Patients undergoing standard of care proton therapy are assessed for toxicities weekly during proton treatment, then from 1 to 3 times up to 90 days from the start of treatment and annually thereafter. Patients also complete questionnaires over 15-20 minutes at baseline, weekly during treatment, and every 2 weeks during follow up for up to 3 months.

Complete questionnaires

Data will be collected from all organs receiving non-negligible proton dose during treatment. For each acute normal-tissue endpoint considered, analyses will be performed using the maximum toxicity score per patient within the 90-day follow-up. Upon analysis, the distribution of observed toxicity scores for each acute normal-tissue endpoint will be reported.

Data will be collected from all organs receiving non-negligible proton dose during treatment. For each late normal-tissue endpoint considered, analyses will be performed using censored time-to-event data corresponding to specified levels of injury (e.g. time to a specific grade >= 2 late toxicity). The log-rank test will be used to compare event times in subgroups of patients with different dosimetric characteristics of treatment (e.g. portion of lung receiving > 20 Gy: <= 40% vs > 40%). In addition, normal-tissue complication probability models will be fitted to data corresponding to incidence of a late endpoint within a specified time frame, with analysis limited to patients having the specified follow-up (e.g. incidence of grade >= 2 late rectal bleeding within 2 years among patients followed for 2 years post-treatment).

Relevant dose-volume response models from the literature, such as the Lyman model, the parallel model, or the critical-element model will be fitted to the data when possible, and model parameter estimates reported. Confidence intervals for model parameter estimates will be derived using the profile-likelihood method.

Will be documented and compared using the M.D. Anderson Symptom Inventory (MDASI). The initial analysis for symptom data will be primarily descriptive in nature, and will include frequencies, proportions, means, medians, standard deviations, ranges, interquartile ranges, confidence intervals for the means, and box and whisker plots. Potential differences in symptom development between patients and between treatment variables will be explored using longitudinal analysis performed with mixed models.