The following variables were tested and did not make a significant contribution to model fit: age squared, vitamin D intake,
and total fat intake as a % of total caloric intake, determined using
the residual method. There were no more than 4.1% of missing
cases for predictor or any covariate. A sensitivity analysis investigating
the effect of adding several factors in the causal pathway
between chocolate intake and diabetes risk is presented in
“Sensitivity Analyses.”