点击显示 收起
1 From the Division of Nephrology, University of California San Francisco (KLJ, BY, and GMC); the San Francisco VA Medical Center, San Francisco (KLJ); and the Division of Nephrology, University of California, Davis (GAK)
2 The authors are responsible for the content of this article. The article does not represent government policy. 3 Supported by contract N01-DK-1-2450 from the National Institutes of Health, National Institute of Diabetes, Digestive and Kidney Diseases. 4 Address reprint requests to GM Chertow, Department of Medicine Research, University of California, San Francisco, UCSF Laurel Heights Suite 430, 3333 California Street, San Francisco, CA 94118-1211. E-mail: chertowg{at}medicine.ucsf.edu. 5 Address correspondence to KL Johansen, Nephrology Section, Box 111J, San Francisco VA Medical Center, 4150 Clement Street, San Francisco, CA 94121. E-mail: johanse{at}itsa.ucsf.edu.
ABSTRACT
Background: Although obesity confers an increased risk of mortality in the general population, observational reports on the dialysis population have suggested that obesity is associated with improved survival. These reports have generally not examined extremely high values of body mass index (BMI; in kg/m2), survival >1 y, or alternative measures of adiposity.
Objective: We sought to clarify the relation between body size and outcomes among a large cohort of patients beginning dialysis.
Design: Data on 418 055 patients beginning dialysis between 1 April 1995 and 1 November 2000 were analyzed by using US Renal Data System data. BMI was divided into 8 categories in increments of 3 units, ranging from <19 to 37, and the relation between survival and BMI was examined by using proportional hazards regression with adjustment for demographic, laboratory, and comorbidity data.
Results: High BMI was associated with increased survival in this cohort, even at extremely high BMI, after adjustment, and over a 2-y average follow-up time. This was true for whites, African Americans, and Hispanics but not for Asians. High BMI was also associated with a reduced risk of hospitalization and a lower rate of mortality in all mortality categories. Alternative estimates of adiposity, including the Benn index and estimated fat mass, yielded similar results, and adjustments for lean body mass did not substantially alter the findings.
Conclusions: High BMI is not associated with increased mortality among patients beginning dialysis. This finding does not appear to be a function of lean body mass and, although modified by certain patient characteristics, it is a robust finding.
Key Words: End-stage renal disease • dialysis • body mass index • adiposity • survival • outcomes
INTRODUCTION
There has been growing concern about the epidemic of obesity in the United States and the potential adverse implications of an "expanding" population (1-3). However, the effect of adiposity on morbidity and mortality with respect to end-stage renal disease (ESRD) is uncertain. Divergent associations have been reported between survival and adiposity in transplant recipients and dialysis patients. While several studies have suggested that outcomes after kidney transplantation are adversely affected by increasing adiposity (4-7), others have shown that dialysis patients with higher body mass index (BMI; in kg/m2) enjoy a survival advantage even in the BMI range that is ordinarily considered harmful (8-14).
It is possible that greater body fat affords a survival advantage to patients on dialysis. However, reports of the association between adiposity and outcomes in dialysis patients have to date focused on BMI, an imperfect measure of adiposity, and have included relatively few categories of BMI (9, 11, 13). These characteristics limit the usefulness of available data. It is unclear whether the improved survival at higher BMI is a result of greater body fatness or of the increased muscle mass that generally accompanies obesity. In addition, the possibility that mild overweight or obesity may be advantageous, whereas greater degrees are associated with an increased hazard of death cannot be excluded. Finally, follow-up periods in such studies have generally been short, typically 1 y (8-10, 12), which makes it possible that high BMI could be a short-term advantage but a longer-term disadvantage.
Given the magnitude of the risks of obesity in the general population, it is important to clarify whether these risks apply to patients on dialysis, who have an overall cardiovascular risk at least 10 times greater. In addition, many other uncertainties about the relation between adiposity and survival among dialysis patients remain. Specifically, the interactions of race or ethnicity and modality with the BMI-survival relation have not been examined carefully. Associations have not been reported between BMI and specific causes of death or the likelihood of hospitalization. The current study was designed to address these uncertainties by using data from an inclusive cohort of patients beginning dialysis between April 1, 1995, and November 30, 2000, and followed for 1 y. We hypothesized that extremely high BMI would not be associated with increased survival time. In addition, if there were a survival advantage at higher BMI, we hypothesized it would be explained in part by the increased lean body mass (LBM) that usually accompanies high BMI.
SUBJECTS AND METHODS
Study population
Data were obtained from the United States Renal Data System (USRDS) Standard Analytic Files. We used data from the Centers for Medicare and Medicaid Services (CMS) Medical Evidence form (Form 2728), the Cause of Death form (Form 2746), and standard files tracking the dialysis modality (ie, hemodialysis or peritoneal dialysis), hospitalization, and mortality (Patients and Hospitalization files) for adult patients beginning dialysis between April 1, 1995, and November 30, 2000. Follow-up extended through November 30, 2001. Cause of death was classified as cardiac, infectious, vascular, gastrointestinal, related to liver disease, other, or unknown. Patients with missing data on height or weight were excluded. In addition, because of suspected height and weight errors, patients with height or weight above or below the 1st percentile were excluded from the analysis.
Predictor variables
BMI was divided into the following categories: <19, 19–37 (in increments of 3 units), and 37. These categories were chosen to cover a broad range of BMIs and to conform approximately to World Health Organization classifications (15). Because BMI is an imperfect estimate of adiposity, we examined other estimates of body fatness for comparison. The Benn index (W/Hp) was calculated for males and females separately so that there would be no correlation between the Benn index value and height (16). Categories similar to BMI categories were created for analysis. Categories of estimated fat mass were also created using demographic information. LBM was estimated by using an ESRD-specific equation for total body water (V; 17):
RESULTS
A total of 462 505 adults began dialysis at least once during the study period. Of these subjects, 3443 (0.7%) were excluded because data on height or weight were missing. An additional 41 007 subjects (8.9%) had height values below the 1st percentile or weight values above or below the 1st percentile, and they were also excluded. The final study cohort consisted of 418 055 persons. Characteristics of the study population by BMI category are presented in Table 1. There were significant differences in most patient characteristics by BMI category, owing to the large sample size. Many of the differences in patient characteristics by BMI category were also clinically significant, defined here as a relative difference of 25% of the lowest BMI category or an absolute difference of 10% between the highest and lowest BMI categories. These characteristics include male sex, Asian race, history of myocardial infarction, cardiac dysrhythmia, pericarditis, prior cerebrovascular accident, peripheral vascular disease, chronic obstructive pulmonary disease, tobacco use, cancer, alcohol dependence, and drug dependence, all of which decreased across categories, and diabetes, hypertension, and African American race, which increased across BMI categories. It is notable that differences in variables potentially related to nutritional status, including serum creatinine and albumin concentrations, did not vary greatly across groups, although differences across categories were statistically significant. Only those in the lowest BMI category (<19) had lower concentrations of both serum creatinine and albumin than did the other groups.
View this table:
TABLE 1. Characteristics of study population by BMI category1
Median follow-up time was 2.0 y. The hazard ratios for men and women for all 3 methods of characterizing adiposity—BMI, the Benn index, and estimated body fat mass—are shown in Figure 1. As expected, low BMI was associated with increased mortality. For the group as a whole, higher BMI was associated with increased survival time, and this positive association extended even to extreme obesity and persisted after adjustment for all of the predictor variables. The Benn index for men was weight/height1.939 and that for women was weight/height1.818, so that there was almost no residual correlation between the index and height (r2 = 0.0001 for men and women). Analyses using the Benn index or estimated body fat mass did not significantly alter the shape of the estimated adiposity-survival relation.
FIGURE 1.. Hazard ratios for death among men () and women () by category of BMI (A), Benn index (B), and estimated fat mass (C). Data were analyzed by using proportional hazards regression. Error bars represent 95% CIs. The interaction between sex and BMI was significant.
Most racial and ethnic groups showed the same pattern of association between BMI and survival as did the whole group, with the notable exception of Asians, who had higher mortality at high BMIs, a finding that was particularly evident in men (Figure 2). The analysis was repeated after excluding Pacific Islanders from the Asian category, which resulted in a considerably more variable relation between BMI and survival in the higher BMI categories.
FIGURE 2.. Hazard ratios for death among men () and women () of different racial and ethnic groups, including whites (A), African Americans (B), all Asians (C), non-Pacific Islander Asians (D), and Hispanics (E). Data were analyzed by using proportional hazards regression. Error bars represent 95% CIs. Interaction terms between all racial and ethnic groups and BMI were significant.
Analysis of the entire cohort by using BMI as a continuous variable found a significant positive association between BMI and survival: an 8% increase in survival for every unit increase in BMI. The squared BMI term was statistically significant and had a positive coefficient, which suggests a U-shaped curve (graphically displayed in Figure 1). However, the upturn was small, and the survival advantage at extremely high BMIs persisted. The interaction term for African American race was significant (P < 0.0001), and there was a stronger association between BMI and survival in African Americans than in the other racial or ethnic groups (see Figure 2). As expected, there was a significant (P < 0.0001) interaction between Asian race and BMI. The association between BMI and survival was somewhat attenuated in women (P < 0.0001). There was no significant interaction between diabetes and BMI or serum creatinine and BMI. Exclusion of diabetes from the model did not materially alter the relation between BMI and survival.
As for overall mortality, the hazard ratios for cardiac death, vascular death, and other-cause death were all significantly lower in the high BMI groups (data not shown). However, the risk of infectious death was not significantly lower in women with extremely high BMI than it was in the groups with normal BMI. There were not enough deaths in the liver or gastrointestinal death categories to allow discernment of a clear association between BMI and mortality due to these causes. The unadjusted annual mortality rate from each cause by BMI category is shown in Table 2. It is evident that, even without adjustment for other risk factors, the risk of death declined with higher BMI. Analysis of hospitalization yielded reductions in risk in the higher BMI categories that paralleled the reductions in mortality observed and shown in the figures (data not shown).
View this table:
TABLE 2. Unadjusted annual mortality rate for different causes of death by BMI category1
The effect of dialysis modality and prior transplantation on the association between BMI and survival is shown in Figure 3. Persons beginning dialysis treatment for the second time, after a failed transplant, had a lower rate of mortality than did those beginning dialysis treatment for the first time. However, the association between BMI and survival in these 2 populations was different. Patients at both extremes of BMI had a higher risk of death than did those with BMI between 22 and 37. The association between high BMI and increased survival was also less evident among patients with new-onset ESRD who were beginning peritoneal dialysis.
FIGURE 3.. Effects of modality and prior transplantation on the relation between BMI and survival. A: Hazard ratios for death among patients beginning hemodialysis () or peritoneal dialysis (). B: Hazard ratios for death among patients who had not previously received a transplant () and patients returning to dialysis after failed transplant (). Data were analyzed by using proportional hazards regression. Error bars represent 95% CIs. The interactions between modality and BMI and between prior transplantation and BMI were significant, P < 0.0001.
To determine whether the association between high BMI and survival was due to higher muscle mass in persons with higher BMI, analyses were performed with and without adjustment for serum creatinine (Figure 4). Because the serum creatinine could be influenced by level of renal function in this group of patients beginning dialysis, a further analysis was performed after adjustment for 24-h creatinine production adjusted for body weight (creatinine index) in patients for whom creatinine clearance and serum creatinine data were available (n = 123 660). Separate analyses of the data from the entire cohort with and without adjustment for estimated LBM were also performed. Adjustment for serum creatinine had little effect on the association between BMI and survival, whereas adjustment for creatinine index tended to strengthen the association (Figure 4). Patients with creatinine clearance data were older and more likely to be white, female, and insured. Patients with creatinine clearance data were also more likely to have comorbidities than were patients with no recorded creatinine clearance (data not shown). Adjustment for estimated LBM attenuated the association between BMI and survival somewhat, but survival was still higher among patients in the higher BMI categories.
FIGURE 4.. Effects of adjustment for serum creatinine and creatinine index (A) and for estimated lean body mass (B) on the relation between BMI and survival. A: Data are shown for the subset (n = 123 660) of patients for whom creatinine clearance data were available. Results without any adjustment related to serum creatinine,
DISCUSSION
The results of these analyses confirm that, for the US ESRD population beginning dialysis between April 1, 1995, and November 30, 2000, higher adiposity was associated with increased survival, even after adjustment for demographics, laboratory values, comorbidities, dialysis modality, employment and insurance status, and even when adiposity was assessed by different methods (BMI, the Benn index, estimate of body fat mass). These results extend previous reports in several ways. It is clear that even extreme degrees of obesity were associated with improved survival in most racial and ethnic groups. Furthermore, this pattern was observed even for cardiovascular death, the most common cause of death in this population and a cause that might be expected to be more frequent in obese patients. The relation between BMI and survival differed between Asians and non-Asians, a point previously highlighted by Wong et al (18). Similarly, patients who had previously received kidney transplants had a U-shaped relation between BMI and survival, whereas patients who started dialysis without prior transplantation did not.
An observational study design of this sort is not optimal for making causal inferences, but some interpretation of the findings is possible. Specifically, these analyses provide little evidence to support our hypothesis that increased survival in obese persons is related to greater muscle mass. Table 1 shows little association between serum creatinine concentrations and BMI, and the relation between BMI and survival was not abrogated by adjustment for serum creatinine, creatinine index, or estimated LBM (Figure 4). Likewise, the lack of strong association between BMI and albumin and the persistence of an association between BMI and survival after adjustment for albumin suggest that inflammation is not likely to be an important factor in establishing the association between BMI and mortality. Rather, it seems possible that the true association is with body fatness and that additional fat is actually protective. In general, end-stage renal disease represents a catabolic and inflammatory state (19-21). Patients undergoing dialysis appear to waste over time (22), and it may be advantageous for a person to begin dialysis treatment at a higher level of adiposity. Patients with greater body fat mass at dialysis initiation may be protected by greater energy reserves in the face of general wasting or may be less prone to wasting processes than are patients with less body fat. The Kidney Disease Outcomes Quality Initiative guidelines have recognized the potential negative effect of predialysis weight loss, and they currently recommend that dialysis be initiated in patients with a low glomerular filtration rate and significant weight loss associated with uremia, even in the absence of other indications (23). Our data tend to support this recommendation even for persons who are above their ideal body weight. Our data do not address the issue of the effect of weight loss or weight gain on survival in the dialysis population, but they do suggest the need for caution and further study before making a strong recommendation of weight loss for those patients awaiting kidney transplantation. Observational studies have suggested that short-term and long-term graft survival rates after kidney transplantation are marginally better in patients with lower BMIs (4, 6, 7). These studies have formed the basis for widespread recommendations that obese patients hoping to be considered for transplantation should lose weight. However, the association between obesity and poorer outcomes after transplantation is not a universal finding (24-26). Whereas peritransplant morbidities and complications associated with obesity are relevant, we do not know how these morbidities might be balanced by the effects of losing weight while awaiting transplant. With an ever-stronger body of literature now supporting the benefit of increased body mass in patients on dialysis and with ever-longer transplant waiting times, the recommendation for weight loss must be considered more carefully.
Creatinine clearance data were missing for most of these patients and were not used in the main analysis (although the main analyses were adjusted for serum creatinine concentration). Patients with creatinine clearance data had significantly different characteristics than did patients for whom this information was not available, and the former group may have had better access to medical care during moderate and advanced stages of chronic kidney disease. However, the relation between BMI and survival in patients with creatinine production data did not differ significantly from that in the whole population, as can be seen by comparing the results for the whole population after adjustmentfor serum creatinine (Figure 2) with the results for the subset with creatinine clearance data (Figure 4). Beddhu et al (14) recently made an important contribution to the literature in this area by using creatinine production as a surrogate for muscle mass and by comparing the relation between BMI 25 and survival among patients in the lowest quartile of creatinine production with that among patients with higher BMIs and higher levels of creatinine production. However, our conclusions do not support those of Beddhu et al, who stated that the survival advantage conferred by high BMI in dialysis patients is limited to patients with normal or high muscle mass. Using many more categories of BMI, we show here that, overall, high BMI is associated with a survival advantage after adjustment for creatinine production or LBM. The reasons for the discrepancy are not certain but increased power in our analysis and differences in data handling may account for the differences. We eliminated patients below the 1st percentile for height in the general population (7.4% compared with an expected 1%) because they were disproportionately represented in the high BMI categories, and in many cases, an error in recording or transcription was likely. However, even with these patients included in the analysis, higher BMI was associated with improved survival (data not shown).
In agreement with Wong et al (18), we noted that Asian patients did not enjoy a survival advantage at higher BMIs. Rather, Asian patients showed a trend toward higher mortality in the higher BMI categories. This trend was not as evident when Pacific Islanders were excluded from the Asian category, which suggests that this subgroup accounted in part for the higher mortality at higher BMI. However, even without the Pacific Islanders included, Asians clearly did not have a positive association between BMI and survival, as did the other racial and ethnic groups. Further study of more specific subgroups of Asian patients could help to clarify this issue. The lessening of the apparent benefit of adiposity in patients who start peritoneal dialysis is consistent with the findings of Snyder et al (27), who observed an improved survival among peritoneal dialysis patients in the first year and a progressive flattening of this relation with time.
The observational nature of this study is a limitation that prevents us from determining whether the improved survival at high BMI is a result of the extra body fat or of some other factor associated with high BMI. Analyses to test alternative interpretations were restricted by the quality of data collected on Form 2728. Thus, adiposity was estimated on the basis of height and weight only. However, an advantage to the approach of this study is that multiple methods of estimation of adiposity were used, and they yielded consistent results across a wide spectrum of categories. The limited data set also precluded examination of the extent to which higher BMI was associated with more sensitive markers of inflammation or nutritional status. However, the large sample size and relative completeness of the available data are strengths of the study. The large sample size allowed us to examine the BMI-survival relation in specific racial and ethnic groups and in other subsets of the population for whom little information was previously available. Although the clinical and demographic data were missing in relatively few cases, it is known that the CMS Form 2728 is insensitive to capture of some comorbid conditions (28) and that it provides no information about the severity of the conditions. Thus, our analyses could not perfectly adjust for the effects of comorbid conditions. However, analysis without any adjustment for comorbidity (Table 2) showed that there were fewer absolute deaths in the high BMI categories. Furthermore, a similar pattern was observed for the 4 major categories of the cause of death in this population.
This study shows that even extremely high BMI is associated with increased survival for most patients beginning dialysis. The relation between BMI and mortality is nonlinear and complex in that there are significant differences in the degree of association between high BMI and survival among different subpopulations, such as racial groups, the sexes, transplant recipients, and patients undergoing different dialysis modalities. At present, our findings suggest that patients who begin dialysis with high BMI should not be advised to lose weight to improve survival. However, there may be other important reasons to encourage weight loss, such as improved quality of life, and the associations between BMI and quality of life in this population remain to be studied.
ACKNOWLEDGMENTS
KLJ contributed to the design of the study, analysis of data, and writing of the manuscript. BY contributed to the analysis of the data and reviewed and approved the final manuscript. GAK contributed to the design of the study and provided advice and consultation on the writing of the manuscript as well as final review and approval. GMC contributed to the design of the study, data analysis, and writing of the manuscript. None of the authors had personal or financial conflicts of interest.
REFERENCES