gms | German Medical Science

Background: Novel high-throughput technology provides detailed information on the biomedical characteristics of each patient's disease. These biomarkers may qualify as predictive factors that distinguish patients who benefit from a particular treatment from patients who do not. Hence, large numbers of biomarkers need to be tested in order to gain evidence for tailored treatment decisions (“personalized medicine”). Tree-based methods divide patients into subgroups with differential treatment effects in an automated and data-driven way without requiring extensive pre-specification. Most of these methods mainly aim for a precise prediction of the individual treatment effect, thereby ignoring interpretability of the tree/random forest. Furthermore, they are mostly only applicable to data from randomized experiments.

Methods: We propose a modification of the model-based recursive partitioning (MOB) approach for subgroup analyses (Seibold et al., 2016), the so-called predMOB, that is able to specifically identify predictive factors (Krzykalla et al., 2020).

For non-randomized data, we investigate the predMOB approach in combination with common methods for confounder adjustment (covariate adjustment, inverse probability of treatment weighting, matching). The performance of these strategies is assessed concerning identification of the predictive factors as well as prediction accuracy of the individual treatment effect and the predictive effects.

For illustration, we apply the investigated versions of confounder-adjusted predMOBs to a real world data set of core-binding factor leukemia patients that were treated with standard intensive chemotherapy plus either placebo or dasatinib in order to identify predictive factors for treatment with dasatinib.

Results and Conclusion: Using simulation studies, we are able to show that predMOB provides a targeted approach to predictive factors by reducing the erroneous selection of biomarkers that are only prognostic. When predMOB is applied to non-randomized data, covariate adjustment achieves an adequate correction concerning the identification of predictive factors. If individual treatment effects or predictive effects are to be estimated, all investigated adjustment methods show satisfying results.

The authors declare that they have no competing interests.

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