Statistical Analysis
Descriptive statistics were summarized for the primary and secondary outcomes as well as demographics, therapeutic exposures, hospitalizations and other clinical variables. These summary measures were compared at baseline for the overall study cohort, at various time periods both before and after treatment completion, and for females and males separately. Mean and standard deviation (SD) or median and interquartile range (IQR) were used for continuous outcomes. Binary outcomes were summarized using frequency and percentages.
Given the retrospective nature of the study, lung function testing was performed at variable time points in coordination with treatments or follow up clinic visits. Therefore, data was summarized in monthly intervals, to allow comparability of the measurements across patients and to allow formal longitudinal investigation of the trajectories for the various outcomes considered. Hence, for each child, the mean outcome values across all visits within a given month were used as the monthly value of the outcome.
The visit date for the last recorded cancer treatment (radiation or chemotherapy) was defined as time zero. Data was summarized into intervals before and after treatment completion, with the post-treatment time-period further subdivided into within one year (immediate effect), between one year and four years (short term effect) and after four years (long term effect). These periods were defined after careful consideration of the distribution of the primary and secondary outcomes and the corresponding time trajectories.
Exploratory analyses were done to investigate lung function trajectories over time for all three pulmonary function outcomes (FEV1, TLC and DLCOadj), and to evaluate similarities and differences in lung function trajectories before and after cancer treatment. Specifically, lung function trajectory was stratified by sex to determine if there were any differences between groups. To determine the potential additive pulmonary toxic effects, sensitivity analyses were performed to explore differences in lung function trajectories among a subgroup of children who received both chemotherapy and radiation therapy.
All analysis was performed using the R statistical software package14.