statistic is simply the usual F (or t) from ANCOVA on ranks. Rank ANCOVA was proposed by Dana Quade (1967) "Rank Analysis of Covariance" in JASA and was popularized after the book by Stokes, Davis, and Koch "Categorical Data Analysis Using SAS" where there was a dedicated chapter to discuss the RANK ANCOVA.

• Aligned‐Rank Stratified Wilcoxon: 

Agligns outcomes within each randomization stratum by subtracting the stratum’s Hodges–Lehmann shift (a median-based location estimate) and then applies a Wilcoxon rank-sum test to the pooled aligned data. Effectively performs an un-stratified Wilcoxon test on aligned values. Yields a Hodges–Lehmann estimate of the overall median shift (with confidence interval) as the treatment effect.

 This method is fundamentally a rank‐sum test controlling for strata or covariates. In a stratified design (or ANCOVA context), one assumes a model Yij=μ+βi+τjY_{ij}=\mu +\beta_i+\tau_j where βi\beta_i are strata/covariate effects and τj\tau_j are treatment effects. The null hypothesis is τ1=τ2=0\tau_1=\tau_2=0 (no treatment shift). To remove (align) the strata effects, one subtracts a location statistic (stratum mean/median/Hodges–Lehmann) from each YijY_{ij}, yielding “aligned” responses. These aligned values (now centered by stratum) are pooled and ranked, and a Wilcoxon test is performed ignoring strata. Equivalently, with a STRATA factor one can perform a Van Elteren (stratified Wilcoxon) test. The resulting null distribution is distribution‐free (asymptotically normal, exact via permutations), requiring only that within each stratum the treated and control distributions differ by a location shift. In summary, Rank ANCOVA assumes a linear rank‐model for covariate adjustment, whereas the Aligned‐Rank Wilcoxon assumes only an additive strata effect (with no specific distribution form) and tests a location‐shift null in a stratified rank framework.

Practical Applications

• Outcome types: Both methods suit continuous or ordinal outcomes that violate parametric assumptions. Rank ANCOVA can handle multiple continuous covariates straightforwardly (since one simply adds them to the ANCOVA on ranks). Aligned‐rank tests naturally handle categorical strata (e.g. randomization factors, center, or baseline strata). If the covariate is continuous (e.g. baseline measure), one may either form strata (e.g. quantiles) or perform alignment via regression residuals before ranking.

• Robustness: Both are robust to outliers and non‐normality because they use ranks. Conover and Iman noted that rank‐transformed regression is robust and powerful even under heavy‐tailed or skewed data. The aligned‐rank approach is fully distribution‐free under its null hypothesis, and by aligning removes nuisance location effects (e.g. site or block means). In practice, Ye and Lai (2023) found that both a covariate‐adjusted rank‐sum (rank ANCOVA) and an aligned‐rank test yielded narrower confidence intervals and maintained type I error across clinical trials, compared to unadjusted tests.

• Small samples: Neither method “magically” solves small‐N issues. Rank ANCOVA relies on large‐sample ANCOVA (normal‐theory) approximations on ranks. In contrast, the aligned‐rank/Wilcoxon method can use exact computations (via PROC NPAR1WAY) if samples are very small. However, note that alignment can behave erratically in very small samples according to some studies. In moderate samples both methods are generally acceptable. Both methods lose some efficiency if there are many ties or if covariate effects are very non‐linear.

Advantages and Disadvantages

• Rank ANCOVA 

– Advantages: It is conceptually simple (just rank the data and run standard ANCOVA). It fully utilizes continuous covariates in a regression framework, and retains much of the power of ANOVA tests while being robust to non‐normal errors. If the linear model is correct (in the rank‐scale), the test is valid and can easily test interactions. Simulations suggest rank‐based ANCOVA is often powerful when parametric assumptions fail. Adjusting for covariates in this way typically yields more precise estimates (narrower CIs) than unadjusted ranks.

– Disadvantages: Its interpretation is subtle: one is testing effects on the ranks of the outcome, not its mean or median in original units. Thus the “treatment effect” corresponds to a location shift in the ranked distribution. Experts caution that rank‐transform ANCOVA is not a direct test of medians in the original scale, and may give misleading inference if reporting medians. It also implicitly ranks the covariates (if one ranks them too) which can distort relationships. If the homogeneity of slopes assumption is violated, the rank‐ANCOVA test can be invalid. In summary, it is not fully nonparametric – it still relies on the linear model structure (albeit on ranks) and on large‐sample approximations.

• Aligned Rank Wilcoxon 

– Advantages: This method is fully nonparametric under a location‐shift null, and can produce an easily interpreted Hodges–Lehmann median shift estimate with confidence limits (e.g. via ALIGN=STRATA(HL) in SAS). It inherently accounts for stratification/baseline effects by alignment, so it naturally handles block effects or randomization strata (van Elteren’s approach). It does not require ranking the covariate itself – only the outcome after centering. In practice it has robust Type I error even under heteroscedasticity or skewness, so long as the alignment model is reasonable. SAS’s PROC NPAR1WAY provides this test directly (see below).

– Disadvantages: One must specify the strata or alignment model in advance. If there is only one continuous covariate, one needs to decide how to align (e.g. subtract predicted baseline effect or stratum median). The choice of alignment statistic (median vs mean vs HL) can affect results slightly. If strata are very small, the within‐stratum rankings may be unstable. Unlike rank‐ANCOVA, this approach cannot easily incorporate arbitrary continuous covariates without discretization or pre‐alignment. In very small samples, alignment procedures can suffer from erratic Type I behavior (especially if alignment assumptions are misspecified). Finally, it is less familiar to many practitioners and thus may be harder to explain.

SAS Implementation

• Rank ANCOVA: SAS does not have a dedicated “rank‐ANCOVA” proc. A common workaround is to rank the data (using PROC RANK) and then run a standard ANCOVA (PROC GLM or PROC REG) on the ranked outcome. For example:

proc rank data=trial out=ranked ties=mean;

  var Y; ranks Y_rank;

run;

proc glm data=ranked;

  class Trt;

  model Y_rank = Trt Baseline;

run;

This fits a linear model to the ranks. Alternatively, one can implement the Hettmansperger–McKean aligned‐rank procedure by first regressing Y on the covariate (e.g. with PROC REG or PROC ROBUSTREG) and then ranking the residuals to test the treatment effect. In short, one must manually rank or residualize and then use standard SAS procs. (There is no built‐in PROC RANKANCOVA or similar in SAS.) 

The book "Categorical Data Analysis Using SAS®, Third Edition" by Stokes, Davis, and Koch contains the sample SAS codes indicating three steps in performing the Rank ANCOVA. 

proc rank nplus1 ties=mean out=ranks;

   by center; 

   var before after;

run;

proc reg noprint; 

   by center;

   model after=before;

   output out=residual r=resid;

run;

proc freq;

   tables center*group*resid / noprint cmh2; 

run; 

 

• Aligned Rank Stratified Wilcoxon: SAS’s PROC NPAR1WAY supports both stratified Wilcoxon and aligned‐rank tests. For a Van Elteren stratified Wilcoxon, use the STRATA statement without alignment. For example:

proc npar1way data=trial wilcoxon;

  class Trt;

  strata Center;  /* stratification variable */

  var Change;

run;

This computes the stratified Wilcoxon (van Elteren) test and provides both one‐sided and two‐sided p-values. To perform an aligned‐rank test, use the ALIGN=STRATA option (with STRATA) in PROC NPAR1WAY. For example:

proc npar1way data=trial wilcoxon align=strata(hl);

class Trt;

strata Center;

var Change;

run;

This subtracts the stratum median (or Hodges–Lehmann shift if (HL) is specified) from each response before ranking, then conducts the Wilcoxon test. The output will include the Hodges‐Lehmann estimate and CI for the location shift. Thus, PROC NPAR1WAY with the STRATA and ALIGN=STRATA options directly implements the aligned‐rank stratified Wilcoxon test. (By default, RANKS=STRATUM is used and weights are by stratum, yielding van Elteren.)

Notice that for both RANK ANCOVA and Aligned Rank Stratified Wilcoxon test, only p-value will be obtained. The treatment difference (so called the difference in medians or location shift in medians) needs to be calculated using Hodges-Lehman estimator.  

Side-by-Side Comparison of RANK ANCOVA and Aligned Rank Stratified Wilcoxon

The table below summarizes these points:

	Rank ANCOVA (Rank‐transform ANCOVA)	Aligned‐Rank Stratified Wilcoxon
Null hypothesis	H₀: no treatment effect on the ranked outcome (treatment coeff=0 in rank‐scale model).	H₀: no treatment effect (no location shift) in stratified model (τ1=τ2=0\tau_1=\tau_2=0).
Model / Assumptions	Linear model on ranks: assume covariate effects are additive and linear on the rank‐scale, with equal slopes across groups. (No distributional form assumed beyond this.)	Additive strata model: Y=μ+βi+τjY=\mu+\beta_i+\tau_j. Assume observations are exchangeable within strata after alignment. Does not assume specific distribution shape.
Test statistic	ANCOVA F or t on the ranked outcome (i.e. usual parametric test applied to ranks).	Wilcoxon rank‐sum on aligned data. (Equivalently, stratified Wilcoxon/Van Elteren statistic.)
Distribution	Uses large‐sample normal/chi‐square approximations (from GLM on ranks). No exact test available in SAS for this.	Asymptotic normal (z) or exact (via permutation) available. SAS PROC NPAR1WAY can compute exact Wilcoxon p-values within strata.
Outcome type	Continuous or ordinal outcomes. Can include multiple continuous covariates in model.	Continuous or ordinal outcomes. Requires (or creates) strata: typically categorical covariates (e.g. randomization strata) or aligned by regression.
Robustness / Outliers	Not fully nonparametric. Robust to outliers and non‐normality (rank‐based). However, if covariate–treatment interactions exist or slope equality fails, Type I error can inflate.	Fully nonparametric. Robust to outliers (rank‐based) and handles non‐normal/heteroscedastic data well. Alignment removes nuisance location shifts.
Small sample	Relies on asymptotic ANCOVA on ranks; no built‐in exact test. May be liberal if sample is very small or distribution very discrete.	PROC NPAR1WAY can use exact Wilcoxon (by strata) for small N. Alignment in tiny samples may have less stable Type I.
Advantages	Easy to implement via standard ANCOVA tools. Uses full continuous covariate information. Retains high power under model correctness. Covariate adjustment usually reduces variance (narrower CIs).	Nonparametric (distribution‐free) test. Directly yields Hodges–Lehmann shift estimate and CI. Naturally incorporates stratification/blocks (van Elteren). Valid under mild assumptions.
Disadvantages	Tests on ranks, not raw scale, so interpretation of effect size is not straightforward. Not a test of medians in original units. Can mislead if model assumptions fail. No simple SAS proc – must manually rank or regress.	Must predefine strata or alignment model. Less flexible for multiple continuous covariates (usually one strata factor). Alignment choice (median vs mean) can affect results. In small samples, alignment may behave poorly.
SAS implementation	No single procedure. Typically use PROC RANK to create ranked Y, then PROC GLM (or PROC REG) with covariates on ranks. Alternatively, regress Y on covariate (PROC REG/ROBUSTREG), rank the residuals, and test group difference. (All manual steps.)	Use PROC NPAR1WAY. For stratified Wilcoxon: STRATA statement (no ALIGN) yields Van Elteren test. For aligned‐rank: add ALIGN=STRATA (and optionally (HL) or (MEAN) option) in PROC statement. E.g.  proc npar1way data=… wilcoxon align=strata(hl);  class Trt;  strata Covar;  var Y; run;.

Some Studies Where the Rank ANCOVA or Aligned Rank Stratified Wilcoxon Were Used:

• Hoeper et al (2023) "Phase 3 Trial of Sotatercept for Treatment of Pulmonary Arterial Hypertension" and details in the attached protocol and SAP - Aligned Rank Stratified Wilcoxon was used. 
• Statistical Analysis Plan for "A Phase 2b, Dose-Ranging, Randomized, Double-Blind, Placebo Controlled, Multicenter Study of Rodatristat ethyl in Patients with Pulmonary Arterial Hypertension" - Aligned Rank Stratified Wilcoxon was used. 
• Van Dyck et al (2022) "Lecanemab in Early Alzheimer’s Disease" and details in the attached protocol and SAP - rank analysis of covariance was used for sensitivity analysis to assess the robustness of the primary efficacy analysis using MMRM
• NDA 22,535/0045 Pirfenidone in the treatment of IPF - Rank ANCOVA was used
• NDA 207533/ O-1 ARISTADA for Treatment of Schizophrenia - Rank ANCOVA was used

ICH E20 "Adaptive Designs For Clinical Trials" and FDA Guidance "Adaptive Designs for Clinical Trials of Drugs and Biologics"

The International Council for Harmonisation (ICH) has released its highly anticipated draft guideline that harmonizes principles for the design, analysis, and reporting of adaptive clinical trials in new drug development. ICH E20 "Adaptive Designs for Clinical Trials" was endorsed by ICH last week on June 25, 2025. This was 5 1/2 years after FDA's guidance for industry "Adaptive Designs for Clinical Trials of Drugs and Biologics".

Both documents aim to provide guidance on the use of adaptive designs in clinical trials. The ICH E20 guideline is a draft version, and is currently under public consultation. It emphasizes that until a final guideline is agreed upon, the draft should not be understood as confirming full regulatory acceptance or superseding current regional guidance. The FDA guidance, issued in November 2019, represents the FDA's current thinking and provides nonbinding recommendations.

Here's a side-by-side comparison of common topics and key differences:

Side-by-Side Comparison: ICH E20 Draft Guideline vs. FDA Guidance

| 1. Purpose and Scope | 

ICH E20: Provides guidance on confirmatory clinical trials with an adaptive design within the context of its overall development program. Defines adaptive design as "a clinical trial design that allows for prospectively planned modifications to one or more aspects of the trial based on interim analysis of accumulating data from participants in the trial." Emphasizes "prospectively planned". Excludes unplanned modifications or changes based on external information, and routine operational monitoring. Focuses on principles for planning, conduct, analysis, and interpretation of trials to confirm efficacy and support benefit-risk assessment. While primarily for confirmatory trials, principles are relevant to all phases. |

FDA: Provides guidance to sponsors on the appropriate use of adaptive designs for clinical trials to provide evidence of effectiveness and safety of a drug or biologic. Describes important principles for designing, conducting, and reporting results. Advises on information to submit for FDA evaluation, including Bayesian adaptive and complex trials relying on simulations. Primary focus is on adaptive designs for trials supporting drug effectiveness and safety, but concepts are useful for early-phase/exploratory trials and post-marketing commitments. Defines adaptive design as "a clinical trial design that allows for prospectively planned modifications to one or more aspects of the design based on accumulating data from subjects in the trial". 

| 2. Advantages and Challenges |

ICH E20: 

Advantages: Provides flexibility against inaccurate assumptions. Ethical advantages (e.g., early stopping for efficacy reduces exposure to inferior control). Improved efficiency (e.g., increased power for a given expected sample size). Improved understanding of treatment effects and decision-making (e.g., dose selection).

Challenges: Adds complexities and uncertainty to key principles. Logistical difficulties in maintaining confidentiality. Risks to trial integrity. More complex and time-consuming planning and assessment. Conventional analysis methods may lead to increased Type I error and biased estimates. May provide less information about safety. Not beneficial in all settings (e.g., fast enrollment relative to endpoint assessment). Requires clear and compelling justification weighing advantages against uncertainties. 

FDA: 

Advantages: Adjusts to information not available at trial start. Statistical efficiency (greater power or smaller expected sample size/shorter duration). Ethical considerations (e.g., early stopping for futility reduces patient exposure to ineffective treatment). Improved understanding of drug effects (e.g., adaptive enrichment, adaptive dose selection). Acceptability to stakeholders (e.g., sponsors willing to commit, patients willing to enroll in RAR trials).

Limitations: Requires specific analytical methods to avoid increasing erroneous conclusions and bias. Methods may not be readily available for complex designs, requiring simulations. Efficiency gains in some respects may be offset by losses in others (e.g., increased maximum sample size, longer lead times). Logistical challenges in trial conduct and integrity (e.g., limiting access to interim results, timely high-quality data). Limited utility in certain settings (e.g., minimum safety sample size, long outcome ascertainment period). Adaptive changes may lead to differing results before and after adaptation, challenging interpretability. |

| 3. Key Principles |

ICH E20: 

Four key principles for confirmatory trials with adaptive designs: Adequacy Within the Development Program (proper design, conduct, analysis within overall program, limited complexity of adaptations at confirmatory stage).

Adequacy of Trial Planning (pre-specification of adaptations, timing, rules, statistical methods, integrity approaches, documentation).

Limiting the Chances of Erroneous Conclusions (controlling Type I error probability, understanding impact on safety and secondary endpoints, evaluating adequacy for trial objectives).

Reliability of Estimation (reliable and unbiased treatment effect estimates, accurate measures of uncertainty, evaluation of bias and variability, reliability of interim estimates for adaptation decisions).

Maintenance of Trial Integrity (limiting knowledge of unblinded data, use of independent IDMC, confidentiality agreements, minimizing inferred information from adaptations, documentation of processes). |

FDA: 

Four key principles for trials providing substantial evidence of effectiveness: Controlling the Chance of Erroneous Conclusions (assessing probability of incorrect conclusions, addressing Type I error inflation due to multiple tests or adaptive features, use of simulations).

Estimating Treatment Effects (reliable estimates for benefit-risk, addressing bias in estimates and confidence intervals, using adjusted methods when available).

Trial Planning (complete pre-specification of design details, timing, type, statistical methods, adaptation algorithm; flexibility for deviations from anticipated algorithm with appropriate methods).

Maintaining Trial Conduct and Integrity (limiting access to comparative interim results, use of independent personnel like DMC, planning to avoid issues from knowledge of accumulating data, documentation of access control). |

| 4. Types of Adaptations |

ICH E20: 

Discusses Early Trial Stopping (efficacy or futility, group sequential designs, stopping boundaries, impact on safety/secondary endpoints).

Sample Size Adaptation (based on nuisance parameters or interim treatment effect estimates, blinded vs unblinded, Type I error control, bias, maintaining integrity).

Population Selection (targeted subpopulations, enrollment restriction, combining data, Type I error control, bias, scientific rationale).

Treatment Selection (different doses/drugs, interim selection, combining data, Type I error control, bias).

Adaptation to Participant Allocation (Response-adaptive randomization, covariate-adaptive, challenges with time trends, integrity). |

FDA: 

Discusses Group Sequential Designs (interim analyses, stopping for efficacy/futility, Type I error control, binding vs non-binding rules, impact on safety/secondary endpoints, adjusted estimates).

Adaptations to the Sample Size (based on nuisance parameters with/without treatment assignment, or comparative interim results, Type I error control, bias, trial integrity).

Adaptations to the Patient Population (e.g., Adaptive Enrichment) (modifications based on comparative interim results, hypothesis tests in multiple populations, multiplicity, scientific rationale, diagnostic devices).

Adaptations to Treatment Arm Selection (adding/terminating arms, dose-ranging, seamless designs, platform trials, Type I error control).

Adaptations to Patient Allocation (covariate-adaptive and response-adaptive randomization, Type I error, predictability, short-term outcomes).

Adaptations to Endpoint Selection (modification to primary endpoint choice, prespecified rules, multiplicity).

Adaptations to Multiple Design Features (combining features, complexity, simulations often critical). 

| 5. Special Topics and Considerations |

ICH E20: 

Further Considerations on Data Monitoring (IDMC expertise, access to unblinded data, charter, independent statistical group, sponsor access).

Planning, Conducting, and Reporting Simulation Studies (objectives, designs/analysis options, key operating characteristics, scenarios, implementation details, comprehensive report).

Adaptive Designs Using Bayesian Methods (ICH E9 principles, informing adaptations, borrowing external information, prior distribution, false positive control, simulations, sensitivity analyses).

Adaptive Designs in Time-to-Event Settings (number of events vs participants, follow-up time, independence assumption, Type I error control, longitudinal outcomes).

Adaptive Designs in Exploratory Trials (relevance of principles, more adaptations, reliable decision-making, sponsor role in monitoring, safety).

Operational Considerations (maintaining integrity, informed consent, randomization systems, drug supply, data quality/timeliness). |

FDA: 

Simulations in Adaptive Design Planning (critical role, estimating operating characteristics, Type I error probability, nuisance parameters, number of iterations, precision, different random seeds).

Bayesian Adaptive Designs (various uses, frequentist vs Bayesian inference, informative priors, simulations).

Adaptations in Time-to-Event Settings (number of events vs subjects, follow-up, use of surrogate/intermediate outcomes).

Adaptations Based on an Intermediate Endpoint (surrogate/intermediate endpoints, correlation with primary, Type I error inflation, safeguards).

Secondary Endpoints (Type I error inflation, biased estimation, adjustment methods).

Safety Considerations (adaptations on safety endpoints, sufficient safety information, minimum sample size/exposure, risk of escalating doses).

Adaptive Design in Early-Phase Exploratory Trials (usefulness, less rigorous expectations, risk of erroneous conclusions).

Unplanned Design Changes Based on Comparative Interim Results (discouraged without FDA discussion, difficulty in Type I error control).

Design Changes Based on Information From a Source External to the Trial (acceptable, do not affect validity, challenging to ensure solely external). |

| 6. Documentation |

ICH E20: 

Prior to conducting: Rationale (clinical and statistical), description of adaptations (aspects, timing, rule, estimands), statistical analysis methods (interim, primary/secondary, sensitivity), implementation details (who does what, committee roles), steps to maintain confidentiality/integrity (information transfer, access control, records), operating characteristics (simulation report if critical). Core elements in protocol, other details in separate documents (e.g., IDMC charter), details of adaptation rule may be restricted.

After completion (marketing application): All prospective plans, information on how design was implemented (actual timing, conduct variations, interim results, adaptation decisions, deviations), compliance with data access/integrity processes, IDMC deliberations records, appropriate reporting of results. |

FDA: 

Prior to conducting: Rationale, detailed description of adaptation plan (number/timing of analyses, modifications, rule), roles of bodies (DMC/adaptation committee), prespecification of statistical methods (interim/final, software, code), evaluation/discussion of operating characteristics (Type I error, power, sample size, bias, coverage - analytical or simulation). Detailed simulation report if primary technique (trial description, examples, parameter configurations, iterations, results, code, summary). Comprehensive data access plan (who performs, who has access, control, decisions, dissemination, compliance evaluation, meeting minutes). Can be in protocol or separate documents.

Evaluating and Reporting a Completed Trial: All prospective plans, committee charters, supporting documentation, compliance with planned rule and data access, deliberation records, interim analysis results, appropriate reporting in package insert (design, adjusted estimates/cautions). |

| 7. Regulatory Interactions |

ICH E20: 

States that the draft guideline is currently under public consultation and does not yet confirm full regulatory acceptance or supersede current regional guidance. The final guideline will indicate key adaptive design principles and approaches for which discussion at the planning stage is particularly critical. Specific feedback is sought on principles and their impact on industry-regulatory interactions to inform finalization. Sponsors should discuss potential implications of adaptations on trial conduct with regulators at the planning stage. Documentation prior to trial initiation should align with national and regional regulatory requirements. 

FDA: 

Encourages sponsors to explore design options and discuss them with FDA at regulatory meetings (e.g., End-of-Phase-2, Type C). Increased complexity of adaptive trials may warrant earlier and more extensive interactions. FDA's role shifts from safety focus in early development to more extensive evaluation of design and analysis in later phases to ensure reliable results. SPAs for complex adaptive designs only if extensive prior discussion with FDA. FDA review considers whether design and analysis satisfy key principles (Type I error control, reliable estimation, integrity, prespecification). FDA generally not involved in prospectively planned adaptive decision-making; this is sponsor's responsibility, often via a committee (e.g., DMC). Meeting minutes of closed sessions and comparative interim results generally kept confidential until trial conclusion, except for safety risks. All important documentation should be submitted to FDA during design stage for feedback prior to initiation. |

| 8. Definition of Adaptive Design |

ICH E20: 

"a clinical trial design that allows for prospectively planned modifications to one or more aspects of the trial based on interim analysis of accumulating data from participants in the trial." |

FDA: 

"a clinical trial design that allows for prospectively planned modifications to one or more aspects of the design based on accumulating data from subjects in the trial." 

| 9. Definition of Interim Analysis | 

ICH E20: 

Not explicitly defined as broadly as FDA. Focuses on "sequential analyses of accumulating unblinded data with anticipated rules for stopping". 

FDA: 

"any examination of data obtained from subjects in a trial while that trial is ongoing and is not restricted to cases in which there are formal between-group comparisons." Includes baseline, safety, PK/PD, biomarker, or efficacy data. Broader than ICH E9 definition. 

| 10. "Prospectively Planned" |

ICH E20: 

Means "the potential trial adaptations are pre-specified in the clinical trial protocol prior to initiation of the trial." 

FDA: 

Means "that the adaptation is planned and details specified before any comparative analyses of accumulating trial data are conducted. In nearly all situations, potential adaptive design modifications should be planned and described in the clinical trial protocol (and in a separate statistical analysis plan) prior to initiation of the trial." 

---

Key Differences and Nuances:

• Stage of Guidance: The most significant difference is that ICH E20 is a draft guideline currently under public consultation , while the FDA guidance is a finalized document from November 2019. ICH E20 explicitly states it does not supersede current regional guidance until finalized.

• Emphasis on "Confirmatory Trials": Both documents primarily focus on adaptive designs in confirmatory clinical trials. However, ICH E20 consistently uses the phrase "confirmatory clinical trials" throughout its document, defining its scope around them. The FDA guidance states its "primary focus...is on adaptive designs for clinical trials intended to support the effectiveness and safety of drugs," but also acknowledges the usefulness of concepts for early-phase/exploratory trials and post-marketing commitments.

• Definition of Interim Analysis: The FDA guidance provides a broader definition of "interim analysis" compared to the ICH E9 definition, stating it's "any examination of data obtained from subjects in a trial while that trial is ongoing and is not restricted to cases in which there are formal between-group comparisons". ICH E20 uses the term more in the context of "sequential analyses of accumulating unblinded data with anticipated rules for stopping".

• Handling Unplanned Changes: The FDA guidance explicitly addresses "Unplanned Design Changes Based on Comparative Interim Results," strongly discouraging them without prior FDA discussion due to the difficulty in controlling Type I error. It also discusses "Design Changes Based on Information From a Source External to the Trial," which are generally considered acceptable as they don't affect statistical inference validity, but notes the challenge in ensuring the decision was entirely based on external information. ICH E20 directly states its scope "does not include trials with unplanned modifications to the design" or "design changes based entirely on emerging information from a source external to the trial". While both recognize the issue, the FDA guidance delves more into how these situations might be handled if they arise.

• Level of Detail on Specific Adaptations: Both documents cover similar types of adaptations (early stopping, sample size, population, treatment arm, patient allocation). However, ICH E20 includes a dedicated section on "Adaptive Designs in Time-to-Event Settings" and "Adaptive Designs Using Bayesian Methods" as full sections within its "Special Topics and Considerations," suggesting a more structured and perhaps deeper dive into these areas. The FDA also covers these, but they are subsections.

• Trial Integrity and Confidentiality: Both documents strongly emphasize maintaining trial integrity and limiting access to unblinded interim results to independent parties (e.g., IDMC). However, ICH E20 suggests that details of the adaptation rule "could be reserved for a specific document rather than the protocol, such as a confidential appendix to the IDMC charter" , and that sponsor personnel, investigators, and participants could be shielded from knowledge of specific adaptive changes. The FDA also mentions that the protocol could outline only the general approach with details reserved for documents like the DMC charter.

• Documentation: Both guidances provide extensive lists of documentation requirements before and after conducting an adaptive trial. ICH E20's section 6.1 and 6.2 provide a comprehensive outline of what should be documented, including explicit links between clinical and statistical assumptions and simulation results. The FDA also requires detailed simulation reports and data access plans.

• Context of Development Program: ICH E20 places a strong emphasis on the "Adequacy Within the Development Program" , suggesting that increasing the number and complexity of adaptations at the confirmatory stage should generally be limited and not replace a sequence of multiple trials or a proper dose-ranging trial. This principle is woven into the justification of adaptive designs. The FDA also touches upon this by noting that adaptive designs are useful for early-phase exploratory trials.

In summary, while both the ICH E20 draft guidelines and the FDA guidance share a common understanding of adaptive designs and their core principles, the ICH E20 document, being a newer draft, appears to offer a more structured and perhaps slightly more detailed approach to certain advanced topics and their implications within the overall drug development program. The FDA guidance provides a solid foundation of principles that aligns with many aspects of the ICH E20 draft, particularly regarding the need for robust statistical methods, trial integrity, and comprehensive documentation.