Adaptive Design vs. Sequential Design in Clinical Trials

1. Core Concept Comparison

Aspect	Adaptive Design	Sequential Design
Basic idea	Allows pre-planned modifications to the trial based on accumulating data	Conducts planned interim analyses to decide whether to continue or stop the trial
What can change	Sample size, randomization ratio, dose levels, number of arms, population, etc.	Usually only early stopping for efficacy or futility
Flexibility	High (must be pre-specified in the protocol)	Moderate
Statistical complexity	High	Moderate
Regulatory acceptance	Increasing but requires strong control of Type I error	Well established and widely accepted

---

2. Which Clinical Trial Phase Are They Commonly Used In?

Adaptive Design is most commonly used in:

Phase I and Phase II (sometimes Phase II/III seamless trials)

Reasons:

• High uncertainty in early development
• Strong need for learning and exploration
• Smaller sample sizes, making adaptations more manageable

Typical examples:

• Phase I:
• Adaptive dose-escalation designs (e.g., CRM, BOIN)
• Phase II:
• Multi-arm multi-stage (MAMS) designs
• Adaptive enrichment (biomarker-driven population selection)
• Phase II/III seamless designs:
• Dose selection transitioning directly into confirmatory testing

👉 Main goal: improve decision-making efficiency and reduce development risk

---

Sequential Design is most commonly used in:

Phase III (also sometimes Phase II)

Reasons:

• Phase III trials are confirmatory and require stable designs
• Sequential methods have mature theory and strong regulatory acceptance
• Primary objective is early stopping for success or futility

Typical examples:

• Group sequential designs
• O’Brien–Fleming
• Pocock
• Early termination due to:
• Overwhelming efficacy
• Lack of efficacy (futility)

👉 Main goal: save time and sample size while maintaining strict control of Type I error

---

3. One-Sentence Summary

• Adaptive design
  → “learn and modify during the trial,” mainly used in Phase I–II
• Sequential design
  → “analyze early and stop if appropriate,” mainly used in Phase III

Interim Finding	Is Sample Size Increase Allowed?	Explanation
Variance larger than expected	✅ Yes	Blinded sample size re-estimation (SSR)
Effect size point estimate slightly smaller but unstable	⚠ Possibly	Must be based on pre-specified rules
Effect size clearly smaller	❌ Not recommended	Not supported scientifically or by regulators
Conditional power very low	❌ No	Trial should stop for futility

Dimension	3+3	BOIN	CRM
Real-world usage frequency	⭐⭐⭐⭐⭐	⭐⭐⭐⭐	⭐⭐
Statistical efficiency	⭐	⭐⭐⭐⭐	⭐⭐⭐⭐⭐
Accuracy of MTD selection	Low	High	Highest
Operational complexity	Very low	Low	High
Clinician acceptance	Very high	High	Moderate
Regulatory communication	Easiest	Easy	Requires preparation
Truly adaptive design	❌	✅ (rule-based adaptive)	✅ (model-based adaptive)