Accounting for treatment by center interaction in sample size determinations and the use of surrogate outcomes in the pessary for the prevention of preterm birth trial: a simulation study

Andrew R Willan

doi:10.1186/s13063-016-1433-y

Abstract

BackgroundThe Pessary for the Prevention of Preterm Birth Study (PS3) is an international, multicenter, randomized clinical trial designed to examine the effectiveness of the Arabin pessary in preventing preterm birth in pregnant women with a short cervix. During the design of the study two methodological issues regarding power and sample size were raised. Since treatment in the Standard Arm will vary between centers, it is anticipated that so too will the probability of preterm birth in that arm. This will likely result in a treatment by center interaction, and the issue of how this will affect the sample size requirements was raised. The sample size requirements to examine the effect of the pessary on the baby’s clinical outcome was prohibitively high, so the second issue is how best to examine the effect on clinical outcome. The approaches taken to address these issues are presented.ResultsSimulation and sensitivity analysis were used to address the sample size issue. The probability of preterm birth in the Standard Arm was assumed to vary between centers following a Beta distribution with a mean of 0.3 and a coefficient of variation of 0.3. To address the second issue a Bayesian decision model is proposed that combines the information regarding the between-treatment difference in the probability of preterm birth from PS3 with the data from the Multiple Courses of Antenatal Corticosteroids for Preterm Birth Study that relate preterm birth and perinatal mortality/morbidity. The approach provides a between-treatment comparison with respect to the probability of a bad clinical outcome. The performance of the approach was assessed using simulation and sensitivity analysis.Accounting for a possible treatment by center interaction increased the sample size from 540 to 700 patients per arm for the base case. The sample size requirements increase with the coefficient of variation and decrease with the number of centers. Under the same assumptions used for determining the sample size requirements, the simulated mean probability that pessary reduces the risk of perinatal mortality/morbidity is 0.98. The simulated mean decreased with coefficient of variation and increased with the number of clinical sites.ConclusionEmploying simulation and sensitivity analysis is a useful approach for determining sample size requirements while accounting for the additional uncertainty due to a treatment by center interaction. Using a surrogate outcome in conjunction with a Bayesian decision model is an efficient way to compare important clinical outcomes in a randomized clinical trial in situations where the direct approach requires a prohibitively high sample size.

Highlights

The Pessary for the Prevention of Preterm Birth Study (PS3) is an international, multicenter, randomized clinical trial designed to examine the effectiveness of the Arabin pessary in preventing preterm birth in pregnant women with a short cervix
The Pessary for the Prevention of Preterm Birth Study (P3S) is an international, multicenter, randomized clinical trial (RCT) to be conducted in 80 clinical centers by the Centre for Mother, Infant, and Child Research at the Sunnybrook Research Institute in Toronto, and is designed to answer the following primary question: Does the introduction of the Arabin pessary reduce the probability of preterm birth in pregnant women with a short cervix? The Arabin pessary is a simple, non-invasive therapeutic option for preventing preterm birth
Modeling the effect that a treatment by center interaction has on sample size is challenging since it depends on many factors, many of which might be unknown at the time of trial planning

Summary

Introduction

The Pessary for the Prevention of Preterm Birth Study (PS3) is an international, multicenter, randomized clinical trial designed to examine the effectiveness of the Arabin pessary in preventing preterm birth in pregnant women with a short cervix. Since treatment in the Standard Arm will vary between centers, it is anticipated that so too will the probability of preterm birth in that arm This will likely result in a treatment by center interaction, and the issue of how this will affect the sample size requirements was raised. The Pessary for the Prevention of Preterm Birth Study (P3S) is an international, multicenter, randomized clinical trial (RCT) to be conducted in 80 clinical centers by the Centre for Mother, Infant, and Child Research at the Sunnybrook Research Institute in Toronto, and is designed to answer the following primary question: Does the introduction of the Arabin pessary reduce the probability of preterm birth in pregnant women with a short cervix? Since there is no agreed-upon standard of care, women will be randomized between (1) the current management of short cervix as determined by the center’s policy or at the attending physician’s discretion (Standard) and (2) Arabin pessary together with any additional care at the attending physician’s discretion (Treatment)

Methods

Results

Discussion

Conclusion