Marilyn S. Huang

Training, Validating, and Testing Machine Learning Prediction Models for Endometrial Cancer Recurrence

PURPOSE Endometrial cancer (EC) is the most common gynecologic cancer in the United States with rising incidence and mortality. Despite optimal treatment, 15%-20% of all patients will recur. To better select patients for adjuvant therapy, it is important to accurately predict patients at risk for recurrence. Our objective was to train, validate, and test models of EC recurrence using lasso regression and other machine learning (ML) and deep learning (DL) analytics in a large, comprehensive data set. METHODS Data from patients with EC were downloaded from the Oncology Research Information Exchange Network database and stratified into low risk, The International Federation of Gynecology and Obstetrics (FIGO) grade 1 and 2, stage I (N = 329); high risk, or FIGO grade 3 or stages II, III, IV (N = 324); and nonendometrioid histology (N = 239) groups. Clinical, pathologic, genomic, and genetic data were used for the analysis. Genomic data included microRNA, long noncoding RNA, isoforms, and pseudogene expressions. Genetic variation included single-nucleotide variation (SNV) and copy-number variation (CNV). In the discovery phase, we selected variables informative for recurrence ( P < .05), using univariate analyses of variance. Then, we trained, validated, and tested multivariate models using selected variables and lasso regression, MATLAB (ML), and TensorFlow (DL). RESULTS Recurrence clinic models for low-risk, high-risk, and high-risk nonendometrioid histology had AUCs of 56%, 70%, and 65%, respectively. For training, we selected models with AUC >80%: five for the low-risk group, 20 models for the high-risk group, and 20 for the nonendometrioid group. The two best low-risk models included clinical data and CNVs. For the high-risk group, three of the five best-performing models included pseudogene expression. For the nonendometrioid group, pseudogene expression and SNV were overrepresented in the best models. CONCLUSION Prediction models of EC recurrence built with ML and DL analytics had better performance than models with clinical and pathologic data alone. Prospective validation is required to determine clinical utility.

The Vaginal Microbiome is Associated with Endometrial Cancer Grade and Histology

The human microbiome has been strongly correlated with disease pathology and outcomes, yet remains relatively underexplored in patients with malignant endometrial disease. In this study, vaginal microbiome samples were prospectively collected at the time of hysterectomy from 61 racially and ethnically diverse patients from three disease conditions: (i) benign gynecologic disease (controls, n = 11), (ii) low-grade endometrial carcinoma (n = 30), and (iii) high-grade endometrial carcinoma (n = 20). Extracted DNA underwent shotgun metagenomics sequencing, and microbial α and β diversities were calculated. Hierarchical clustering was used to describe community state types (CST), which were then compared by microbial diversity and grade. Differential abundance was calculated, and machine learning utilized to assess the predictive value of bacterial abundance to distinguish grade and histology. Both α- and β-diversity were associated with patient tumor grade. Four vaginal CST were identified that associated with grade of disease. Different histologies also demonstrated variation in CST within tumor grades. Using supervised clustering algorithms, critical microbiome markers at the species level were used to build models that predicted benign versus carcinoma, high-grade carcinoma versus benign, and high-grade versus low-grade carcinoma with high accuracy. These results confirm that the vaginal microbiome segregates not just benign disease from endometrial cancer, but is predictive of histology and grade. Further characterization of these findings in large, prospective studies is needed to elucidate their potential clinical applications. Significance: The vaginal microbiome reliably segregates not just benign gynecologic condition from endometrial cancer, but also predicts cancer grade and histology. Patterns of microbial abundance and gene expression should be increasingly considered as a factor in the evolution of precision medicine approaches, especially as they relate to cancer screening, disease pathogenesis, and patient-centered outcomes.

Ofranergene Obadenovec (Ofra-Vec, VB-111) With Weekly Paclitaxel for Platinum-Resistant Ovarian Cancer: Randomized Controlled Phase III Trial (OVAL Study/GOG 3018)

PURPOSE To evaluate the addition of ofranergene obadenovec (ofra-vec, VB-111), a novel gene-based anticancer targeted therapy, to once a week paclitaxel in patients with recurrent platinum-resistant ovarian cancer (PROC). METHODS This placebo-controlled, double-blind, phase III trial (ClinicalTrials.gov identifier: NCT03398655 ) randomly assigned patients with PROC 1:1 to receive intravenous ofra-vec every 8 weeks with once a week IV paclitaxel or placebo with paclitaxel until disease progression. The dual primary end points were overall survival (OS) and progression-free survival (PFS) as assessed by Blinded Independent Central Review. RESULTS Between December 2017 and March 2022, 409 patients were randomly assigned. The median PFS was 5.29 months in the ofra-vec arm and 5.36 months in the control arm, hazard ratio (HR) 1.03 (CI, 0.83 to 1.29; P = .7823). The median OS with ofra-vec was 13.37 months versus 13.14 months, HR 0.97 (CI, 0.75 to 1.27; P = .8440). Objective response rates (ORRs) per RECIST 1.1 were similar in both arms: 28.9% with ofra-vec versus 29.6% with control. In both treatment arms, response to CA-125 was a substantial prognostic factor for both PFS and OS. In the ofra-vec arm, the HR in CA-125 responders compared with that in nonresponders for PFS was 0.2428 (CI, 0.1642 to 0.3588), and for OS, the HR was 0.3343 (CI, 0.2134 to 0.5238). Safety profile was characterized by common transient flu–like symptoms such as fever and chills. CONCLUSION The addition of ofra-vec to paclitaxel did not improve PFS or OS. The PFS and ORR in the control arm exceeded the results that were anticipated on the basis of the AURELIA chemotherapy control arm. CA-125 response was a substantial prognostic biomarker for PFS and OS in patients with PROC treated with paclitaxel.