Constructing treatment decision rules based on scalar and functional predictors when moderators of treatment effect are unknown

Abstract

Treatment response heterogeneity poses serious challenges for selecting treatment for many diseases. To better understand this heterogeneity and to help in determining the best patient-specific treatments for a given disease, many clinical trials are collecting large amounts of patient-level data prior to administering treatment in the hope that some of these data can be used to identify moderators of treatment effect. These data can range from simple scalar values to complex functional data such as curves or images. Combining these various types of baseline data to discover "biosignatures" of treatment response is crucial for advancing precision medicine. Motivated by the problem of selecting optimal treatment for subjects with depression based on clinical and neuroimaging data, we present an approach that both (1) identifies covariates associated with differential treatment effect and (2) estimates a treatment decision rule based on these covariates. We focus on settings where there is a potentially large collection of candidate biomarkers consisting of both scalar and functional data. The validity of the proposed approach is justified via extensive simulation experiments and illustrated using data from a placebo-controlled clinical trial investigating antidepressant treatment response in subjects with depression.

Keywords: Depression; Functional data; Penalized estimation; Precision medicine; Treatment regime.

Figures

Fig. 1

Simulated 1-D covariates.

Fig. 2

Simulation Set A (with error in functional predictors) Scenarios 2 and 4. Boxplots of expected values of response in test sets under estimated optimal regime for the 100 experiments in each setting. First Row Settings with p = 5, q = 3. Second Row Settings with p = 100, q = 10. Sample sizes and treatment regime estimation methods are on the vertical axis. Expected values of the decision rule is on the horizontal axis with E{Y*(−1)},E{Y*(1)}, and E{Y*(gopt)} marked. Mean (sd) PCD for each method and sample size combination shown on the right of each plot.

Fig. 3

Simulation Set B (with error in functional predictors) Scenarios 2 and 4. Boxplots of expected values of response in test sets under estimated optimal regime for the 100 experiments in each setting. First Row Settings with p = 5, q = 3. Second Row Settings with p = 100, q = 10. Sample sizes and treatment regime estimation methods are on the vertical axis. Expected values of the decision rule is on the horizontal axis with E{Y*(−1)},E{Y*(1)}, and E{Y*(gopt)} marked. Mean (sd) PCD for each method and sample size combination shown on the right of each plot.

Fig. 4

Left: EEG scalp electrodes used in the analysis. Right: Relative CSD amplitude curves for test set subjects (rows 1 and 3) and the corresponding estimated contrast coefficients (rows 2 and 4).

Fig. 5

Left: Boxplots of week 8 HAMD scores for test set subjects under different regimes. Right: Mean response (lower values are better) and 95% bootstrap confidence intervals under different regimes in the test set.