Predictors of PEG dependence after IMRT ± chemotherapy for oropharyngeal cancer

Purpose: To prospectively assess predictors of PEG dependence after IMRT with/without concomitant chemotherapy (CHT). Methods and materials: One-hundred-seventy-one patients were considered (exclusive RT: 58, RT+CHT: 113; 159/171 treated at a median dose of 70 Gy, 2 Gy/fr). Patients treated with RT+CHT underwent prophylactic PEG insertion; PEG was as needed for the others. A number of clinical factors and dose–volume information concerning oral mucosa (OM), constrictors, masticatory muscles, larynx, esophagus and parotids were available. The 25th/10th percentiles of the duration of PEG dependence were our end-points (respectively 3.3 and 7 months, PEG3/PEG7). Logistic uni and multi-variate (MVA) analyses were performed. Results: Concerning PEG3, the independent predictors at MVA were: CHT/PEG policy (OR: 6.8, p = 0.001), V9.5GOM Gy/week (OR: 1.017, p = 0.01), larynx V50 (OR: 1.018, p = 0.01) and superior constrictor (SC) Dmean (OR: 1.002, p = 0.005); the predictive value of the model (AUC) was 0.818 (95% CI: 0.751–0.873). The independent predictors of PEG7 were: larynx V50 (OR: 1.042, p = 0.0005) and SC Dmean (OR: 1.003, p = 0.02), symptoms at diagnosis (yes vs no, OR: 3.6, p = 0.08) and sex (male vs female, OR: 0.25, p = 0.07); AUC was 0.897 (95% CI: 0.841–0.939). Conclusions: OM V9.5 Gy/week and CHT/PEGpolicy modulate the risk of early PEG dependence. For longer PEG dependence, larynx V50 (or Dmean) and SC Dmean are highly predictive, suggesting that the fibrosis of constrictors and larynx is the main cause