gms | German Medical Science

Background: In solid tumor clinical development, a standard measure of drug-induced anti-tumor activity is the best percentage change from baseline in tumor size during the treatment period. This measure is often displayed in the form of waterfall plots. Unfortunately, both estimand and display are sources of interpretation error.

The goal of this presentation is to highlight these errors and offer alternative solutions to assess the drug effect on solid tumors.

Methods: In contrast to waterfall plots, simple spaghetti plots are introduced. Furthermore, several nonlinear mixed-effects models are presented as options to describe typical trends in drug-induced anti-tumor activity.

Results: Despite imaging technology informing on the 3D mesh of the primary tumor and metastases, only the longest diameter is usually retained in the anti-tumor response assessment. Information on the actual tumor burden is lost during this dimension reduction step.

As differences in percentage change from baseline can be imputed to either the baseline or the amplitude of response to treatment, comparisons across non-randomized trials or between subgroups can be misleading.

Guidelines (such as RECIST) suggesting dichotomization in responder/non-responder based on the percentage change from baseline add another layer of confusion, due to the risk of misclassification associated with the error in measurement (approximately 10%).

To address these limitations, we recommend a simple display of the raw data, i.e. tumor volume over time. We show that bi-exponential models can easily describe typical trends in drug-induced anti-tumor activity. We discuss various metrics used to summarize these trends.

Simulations and real clinical trial data applications are used to support this presentation.

Conclusion: Our work supports the idea that adequate analysis of tumor size data in their simplest form will ensure robustness of the statistical results.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.