Ethical and Scientific Features of Cutoff-based Designs of Clinical Trials

Abstract

Cutoff-based clinical trial designs are geared towards balancing ethical and scientific con cerns when it is deemed unethical or infeasible to randomize all patients to study treatments. In a cutoff-based design with randomization, patients who are the least sick based on a quantitative baseline indicator are assigned the control treatment, patients who are the most sick based on the same indicator are assigned to test treatment, and patients who are moderately sick based on the indicator are randomly assigned. Simulations were conducted to examine statistical efficiency and potential bias for designs with varying amounts of cutoff- based assignment and randomization. All design variations yielded unbiased estimates of a main treatment effect and a linear interaction effect. While randomization tends to lead to greater efficiency (or lower standard errors of treatment effect), the correlation between the binary treatment variable and baseline assignment variable completely determines the ef ficiency of a design. Key words: clinical trials; randomized control trials; randomization; cutoff- based designs; experimental designs; quasi-experimental designs; methodology; statistical efficiency; bias. (Med Decis Making 1995;15:387-394)