Using Surrogate Endpoints to Improve the Design and Analysis of Randomized Screening Trials

Principal Investigator

Name
Yingqi Zhao

Degrees
Ph.D.

Institution
Fred Hutchinson Cancer Center

Position Title
Professor

Email
yqzhao@fredhutch.org

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
NLST-1511

Initial CDAS Request Approval
May 4, 2026

Title
Using Surrogate Endpoints to Improve the Design and Analysis of Randomized Screening Trials

Summary
As cancer screening techniques evolve rapidly, there is increasing demand for quicker evaluation of effectiveness in randomized screening trials while maintaining rigor in estimating mortality benefits. The traditional screening trial, such as the NLST, requires many thousands of participants and several years of follow-up. This resource-intensive approach means trials are rare, can only assess one or two novel screening methods at once, and their screening technology may be outdated by the time endpoints are available. Consequently, statistical methods to run trials for fewer years with fewer participants could have enormous impact, helping deliver life-changing technologies to patients faster. On the other hand, rushing ineffective new products to market based on incomplete evidence could have negative consequences for patients and the healthcare system at large. Careful, rigorous development and validation of these methods is essential. We have developed statistical methods for (1) studying the reliability of short-term surrogate endpoints to aid the design of future trials and (2) leveraging many short-term surrogate outcomes to estimate the ultimate endpoint of interest – cancer-specific mortality – with greater precision. We propose validating the utility of our new methods on NLST data.

Aims

The aims of this project are to:

1. Predict the surrogate value of candidate short-term endpoints in future trials using microsimulation models calibrated to NLST data.
2. Evaluate the surrogate value of candidate short-term endpoints by estimating it from NLST data.
3. Validate our surrogate-assisted mortality benefit estimator by applying it to NLST data and evaluating the fit of the model.
4. Develop an adaptive trial design that uses information from short-term surrogates to adjust recruitment or even provide evidence for early termination of the trial. Subsequently, investigate whether the new design could have led to a smaller, shorter NLST if adopted.

Collaborators

Yingqi Zhao Fred Hutchinson Cancer Center
Arnold Johnsen Fred Hutchinson Cancer Center
Yingye Zheng Fred Hutchinson Cancer Center
Ting Ye Fred Hutchinson Cancer Center