A Meta-Analysis-Based Risk Score for Predicting Radiation Pneumonitis After Thoracic Radiation Therapy for Lung...

To develop and validate a risk score for predicting radiation pneumonitis (RP) in lung cancer patients following thoracic radiation therapy.

The linear predictor was first derived from a meta-analysis. Risk factors with consistent definitions and statistical significance were identified, and their pooled odds ratios were converted to β-coefficients and then linearly transformed to construct the initial scoring items. An independent Asian calibration cohort (cohort 1, n = 218) was used to estimate the model's intercept and to determine the optimal risk stratification cut-off via receiver operating characteristic analysis. The discrimination capability and clinical applicability of the model were subsequently evaluated in 2 independent Asian external validation cohorts (cohort 2, n = 69; cohort 3, n = 81). The primary endpoint was grade ≥2 RP (National Cancer Institute Common Terminology Criteria for Adverse Events version 5.0).

The final score for predicting RP included 7 variables: smoking, chemotherapy, tumor location, pulmonary comorbidities, use of renin-angiotensin system inhibitors, percentage of lung volume receiving ≥5 Gy (V5), and percentage of lung volume receiving ≥20 Gy (V20), with a total possible score of 107. It demonstrated an area under the curve of 0.831 (95% CI, 0.768-0.894) in the calibration cohort, establishing 53.5 points as the optimal cut-off value for risk stratification. Discriminative performance remained robust in external validation, with areas under the curve of 0.881 (95% CI, 0.799-0.962) and 0.845 (95% CI, 0.749-0.942), respectively. At this threshold, positive likelihood ratios (LR+) across the 3 cohorts were 3.097, 2.934, and 3.492, and negative likelihood ratios (LR-) were 0.224, 0.159, and 0.212. The high-risk group showed significantly elevated relative risks for RP (7.270, 5.143, and 5.884; all P < .001). To facilitate clinical use, this scoring system has been made available as a web-based calculator.

We developed a meta-analysis-based risk score for RP and independently validated it in Asian cohorts. This tool provides reliable risk stratification that may aid in personalized management and optimization of resource allocation for this population.