Statistical Analysis
Statistical analyses were performed using STATA Version 15 (Statistical
Software, College Station, TX) and R version 3.4.3
(https://www.R-project.org/). Preoperative and operative
characteristics and postoperative outcomes were summarized. For
continuous variables, comparisons between groups were conducted based on
normality with either a two-sample independent t-test or Wilcoxon rank
sum test. Categorical comparisons were by Chi square or Fisher’s exact
tests. Overall survival analyses were conducted by Kaplan-Meier methods,
with log-rank comparison between surgical treatment groups. Cox
proportional hazard modeling was conducted to identify pre and
intraoperative factors associated with overall mortality. All variables
with a univariable p value <0.2 were considered in creating a
multivariable Cox proportional hazard model of overall mortality. The
independent variable of interest was type of aortic arch repair
(hemiarch versus total arch). Model diagnostics included use of
Schoenfeld residuals to examine the proportional hazard assumption,
martingale residuals to assess linearity, and Cox-Snell residuals to
examine overall model fit. Influential observations were identified
through use of delta beta and likelihood displacement. The proportional
hazard assumption was found to be violated by the “other” category of
surgical indication, therefore the final multivariable model is
stratified by this covariate. A sensitivity analysis excluding key
influential observations, and a separate analysis excluding the
“other” category of surgical indication, confirmed the primary
findings (see supplemental material). Results are presented as hazard
ratio (HR) and 95% confidence intervals. Non-parametric estimates of
the cumulative incidence of aortic reintervention were
calculated9. Statistical analyses of need for aortic
reintervention used competing risks and considered death as a competing
outcome and are presented as sub-hazard ratios (SHR) and 95% confidence
intervals10. Statistics with a two-sided p value less
than 0.05 were considered statistically significant.