Prediction models for platinum-based chemotherapy response and toxicity in advanced NSCLC patients

Abstract

In this study, we aimed to establish a platinum-based chemotherapy response and toxicity prediction model in advanced non-small cell lung cancer (NSCLC) patients. 416 single nucleotide polymorphisms (SNPs) in 185 genes were genotyped, and their association with drug response and toxicity were estimated using logistic regression. Nine data mining techniques were employed to establish the prediction model; the sensitivity, specificity, overall accuracy and receiver operating characteristic (ROC) curve were used to assess the models' performance. Finally, selected models were validated in an independent cohort. The models established by naïve Bayesian algorithm had the best performance. The response prediction model achieved a sensitivity of 0.90 and a specificity of 0.47 with the ROC area under curve (AUC) of 0.80. The overall toxicity prediction model achieved a sensitivity of 0.86 and a specificity of 0.46 with the ROC AUC of 0.73. The hematological toxicity prediction model achieved a sensitivity of 0.89 and a specificity of 0.39 with the ROC AUC of 0.76. The gastrointestinal toxicity prediction model achieved a sensitivity of 0.93 and a specificity of 0.35 with the ROC AUC of 0.80. In conclusion, we provided platinum-based chemotherapy response and toxicity prediction models for advanced NSCLC patients.

Keywords: Data mining; NSCLC; Platinum; Response; SNP; Toxicity.