In Reply to 'Etiological versus prognostic models in cohort studies' and 'nutritional supplement use in hemodialysis patients'

摘要

We thank Dr Imamovic and Drs Bossola and Tazza for their comments on our report, and acknowledge their stated limitations and caveats associated with cohort data. In our analyses using clinical data, we focused on the most reliable data-specifically, laboratory results and demographic data. We incorporated available indicators of nutrition, inflammation, and health status into models, including some of the most important predictors of outcomes in hemodialysis patients; variables in the propensity match, as well as in adjusted models, included serum albumin level, white blood cell count, protein catabolic rate, serum creatinine levels, iron study results, vascular access type, and recent hospitalizations. We used multiple models, serially including possible confounders and mediators to explore whether results were robust. Additionally, we included time-dependent serum albumin values in the models, which showed only a small improvement in albumin levels for patients receiving supplements, suggesting that the survival benefit likely was not explained entirely by change in serum albumin level. A plausible alternative explanation for the benefit seen with protein intake during dialysis is a favorable change in anabolic-catabolic balance that precedes any substantial change in serum albumin levels, demonstrated elegantly over the past decade by Ikizler and colleagues.