NTCP model for postoperative complications and one-year mortality after trimodality treatment in oesophageal cancer.

Abstract

To develop normal tissue complication probability (NTCP) models for postoperative pulmonary and cardiac complications and one-year mortality after preoperative chemoradiotherapy and surgery in oesophageal cancer patients. 691 patients from two institutions (2002-2017) were included; 134 treated with protons. Multivariable logistic regression analyses on 601 patients studied the predictive value of clinical/treatment-related (gender, age, body mass index (BMI), smoking, cardiac comorbidity, chronic obstructive pulmonary disease, histology, cT/N) and dosimetric variables (absolute/relative lung/heart volumes receiving or spared from xGy, mean doses, planning target volume) for the presence of pulmonary complications, cardiac complications and one-year mortality. Model validation was performed using a nonrandom split-sample of 90 patients. Model performance was assessed by AUC and calibration plots. Respectively 144/601 (24.0%) and 165/601 (27.5%) patients developed a pulmonary or cardiac complication. For pulmonary complications, an NTCP model with optimism-corrected AUC of 0.75 (95%CI = 0.73-0.76) was obtained. The model contained mean lung dose (OR = 1.15, 95%CI = 1.09-1.22, p < 0.001), increasing age (OR = 1.03, 95%CI = 1.01-1.06, p = 0.002), BMI (OR = 1.04, 95%CI = 0.99-1.08, p = 0.084) and squamous cell carcinoma (OR = 3.22, 95%CI = 1.97-5.24, p < 0.001) as predictors. In validation, AUC of 0.79 was obtained (calibration slope 1.26). For cardiac complications, only age (OR = 1.06, 95%CI = 1.04-1.09, p < 0.001) with optimism-corrected AUC of 0.67 (95%CI = 0.65-0.68) was selected. For one-year mortality, an NTCP model with optimism-corrected AUC of 0.63 (95%CI = 0.58-0.66) was obtained. Lung absolute V₃₅ (OR = 1.0016, 95%CI = 1.0007-1.0026, p = 0.001), cN (OR = 2.45, 95%CI = 1.18-5.09, p = 0.017), cT4 (OR = 2.51, 95%CI = 1.10-5.74, p = 0.029) and cardiac comorbidity (OR = 2.91, 95%CI = 1.46-5.77, p = 0.002) were selected as predictors. At validation, AUC of 0.57 was obtained (calibration slope 0.75). We were able to build and validate NTCP models for the presence of a postoperative pulmonary complication and for one-year mortality after trimodality treatment in oesophageal cancer.