Integrating photon with proton dose-response data alters a pulmonary complication model-based patient selection in esophageal cancer.

Abstract

To develop a general lung dose-response curve for postoperative pulmonary complications (PPCs) following neoadjuvant chemoradiotherapy (nCRT) and surgery in esophageal cancer (EC), and to assess the impact of including proton therapy (PT) data in normal tissue complication probability (NTCP) model development. A comprehensive PubMed search (2004-2024) identified studies reporting mean lung dose (MLD) and PPCs after esophagectomy with nCRT. Logistic and probit NTCP models were fitted to study-level or dose-bin data using weighted maximum likelihood optimization, with and without PT data (XT + PT and XT model, respectively). Meta-analysis evaluated other PPCs risk factors. A simulation study on 40 patients with comparative XT and PT plans assessed model-based patient selection, assuming PT referral when Δ NTCP ≥ 10 %. 16 non-overlapping studies were included for review. PPCs incidence was significantly lower in PT versus XT cohort (15.6 % versus 30.7 %, p < 0.001). Both NTCP models showed increasing PPCs risk with rising MLD, with the XT model predicting higher PPCs rate in the low-dose region. Meta-analysis identified histology, surgery approach and pre-RT FEV₁ as significant predictors (p < 0.05). In the simulation study, the XT + PT model predicted lower PPCs risk and subsequently higher Δ NTCP than the XT model (average 16.3 % versus 8.0 %, p < 0.001). Referral decisions differed in 15 of 40 patients (37.5 %), who were redirected to PT only by the XT + PT model. Including PT dose-response data into thoracic NTCP models alters pulmonary complication model-based patient selection in EC.