Usage Examples
- The adaptive design pre-specifies one sample size re-estimation at 50% of planned events, on blinded variance only.
- That interim look was not in the protocol, so we cannot call it an adaptation and the result is exploratory.
- The SAP has to fix the alpha-spending function before the blind breaks, or the type I error claim does not hold.
What is Adaptive Trial Design (Adaptive Design)?
Adaptive trial design is a clinical trial design whose prespecified rules let interim data change sample size, dose, arms, or enrollment mid-study, unlike a fixed design, which locks every parameter before the first subject is dosed.
Adaptive trial design exists because a fixed protocol forces every assumption to be locked before a single subject is dosed: the effect size, the variance, the event rate, the right dose. Those assumptions are frequently wrong. Adaptive trial design lets accumulating data correct them without discarding the trial, provided the correction rules were written down before anyone saw comparative results.
Adaptive trial design covers the modifications a protocol can pre-commit to: group sequential stopping for efficacy or futility, sample size re-estimation, dose selection, arm dropping, and population enrichment. Adaptive trial design does not cover changes invented after an unplanned look at comparative data. ICH E9 is explicit that only analyses envisaged in the protocol, including amendments, can be regarded as confirmatory.
Adaptive trial design is applied through three documents and one firewall. The protocol fixes the adaptation rules and the schedule of interim analyses; the statistical analysis plan carries the elaboration and is finalised before the blind is broken; the IND amendment carries anything added later. The firewall is confidentiality, because E9 treats the execution of an interim analysis as a completely confidential process.
Not to be confused with
- Protocol amendment
- an amendment changes a trial after it is underway; an adaptation is a change the protocol already authorised in advance. Both may reach FDA through the IND, but only the pre-planned one supports a confirmatory claim without argument.
- Group sequential design
- group sequential is one species of adaptive design, confined to planned interim looks with stopping boundaries. It says nothing about dose selection, arm dropping, or population enrichment, so the two terms are not interchangeable.
- Master protocol
- a master protocol is an organising structure that runs multiple sub-studies under shared infrastructure; an adaptive design is a rule set operating inside a single trial. A master protocol can be adaptive, and an adaptive trial need not be a master protocol.
- Interim analysis
- the interim analysis is the look at the data; the adaptive design is the pre-written rule stating what that look is permitted to change. An interim look with no pre-specified consequence is monitoring, not adaptation.
The obligations sit in two places: the statistical guideline that governs prespecification, and the IND regulations that govern what FDA sees.
What you must do
- 1Specify all important details of the trial's design and conduct, and the principal features of the proposed statistical analysis, in a protocol written before the trial beginsICH E9 §1.2
- 2Plan every interim analysis in advance and describe it in the protocol, including the schedule of looks and the stopping guidelines, using procedures that ensure the overall probability of type I error is controlledICH E9 §4.5
- 3Justify and document any revised sample size in a protocol amendment and in the clinical study report, explaining the steps taken to preserve blindness and the consequences for type I error and confidence interval widthICH E9 §4.4
- 4Describe in the protocol the study design and the methods to be used to minimize bias on the part of subjects, investigators, and analysts21 CFR 312.23(a)(6)(iii)(d)
- 5Amend the IND as needed so that the clinical investigations are conducted according to the protocols included in the application21 CFR 312.30
Common mistakes
Putting the adaptation rule in the SAP but not the protocol
E9 places the schedule of interim analyses and the stopping guidelines in the protocol, with the SAP holding the technical elaboration. A rule that exists only in the SAP is hard to defend as prespecified, and the analysis it produces stops counting as confirmatory.
Treating a blinded sample size check as a free option
even a re-estimation done without unblinding must be justified and documented in a protocol amendment and in the clinical study report, with the effect on type I error and confidence interval width explained. Teams run the check, resize enrollment, and never paper it.
Letting sponsor staff see interim comparative results
E9 treats interim execution as a completely confidential process precisely because knowledge of the result changes who gets recruited and how comparisons are made. Once operational bias is plausible, no statistical correction repairs it and the reviewer discounts the trial.
When This Matters
- The adaptive design pre-specifies one sample size re-estimation at 50% of planned events, on blinded variance only.
- That interim look was not in the protocol, so we cannot call it an adaptation and the result is exploratory.
- The SAP has to fix the alpha-spending function before the blind breaks, or the type I error claim does not hold.
Frequently Asked Questions
Yes, but only through a protocol amendment completed before anyone sees unblinded comparative data. ICH E9 requires an interim analysis not defined at the start of a trial to be described in an amendment completed prior to unblinded access, and 21 CFR 312.30 requires the sponsor to amend the IND so investigations follow the protocols in the application.
Related Use Cases
- Pharma Use Cases
Cut NDA and sNDA prep time by 60% with AI-assisted drafting and automated readiness checks
- Biotech Use Cases
Compress IND prep from 8-12 weeks to under 3 weeks with AI-assisted drafting and validation
- Clinical Workflows
Track clinical trial regulations across global markets
- Regulatory Affairs Workflows
Cut regulatory intelligence tracking from 10+ hours/week to automated, real-time alerts
