Extremes of body mass index and mortality among Asian peritoneal dialysis patients.

Abstract

Weight is an important determinant of health in the general population, as well as among patients with end-stage renal disease (ESRD) on dialysis. The literature to date suggests a discrepancy in the impact of weight on mortality among hemodialysis (HD) vs peritoneal dialysis (PD) patients. Several studies in HD patients suggest that a higher body mass index (BMI) confers a survival advantage—the so-called obesity paradox, and a prime example of the phenomenon of reverse epidemiology (1-6). In contrast, the literature among PD patients is inconsistent, with various studies reporting an increased risk of death (7-9), no difference (10-15), or a decreased risk of death among obese patients (5,16). At the other end of the spectrum, among PD patients with low BMI, there are more consistent data showing adverse outcomes (5,13,14,17-20). In this issue of Peritoneal Dialysis International, three single-center observational studies are reported which add to our current understanding of the relationship between BMI and survival among Asian PD patients. In a study from Hong Kong by Kiran et al., based on 274 incident PD patients followed for a mean of 44 months, there was a higher mortality among both underweight and obese patients after multivariable adjustment (21). For the latter group, there was an interaction with both diabetes and cardiovascular disease (CVD), such that there was an increased risk of death among obese diabetics but not among obese non-diabetics, and among obese patients with CVD but not among those without CVD. There was no association between BMI and peritonitis. In another study from India by Prasad et al., based on 328 incident PD patients followed for a mean of 20 months, after adjustment for confounders, there was a doubling of mortality among underweight PD patients (22). The higher mortality was entirely driven by the subgroup of underweight diabetic patients. There was no difference in mortality among obese PD patients despite an increased risk of peritonitis in those who were obese. In a third study from Korea by Kim et al., the influence of BMI on mortality is reported among 900 prevalent PD patients followed for a median of 24 months (23). Similar to the study from India, after multivariable adjustment, there was an increased risk of death among underweight patients, but not among obese patients. Of note, all 3 studies adjusted for a limited number of important potential confounders, including age, diabetes, and some assessment of co-morbidity burden, with 2 of the 3 studies also adjusting for residual renal function. Each of the studies used BMI at baseline, and did not account for changes in BMI over time. How were the extremes of BMI defined? In the first 2 studies described above, a modified BMI classification for Asian patients was used, such that ‘underweight’ was defined as a BMI 25.4 kg/m2. The reason for using a modified BMI classification among Asian patients as compared with the standard World Health Organization (WHO) BMI classification is threefold. First, the BMI distribution among Asians is shifted to the left relative to Caucasians, with lower average BMIs (24). Second, Asian patients tend to have a different body composition as compared with Caucasian patients, namely a higher percent body fat for any given BMI (24,25). Finally, the risk of diabetes and CVD are higher among some Asian populations than Caucasian patients at BMI levels that would be considered normal by standard WHO definitions (26-28). The most consistent finding across all 3 studies was an increased risk of death among underweight PD patients. This finding is in line with several prior studies from North America, Europe and Asia (13,14,17-20). One commonly cited example is the CANUSA study, which prospectively followed 680 incident PD patients for 2 years, and identified poor nutritional status as measured by subjective global assessment as a predictor of death (19). A subsequent study by Chung et al. attempted to dissociate the influence of comorbid disease on mortality in PD patients from that of malnutrition (18). Compared to patients with normal nutritional status and no comorbid disease, a 9-fold increased risk of death was seen among patients with both malnutrition and comorbid disease. Determination of the mortality risk among patients with malnutrition without comorbid disease was limited by a small sample size, as a large majority of the malnourished patients (78%) had comorbidities. The increased risk of death among malnourished PD patients has also been shown among underweight HD patients (29-31). The basis for the consistently higher mortality among underweight dialysis patients cannot be definitively elucidated from these observational studies but, as alluded to above, it is likely that low BMI is a surrogate for malnutrition, inflammation, and comorbidity, all of which can contribute to adverse outcomes. While the 3 studies in this issue of PDI adjusted for some potential confounding variables (including CVD and/or degree of comorbidity), it is unlikely that any observational study would be able to completely adjust for all of the potential confounding features of underweight patients. Interestingly, it was recently shown that weight gain > 1% from baseline among underweight HD patients was associated with improved survival (32), suggesting that the increased mortality risk among underweight dialysis patients may be modifiable if the underlying cause of the malnutrition is addressed. In contrast to the consistent finding of an increased risk of death among underweight Asian PD patients in the 3 studies in this issue, only the study from Hong Kong (21) reported an increased risk of death among obese patients, and this finding was only seen among obese diabetics and obese patients with a history of CVD. Part of the inconsistency of the association between obesity and death between studies may be due to the wide variability in patient characteristics from one study to the next. For example, in an ANZDATA study of 9,679 PD patients which described a higher mortality risk among obese patients, 17% of the patients had a BMI ≥ 30 kg/m2, and 5% had a BMI ≥ 35 kg/m2, with high proportions of diabetes (56%) and CVD (45%) in the obese subgroup (7). In contrast, some of the other studies that did not support an association between obesity and mortality included patients with less extreme BMIs and/or with fewer associated comorbidities (13,14,16). In addition to variable patient populations between studies, the literature to date includes studies with variable durations of follow-up, and it is plausible that an increased risk of mortality among obese patients is more likely to be identified in studies with a longer follow-up time. For example, in the ANZDATA study that reported a higher risk of death among obese PD patients, the Kaplan-Meier survival curves only started to separate between 2 and 3 years of follow-up (7). With this in mind, it is worth noting that the study from Hong Kong that identified an increased risk of death among obese PD patients had a mean follow-up time of nearly 4 years, as opposed to the other 2 studies in this issue of PDI, in which the mean or median follow-up was for 2 years or less. A further important consideration in explaining the inconsistent results with regard to obesity and mortality risk among PD patients is the inability of BMI to distinguish between fat mass and muscle mass. This raises the possibility that 2 subpopulations of patients make up the high BMI category. The influence of body composition on outcomes among PD patients has previously been shown in a study by Ramkumar et al., in which PD patients with high BMI due to a higher muscle mass had a clear survival advantage over those with a high BMI due to increased fat mass (33). The relative absence of the confounding issue of body composition among underweight patients may render this an easier subgroup in which to identify a more consistent degree of risk across studies. So what have we learned? The 3 studies in this issue describing the influence of BMI on outcomes among PD patients are all concordant with the existing body of evidence showing a higher risk of death among underweight dialysis patients. Regardless of the basis for the link between low BMI and increased mortality, it is clear that this represents a group of patients in whom particular attention is required to try to identify potentially treatable factors contributing to their underweight status. With regard to obesity and mortality, that an increased risk of mortality with obesity was only seen in one of the 3 studies is in line with the inconsistent association of high BMI and mortality in the literature to date. The variable relationship between obesity and outcomes suggests a more complex relationship that is likely very much dependent on the demographics of the populations being studied. The 3 studies from different parts of Asia in this issue highlight the limited generalizability of high BMI as a modifier of mortality risk from one PD population to another. In fact, BMI may be just one of several variables for which the ethnicity and geography of the patient population may make extrapolation of associations to other patient populations difficult.