A Nomogram Based on Radiomics and Genomics Factors for Predicting Radiation Pneumonia in Patients with Limited-Stage Small Cell Lung Cancer after Definitive Chemoradiotherapy.

PURPOSE/OBJECTIVE(S): Although definitive chemoradiotherapy (dCRT) is currently the most effective treatment for limited-stage small cell lung cancer (SCLC), some patients may develop severe radiation pneumonia (RP). We aimed to develop a nomogram by combining the clinical, radiomics and genomics factors to predict the RP.

MATERIALS/METHODS: Totally 148 limited-stage SCLC patients treated with dCRT were retrospectively enrolled and divided into the training (n = 103) and validation (n = 45) cohorts. All tumor tissue samples were detected by 474 Panel gene sequencing technology. Eight hundred fifty-one radiomics features of lung other than hilum, atelectasis and tumor were extracted from the initial CT images. Least absolute shrinkage and selection operator (LASSO) method was applied to select and build radiomic signature. Combined with genomic, radiomic and clinical features, a nomogram was established by univariate and multivariate analysis. Concordance index (C-index), calibration curve and decision curve were used to evaluate the predictive ability and application value of the nomogram.

RESULTS: Six selected radiomic features was used to construct radiomic signature. Multivariate analysis showed that NQO1 gene (OR, 8.354; 95% CI, 1.061-65.765; P = 0.044), radiotherapy technique (OR, 9.586; 95% CI, 1.168-78.70; P = 0.035) and radiomic signature (OR, 5.745; 95% CI, 1.879-17.56; P = 0.002) were independent risk factors for RP. Compared with the single clinical (C-index = 0.621 and 0.533), radiomic (C-index = 0.696 and 0.550) or genetic (C-index = 0.561 and 0.433) factor, the C-index of nomogram was the highest, reaching 0.764 and 0.622 in the training group and validation group, respectively. The calibration curves indicated good agreement between the predictions and actual observations. Moreover, decision curves revealed that the nomogram had high clinical practicability.

CONCLUSION: This study further reveals the potential of radiomics and genomics in predicting the occurrence of RP, and may provide individualized prediction for patients with limited-stage SCLC, as well as theoretical and technical support for physicians to optimize clinical decisions.