Enrichment designs using placebo nonresponders.

Area of Science:

• Clinical trial design
• Psychopharmacology
• Statistical modeling

Background:

• Enrichment designs selecting placebo nonresponders are increasingly used in high placebo-response areas like depression.
• Sequential Parallel Design (SPD) re-randomizes nonresponders to placebo into a second phase with an enriched population.

Purpose of the Study:

• To analyze selection bias in enrichment designs that combine data from overall and enriched populations.
• To investigate the impact of subject-by-treatment interaction variability on bias.
• To provide sample sizes for detecting effects under specific conditions.

Main Methods:

• Theoretical analysis of bias in combined populations.
• Examination of the relationship between bias and subject-by-treatment interaction variability.
• Calculation of sample sizes for significance testing.

Main Results:

• Selection bias in enrichment designs is linked to the variability of subject-by-treatment interactions.
• If no subject-by-treatment interaction exists, enrichment designs are inefficient.
• Sample size calculations are provided for scenarios with zero average effect in the overall population.

Conclusions:

• Enrichment designs using placebo nonresponders cannot claim positive average effects in the overall population if subject-by-treatment interactions are present.
• These designs are unsuitable for pivotal Phase III trials aiming to demonstrate overall population efficacy.
• Enrichment designs may be valuable in early-phase trials for identifying heterogeneous treatment effects.