Statistical analysis
Descriptions of continuous variables (time between transplant and QoL survey) were expressed as the means and standard deviations. Discrete qualitative variables (age at transplant, QoL scores via PedsQL TM 4.0) were expressed as numbers and percentages. Characteristics of pediatric HSCT patients were grouped by patient race. Fisher’s exact tests were then used to examine potential differences among categorical variables (including sex, type of diagnosis, conditioning regimen, donor type, donor source, CMV status, insurance type, and household income classification) and one way analysis of variance (ANOVA) were used to test for differences in age at transplant and QoL scores. SAS version 9.4 (SAS Institute Inc; Cary, NC) was used for data summary and analysis.
Proc Regression in SAS was used to estimate unadjusted and multivariate adjusted models of the four QoL and overall functioning scores as dependent variables. Independent predictors included patient race, age at the time of transplant, sex, diagnosis (malignant or non-malignant), conditioning regimen, and time since transplant. Models were additionally run assessing the interaction of race and estimated income level (Table 3) and the interaction of race and type of insurance (Table 4).
In multivariate modeling, non-White (including Hispanic, Black, or Native American) patient QoL outcomes were compared to those of non-Hispanic White patients in separate models for each QoL outcome. In these supplementary analyses, multivariate models compared Hispanic, Black, or Native American patient QoL outcomes individually to those of non-Hispanic White patients. For the sets of multivariate analyses in both the manuscript and the supplement, the primary multivariate analysis was adjusted for age, sex, type of disease, and conditioning (Supplemental Table S2) while a secondary analysis further adjusted for insurance type and estimated household income (Supplemental Table S3). An a priori alpha level was set to 0.05.