A sparse additive model for treatment effect-modifier selection

A RELUCTANT ADDITIVE MODEL FRAMEWORK FOR INTERPRETABLE NONLINEAR INDIVIDUALIZED TREATMENT RULES

Individualized treatment rules (ITRs) for treatment recommendation is an important topic for precision medicine as not all beneficial treatments work well for all individuals. Interpretability is a desirable property of ITRs, as it helps practitioners make sense of treatment decisions, yet there is a need for ITRs to be flexible to effectively model complex biomedical data for treatment decision making. Many ITR approaches either focus on linear ITRs, which may perform poorly when true optimal ITRs are nonlinear, or black-box nonlinear ITRs, which may be hard to interpret and can be overly complex. This dilemma indicates a tension between interpretability and accuracy of treatment decisions. Here we propose an additive model-based nonlinear ITR learning method that balances interpretability and flexibility of the ITR. Our approach aims to strike this balance by allowing both linear and nonlinear terms of the covariates in the final ITR. Our approach is parsimonious in that the nonlinear term is included in the final ITR only when it substantially improves the ITR performance. To prevent overfitting, we combine crossfitting and a specialized information criterion for model selection. Through extensive simulations we show that our methods are data-adaptive to the degree of nonlinearity and can favorably balance ITR interpretability and flexibility. We further demonstrate the robust performance of our methods with an application to a cancer drug sensitive study.

An Innovative Non-Linear Prediction Model for Clinical Benefit in Women with Newly Diagnosed Breast Cancer Using Baseline FDG-PET/CT and Clinical Data

Objectives: We aimed to develop a novel non-linear statistical model integrating primary tumor features on baseline [18F]-fluorodeoxyglucose positron emission tomography/computed tomography (FDG-PET/CT), molecular subtype, and clinical data for treatment benefit prediction in women with newly diagnosed breast cancer using innovative statistical techniques, as opposed to conventional methodological approaches. Methods: In this single-center retrospective study, we conducted a comprehensive assessment of women newly diagnosed with breast cancer who had undergone a FDG-PET/CT scan for staging prior to treatment. Primary tumor (PT) volume, maximum and mean standardized uptake value (SUVmax and SUVmean), metabolic tumor volume (MTV), and total lesion glycolysis (TLG) were measured on PET/CT. Clinical data including clinical staging (TNM) but also PT anatomical site, histology, receptor status, proliferation index, and molecular subtype were obtained from the medical records. Overall survival (OS), progression-free survival (PFS), and clinical benefit (CB) were assessed as endpoints. A logistic generalized additive model was chosen as the statistical approach to assess the impact of all listed variables on CB. Results: 70 women with newly diagnosed breast cancer (mean age 63.3 ± 15.4 years) were included. The most common location of breast cancer was the upper outer quadrant (40.0%) in the left breast (52.9%). An invasive ductal adenocarcinoma (88.6%) with a high tumor proliferation index (mean ki-67 expression 35.1 ± 24.5%) and molecular subtype B (51.4%) was by far the most detected breast tumor. Most PTs displayed on hybrid imaging a greater volume (12.8 ± 30.4 cm3) with hypermetabolism (mean ± SD of PT maximum SUVmax, SUVmean, MTV, and TLG, respectively: 8.1 ± 7.2, 4.9 ± 4.4, 12.7 ± 30.4, and 47.4 ± 80.2). Higher PT volume (p < 0.01), SUVmax (p = 0.04), SUVmean (p = 0.03), and MTV (<0.01) significantly compromised CB. A considerable majority of patients survived throughout this period (92.8%), while five women died (7.2%). In fact, the OS was 31.7 ± 14.2 months and PFS was 30.2 ± 14.1 months. A multivariate prediction model for CB with excellent accuracy could be developed using age, body mass index (BMI), T, M, PT TLG, and PT volume as predictive parameters. PT volume and PT TLG demonstrated a significant influence on CB in lower ranges; however, beyond a specific cutoff value (respectively, 29.52 cm3 for PT volume and 161.95 cm3 for PT TLG), their impact on CB only reached negligible levels. Ultimately, the absence of distant metastasis M displayed a strong positive impact on CB far ahead of the tumor size T (standardized average estimate 0.88 vs. 0.4). Conclusions: Our results emphasized the pivotal role played by FDG-PET/CT prior to treatment in forecasting treatment outcomes in women newly diagnosed with breast cancer. Nevertheless, careful consideration is required when selecting the methodological approach, as our innovative statistical techniques unveiled non-linear influences of predictive biomarkers on treatment benefit, highlighting also the importance of early breast cancer diagnosis.

Functional additive models for optimizing individualized treatment rules

A novel functional additive model is proposed, which is uniquely modified and constrained to model nonlinear interactions between a treatment indicator and a potentially large number of functional and/or scalar pretreatment covariates. The primary motivation for this approach is to optimize individualized treatment rules based on data from a randomized clinical trial. We generalize functional additive regression models by incorporating treatment-specific components into additive effect components. A structural constraint is imposed on the treatment-specific components in order to provide a class of additive models with main effects and interaction effects that are orthogonal to each other. If primary interest is in the interaction between treatment and the covariates, as is generally the case when optimizing individualized treatment rules, we can thereby circumvent the need to estimate the main effects of the covariates, obviating the need to specify their form and thus avoiding the issue of model misspecification. The methods are illustrated with data from a depression clinical trial with electroencephalogram functional data as patients' pretreatment covariates.