Haskell-Mendoza, Aden

Therapeutic clinical trial enrollment does not match glioma incidence across demographics. Traditional statistical methods have identified independent predictors of trial enrollment; however, our understanding of the interactions between these factors remains limited. To test the interactive effects of demographic, socioeconomic, and oncologic variables on trial enrollment, we designed boosted neural networks (BNNs) for all glioma patients ( n = 1042), women ( n = 445, 42.7%), and minorities ( n = 151, 14.5%) and externally validated these models [whole cohort, n = 230; women, n = 89 (38.7%); minority, n = 66 (28.7%)]. For the whole-cohort BNN, the most influential variables on enrollment were oncologic variables, including KPS [total effect (TE), 0.327], chemotherapy (TE, 0.326), tumor location (TE, 0.322), and seizures (TE, 0.239). The women-only BNN exhibited a similar trend. Conversely, for the minority-only BNN, socioeconomic variables [insurance status (TE, 0.213), occupation classification (TE, 0.204), and employment status (TE, 0.150)] were most influential. These results may help prioritize patient-specific initiatives to increase accrual.

Abstract BACKGROUND Despite changes in national policy, women and minority patients remain under-accrued in glioma clinical trials. Supervised machine learning may serve as a novel approach to understanding which variables and their complex interplay most influence enrollment for underrepresented patients. We aimed to design and externally validate machine learning models to predict therapeutic clinical trial enrollment for a) all patients, b) women, and c) NIH-designated minority patients with low- and high-grade glioma. METHODS In a 20-year retrospective cohort of 1042 adult patients with low- and high-grade glioma, boosted neural network (BNN) models for the whole, women, and minority cohorts were used to analyze the combined effects of demographic, socioeconomic, and oncologic variables on clinical trial enrollment. These models were validated using an external cohort (n=103). RESULTS In the development cohort (445 [42.7%] women; 151 [14.5%] minorities; median age, 51.5 [interquartile range: 39.4-62.5] years), 350 patients (33.6%) enrolled in a therapeutic clinical trial. The whole cohort BNN (AUC: 0.827, misclassification rate (MR): 0.175), the women cohort BNN (AUC: 0.895, MR: 0.125), and the minority cohort BNN (AUC: 0.983, MR: 0.053) models accurately predicted trial enrollment on external validation. For the whole cohort and women models, several demographic (race, ethnicity, minority status, sex), socioeconomic (insurance, employment status), and oncologic variables (seizure, chemotherapy, KPS, WHO grade, tumor lobe) were equally influential (Main effect (ME): 1.000, Total effect (TE): 1.000) on enrollment. For the minority model, socioeconomic variables including insurance status (ME: 0.089, TE: 0.417) and occupation classification (ME: 0.087, TE: 0.381) drove enrollment. CONCLUSION We designed and externally validated the first machine learning model for predicting therapeutic clinical trial enrollment for patients with diffuse glioma. Model features driving trial enrollment varied for underrepresented patients. These results may help prioritize personalized patient-specific initiatives to address barriers to equitable patient accrual.