Sample Size Determination for Stratified Phase II Cancer Trials With Monotone Order Constraints

Abstract

Abstract–It is common for clinical trials to include strata of patients with different risks of disease. In cancer trial settings, patients with different prior treatment history, tumor stages, metastatic sites may be included in a single trial. This is especially so for early phase studies that are exploratory in nature. In many cases, patient responses from these strata may be ordered. For example, earlier stage patients should have better outcomes than later stage patients. Incorporating such stratum information can lead to increased statistical efficiency and therefore possibly reduce sample sizes. Recent works have begun to deal with this issue for single-arm phase II cancer trials in terms of sample size determination. However, no approaches yet exist to explicitly consider the order constraint. In this article, we propose to use likelihood ratio tests with order constraints for both one-stage and two-stage designs. Our numerical results show that the proposed designs tend to have better power profiles compared with existing methods. A real application of our method to an ongoing phase II study is also included.

KEYWORDS:

• Constrained optimization
• Likelihood ratio test
• Patient heterogeneity
• Phase II clinical trials
• Sample size determination

Supplementary Materials

R code:R code to perform the one-stage and two-stage designs described in the article.

Additional information

Funding

Jun-jun Zhao’s work was supported by the Shanghai Municipal Commission of Health and Family Planning Science Funds [201440532]. Menggang Yu gratefully acknowledges partial support from the University of Wisconsin Carbone Cancer Center Support Grant [P30 CA014520] and the University of Wisconsin Specialized Program of Research Excellence (SPORE) Head and Neck NIH program grant [P50 DE026787]. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.