Background
Frailty is more prevalent among patients with CKD than among individuals with normal kidney function [
1‐
4], and frailty is also highly prevalent among non-elderly patients with ESRD [
3,
4]. The association between frailty and adverse clinical outcomes is established in the ESRD population. Most studies of frailty and CKD have been among patients with ESRD, in community cohorts not specifically enriched for CKD [
2,
5], or in CKD cohorts that employ estimates of kidney function limiting the ability to ascertain the strength of association between directly measured kidney function and frailty, independent of comorbidity and other factors.
The use of serum creatinine to estimate kidney function could also complicate the problem of assessing the association of kidney function with frailty. Because muscle wasting associated with frailty, there is a potential for bias using creatinine based estimates of glomerular filtration rate (eGFR). To our knowledge there have been no evaluations of the relationship of frailty with kidney disease using direct measurement of GFR. The Modification of Diet in Renal Disease (MDRD) study affords a unique opportunity to study frailty by self-report among a cohort of healthier patients who have stage 3 to 5 CKD not on dialysis. Within this cohort there is a wide range of directly-measured GFR to assess the association of kidney function with frailty. The purpose of this study was twofold; first to examine the association of kidney function with self-reported frailty, using a direct measure of kidney function and estimated GFR as a comparison, and second, to determine whether self-reported frailty was associated with death in this cohort with few comorbidities.
Discussion
We found a strong association between mGFR and self-reported frailty, such that patients with better kidney function were less likely to be frail by self-report even in healthy clinical trial study participants with relatively advanced CKD. Although similar when kidney function was modelled with mGFR or eGFR, the point estimate was strongest with mGFR. Regardless of which GFR variable was used as a covariate, the hazard ratios for the association of self-reported frailty with death were similar.
Our examination of associations using mGFR, eGFRcr and eGFRcys was novel. A common limitation of previous studies of frailty in the CKD population was the use of creatinine-based measures of kidney function [
2,
5]. Because muscle mass is associated with serum creatinine and inversely with frailty, use of creatinine-based measures of kidney function may produce inaccurate associations. However, in this study we found similar associations using mGFR and eGFR. Our results suggest that the relationship between self-reported frailty and kidney function may not be particularly sensitive to method of measurement or estimation of GFR. Informally comparing the association of self-reported frailty with all three renal function measures suggests there was no difference between the three measures of kidney function. It is possible, however, that eGFRcr might not perform as well in a more heterogeneous group of patients with CKD as might be encountered in a community-based cohort [
13] or in clinical practice and that eGFRcys may have non-GFR-related variability not observed in these analyses [
14].
Direct comparisons of the prevalence of frailty across CKD populations are hampered by different characteristics of the cohorts (e.g., age range) and different definitions of frailty. Specifically, the prevalence of frailty is higher using definitions based on self-reported physical function compared to frailty defined by direct measures of physical performance [
15]. Nevertheless, it is interesting that our estimate of frailty based on self-report is similar to that reported in the Seattle Kidney Study (SKS), which used an adapted fried criteria [
1,
10]. We found a prevalence of frailty among MDRD participants of 16 % using a self-reported physical function based definition, whereas the frailty prevalence in the Seattle Kidney Study (SKS) was 14 % (95 % CI 10.5–18.2 %) based on direct measures of physical performance in a population with a larger comorbidity burden. Surprisingly, although MDRD clinical trial participants were healthier and younger than individuals in typical general elderly and ESRD cohorts, the prevalence of self-reported frailty among MDRD participants was double that of community elders [
5,
10] and similar to that of other CKD cohorts with higher comorbidity burden [
2]. As anticipated, the prevalence of self-reported frailty among MDRD participants was less than among ESRD cohorts, in which estimates range from 42 % to 73 % [
3,
4,
15,
16].
The association of the severity of CKD with frailty has been previously evaluated using estimates of renal function. Similar to our findings with GFR modeled continuously, these studies have shown patients with more severe renal disease to have a higher likelihood of frailty [
1,
13,
17]. To our knowledge, ours is the first study to test association of GFR with frailty based on three different methods for assessing renal function. Our study suggests associations of eGFRcys with frailty may have a similar magnitude of association using directly measured renal function. Thus studies that have used cystatin C as a measure of renal function may have provided reasonably calibrated findings to estimates using direct renal function measures [
1,
13].
The expectation that frailty would be associated with a higher risk of death was based on a conceptual framework in which frailty is a marker of accelerated loss of functional reserve above and beyond what occurs as a result of kidney disease. Although other cohort studies have evaluated the association of frailty with risk of death among patients with CKD [
1,
4,
16], the long period of follow-up in our study was unique, and our results show that frailty even by self-report is predictive of adverse outcomes even over longer term follow up. Self-reported frailty was associated with higher mortality independent of kidney function, raising the possibility that frailty operates through mechanisms not related to CKD such as inflammation, oxidative stress, or endothelial dysfunction related to CKD or its sequelae [
18‐
21]. Further research into the role of these processes will be important to understand the pathophysiology of frailty in the CKD population.
A number of limitations of our study should be acknowledged. Our definition of frailty was based on patients’ self-reported physical function rather than direct measures of physical performance. We and others have shown that use of self-reported function identifies more individuals as frail, but such definitions have been associated with mortality as in this report [
3‐
5,
8]. We assessed self-reported frailty at baseline and did not update frailty status during the long follow-up period, which may have biased our findings towards less of an association with death. We used underweight based on standard criteria in place of the weight loss criterion in our frailty definition, which likely underestimates the contribution of this component of frailty [
15] MDRD participants were younger and healthier than the general CKD population. Although the relative health of the cohort helped to minimize potential confounding due to burden of comorbid disease, the finding that creatinine-based eGFRcr and mGFR were similarly associated with self-reported frailty must be considered with caution as greater differences might be observed in less healthy or older cohorts. For example, in the CHS, findings were much stronger with cystatin C than creatinine.
Acknowledgements
The data reported here have been supplied by the United States Renal Data System (USRDS). The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as official policy or interpretation of the US government.
Cynthia Delgado MD’s work was supported by Career Development Award # 1IK2CX000527-01A2 from the United States Department of Veteran Affairs, Clinical Science Research and Development Program
Dr. Johansen is supported by 1K24DK085153 from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK).
Drs. Shlipak and Sarnak are supported by R01AG027002 from the National Institute of Aging (NIA).
Competing interests
Authors have no disclosures or conflicts of interest.
Authors’ contributions
Study Design &Data Analysis: CD, KLJ, DG, BG. Manuscript Preparation: CD, KJL. Editing Manuscript: CD, KLJ, MS, MJS. All authors read and approved the final manuscript.