Predicting Intolerance to Deep-Inspiration Breath-Hold for Abdominal Cancer Radiation Therapy

This study aimed to predict deep-inspiration breath-hold (DIBH) intolerance during abdominal cancer radiation therapy and its impact on residual motion, based on patient characteristics and DIBH performance at simulation.

We retrospectively studied 82 patients on voluntary DIBH for abdominal cancers between January and July 2022. DIBH was performed using infrared markers to track abdominal wall motion. The anteroposterior motion during simulation was recorded as a reference trace based on which a gate window was set. DIBH intolerance was defined as modification of the gate window or conversion to free breathing during the treatment course. Predictive features included reference trace characteristics (breath-hold level [BHL], DIBH stability and reproducibility, free-breathing amplitude, period, baseline drift) and baseline clinical characteristics (age, performance status, comorbidities, anxiety, preferred language). Univariate tests and random forest multivariate modeling were used for analysis, with model performance assessed using 10-fold cross-validation. The association between predicted intolerance and residual fiducial motion during treatment in a subset of 15 patients, was also investigated.

A statistically significant association was found between BHL at simulation and DIBH intolerance (4% false discover rate). A 1-mm increase in BHL increased intolerance risk by 7%. The multivariate model, combining trace and clinical characteristics, predicted DIBH intolerance with a median area under the curve of 0.63. Reference trace characteristics showed higher predictive importance than clinical characteristics. The predicted intolerance risk increased sharply for BHL >25 mm. Residual motion was significantly higher for the patient intolerant of DIBH (P = .02). A moderate positive correlation (r = 0.46, P = .08) was observed between the model prediction and residual motion during treatment.

Coaching patients to reach appropriate BHL is critical to DIBH tolerability and to reduce residual motion. The multivariate model suggests a limited but potential framework for predicting intolerance risk.