2.8. Statistical analysis

Longitudinal data were collected from 11 visits, including the baseline visit. Patients with RP CTCAE grade ≥ 2 were defined as the RP group while patients with RP CTCAE grade < 2 were allocated to the non-RP group. Patient characteristics between the groups were compared using the t-test, Mann-Whitney’s U test or chi-square tests depending on whether the variables were continuous or categorical. The mean FeNO_{(visit 0-6)} for each patient in the RP and NRP groups was compared using the t-test after logarithmic transformation. Univariate logistic regression was used to test associations between clinical variables and RP. Statistically significant variables from the univariate analyses were used in the multivariate logistic regression analysis. The optimal cut-off value of FeNO_{(visit 0-6)} as a predictor of symptomatic RP was obtained as the value that maximized the sum of sensitivity and specificity (i.e., Youden’s index) and visualized by the area under the receiver operating curve (AUC). Cross-validation of the sensitivity, specificity and positive and negative predictive values with leave-one-out was performed to test the generalizability of Youden’s index as a predictor for symptomatic RP. Tests with a two-sided p < 0.05 were deemed statistically significant. All computations were performed using the statistical program Stata, version 14.2.