Can Mean Pericardial Irradiated Dose (MPD) Predict Pericardial Effusion after Chemoradiotherapy in Esophageal Cancer Patients? Normal Tissue Complication Probability (NTCP) Model-Based Analyses

Abstract

We previously reported MPD is the most useful predictor for the incidence of symptomatic pericardial effusion in esophageal cancer patients after concurrent chemoradiotherapy (CCR). The purpose of this study is to validate the usefulness of the MPD to predict the incidence using the Lyman–Kutcher–Burman (LKB) NTCP model. We retrospectively reviewed consecutive primary esophageal cancer patients treated with CCR between 2001 and 2014. Inclusion criteria were as follows: newly diagnosed and first treatment for esophageal cancer; disease including thoracic esophagus; follow-up ≥6 months; conventional fractionation and irradiated larger than 50 Gy; WHO performance status ≤2; available computed tomography (CT) data to analyze dose volume of pericardium and MPD; intact of malignant pericardial diseases. Symptomatic effusion was defined as pericardial effusion ≥ grade 3, according to the CTCAE version 4.0. Pericardial effusion was reviewed on follow-up chest CT. Generalized equivalent uniform dose (gEUD) based LKB NTCP parameters (n, m, and TD50) were estimated using maximum likelihood method. The optimum parameters were determined by the gradient descent method. Linear and probit regression was performed for MPD and estimated NTCP for each patient. In the probit regression using MPD, LKB NTCP parameters um and TD50 were estimated under the situation of n = 1, presuming pericardium as an organ with large volume effect. In this case, TD50 means MPD, instead of gEUD. Overall, 229 patients were met the criteria. The median follow up was 37 months (ranged 6 to 178 months). Pericardial effusion in any grade and symptomatic effusion were observed in 100 (44%) and 18 (8%) patients, respectively. Onset of pericardial effusion ranged from 2 to 75 months. The aggregated dose-volume data were preferably fitted by NTCP curve, in particular for symptomatic effusion. The estimated TD50 for symptomatic effusion and pericardial effusion in any grade were 56.0 Gy (m = 0.19, n = 0.43) and 41.0 Gy (m = 0.47, n = 0.38), respectively. The calculated NTCP curve for the regression using the MPD for symptomatic effusion showed similar sigmoid curve with a cut-off. The estimated TD50 (MPD) was 54.8 Gy (m = 0.25). While the NTCP curve for the regression using the MPD for pericardial effusion in any grade were approximated by the slope of 0.014 per Gy and intercept of 0.04 for linear regression. The estimated TD50 (MPD) was 34.4 Gy (m = 0.76). The MPD based NTCP model for symptomatic effusion showed similar results as with the original model and supported MPD as a practical predictor. Meanwhile, the model for pericardial effusion in any grade displayed different shape and parameter, as with the original model, might be indicating influence of other factors than MPD.