ABSTRACT
Department of Biostatistics, School of Public Health, University of Michigan,
Ann Arbor, Michigan, USA
Kalyanee Viraswami-Appanna
Novartis Pharmaceuticals Corporation, Florham Park, New Jersey, USA
Bharani Dharan
Novartis Pharmaceuticals Corporation, Florham Park, New Jersey, USA
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
10.1.1 Overview of Progression-Free Survival . . . . . . . . . . . . . . . . . . 272
10.1.2 Potential Bias in PFS Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 274
10.1.3 Conventional PFS Analysis and Interval-Censored
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
10.2 Statistical Methods for PFS Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
10.2.1 Conventional Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
10.2.2 Finkelstein’s Method for Interval-Censored Data . . . . . . . 279
10.3 Monte Carlo Simulation Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
10.3.1 Simulation Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
10.3.2 Results on Point Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
10.3.3 Results on Hypothesis Testing . . . . . . . . . . . . . . . . . . . . . . . . . . 289
272 Interval-Censored Time-to-Event Data: Methods and Applications
10.3.4 Summary of Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . 295
10.4 Discussions and Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
Appendix: SAS Programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
In oncology clinical trials, the primary goal of anti-cancer drugs is to prolong
the survival of patients. Hence, overall survival (OS: time from randomiza-
tion to death due to any cause) thus becomes the most desirable endpoint for
evaluating treatment effect because it is the most objective, least biased, and
precise to measure endpoint. In addition, OS is also the best measure of clinical
benefit in any disease indication. However, in some solid tumor settings, OS
may not always be the most appropriate endpoint as it takes a longer time
to follow up patients, requires a larger sample size, and, in addition, treat-
ment effect can be confounded due to post-treatment antineoplastic therapies
received by the patients.
