Mortality predictability of body size and muscle mass surrogates in Asian vs white and African American hemodialysis patients

Abstract

Objective: To determine whether the association of body size and muscle mass with survival among patients undergoing long-term hemodialysis (HD) is consistent across race, especially in East Asian vs white and African American patients.

Patients and methods: Using data from 20,818 patients from South Korea who underwent HD from February 1, 2001, to June 30, 2009, and 20,000 matched patients from the United States (10,000 whites and 10,000 African Americans) who underwent HD from July 1, 2001, to June 30, 2006, we compared mortality associations of baseline body mass index (BMI) and serum creatinine level as likely surrogates of obesity and muscle mass across the 3 races.

Results: In Korean HD patients, higher BMI together with higher serum creatinine levels were associated with greater survival, as previously reported from US and European studies. In the matched cohort (10,000 patients from each of the 3 races), mortality risks were lower across higher BMI and serum creatinine levels, and these associations were similar in all 3 races (reference groups: patients with BMI >25.0 kg/m(2) or serum creatinine >12 mg/dL in each race). White, African American, and Korean patients with BMI levels of 18.5 kg/m(2) or less (underweight) had 78%, 79%, and 57% higher mortality risk, respectively, and white, African American, and Korean patients with serum creatinine levels of 6.0 mg/dL or less had 108%, 87%, and 78% higher mortality, respectively.

Conclusion: This study shows that race does not modify the association of higher body size and muscle mass with greater survival in HD patients. Given the consistency of the obesity paradox, which may be related to a mitigated effect of protein-energy wasting on mortality irrespective of racial disparities, nutritional support to improve survival should be tested in HD patients of all races.

Figure 1

Association between baseline body mass index and all–cause mortality in 20,818 Korean hemodialysis patients^{† †}Hazard ratios were estimated with Cox regression using restricted cubic splines. Values of <1 percentile and >99 percentile in body mass index were excluded from the model to minimize an influence from the outlier values. The model was adjusted for case–mix and nutrition covariates including age, gender, diabetes mellitus, dialysis vintage, single–pool Kt/V, hemoglobin, serum albumin and normalized protein nitrogen appearance. Dashed lines are 95% point–wise confidence bands. Number of patients is displayed simultaneously.

Figure 2

Adjusted hazard ratios (95% confidence intervals) for all–cause death with baseline body mass index (A) and serum creatinine (B) categories across 3 different races in the matched cohort (n = 10,000 in each race)^{† †}Patients with BMI >25.0 kg/m² and serum creatinine level >12.0 mg/dL in each race were reference groups. The model was adjusted for case–mix and nutrition covariates including age, gender, diabetes mellitus, dialysis vintage, single–pool Kt/V, hemoglobin, serum albumin and normalized protein nitrogen appearance. Propensity score which was the probability of having the same characteristics as Korean HD patient was calculated by logistic regression including all variables in the survival analysis. The matched cohort was created by 1:1 matching based on propensity score.

Figure 2

Adjusted hazard ratios (95% confidence intervals) for all–cause death with baseline body mass index (A) and serum creatinine (B) categories across 3 different races in the matched cohort (n = 10,000 in each race)^{† †}Patients with BMI >25.0 kg/m² and serum creatinine level >12.0 mg/dL in each race were reference groups. The model was adjusted for case–mix and nutrition covariates including age, gender, diabetes mellitus, dialysis vintage, single–pool Kt/V, hemoglobin, serum albumin and normalized protein nitrogen appearance. Propensity score which was the probability of having the same characteristics as Korean HD patient was calculated by logistic regression including all variables in the survival analysis. The matched cohort was created by 1:1 matching based on propensity score.