Dr Zhan also raises the concern that diabetes, alcohol use, and chronic heart failure may lie within the causal pathway of the ABO and ARDS relationship and, therefore, should not be included in adjusted analyses. Of these three covariates, we included only diabetes in the final multivariable models, because ABO blood types have been associated with risk of diabetes in prior studies.5 Removing diabetes from the logistic regression model does not significantly alter any effect estimates of the association of blood type A and ARDS (ARR 1.53 in trauma, 1.39 in sepsis), suggesting that diabetes is neither a confounder nor a causal mediator. This was also true when all three variables were either included or excluded from multivariable models. Therefore, although ABO blood types have been associated with risk of vascular disease and diabetes,5,6 we do not believe the increased risk of ARDS is mediated through these diseases. ARDS develops in the setting of significant environmental insult (eg, trauma or sepsis), resulting in dramatic activation of endothelium and the inflammatory system. Endothelial-derived glycoproteins linked to ABO blood type, including von Willebrand factor, ICAM-1, and the selectins, are released and/or upregulated by critical illness and may provide the mechanistic link between ABO blood type and ARDS.