Summary
Background and objectives Many studies show that obesity in dialysis patients is not strongly associated with mortality but not whether this modest association is constant over age. This study investigated the extent to which the relation of body mass index (BMI) and mortality differs between younger and older dialysis patients.
Design, setting, participants, & measurements Adult dialysis patients were prospectively followed from their first dialysis treatment for 7 years or until death or transplantation. Patients were stratified by age (<65 or ≥65 years) and baseline BMI (<20, 20–24 [reference], 25–29, and ≥30 kg/m2).
Results The study sample included 984 patients younger than 65 years and 765 patients 65 years or older; cumulative survival proportions at end of follow-up were 50% and 16%. Age-standardized mortality rate was 1.7 times higher in obese younger patients than those with normal BMI, corresponding to an excess rate of 5.2 deaths/100 patient-years. Mortality rates were almost equal between obese older patients and those with normal BMI. Excess rates of younger and older patients with low compared with normal BMI were 8.7 and 1.1 deaths/100 patient-years. After adjustment for age, sex, smoking, comorbidity, and treatment modality, hazard ratios by increasing BMI were 2.00, 1, 0.95, and 1.57 for younger patients and 1.07, 1, 0.88, and 0.91 for older patients, implying that obesity is a 1.7-fold (95% confidence interval, 1.1- to 2.9-fold) stronger risk factor in younger than older patients.
Conclusions In contrast to older dialysis patients, younger patients with low or very high BMI had a substantially elevated risk for death.
Introduction
Obesity has reached global epidemic proportions. Trends in obesity prevalence among dialysis patients mirror trends in the general population. Over the past decade the prevalence of obesity, measured as a body mass index (BMI, calculated as weight in kilograms divided by the square of height in meters) of 30 kg/m2 or greater, has increased among adults with ESRD to 10% in the Netherlands and to 30% in the United States (1,2). Obesity increases the risk for diabetes and hypertension, the primary causes of ESRD in the Netherlands and the United States (2,3). Most likely, obesity is also an independent risk factor for ESRD as a result of glomerular hyperfiltration and activation of the renin-angiotensin system (4). Finally, it has become evident that obesity changes adipose tissue to a pro-inflammatory state that could contribute to focal and segmental glomerulosclerosis (5).
Obesity is an established risk factor for cardiovascular disease and death in the general population (albeit a less important one in the elderly) and kidney transplant recipients (3,6–9). Several studies, but not all, have suggested that the effect of obesity in patients with ESRD undergoing maintenance dialysis is paradoxically in the opposite direction, showing that a high BMI is associated with improved survival (the so-called obesity paradox) (10–13).
Patients with ESRD who require maintenance dialysis have a high annual mortality rate of about 20%; about 40% of this figure reflects cardiovascular disease (10). Identification of potential modifiable cardiovascular risk factors is important for targeted prevention and improved life expectancy. Nevertheless, in the general population the association of such risk factors as hypercholesterolemia, hypertension, smoking, and diabetes with cardiovascular disease and mortality, when measured on a relative scale, decreases with age (14–18). Differences in mortality between younger and elderly obese dialysis patients have not been well studied, prompting us to investigate the extent to which the relation between obesity and mortality differs between younger (<65 years) and older (≥65 years) dialysis patients.
Materials and Methods
Study Design and Population
The Netherlands Cooperative Study on the Adequacy of Dialysis-2 (NECOSAD) is a prospective multicenter cohort study of patients with incident ESRD older than 18 years. Patients were followed starting with their first renal replacement therapy. Enrollment occurred between January 1997 and April 2004 throughout the Netherlands. All patients gave informed consent, and all local medical ethics committees approved the study. We followed patients at 3 months after the start of dialysis until death or censoring, which occurred because of a transfer to a nonparticipating dialysis center, withdrawal from the study, kidney transplantation, or the end of the follow-up period, which was a maximum of 7 years. All deaths during follow-up were immediately ascertained by a nephrologist. Causes of death were classified according to the coding system of the European Renal Association–European Dialysis and Transplantation Association as previously described elsewhere (3).
The cohort comprised 1957 patients who were at least 18 years of age, had not had previous renal replacement therapy, and survived at least 90 days after starting hemodialysis (HD) or peritoneal dialysis (PD). We began follow-up 90 days after the start of dialysis to avoid the problem that fluid overload at the start of dialysis would distort BMI measurements.
BMI
We used BMI measured at baseline as an index of adiposity. Weight was measured in HD patients after a dialysis session. In PD patients the volume of the intraperitoneal fluid, if present, was recorded and subtracted from the measured weight. According to the World Health Organization guidelines, obesity was defined as a BMI of 30 kg/m2 or greater, and overweight was defined as a BMI of 25–29 kg/m2. We defined normal weight as a BMI of 20–24 kg/m2, which is within the normal range according to the WHO, and we considered this a priori as the referent category (7). We defined underweight as a BMI less than 20 kg/m2.
Other Demographic and Clinical Data
Data on age, sex, primary kidney disease, diabetes, history of cardiovascular disease (myocardial infarction, ischemic stroke, or limb amputation due to peripheral arterial disease), history of malignancy, and chronic lung disease were collected at the start of dialysis treatment. Primary kidney disease was classified according to the codes of the European Renal Association–European Dialysis and Transplant Association (19). We grouped patients into four classes of primary kidney disease: GN, diabetes mellitus, renal vascular disease, and other kidney diseases. Diabetes was defined on the basis of diabetes mellitus registered as the primary kidney disease or as a comorbid condition. Patients were classified as current smokers of cigarettes (including those who had quit smoking within the past 6 months), former smokers, or never-smokers. Serum albumin, creatinine, and urea were determined from the blood samples. Urea and creatinine levels were also measured in the urine samples. Renal function, expressed as GFR, was calculated as the mean of creatinine and urea clearance corrected for body surface area (ml/min per 1.73 m2).
Statistical Analyses
Variables are presented as mean ± SD, median (interquartile range), or number (proportion) where appropriate. We assessed the relation between BMI and the primary endpoint (mortality observed over 7 years of follow-up) using several methods. In all analyses, survival was measured from 90 days after the start of dialysis. We used life tables to calculate cumulative proportions surviving during follow-up. Because obesity shows its detrimental effects after long-term exposure, we studied BMI as a fixed risk factor at baseline (20). “Elderly” is often defined as age 65 years or older; therefore, we chose that boundary point a priori for all our analyses. To compare the association of BMI and mortality between younger (<65 years) and elderly (≥65 years) dialysis patients, patients were divided into eight categories on the basis of their age (<65 or ≥65 years) and baseline BMI (<20, 20–24, 25–29, and ≥30 kg/m2).
First, absolute mortality rates were calculated within each age and BMI category. In addition, age-standardized mortality rates were calculated for each BMI category using 5-year age groups. For younger patients the age standard was the age distribution of all study dialysis patients younger than 65 years; for older patients it was the age distribution of study patients 65 years or older. Within each of the two broad age categories, age distribution differences across the four BMI categories were thus taken into account. We did not standardize for sex because we found that sex was not an important confounder. The age-standardized mortality rates can be compared only within each of the two broad age strata (younger or older).
Second, we conducted a proportional hazards regression analysis, obtaining hazard ratios (HRs) estimating the effect of various BMI and age categories. We performed the analyses within two age groups (<65 or ≥65 years) using a BMI of 20–24 kg/m2 in younger patients as the reference. Analyses were adjusted for potential confounders: sex, smoking, comorbidity (history of cardiovascular disease, chronic lung disease, and malignancy) and treatment modality (HD or PD). To avoid bias toward the null, any factor postulated to be in the causal pathway between elevated BMI and death, such as hypertension, hypercholesterolemia, and insulin resistance, should not be treated as a confounder (7). To remove the effect of age on the association between BMI and mortality, we performed an additional analysis using a BMI of 20–24 kg/m2 within each age group as the referent category and adjusted for all potential confounders, including age within the two broad age categories (using both linear and squared terms for age). Statistical interaction between age and BMI was modeled by including a product term for age and BMI in the proportional hazards model. In a separate analysis we also adjusted for primary kidney disease, including diabetes. GFR was missing for 274 (16%) patients. We used multiple imputation to impute missing values of GFR (21).
Third, we evaluated biologic, as opposed to statistical, interaction between age and obesity by assessing departures from additive effects, as described by Rothman and colleagues (22–24). Basically, this approach examines the extent to which the effect of both risk factors when present jointly departs from the sum of their effects when present singly. We quantified the amount of biologic interaction by assessing departures from additivity using the synergy index (S):
where HR designates hazard ratio and the subscripts represent presence or absence of each of the two risk factors (25). S is the ratio of the joint effect of the two risk factors to the sum of the effects of each factor in the absence of the other. With no biologic interaction, S = 1. S > 1 measures synergy, and S < 1 measures antagonism. Ninety-five percent confidence intervals (CIs) were calculated as described elsewhere (26).
Fourth, to study the continuous relation between BMI and mortality, we modeled BMI by a four-knot restricted cubic spline (27). The knots were chosen at the 5th, 35th, 65th, and 95th percentiles of the BMI distribution, corresponding to 19.0, 22.9, 25.6, and 32.6 kg/m2. To account for age, which is the strongest determinant of mortality and a potential confounder, we used a five-knot restricted cubic spline with knots at the following percentiles: 5th, 28th, 50th, 72nd, and 95th. To investigate the modifying effect of age we added product terms between age (<65 years or ≥65 years) and the BMI spline covariates. In an additional analysis, we created a product term for quartiles of age, effectively allowing the coefficients of BMI to vary freely for each age quartile. In the proportional hazards regression models, the proportionality assumption for each covariate was checked by adding a product term between that covariate and the logarithm of follow-up time. All analyses were done using SPSS, version 19.0 (SPSS, Inc., Chicago, IL).
Results
Weight or height was missing for 302 patients. We included in the main analyses the 1749 patients (89%) for whom there were sufficient data to compute baseline BMI.
The mean ± SD age of the study cohort (n=1749) was 60±15 years, 62% of all patients were men, 62% were undergoing HD, and 92% were white. The mean ± SD BMIs at baseline for younger and elderly patients were 24.8±4.4 and 24.7±3.9 kg/m2, respectively, with an approximately normal distribution. Table 1 presents the baseline characteristics of the younger (<65 years) and older (≥65 years) dialysis patients.
Baseline characteristics of a cohort of incident dialysis patients at 3 months after start of dialysis
We used life-table methods to estimate the cumulative proportions of younger and older patients who had died at the end of the 7-year follow-up period: 50% and 84%, respectively. Approximately 46% of all deaths were due to cardiovascular causes. The median survival times were 6.0 years for younger patients and 2.9 years for older patients. Overall, the crude mortality rates were 8.2 (95% CI, 7.3–9.3) per 100 person-years in younger and 25.2 (95% CI, 23.4–27.2) per 100 patient-years in older dialysis patients.
Figure 1 shows the crude and age-standardized mortality rates per 100 patient-years according to the four BMI categories at baseline for younger and older dialysis patients (Table 2). Age standardization had only a small effect on the rates except for the younger patients with low BMI. As expected, at all levels of BMI, the absolute risk for death was higher in older than in younger dialysis patients. The crude mortality rate was 1.9 times higher in younger obese patients than in younger patients with a normal BMI. The age-standardized mortality rate was 1.7 times higher in younger obese patients than in younger patients with a normal BMI, which corresponded to an excess rate of 5.2 deaths/100 patient-years. In older dialysis patients, the mortality rates were almost equal between obese and patients with a normal BMI. Thus, obesity appears to be a risk factor for mortality among younger but not among older dialysis patients. The crude mortality rate was 1.4 times higher in younger patients with a low BMI than in patients with a normal BMI, and the age-standardized mortality rate was 2.2 times higher in younger patients with a low BMI than in patients with a normal BMI, which corresponds to excess rates of 3.0 and 8.7 deaths/100 patient-years, respectively. In older dialysis patients the mortality rates were almost equal between patients with a low BMI and patients with a normal BMI.
Crude and age-standardized mortality rates by body mass index (BMI) category. Crude and age-standardized mortality rates (per 100 person-years) with 95% confidence intervals in four categories of baseline BMI during 7 years of follow-up of dialysis patients younger than 65 years and older than 65 years. py, patient-years.
Crude and age-standardized mortality rates per body mass index category in 7-year follow-up of patients <65 and ≥65 years of age
Stratification by treatment modality (HD or PD) showed no evidence of confounding by indication. Therefore, we present results for both treatment modalities combined. We conducted a proportional hazards regression analysis, obtaining HRs estimating the associations of various BMI categories for younger and older patients. After checking the proportional hazards assumption we found no sign of violation. After stratification by age (<65 years or ≥65 years) and adjustment for sex, smoking, comorbidity, and treatment modality, we observed a U-shaped relation between BMI and mortality among younger, but not among older, dialysis patients (Table 3). Additional adjustment for serum albumin and residual renal function slightly attenuated the relation between BMI and survival. Additional adjustment for primary kidney disease, including diabetic nephropathy, attenuated the relation between obesity and mortality, as expected. An analysis restricted to patients with diabetes showed a similar trend (data not shown).
Mortality rate ratios during 7-year follow-up from proportional hazards regression for body mass index at baseline for younger (<65 years) and older (≥65 years) dialysis patients
After multivariable adjustment, the synergy index was 0.57 (95% CI, 0.33–0.98) (Table 3). This result indicates an estimated effect of joint exposure that is less than the sum of the separate effects, which corresponds to biologic antagonism. It implies that older age attenuates the association between obesity and mortality risk.
Table 4 shows the association between BMI and mortality after multivariable adjustment, including age, for younger and older dialysis patients. The ratio of the HRs of older and younger obese patients was 0.58 (95% CI, 0.35–0.95), implying that obesity among younger dialysis patients is a 1.7-fold (95% CI, 1.1- to 2.9-fold) stronger risk factor compared with older obese dialysis patients.
Ratio of mortality during 7-year follow-up from proportional hazards regression for body mass index at baseline for younger (<65 years) and older (≥65 years) dialysis patients
Figure 2 shows the relation between BMI and mortality rates within age quartiles expressed by the HR ratio after multivariable adjustment. We found a U-shaped relation between BMI and mortality among patients younger than 62 years of age (second quartile), which became more pronounced among patients younger than 49 years (first quartile), with a nadir around a BMI of 25 kg/m2. The relation was weaker in the two highest age quartiles. Figure 3 depicts the continuous relation between BMI and mortality for younger and older patients, along with the 95% CIs, and clearly shows the different shape for each age category.
Relation between body mass index (BMI) and mortality rates within age quartiles. Hazard ratios for mortality depending on BMI were modeled by separate restricted cubic splines within age quartiles in a Cox regression model. Patients with extreme values of BMI (<18 kg/m2 [2%] and >40 kg/m2 [0.4%]) were excluded from the figure. The reference was a BMI of 25 kg/m2. The model was adjusted for age, sex, product term between age and BMI, smoking, comorbidity (history of cardiovascular disease, chronic lung disease, and malignancy), and treatment modality (hemodialysis or peritoneal dialysis).
Relation between body mass index (BMI) and mortality rates with 95% confidence intervals (dotted curves) for younger (<65 years) and older (≥65 years) dialysis patients. Hazard ratios for mortality depending on BMI were modeled by separate restricted cubic splines for younger (<65 years) and older (≥65 years) dialysis patients in a Cox regression model. Patients with extreme values of BMI (<18 kg/m2 [2%] and >40 kg/m2 [0.4%]) were excluded from the figure. The reference was a BMI of 25 kg/m2. The model was adjusted for age, sex, smoking, comorbidity (history of cardiovascular disease, chronic lung disease, and malignancy), treatment modality (hemodialysis or peritoneal dialysis), and product term between age and BMI.
Discussion
This cohort study shows that among younger (<65 years) incident dialysis patients, obesity (BMI ≥30 kg/m2) compared with normal weight is associated with an almost 2-fold increased mortality rate. Independent of important confounders, patients younger than 65 years at the start of dialysis with a BMI ≥30 kg/m2 have a 70% higher risk for death compared with patients with a normal BMI. The association between obesity and mortality was even more pronounced among patients younger than 50 years. After adjustment for diabetic nephropathy, the relation between obesity and mortality was slightly attenuated but persisted, indicating that obesity exerts its detrimental effects also via mechanisms independent of diabetes. A low BMI was also associated with increased mortality in this age group, but this may in part be due to reverse causation. Finally, among older (≥65 years) dialysis patients we found no association between obesity and mortality.
In the general population, the negative association of obesity with life expectancy is greater at a young age and less pronounced at age 70 years (28,29). We found a similar continuous age-dependent association between obesity and mortality among dialysis patients. A possible explanation for why obesity is not a risk factor for mortality among older dialysis patients might be that they have a high mortality rate and that short-term effects of competitive risk factors, such as underweight and infection, may preempt long-term effects of obesity.
Our finding is consistent with large studies (>1000 participants) among incident dialysis patients with a long-term follow-up (5–10 years) that showed a U-shaped association between BMI and mortality or an increased mortality risk among obese patients (30–32). Our results contrast with those of previous large studies, which demonstrated a neutral or reduced risk for death among obese maintenance dialysis patients with a relative short median follow-up of 1–5 years (10–13,32–39). Leavey et al. stratified maintenance dialysis patients (of whom about 25% were black) into three age groups (<45, 45–64, ≥65 years), showing a lower mortality risk with obesity in all three age groups (37).
Part of the disparity between the findings of these observational studies and our results may be explained by the following: (1) too-short duration of follow-up, (2) survivor bias due to inclusion of maintenance dialysis patients, (3) differences in statistical approach of BMI as a variable (for instance, including low BMI [<20 kg/m2] in the reference category might result in underestimation of the association between obesity and mortality), (4) combining results of both younger and older dialysis patients, (5) misclassification of obesity due to fluid overload by calculating BMI at the time of the initiation of dialysis, (6) inadequate control for confounding (such as smoking), and (7) not taking into account differences of body composition between races. For instance, blacks have a greater muscle mass at any BMI than do persons of other races (39).
Several aspects of our study set it apart from others. Because we included only incident dialysis patients, our study, unlike those that included patients on maintenance dialysis, is not susceptible to survivor bias. Survivor bias is a form of selection bias that occurs when the risk for an outcome is estimated from data collected at a given time point among survivors rather than on data gathered in a group of incident cases (40). As with other biases, an increased study size cannot compensate for survivor bias (41). Second, we have a relatively long follow-up period. Third, we adjusted for chronic disease and smoking. Reverse causation owing to inadequate control for smoking status and preexisting chronic disease can distort the true relation between BMI and risk for death because smoking and chronic illness are associated with both decreased BMI and increased risk for death.
Nevertheless our study has several limitations. We combined patients treated with HD or PD. Differences in outcome between these two patient groups have been suggested, although the determinants of dialysis modality are largely nonmedical. Two large studies among HD and PD patients showed little difference in risk for death during 2–5 years of follow-up (42,43). Second, we used BMI as a proxy for obesity, but BMI only indirectly reflects the risk stemming from the metabolic effects of an increased fat mass (44). Moreover, BMI does not fully reflect some age-related changes, namely that the proportion of body fat increases with age while muscle mass decreases. Third, in the Netherlands wait-listing for kidney transplantation is suspended for patients with a BMI of 30 kg/m2 or greater, resulting in differential selection of healthy patients with a BMI less than 30 kg/m2 for transplantation. This differential selection may have resulted in bias to the extent that we could not completely control for covariates related to the selection for kidney transplantation. Finally, failure to control for unintentional weight loss or cigarette smoking could result in residual confounding (7). Loss of body mass, even in obese dialysis patients, might portend greater mortality (12). We had no information about the number of pack-years smoked by patients, and therefore we cannot rule out residual confounding due to insufficient adjustment for smoking. Such residual confounding would tend to underestimate the effect of obesity.
The clinical implications of this study are as follows: first, that potential kidney recipients should be encouraged to lose weight if obese. Given the suggestion that obesity might be advantageous, some argue for caution in recommending that patients strive for a BMI less than 30 kg/m2 (32). Second, for patients younger than 65 years who are not eligible for kidney transplantation, identification of potential modifiable cardiovascular risk factors, such as obesity, is important to improve life expectancy.
In conclusion, obesity was not materially associated with an increased mortality in older dialysis patients, whereas among younger dialysis patients, having a BMI that was low or very high was associated with an almost twofold increased mortality rate.
Disclosures
None.
Acknowledgments
We thank the investigators and study nurses of the participating dialysis centers and the data managers of The Netherlands Cooperative Study on the Adequacy of Dialysis (NECOSAD) for collection and management of data. The members of the NECOSAD Study Group include A.J. Apperloo, J.A. Bijlsma, M. Boekhout, W.H. Boer, P.J.M. van der Boog, H.R. Büller, M. van Buren, F.Th. de Charro, C.J. Doorenbos, M.A. van den Dorpel, A. van Es, W.J. Fagel, G.W. Feith, C.W.H. de Fijter, L.A.M. Frenken, W. Grave, J.A.C.A. van Geelen, P.G.G. Gerlag, J.P.M.C. Gorgels, R.M. Huisman, K.J. Jager, K. Jie, W.A.H. Koning-Mulder, M.I. Koolen, T.K. Kremer Hovinga, A.T.J. Lavrijssen, A.J. Luik, J. van der Meulen, K.J. Parlevliet, M.H.M. Raasveld, F.M. van der Sande, M.J.M. Schonck, M.M.J. Schuurmans, C.E.H. Siegert, C.A. Stegeman, P. Stevens, J.G.P. Thijssen, R.M. Valentijn, G.H. Vastenburg, C.A. Verburgh, H.H. Vincent, and P.F. Vos. We thank the nursing staff of the participating dialysis centers and the staff of the NECOSAD trial office for their invaluable assistance in the collection and management of data for this study. We are indebted to Dr. Eric Melse for graphic art work.
This work was supported in part by unrestricted grants from the Dutch Kidney Foundation. The funding source was not involved in the collection, interpretation, or the analysis of the data or in the decision to write or submit this report for publication.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
- Received June 11, 2011.
- Accepted November 20, 2011.
- Copyright © 2012 by the American Society of Nephrology
References
In this issue
Jump to section
More in this TOC Section
Original Articles
Epidemiology and Outcomes
Cited By...
- Management of Obesity in Adults with CKD
- The Obesity Paradox and the Role of Inflammation
- Association of Height with Elevated Mortality Risk in ESRD: Variation by Race and Gender
- Effect of Omega-3 Fatty Acids on Kidney Function after Myocardial Infarction: The Alpha Omega Trial
- Body Mass Index and Mortality Risk in Asian Peritoneal Dialysis Patients in Hong Kong--Impact of Diabetes and Cardiovascular Disease Status
- Extremes of Body Mass Index and Mortality among Asian Peritoneal Dialysis Patients
- The Association Between Body Mass Index and Mortality on Peritoneal Dialysis: A Prospective Cohort Study
- Obesity in CKD--What Should Nephrologists Know?
- Influence of Body Mass Index on the Association of Weight Changes with Mortality in Hemodialysis Patients