Date of Award

2021

Document Type

Dissertation - MUSC Only

Degree Name

Doctor of Philosophy (PhD)

Department

Public Health Sciences

College

College of Graduate Studies

First Advisor

Renee’ H. Martin

Second Advisor

Caitlyn Meinzer

Third Advisor

Bethany J. Wolf

Fourth Advisor

Hong Li

Fifth Advisor

Hooman Kamel

Sixth Advisor

Matthew J. Carpenter

Abstract

For multisite time-to-event clinical trials, the site initiation and resulting subject accrual process are important but often difficult parameters to predict in the design phase. Multiple factors influence the accrual of subjects into a trial including number of sites, recruitment rate of each site, and the site initiation pattern. An improvement in prediction of the overall subject accrual may positively impact many aspects of the design and conduct for this type of trial, such as sample size calculation, logistics planning, and interim analyses. While many improvements have been made for predicting subject accrual, methodology that accounts for a non-uniform site start-up process that includes a lag in site initiation has not yet been developed. Being able to account for this lag at the beginning of the start-up period will allow for a better prediction of when subjects can be enrolled in the study. These updates are particularly useful in maximum duration time-to-event clinical trials, which assume that subjects are followed from enrollment to end of study with a minimum follow-up specified. Under a maximum duration design, the maximum information is influenced by subject follow-up time as the trial is designed to continue until a specified time. Neglecting differential site initiation can lead to an underestimate in variability of subject accrual, which impacts the predicted subject follow-up time and number of events observed. Thus, this dissertation uses the Program Evaluation and Review Technique (PERT) distribution to account for differential site initiation. The first aim uses the PERT distribution as a model for site initiation to update both the sample size formula and the Poisson-Gamma accrual model for logistics planning. Similarly, the second aim uses the PERT distribution as a model for site initiation to update the information fraction used to determine statistical significance within a group sequential design. The third aim uses the PERT-adjusted sample size formula, PERT-adjusted subject accrual models, and PERT-adjusted information fractions from the first and second aims in a sample size re-estimation. Overall, the developed methodologies allow for additional flexibility to account for the previously underestimated variability in subject accrual during the site initiation period.

Rights

All rights reserved. Copyright is held by the author.

Download

Share

COinS