Contacts
An article by Marco Bonetti and colleagues describes a statistical tool which analyses data from large randomized clinical trials to make it possible to tailor treatment decisions to the individual patient

Modern large-scale cancer clinical trials include the quantitative assessment of selected biomarkers thought to be associated with clinical outcome and treatment effectiveness, thus providing great potential for improving medical care. These biomarkers are increasingly reported on a continuous scale, allowing investigators to explore how treatment efficacy changes as the biomarker values vary.

Typical analytical approaches in cancer clinical trials evaluate treatment effect modification, also called interaction or treatment-effect heterogeneity, by first defining (often arbitrary) patient subgroups based on biomarker expression levels. Treatment comparisons are then performed within each subgroup, and the results are assessed for heterogeneity. Regression methods are also typically used to evaluate whether biomarker status is associated with treatment efficacy. The approach of categorizing biomarker expression may fail to fully identify the worth of the biomarker as a predictor of treatment efficacy because categorization results in a loss of information. Alternatives to such dichotomized analyses should be applied. However, full-scale statistical modeling may be difficult to accept, as it typically implies rather strong assumptions about the structure of the phenomenon and as it moves away from customary, accepted analyses.

Marco Bonetti (Department of Decision Sciences) provides an overview of an intuitive statistical approach, the Subpopulation Treatment Effect Pattern Plots (STEPP), to be used, together with other statistical methods, for evaluating treatment-effect heterogeneity when the biomarker is measured on a continuous scale in Evaluation of Treatment-Effect Heterogeneity Using Biomarkers Measured on a Continuous Scale: Subpopulation Treatment-Effect Pattern Plot (STEPP) (with Ann Lazar, Dana-Farber Cancer Institute and Harvard School of Public Health, Bernard Cole, Dana-Farber Cancer Institute and University of Vermont and Richard Gelber, Dana-Farber Cancer Institute, Harvard School of Public Health and Harvard Medical School, Journal of Clinical Oncology (Statistics in Oncology), 2010 vol. 28 no. 29 4539-4544, doi: 10.1200/JCO.2009.27.9182).

This approach, originally introduced and developed by Bonetti and Gelber, is meant to aid in the identification of subgroups of patients most likely to benefit from a particular treatment modality by exploring the patterns of treatment effect across subgroups as a function of the biomarker values.

The STEPP methodology examines treatment-effect heterogeneity by estimating treatment effect within overlapping subpopulations of patients, where the subpopulations are defined with respect to values of the variable of interest along its range. After the overlapping subpopulations are constructed, the treatment effect is estimated within each subpopulation using a standard approach, with the treatment effect represented, say, by the absolute difference between two survival curves at a specific time point.

Typically, very large studies are needed to explore interactions between treatment and covariates. The paper illustrates STEPP with an analysis of data extracted from the BIG 1-98 study, an international, double-blind, phase III clinical trial of 8,010 women with early stage invasive breast cancer, who were randomly assigned to one of four adjuvant endocrine therapy arms to compare adjuvant letrozole versus tamoxifen for the treatment of postmenopausal women with hormone-receptor-positive breast cancer. In particular, the prognostic and predictive value of a well-known predictor of breast cancer prognosis (Ki-67, or tumor proliferation fraction), which is associated with the degree of effectiveness of chemotherapy, is studied. The STEPP analyses showed patients with higher Ki-67 values who were assigned tamoxifen had poorest prognosis and may benefit most from letrozole.

The study of interactions between treatment effect and one or more relevant variables has clear implications for the design of later trials and for clinical practice, since the tailoring of treatment decisions to the individual patient becomes possible. As many clinical trials are extremely expensive and time consuming, it is also important that the largest amount of information be obtained from them. One may argue that it would indeed not be ethical to do otherwise.

Research