Hypofractionated Stereotactic Radiation Therapy Dosimetric Tolerances for the Inferior Aspect of the Brachial Plexus: A Systematic Review.

We sought to systematically review and summarize dosimetric factors associated with radiation-induced brachial plexopathy (RIBP) after stereotactic body radiation therapy (SBRT) or hypofractionated image guided radiation therapy (HIGRT). From published studies identified from searches of PubMed and Embase databases, data quantifying risks of RIBP after 1- to 10-fraction SBRT/HIGRT were extracted and summarized. Published studies have reported <10% risks of RIBP with maximum doses (Dmax) to the inferior aspect of the brachial plexus of 32 Gy in 5 fractions and 25 Gy in 3 fractions. For 10-fraction HIGRT, risks of RIBP appear to be low with Dmax < 40 to 50 Gy. For a given dose value, greater risks are anticipated with point volume-based metrics (ie, D0.03-0.035cc: minimum dose to hottest 0.03-0.035 cc) versus Dmax. With SBRT/HIGRT, there were insufficient published data to predict risks of RIBP relative to brachial plexus dose-volume exposure. Minimizing maximum doses and possibly volume exposure of the brachial plexus can reduce risks of RIBP after SBRT/HIGRT. Further study is needed to better understand the effect of volume exposure on the brachial plexus and whether there are location-specific susceptibilities along or within the brachial plexus structure.

Radiation-induced inferior brachial plexopathy after stereotactic body radiotherapy: Pooled analyses of risks.

INTRODUCTION: Radiation-induced brachial plexopathy (RIBP), resulting in symptomatic motor or sensory deficits of the upper extremity, is a risk after exposure of the brachial plexus to therapeutic doses of radiation. We sought to model dosimetric factors associated with risks of RIBP after stereotactic body radiotherapy (SBRT). METHODS: From a prior systematic review, 4 studies were identified that included individual patient data amenable to normal tissue complication probability (NTCP) modelling after SBRT for apical lung tumors. Two probit NTCP models were derived: one from 4 studies (including 221 patients with 229 targets and 18 events); and another from 3 studies (including 185 patients with 192 targets and 11 events) that similarly contoured the brachial plexus. RESULTS: NTCP models suggest ≈10% risks associated with brachial plexus maximum dose (Dmax) of ∼32-34 Gy in 3 fractions and ∼40-43 Gy in 5 fractions. RIBP risks increase with increasing brachial plexus Dmax. Compared to previously published data from conventionally-fractionated or moderately-hypofractionated radiotherapy for breast, lung and head and neck cancers (which tend to utilize radiation fields that circumferentially irradiate the brachial plexus), SBRT (characterized by steep dose gradients outside of the target volume) exhibits a much less steep dose-response with brachial plexus Dmax > 90-100 Gy in 2-Gy equivalents. CONCLUSIONS: A dose-response for risk of RIBP after SBRT is observed relative to brachial plexus Dmax. Comparisons to data from less conformal radiotherapy suggests potential dose-volume dependences of RIBP risks, though published data were not amenable to NTCP modelling of dose-volume measures associated with RIBP after SBRT.

Explainable Artificial Intelligence to Identify Dosimetric Predictors of Toxicity in Patients with Locally Advanced Non-Small Cell Lung Cancer: A Secondary Analysis of RTOG 0617.

PURPOSE: Dosimetric predictors of toxicity in patients treated with definitive chemoradiation for locally advanced non-small cell lung cancer are often identified through trial and error. This study used machine learning (ML) and explainable artificial intelligence to empirically characterize dosimetric predictors of toxicity in patients treated as part of a prospective clinical trial. METHODS AND MATERIALS: A secondary analysis of the Radiation Therapy Oncology Group (RTOG) 0617 trial was performed. Multiple ML models were trained to predict grade ≥3 pulmonary, cardiac, and esophageal toxicities using clinical and dosimetric features. Model performance was evaluated using the area under the curve (AUC). The best performing model for each toxicity was explained using the Shapley Additive Explanation (SHAP) framework; SHAP values were used to identify relevant dosimetric thresholds and were converted to odds ratios (ORs) with confidence intervals (CIs) generated using bootstrapping to obtain quantitative measures of risk. Thresholds were validated using logistic regression. RESULTS: The best-performing models for pulmonary, cardiac, and esophageal toxicities, outperforming logistic regression, were extreme gradient boosting (AUC, 0.739), random forest (AUC, 0.706), and naive Bayes (AUC, 0.721), respectively. For pulmonary toxicity, thresholds of a mean dose >18 Gy (OR, 2.467; 95% CI, 1.049-5.800; P = .038) and lung volume receiving ≥20 Gy (V20) > 37% (OR, 2.722; 95% CI, 1.034-7.163; P = .043) were identified. For esophageal toxicity, thresholds of a mean dose >34 Gy (OR, 4.006; 95% CI, 2.183-7.354; P < .001) and V20 > 37% (OR, 3.725; 95% CI, 1.308-10.603; P = .014) were identified. No significant thresholds were identified for cardiac toxicity. CONCLUSIONS: In this data set, ML approaches validated known dosimetric thresholds and outperformed logistic regression at predicting toxicity. Furthermore, using explainable artificial intelligence, clinically useful dosimetric thresholds might be identified and subsequently externally validated.