Methods and Procedures
Subjects and Patients
This is a prospective, observational study performed in two patient cohorts. The first cohort was composed of 370 prevalent patients undergoing maintenance HD [mean age 60 ± 12 years; 162 females; median HD vintage 4.1 years (range: 0.8–19.5 years)]. Information on these patients has been described elsewhere in more detail. Among these 370 patients, 347 had complete data on WC, BMI, lipid profiles and high sensitive C-reactive protein (hs-CRP). The second cohort was composed of 216 prevalent HD patients [mean age: 60 ± 12 years; 103 females; median HD vintage 6.1 years (range: 0.6–25.5 years)]. All 216 patients in this cohort had complete data on WC, BMI, lipid profiles and hs-CRP. The exclusion criteria for entry into the current study in both cohorts were: (1) active infection; (2) recent hospitalization within 3 months; (3) psychotic illness or other communication problems; (4) active malignancy; (5) aged less than 20 years; and (6) receiving HD for less than 3 months. There were 99 patients in both the first and second cohort, and therefore 464 patients (mean age:60 ± 12 years; 235 females) who received prevalent HD at the Far Eastern Memorial Hospital, Taiwan, were enrolled from February 2007 (the first cohort) to October 2011 (the 1st participant enrolled in the second cohort) into the analysis. The study design of the first cohort has been reported in the previously published articles. In the second cohort, all subjects gave written informed consent, and the local ethics committees of the involved hospitals approved the study protocol (Far Eastern Memorial Hospital Research Ehics Review Committee, FEMH-IRB-099090-E; chairman Shih-Hong Huang; Oct. 12, 2010; ClinicalTrials.gov; NCT01457625; Oct. 20, 2011). The authors confirm that all ongoing and related trials for these cohorts are registered.
Measurements of Clinical Parameters, Nutritional and Inflammatory Status
The demographic data and concurrent medical history of CV disease were recorded. WC was measured at the umbilical level over light clothing, using an un-stretched tape meter, without any pressure to the body surface. BMI was calculated as weight (kg) divided by the square of the height (m). WHtR was calculated as WC (cm) divided by height (cm). Venous blood was sampled in the morning after an overnight fast of more than 8 hours before dialysis.
The nutritional status of the participants was calculated using the geriatric nutritional risk index (GNRI). This index is calculated from serum albumin level and body weight as follows: GNRI = [14.89 × albumin (g/dL)] + [41.7 × body weight/WLo], where WLo is the ideal body weight calculated from the Lorentz equation. The GNRI has been validated in dialysis patients, and a higher GNRI score indicates better nutritional status. We used the immuno-nephelometric method with a Tina-quant CRP (Latex) ultra-sensitive assay (D & P Modular Analyzer, Roche Diagnostics GmbH, Mannheim, Germany) to determine high sensitive C-reactive protein (hs-CRP) levels.
Visceral Adiposity Index (VAI)
VAI is a sex-specific index based on WC, BMI, TG, and HDL-C, and estimates the visceral adiposity functionality. The VAI was calculated as follows: (TG and HDL-C were in mmol/l and WC in cm).
Outcomes
The outcomes were a composite of all-cause mortality and CV events, considered jointly or separately. The CV events were defined as the new occurrence of CV events including coronary events (non-fatal myocardial infarct, unstable angina and coronary re-vascularization), hospitalized heart failure, incident hospitalized stroke (either ischemic or hemorrhagic stroke), and incident peripheral arterial occlusion disease requiring surgical intervention. The observation period for outcomes was from February 2007 for the patients in the first cohort, from March 2011 for those in the second cohort, and February 2007 for the 99 patients who were recruited in both cohorts. Follow-up was censored on the date of the first CV event, the end of the study (November 1, 2013), the date of death or undergoing renal transplantation, or at the time the patients were transferred to other dialysis facilities and were no longer followed up, whichever came first. Initially, we constructed plots of the VAIs and hazard ratios (HRs) of the outcomes using the Lowess function. The results revealed their non-linear relationship, suggesting the need for stratification of the patients into tertiles according to their VAI scores for outcome analysis. Therefore, we stratified the patients into tertiles according to their VAI score VAI tertile 1 represented the patients who had VAI values within 0.32–1.41, VAI tertile 2 had VAI values 1.42–3.24, and VAI tertile 3 had VAI values 3.25–31.66. The VAI is a gender-specific index, and there has been reported to be a remarkable difference between genders in survival. In addition, the VAI, WC and WHtR have been shown to interact with nutritional status in predicting mortality. Therefore, we performed pre-specified subgroup analysis (gender and nutritional status) when assessing all-cause mortality.
Statistical Analysis
Continuous data were presented as mean ± SD or median (interquartile range), and categorical data were reported as percentages. Differences in baseline characteristics and biochemical parameters among the patients in tertiles of VAI were compared by ANOVA for continuous variables and the chi-square test for categorical variables. The non-parametric Kruskal-Wallis test was used for non-normally distributed continuous variables.
Outcome analysis was done with using a Cox proportional hazard model, in which the primary predictor variable was either VAI tertile or VAI as continuous variable, and the covariates included age, gender, vintage of HD, presence of DM, hypertension and concurrent CV disease, hemoglobin, calcium phosphate product, hs-CRP, intact parathyroid hormone and nutritional status (GNRI). We also selected WC and WHtR as the primary predictor variables and repeated the outcome analysis. We used the "Enter" method to analyze the hazard ratio of each primary predictor variable in the multivariate Cox regression model, and in order to differentiate the superiority of the predictive power of each primary predictor variable, we also used the "stepwise forward likelihood ratio test" method to analyze the outcomes in the multivariate Cox regression model.
To further evaluate and compare the predictive performance of the VAI, WC and WHtR, we used time-dependent receiver operating characteristic (ROC) curves for censored data, and the area under the ROC curve (AUC) as the criterion. The time-dependent ROC curve estimation was analyzed using open-source statistical software R. All other statistical analyses were performed using SPSS software, version 19.0 (SPSS, Inc., Chicago, IL). A P value of less than 0.05 was considered to be statistically significant.