Significance statement
What was previously known? Obesity and CKD are both common in the general population. People with CKD and obesity have lower adjusted mortality and a higher risk of kidney failure compared to otherwise similar people without obesity.
What were the most important findings? The excess risk of death at lower eGFR was modestly lower in people with obesity than in otherwise similar people without obesity. In contrast, the excess risk of placement in long-term care associated with lower eGFR was modestly greater among people with obesity than without obesity, and the excess risk of progression to kidney failure among people with more severe albuminuria was attenuated in those who also had obesity.
Why is the new information significant? Since obesity is common in the general population, tools that use eGFR and/or albuminuria to estimate prognosis in people with CKD might benefit from including information on BMI or other proxies for body size.
Background
Obesity is increasing in prevalence worldwide [
1] and is correlated with an increased risk of various health outcomes including death, progression to end-stage renal disease (ESRD), and myocardial infarction [
2,
3]. Lower estimated glomerular filtration rate (eGFR) and more severe albuminuria are also associated with these adverse outcomes [
4], and people with obesity appear to have an increased risk of chronic kidney disease, including albuminuria and ESRD [
5‐
8]. However, obesity has been associated with favorable outcomes in a number of chronic diseases, including coronary disease, heart failure and stroke [
9‐
11]. Despite these interrelationships between obesity, eGFR and albuminuria, few large studies examine the joint associations between obesity, markers of kidney function and adverse clinical outcomes. This information is potentially important because current tools for estimating prognosis in people with CKD generally do not include information on the presence or absence of obesity [
12].
We did this large population-based study of more than 1 million people treated in a universal healthcare system to examine the joint associations between obesity, eGFR and albuminuria on four clinical outcomes (death, ESRD, myocardial infarction [MI], and placement in a long-term care facility). Our primary goal was to determine whether the excess risk associated with low eGFR and albuminuria was increased or reduced by the simultaneous presence of obesity.
Discussion
In this observational study of more than 1 million people, we examined the joint associations between markers of kidney disease (eGFR, albuminuria) and obesity with a range of clinical outcomes including death, ESRD, MI, and placement in a long-term care facility. In all analyses, obesity was defined by BMI ≥35 kg/m2 as defined by a fee modifier applied to an eligible procedure. There was a significant interaction between obesity and eGFR on the odds of all-cause mortality – after adjustment for potential confounders, the excess risk of death at lower eGFR was modestly lower in people with obesity than in otherwise comparable people without obesity. A similar interaction between obesity and albuminuria was seen on the odds of ESRD -- the excess risk of progression to kidney failure among people with more severe albuminuria was slightly attenuated in those who also had obesity. In contrast, the excess risk of placement in long-term care associated with lower eGFR was greater among people with obesity than without obesity. Finally, although the risk of myocardial infarction appeared to be modestly greater at lower eGFR for both people with and without obesity, there was no evidence that obesity (as compared to the absence of obesity) was associated with an excess risk of myocardial infarction among people with eGFR < 60 mL/min*1.73m2.
Although others have reported that people with CKD and obesity have lower adjusted mortality compared to otherwise similar people without obesity [
25,
26], we are not aware of prior studies suggesting that obesity may attenuate the higher adjusted risk of death associated with CKD. The latter finding is consistent with the hypothesis that obesity acts as a potential buffer against the risk of death associated with acute illnesses, which in turn may be more common at lower eGFR or with more severe albuminuria. Other potential explanations for the apparent protective effects of obesity in CKD have been described elsewhere [
3], and include neutralization of uremic toxins or harmful cytokines by adipose tissue, enhanced antioxidant status, the inclusion of chronically ill (underweight) people in the comparator group (classified as not having obesity), and a more favorable hemodynamic response to stressful stimuli.
Similarly, although faster kidney function loss among people with higher vs lower BMI has been previously reported [
7,
8,
27], we are not aware of other work demonstrating that obesity attenuates the excess risk of kidney failure associated with more severe albuminuria. Results were consistent in a sensitivity analysis that used a more objective definition of kidney failure (initiation of dialysis or sustained eGFR < 10 mL/min*1.73m
2), reducing the possibility that this finding was due to bias resulting from differential initiation of dialysis among people with vs without obesity. Using a model that explicitly accounted for competing risks (Fine and Gray regression) did not change results, suggesting that differential mortality between groups does not explain our findings.
We found that obesity also accentuates the known association between lower eGFR and the likelihood of placement in a long-term care facility [
21], perhaps because obesity may further impede activities of daily living among people who already have reduced functional status due to chronic illness. The lack of an association between obesity and the adjusted risk of myocardial infarction among people with reduced eGFR (in contrast to the significant association observed among those with eGFR > 60 mL/min*1.73m
2) may be attributable to confounding by inflammation among the subset of non-obese people with evidence of wasting and/or malnutrition, which have been associated with a higher risk of vascular disease and other adverse outcomes [
28]. However, since we did not have data on inflammation, wasting or malnutrition, this suggestion is speculative.
Previous studies have consistently shown that obesity modifies the association between eGFR and the risk of death but generally did not examine how obesity interacts with albuminuria to influence the likelihood of death or other clinical outcomes. For example, prior reports indicate that obesity is associated with excess mortality in people with apparently normal kidney function, and in those with mild to moderate reductions in eGFR. Lu et al’s study of 453,946 US Veterans [
26] demonstrated a significantly higher adjusted risk of death among participants with BMI ≥35 kg/m
2 but only among those with eGFR > 30 mL/min*1.73m
2. Below this level of eGFR, the adjusted risk of death among participants with and without obesity did not significantly differ. Differences between our study and previous work may relate to our use of a cohort drawn from the general population as compared to a subset thereof (e.g., U. S Veterans; Israeli adolescents) [
8,
27], our ability to adjust for both eGFR and albuminuria, the comprehensive panel of comorbidities for which we adjusted, or greater statistical power due to the large sample size of the current cohort.
Existing tools for estimating prognosis in people with CKD incorporate information on eGFR, albuminuria, age, and other clinical characteristics – but generally do not consider obesity. Our findings suggest that the relations between eGFR, albuminuria and adverse clinical outcomes may vary by the presence of obesity, which in turn indicates that there may be value in incorporating information on BMI into these prognostic tools. Future studies should examine the incremental value of including obesity as both a main effect and an interaction term in models linking eGFR and albuminuria to the risk of clinical outcomes.
Like other reports from our group, this study has several important strengths [
21]. However, our study also has certain limitations that should be considered when interpreting results. First, although the algorithms used to classify patients with respect to obesity have been previously used [
18], they have not been externally validated and thus some misclassification is possible. In addition, we used a relatively crude definition of obesity (BMI ≥35 kg/m
2), and therefore some participants with BMI between 30 and 34.9 would have been incorrectly classified as not having obesity. Second, BMI is known to be an insensitive predictor of obesity as assessed by measures of body composition among people with CKD specifically [
29], and so some differential misclassification of kidney function may have occurred at lower levels of eGFR. Third, only Alberta residents who underwent an eligible procedure and had information on eGFR or albuminuria were included in the study. Thus, whether our findings are generalizable to a broader population is uncertain. Fourth, like many prior studies using administrative data, we required only a single measure of eGFR to define CKD, which might have led to misclassification of CKD status as compared to requiring at least two measurements that were > 90 days apart. Finally, although we adjusted for a panel of more than 40 potential confounders (including 32 comorbidities), we did not have data on tobacco use, dietary habits, physical activity, insulin resistance, or family history of chronic disease, and thus residual confounding by these characteristics is possible.