Multivariable Normal Tissue Complication Modeling of Acute Esophageal Toxicity in Esophageal Squamous Cancer Patients Treated with Radio(Chemo)Therapy

Abstract

To establish a multivariable normal-tissue complication probability (NTCP) model for the Grade ≧2 acute esophageal toxicity (AET) after curative radio(chemo)therapy in patients with esophageal squamous cell carcinoma (ESCC). A cohort of 184 ESCC patients who were treated with definitive radio(chemo)therapy between April 2008 and December 2011 was enrolled in this study. The analysis included clinical and esophageal dose–volume histogram parameters. Clinical parameters included age, sex, weight loss during treatment, alcohol history, tumor site, stage, the length of gross tumor volume, the length of planning target volume (LPTV), the LPTV of the primary tumor (LPTV-P). Dose–volume histogram parameters included V5 to V65 with step of 5Gy, Dmean, and Dmax of the esophagus. Univariate logistic analysis was performed for each available variable.The multivariable logistic regression model was achieved by using of least absolute shrinkage and selection operator (LASSO) with 10-folder cross-validation. Model performance was assessed by the area under the receiver operating characteristic curve. A total of 42 patients (22.83%) developed Grade ≧2 AET. In the univariate analysis, age (>65, OR=0.323, 95%CI=0.144-0.725, p=0.006) and LPTV-P (OR=1.12, 95%CI=1.007-1.246,p=0.036) were associated with Grade≧2 AET. The final LASSO model had enrolled age (>65, OR=0.65), LPTVP (OR=1.02), and tumor site (lower, OR=0.63), The multivariable NTCP model for Grade ≧2 AET was expressed by the multivariable logistic regression formula: NTCP=(1+exp(-S))-1, and S=0.0209×LPTV-P-0.4260×age-0.4586×site-1.2560.The area under the curve was 0.68 (CI=0.5925-0.7688). A multivariable NTCP model including age, LPTVP, tumor site could predict Grade≧2 AET after radio(chemo)therapy for ESCC.