Predicting Hearing Loss in Testicular Cancer Patients after Cisplatin-Based Chemotherapy

Abstract

Testicular cancer is predominantly curable, but the long-term side effects of chemotherapy have a severe impact on life quality. In this research study, we focus on hearing loss as a part of overall chemotherapy-induced ototoxicity. This is a unique approach where we combine clinical data from the acclaimed nationwide Danish Testicular Cancer (DaTeCa)-Late database. Clinical and genetic data on 433 patients were collected from hospital files in October 2014. Hearing loss was classified according to the FACT/GOG-Ntx-11 version 4 self-reported Ntx6. Machine learning models combining a genome-wide association study within a nested cross-validated logistic regression were applied to identify patients at high risk of hearing loss. The model comprising clinical and genetic data identified 67% of the patients with hearing loss; however, this was with a false discovery rate of 49%. For the non-affected patients, the model identified 66% of the patients with a false omission rate of 19%. An area under the receiver operating characteristic (ROC-AUC) curve of 0.73 (95% CI, 0.71-0.74) was obtained, and the model suggests genes SOD2 and MGST3 as important in improving prediction over the clinical-only model with a ROC-AUC of 0.66 (95% CI, 0.65-0.66). Such prediction models may be used to allow earlier detection and prevention of hearing loss. We suggest a possible biological mechanism for cisplatin-induced hearing loss development. On confirmation in larger studies, such models can help balance treatment in clinical practice.

Keywords: chemotherapy regimen; genetics; hearing loss; machine learning; testicular cancer.

Figure 1

ROC-AUC mean (30 random data splits) performances in each step of the forward feature selection. (A): Model with clinical data with forward feature selection up until nine features. Shaded blue area indicates 95% CI. Exact ROC-AUC mean and 95% CI in (C). (B): Model with clinical and genetic data with forward feature selection up until 28 features. Shaded areas indicate 95% CI, blue color indicates that only clinical data were added, green color that clinical and genetic data were added, and red color that ROC-AUC reached a plateau. Exact ROC-AUC mean and 95% CI in (D). For illustration purposes, exact ROC-AUC mean and 95% CI are not indicated in (D) from 13 features. From 13 to 28 features, ROC-AUC mean (95% CI) was 0.73 (0.71–0.75) (13–15 features); 0.73 (0.71–0.75) (14–15 features); 0.73 (0.72–0.75) (16–17 features); 0.73 (0.71–0.75) (18–21 features); 0.72 (0.70–0.74) (22–25 features); 0.72 (0.70–0.73) (26 features); and 0.71 (0.69–0.73) (27–28 features). ROC-AUC = area under the receiver operating characteristic curve; No. = number; CI = confidence interval.

Figure 2

Final model performance measures and prediction scores. (A): Model performance shown as ROC-AUC curve. Solid blue line and shaded area indicate the mean and standard deviation across 30 data splits. Dashed red line indicates a random classifier. (B): ROC-AUC and other performance measures, i.e., MCC, sensitivity, specificity, PPV, and NPV using a cut-off of 0.50. (C): Final prediction scores (x-axis) for each patient, represented by a dot. Orange dots represent controls or non-affected patients (FACT/GOG-Ntx6 score 0–1), while blue dots represent cases or affected patients (FACT/GOG-Ntx6 score 2–4). Dashed vertical line represents a cut-off of 0.50, where patients with a prediction score of 0.50 or higher are considered cases. ROC-AUC = area under the receiver operating characteristic curve; MCC = Matthews correlation coefficient; PPV = positive predictive value; NPV = negative predictive value.

Figure 3

SHAP value feature importance. Individual features are ranked by importance, where age at diagnosis is the most important feature. The color represents the feature value (red: high; blue: low). Negative SHAP values (x-axis) contribute toward a negative model outcome (control or non-affected), while positive SHAP values contribute toward a positive model outcome (case or affected).