Surrogate endpoints for cancer screening trials
Individual-participant data from NLST will be used to contribute to an analysis of surrogate endpoints for cancer screening trials, and the development of new statistical methods to predict mortality results from trials. This research is intended to help improve trial efficiency, including new single or multi-cancer screening programs.
We aim to expand on current stage-based measures of surrogacy, including a two-group stage shift model (doi: 10.1158/1055-9965.EPI-22-0024). We will do this by considering models with more than two stages and exploring assumptions underlying the method, in comparison with analysis based on individual-level data which has not previously been assessed in the NLST trial. We will then use data from a number of historic trials and new survival rates to evaluate changes in the projected effect of screening through time and to estimate the current effect of screening, in absence of modern treatments. Further we would like to estimate predicted mortality, which is derived by weighting cancer cases diagnosed during a screening trial (including screen-detected cases and those diagnosed following other pathways), by the cancer- and stage-specific survival rate from diagnosis to a pre-specified time since randomisation. We will explore the role of lead time in and the sensitivity of predicted mortality (to changes in stage-specific survival assumptions) and evaluate the impact of considering net and cause-specific survival on predicted mortality.
We request individual level data on trial arm allocation, year of enrollment, basic sociodemographic information, inclusion and exclusion criteria, screening participation, cancer diagnosis (including relevant prognostic variables), model of detection, time of diagnosis from randomisation and reason for and time at censoring. We will use these data for the following:
1. Evaluation of the two-stage model to calculate the expected reduction in mortality from screening in the NLST and compare with the difference in observed mortality between the screening arms.
2. Development and evaluation of a stage shift model that considers three and four stages on data from the NLST.
3. Evaluation of stage-specific predicted mortality, using registry data, based on information available at time of diagnosis in the NLST and compare with the difference in observed mortality between the screening arms.
4. Using data from the NLST as part of analysis of the projected effect of screening through time, to evaluate the role of timing on utility of surrogate endpoints.
Peter Saseini: Professor of Cancer Epidemiology, Centre Co-Lead for the Centre for Cancer Screening, Prevention and Early Diagnosis, Queen Mary University of London.
Rhian Gabe: Professor of Biostatistics and Clinical Trials, Queen Mary University of London.
Matejka Rebolj: Senior Epidemiologist, Queen Mary University of London.