ICH E20 is the harmonised guideline being developed on adaptive designs for clinical trials. We checked its status on 16 October 2026 and could confirm only a draft. This guide explains the guideline, the design types it covers and the operational demands adaptive trials place on data capture.
Free sandbox · No credit card · 21 CFR Part 11 aligned
Pre-specified adaptation rules
In protocol and SAP
Simulation report for operating characteristics
Interim analysis charter and firewall plan
Blinded and unblinded role matrix
Data cut and lock procedure for interim looks
Change log for design modifications
ICH E20 in one page
The guideline
ICH E20 is a topic in the efficacy series of the International Council for Harmonisation. Its goal is a common set of principles for planning, conducting, analysing and interpreting clinical trials with adaptive designs, so that sponsors and regulators in different regions work from the same expectations. The topic was adopted several years ago and the schedule has slipped more than once.
The draft guideline was endorsed at Step 2b in mid-2025 and released for public consultation by regulators. The FDA's Federal Register notice is dated 30 September 2025 and covers comments to 1 December 2025. The EMA consultation ran from 30 June to 30 November 2025. After consultation the expert working group reviews the comments and, when it reaches agreement, the guideline moves to Step 4 and is then adopted by each regulator. Until that sign-off, treat E20 as a draft and do not describe it as final. The status can change quickly, so check the ICH efficacy guidelines page.
E20 sits alongside the other ICH guidelines you may already use: the statistical principles in E9, the good clinical practice requirements in E6 and the data-integrity expectations our ICH E6(R3) guide explains. It does not replace them.
Public summaries of the draft describe a focus on principles rather than a catalogue of methods: justifying an adaptive design, controlling the chance of false conclusions, obtaining reliable estimates of treatment effect, and protecting the integrity of the trial when results are seen during the study. We have not reproduced the draft text here. Read the guideline itself for its specific terms, and read our statistical analysis plan guide for how pre-specification works in practice.
Design types
These are widely used design families in the statistical literature. Which ones E20 addresses and how is for the guideline text.
| Design | What adapts | What operations must support |
|---|---|---|
| Group sequential | The trial can stop early for efficacy or futility at pre-planned interim looks. | Clean, timely interim data cuts and a restricted-access unblinded review. |
| Sample size re-estimation | The planned sample size is revised using interim information. | A controlled process so the people running the trial are not exposed to unblinded results. |
| Adaptive randomisation | Allocation probabilities shift based on accumulating outcomes. | Randomisation that can be updated under a documented plan, with a record of every version. |
| Seamless phase 2/3 | Dose or arm selection in one stage feeds directly into confirmation in the next. | Data from the first stage is complete and locked for the selection decision. |
| Population enrichment | Eligibility narrows to a subgroup that appears to benefit. | Amended eligibility forms and screening, with clear version control. |
| Bayesian adaptive | Decisions use posterior probabilities updated as data arrives. | Frequent data refreshes and a validated analysis environment. See our Bayesian design guide. |
Adaptive trials punish late surprises in the data. Build the forms and edit checks in the free sandbox and rehearse your data cut with sample data.
Operations
Statistical design is only half of an adaptive trial. The other half is making sure the data at each decision point is complete, queried and trustworthy, and that nobody who should stay blind sees results they should not. Regulators judge adaptive trials partly on that operational discipline.
Think in four layers. First, timeliness: if an interim analysis depends on data up to a cut date, the delay between a visit and its entry must be short and measured. Query resolution and source data verification need to be on a schedule that serves the look. Second, access control: the unblinded team and the blinded operational team need different views. Our guide to blinded and unblinded roles in EDC covers the role design. Third, versioning: any change to eligibility, forms or randomisation must be documented and dated. See eCRF versioning for mid-study changes. Fourth, traceability: you must show what data existed at each interim look and what happened afterwards.
Sample size and power are also moving parts. Start from a defensible plan with our sample size calculator and rely on your statisticians' simulations for the adaptive operating characteristics.
Capture provides the capture side: forms with edit checks, a field-level audit trail, a draft to approved form lifecycle, randomisation and blinded roles in which blinded roles never receive treatment-arm values, enforced by masked database views at the database layer. It also exports datasets with a data dictionary and SDTM as SAS XPT with Define-XML for your statisticians.
Capture does not run your interim analyses or simulations, and we do not claim built-in adaptive allocation algorithms or data monitoring committee tooling. Those belong in your statistical environment under your own validated procedures. Our clinical trial platform with randomization page describes what randomisation covers.
Study coordinator
Subject 01-004
Randomized · R-0012
KitK-10418
Arm columnNot sent
Unblinded administrator
Subject 01-004
Randomized · R-0012
KitK-10418
Arm columnA
Same subject, same screen. The arm never reaches blinded roles.
Checklist
Rules, thresholds and decision makers are written into the protocol and the statistical analysis plan before first enrolment.
Define cut dates, cleaning steps, who authorises the cut and how it is reproduced.
Map who sees treatment-arm information and confirm the system enforces it, not just policy.
Every change to forms, eligibility or allocation is approved, dated and explained.
Export the trail and show it can reconstruct the state at the interim date.
Read the current ICH E20 status, the FDA 2019 guidance and any regional guidance for your product.
ICH E20 is the International Council for Harmonisation guideline in development on adaptive designs for clinical trials, intended to give sponsors and regulators common principles for planning, conducting and interpreting them.
When we checked on 16 October 2026 we could confirm only the draft (Step 2b) after public consultation. A finalisation around October 2026 was planned but we could not confirm it. Check the ICH website for the current step.
Nothing yet. The FDA December 2019 guidance on adaptive designs for drugs and biologics remains the established US reference until E20 is finalised and implemented by the FDA.
A design that allows pre-planned modification of aspects of a trial, such as sample size, arms or allocation, based on accumulating data, under rules set before the data is seen.
It needs reliable, timely data, strict blinding enforcement, version control for any change and a traceable audit trail. Many requirements are about process as much as software.
No. Capture captures and exports the data and enforces blinded roles. Interim analyses and simulations run in your statistical environment under your own procedures.
It is a separate guideline. E6 still governs good clinical practice. Adaptive trials must meet both, plus the statistical principles in E9.
No. Use the current guideline text, your statisticians and your regulatory advisers.
Keep exploring
Free sandbox with every feature. No credit card, and you pay only when you go live with real participants.