Statistical analysis
In order to comparatively assess the effect of center variations, the
induction therapy utilization rate was stratified into equal terciles,
which were classified as low (<29%), intermediate (29-66%)
or high (>67%) induction therapy utilization. The data is
described as proportions (%) for categorical variables and as means or
medians for continuous variables, with standard deviation (SD) or
interquartile range (IQR), respectively. Pearson’s
Chi2 test was used to compare categorical variables
and one-way analysis of variance (ANOVA) or students t-tests were used
to compare continuous data. All statistical tests were 2-sided, and
p<0.05 was considered significant. Multilevel mixed-effects
logistic regression models were used to estimate the relationship of
induction therapy and post-OHT clinical outcomes accounting for
inter-center variability. The primary endpoint was drug-treated acute
rejection within the first year following OHT. Model A included only the
individual transplant centers as random effect parameters. Model B added
the recipient and donor risk factors identified as fixed effect
parameters to Model A. Finally, Model C, added induction therapy
variable as a fixed effect parameter to Model B. Changes in the estimate
of the between-centers variance may suggest the distribution of
variables across centers. The Variance Partition Coefficient (VPC) was
estimated to evaluate the proportion of variation attributable to
systematic differences between centers characteristics in the context of
a binary dependent variable. This contribution as a percentage was
calculated as [X /X+3.29] * 100, where X is the variance estimate
from the random effect component of the model (15). Risk-adjusted Cox
proportional hazards models were used to evaluate the adjusted impacts
of induction therapy on mortality. Kaplan-Meier estimates were used to
evaluate the longitudinal impacts of the use of induction therapy on
survival. Pearson’s correlation coefficient (R) was utilized to assess
the correlation between pre-OHT risk scores and the use of induction
therapy. All analyses were performed with version 15 STATA statistical
software (StataCorp LP, College Station, TX).