Development and application of an elastic net logistic regression model to investigate the impact of cardiac substructure dose on radiation-induced pericardial effusion in patients with NSCLC

Abstract

Background: Typically, cardiac substructures are neither delineated nor analyzed during radiation treatment planning. Therefore, we developed a novel machine learning model to evaluate the impact of cardiac substructure dose for predicting radiation-induced pericardial effusion (PCE). Materials and methods: One-hundred and forty-one stage III NSCLC patients, who received radiation therapy in a prospective clinical trial, were included in this analysis. The impact of dose-volume histogram (DVH) metrics (mean and max dose, V5Gy[%]-V70Gy[%]) for the whole heart, left and right atrium, and left and right ventricle, on pericardial effusion toxicity (!grade 2, CTCAE v4.0 grading) were examined. Elastic net logistic regression, using repeat cross-validation (n ¼ 100 iterations, 75%/ 25% training/test set data split), was conducted with cardiac-based DVH metrics as covariates. The following model types were constructed and analyzed: (i) standard model type, which only included whole-heart DVH metrics; and (ii) a model type trained with both whole-heart and substructure DVH metrics. Model performance was analyzed on the test set using area under the curve (AUC), accuracy, calibration slope and calibration intercept. A final fitted model, based on the optimal model type, was developed from the entire study population for future comparisons. Results: Grade 2 PCE incidence was 49.6% (n ¼ 70). Models using whole heart and substructure dose had the highest performance (median values: AUC ¼ 0.820; calibration slope/intercept ¼ 1.356/ À0.235; accuracy ¼ 0.743) and outperformed the standard whole-heart only model type (median values: AUC ¼ 0.799; calibration slope/intercept ¼ 2.456/À0.729; accuracy ¼ 0.713). The final fitted elastic net model showed high performance in predicting PCE (median values: AUC ¼ 0.879; calibration slope/intercept ¼ 1.352/À0.174; accuracy ¼ 0.801). Conclusions: We developed and evaluated elastic net regression toxicity models of radiation-induced PCE. We found the model type that included cardiac substructure dose had superior predictive performance. A final toxicity model that included cardiac substructure dose metrics was developed and reported for comparison with external datasets.

Prediction models for treatment-induced cardiac toxicity in patients with non-small-cell lung cancer: A systematic review and meta-analysis

“…The key characteristics of the included studies are listed in Table 1. Of the 28 included papers, seven studies developed prediction models for late cardiac toxicity after (chemo-)radiotherapy [15][16][17][18][19][20][21]. We identified 20 prediction models for acute cardiac events after surgery [22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40][41] and one study estimated the risk of developing cardiovascular events after immunotherapy (nivolumab and pembrolizumab) [42].…”

“…Five studies (17.9%) considered radiotherapy (3D-conformal radiotherapy, intensity-modulated radiotherapy, or radioactive implant) with or without chemotherapy as the eligibility criteria of their study samples [15,[17][18][19][20]. Moreover, two studies (7.1%) included the patients in clinical trials who received either photon (intensity-modulated) or proton beam therapy [16,21].…”

“…The cardiac event rate in the study samples ranged between 3.9% and 49.6%. Atrial fibrillation, arrhythmia, pericardial effusion, and ischemic heart diseases were considered as the primary end point in 12 (42.9%) [27,28,[30][31][32][34][35][36][37][38][39][40], 4 (14.3%) [22][23][24]29], 2 (7.1%) [16,21], and 1 (3.6%) studies [15], respectively. Nine studies (32.1%) used all types of cardiac events as the outcome.…”

“…Nineteen studies (67.9%) measured acute cardiac side effects after surgical resection and the remaining nine studies recorded late cardiac toxicities occurring at least three months after treatment completion. Electrocardiography (n = 16, 57.1%) [22,24,[26][27][28][29][30][31][32][33][34][35][36][38][39][40], Common Terminology Criteria for Adverse Events (CTCAE) (n = 3, 10.7%) [17,19,21], and the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) (n = 2, 7.1%) [15,42] were the most frequently used instruments for assessing or grading the cardiac events.…”

“…All included studies used regression analysis as the machine learning algorithm. Four studies documented the reason for method selection [21,33,37,38] and only one study performed systematic review as the predictor selection technique before modeling [37].…”