The design and statistical aspects of VIETNARMS: a strategic post-licensing trial of multiple oral direct acting antiviral Hepatitis C treatment strategies in Vietnam

TL;DR

This paper details the statistical design of the VIETNARMS trial, a multi-arm study testing hepatitis C treatment strategies in Vietnam, with Bayesian monitoring to efficiently identify and stop ineffective treatments.

Contribution

It introduces a Bayesian interim monitoring approach for a multi-arm hepatitis C trial, optimizing timing and criteria for stopping ineffective strategies.

Findings

Bayesian stopping rules effectively identify low cure rate strategies.

Interim analyses at 7, 10, 13, and 18 months are optimal.

High power (>90%) to detect non-inferiority with specified margins.

Abstract

Background Achieving hepatitis C elimination is hampered by the costs of treatment and the need to treat hard-to-reach populations. Treatment access could be widened by shortening treatment, but limited research means it is unclear which strategies could achieve sufficiently high cure rates to be acceptable. We present the statistical aspects of a multi-arm trial designed to test multiple strategies simultaneously with a monitoring mechanism to detect and stop those with unacceptably low cure rates quickly. Methods The VIETNARMS trial will factorially randomise patients to three randomisations. We will use Bayesian monitoring at interim analyses to detect and stop recruitment into unsuccessful strategies, defined as a >0.95 posterior probability of the true cure rate being <90%. Here, we tested the operating characteristics of the stopping guideline, planned the timing of the interim analyses and explored power at the final analysis. Results A beta(4.5, 0.5) prior for the true cure rate produces <0.05 probability of incorrectly stopping a group with true cure rate >90%. Groups with very low cure rates (<60%) are very likely (>0.9 probability) to stop after ~25% patients are recruited. Groups with moderately low cure rates (80%) are likely to stop (0.7 probability) before the end of recruitment. Interim analyses 7, 10, 13 and 18 months after recruitment commences provide good probabilities of stopping inferior groups. For an overall true cure rate of 95%, power is >90% to detect non-inferiority in the regimen and strategy comparisons using 5% and 10% margins respectively, regardless of the control cure rate, and to detect a 5% absolute difference in the ribavirin comparison. Conclusions The operating characteristics of the stopping guideline are appropriate and interim analyses can be timed to detect failing groups at various stages.