Adaptive Trial Designs Explained: Concepts, Benefits, and Lessons from the SODa-BIC Trial

We briefly touched on adaptive trials in the latest episode of our podcast (https://youtu.be/h9G5gNBQTAk?si=n3_U2xN7b_ZiaRgh). In this post, we build on that discussion- unpacking key concepts in more detail and taking a closer look at the specific adaptations used in the SODa-BIC trial.

What are adaptive trials?

Adaptive trial designs allow for planned modifications to elements of a trial protocol while the trial is ongoing—without compromising its scientific integrity. These adaptations are pre-specified and conducted in a way that preserves the internal validity, credibility, and interpretability of the trial.

Why do we need adaptive trials?

Traditional (fixed) trial designs come with several important limitations. All key design elements—sample size, interventions, eligibility criteria, and outcomes—must be specified upfront. This rigidity can create challenges:

Uncertain assumptions: Sample size calculations rely on assumptions (e.g., event rates, effect sizes) that may turn out to be inaccurate. In a fixed design, there is no formal mechanism to revisit these assumptions.
Limited flexibility: Fixed trials typically evaluate a small number of interventions and cannot easily adapt to emerging evidence—such as adding a promising new treatment or dropping an ineffective one.
Inefficiency: Running multiple sequential or parallel fixed trials is often resource-intensive and time-consuming.

Adaptive designs attempt to address these limitations by introducing flexibility- while maintaining methodological rigour.

What elements can be adapted?

A wide range of trial components can be adapted. Common examples include:

Sample size re-estimation
Adding or dropping treatment arms
Modifying the primary outcome
Refining eligibility criteria to enrich the study population (e.g., selecting patients at higher risk or more likely to benefit)
Changing randomisation ratios (e.g., response-adaptive randomisation)
Shifting the hypothesis (e.g., from superiority to non-inferiority)

Core principles of adaptive designs

Regardless of the type of adaptation, a few key principles apply:

Pre-specification: All planned adaptations must be clearly defined in the protocol and approved by ethics and regulatory bodies.
Independence and blinding: Wherever possible, adaptations should be conducted in a blinded manner by an independent team. This often involves an independent statistician and oversight from the Data Safety Monitoring Board (DSMB).
Preservation of integrity: The design and execution of adaptations must ensure that the trial’s validity and credibility are not compromised.

What did the SODa-BIC trial do?

The SODa-BIC trial (https://www.nejm.org/doi/full/10.1056/NEJMoa2600526) incorporated a blinded sample size re-estimation after the first 100 patients were enrolled. This decision was driven by concerns that the original assumptions- particularly event rates derived from the BICAR-ICU trial (2018)- might not hold true in the current study population.

Here’s how the process worked:

An independent statistician re-estimated the required sample size using pooled (combined) event rates across both treatment groups.
These results were presented to the DSMB, which then advised the principal investigator and steering committee.
To maintain feasibility, the investigators capped any increase at an additional 200 patients (raising the total from 500 to a maximum of 700).
Importantly, they pre-specified that the sample size would not be reduced, even if event rates were higher than expected.

This approach strikes a careful balance- introducing flexibility while preserving trial integrity. The use of an independent statistician and DSMB oversight minimises bias, while the cap on sample size ensures the study remains logistically and financially feasible. In the end, they did not have to increase the sample size and completed the trial based on the original plan.

A note on unblinded adaptations

In some situations, unblinded sample size re-estimation may be justified. However, this introduces additional risks, particularly the potential for investigators to infer emerging treatment effects, which could influence trial conduct. Even with blinded approaches, an increase in sample size can lead to speculation about the likely direction of results.

These risks are mitigated—but not eliminated—through blinding, independent oversight, and pre-specified limits on adaptation.

Limitations of adaptive trial designs

Despite their advantages, adaptive designs are not without challenges:

Statistical complexity: They require more sophisticated methods and can make results harder to interpret.
Operational demands: Adaptive trials often need real-time data cleaning, monitoring, and decision-making infrastructure.
Regulatory considerations: Acceptance by regulatory agencies can vary and may require additional justification.
Uncertain efficiency gains: The logistical and analytical complexity may offset some of the anticipated efficiency benefits.