Using Contour Plots to Assess the Sensitivity of Clinical Trial Design Assumptions

Sample size calculations are an important part of the design of any clinical trial. These calculations ensure a sufficient number of patients to detect a clinically meaningful difference between 2 treatments with high probability (ie, power). Perhaps less-often discussed, the sample size exercise is important so that resources are not wasted studying too many observations to test a particular hypothesis. Since patients may be randomized to doses of a novel treatment with a limited safety profile, or a placebo which provides no therapeutic benefit, sample size calculations in clinical trials come with an ethical burden not experienced in many subject-matter areas.

The sample size of a clinical trial should be determined using as much data as is available, over a range of assumptions, and with input from clinical colleagues. Graphical techniques are often utilized to summarize power and sample size calculations. In this manuscript, we propose the use of contour plots to better assess, report, and communicate the sensitivity of clinical trial design assumptions.

Through several examples, we illustrate that contour plots are applicable to binary, continuous, and time-to-event endpoints for a variety of study design scenarios.

Contour plots are a useful tool for the study team in designing clinical trials, and they can be included in study documents to better communicate the rationale for sample size for clinicians and regulators. Contour plots provide greater transparency as to the uncertainty of the currently available information, and can be useful in deciding whether to consider adaptive designs.