Association of Frailty based on self-reported physical function with directly measured kidney function and mortality

Background Use of serum creatinine to estimate GFR may lead to underestimation of the association between self-reported frailty and kidney function. Our objectives were to evaluate the association of measured GFR (mGFR) with self-reported frailty among patients with CKD and to determine whether self-reported frailty was associated with death after adjusting for mGFR. Methods Participants in the Modification of Diet in Renal Disease study (1989–1993) had GFR measured using iothalamate clearance (mGFR), and GFR was estimated based on the CKD-EPI creatinine (eGFRcr) and cystatin C (eGFRcys) equations. We defined self-reported frailty as three or more of: exhaustion, poor physical function, low physical activity, and low body weight. Death was ascertained through 2007 using the National Death Index and the United States Renal Data System. Results Eight hundred twelve MDRD participants (97 %) had complete data on self-reported frailty (16 % prevalence, N = 130) and mGFR (mean (SD) 33.1 ± 11.7 ml/min/1.73 m2). Higher GFR was associated with lower odds of self-reported frailty based on mGFR, (OR 0.71, 95 % CI 0.60–0.86 per 10 ml/min/1.73 m2), eGFRcr (OR 0.80, 95 % CI 0.67–0.94 per 10 ml/min/1.73 m2), and eGFRcys (OR 0.75, 95 % CI 0.62–0.90 per 10 ml/min/1.73 m2). Median follow-up was 17 (IQR 11–18) years, with 371 deaths. Self-reported frailty was associated with a higher risk of death (HR 1.71, 95 % CI 1.26–2.30), which was attenuated to a similar degree when mGFR (HR 1.48, 95 % CI 1.08–2.00), eGFRcr (HR 1.57, 95 % CI 1.15–2.10), or eGFRcys (HR 1.51, 95 % CI 1.10–2.10) was included as an indicator of kidney function. Conclusions We found an inverse association between kidney function and self-reported frailty that was similar for mGFR, eGFR and eGFRcys. In this relatively healthy cohort of clinical trial participants with CKD, using serum creatinine to estimate GFR did not substantially alter the association of GFR with self-reported frailty or of self-reported frailty with death.


Background
Frailty is more prevalent among patients with CKD than among individuals with normal kidney function [1][2][3][4], and frailty is also highly prevalent among nonelderly 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.

Study participants
The MDRD Study was a multicenter cooperative clinical trial designed to determine whether restriction of dietary protein and a low target blood pressure (mean arterial pressure <92 mmHg vs. usual target blood pressure (<107 mmHg)) reduced the rate of progression of CKD, irrespective of the nature of the primary underlying process. Patients with diabetes requiring insulin were excluded [6]. At baseline, the severity of kidney disease was assessed by measurement of GFR using iothalamate. Data on physical function and exercise were collected by questionnaires. Data from ancillary studies of this cohort that evaluated C-reactive protein (CRP) as a predictor of cardiovascular outcomes were included in this analysis [7]. Mortality outcomes were acquired via direct data collection within the MDRD study and through linkage with the National Death Index and the United States Renal Data System (USRDS) through December 12, 2007. Participants from MDRD study A and B were combined for the purpose of this analysis. Only participants with complete measures of physical functioning and linkage information were included in analyses (n = 812; 97 %). Informed consent was obtained from all participants as part of the original study. The Committee on Human Research at the University of California -San Francisco and the Research and Development Committee of the San Francisco Veterans Affairs Medical Center deemed this study exempt.

Self-report frailty definition
Our frailty definition was an adaptation of the Fried Frailty Index that substituted patients' self-report of physical function for the direct measures of physical performance (gait speed and grip strength) that are part of the original definition. This approach is similar to that originally developed by Woods et al. [8] and subsequently applied among patients with endstage renal disease [9] and is (henceforth referred to as "self-reported frailty") (Appendix). For the exhaustion criterion, participants responded to a symptom questionnaire that asked "In the past month how often have you felt lack of pep and energy? Tiring easily, weakness?" Patients were given one point in the exhaustion domain if their response to both questions corresponded to a moderate amount of time or more. The physical function domain was ascertained using the MDRD quality of wellbeing measure, which assessed participants' ability to complete activities of daily living (ADLs) similar to the Rand SF-36 physical function scale. Individuals who scored in the lowest quartile based on normative data were allocated two points in the physical function domain of self-reported frailty. Physical activity was assessed using the MDRD Leisure Time Physical Activity Questionnaire, which measured the number of times per week each individual performed walking or other moderate and vigorous activities during leisure time. After converting physical activity into total kilocalories (kcal) of activity per week, individuals in the lowest quintile based on normative data [10] were allocated 1 point in the physical activity domain of self-reported frailty. The baseline assessment of standard body weight was used as a surrogate for weight loss. Individuals who were less than 95 % of standard body weight for sex and height were allocated 1 point in the weight loss/ underweight domain of self-reported frailty. Selfreported frailty was defined by a score of ≥3 points, and patients scoring 1 or 2 points were considered intermediate frail by self-report from a total of 5 possible points [10].

Measures of kidney function
Assessment of kidney function using clearance of 125 I-iothalamate was completed twice during the baseline period for all participants. We used the average of the 2 baseline 125 I-iothalamate clearance measures in our analyses (averaged measured GFR; mGFR). We modeled GFR as a continuous variable and as categories (≥45 ml/min/1.73 m 2 , 30 to 44 ml/min/ 1.73 m 2 , ≤ 29 ml/min/1.73 m 2 ). To compare the association of the direct measurement of kidney function with self-reported frailty to the association of a creatinine-based estimate (eGFRcr) and cystatin Cbased estimate (eGFRcys) of GFR with self-reported frailty, we used the Chronic Kidney Disease Epidemiology Collaboration [11] (CKD-EPI) formulas [12] using the same modeling strategies. We chose to use the CKD-EPI equation rather than the MDRD equation to evaluate the association between creatinine-based eGFR and self-reported frailty to avoid overestimating the similarity between mGFR and eGFRcr in the derivation cohort for the MDRD equation.

Statistical analysis
Characteristics of participants who were frail by selfreport, intermediate frail by self-report and non-frail were compared by Wilcoxon rank sum for continuous variables and chi square for categorical variables. We treated self-reported frailty as a dichotomous variable in the analyses comparing the different methods of measuring GFR with self-reported frailty, similar to previous studies in CKD and ESRD [2,4]. In our survival analysis, self-reported frailty was treated as a three-level categorical variable (0 = not frail; 1-2 = intermediate frail, and 3 = frail by self-report) as Fried and others have done [2,10,13]. We used logistic regression modeling to estimate the association of mGFR, eGFRcr, and eGFRcys with self-reported frailty. With the exception of albumin and age, covariates included in the model were treated as continuous variables. Covariates included age (≤40 years, 41-59 years, ≥ 60 years), sex, quartiles of serum albumin concentration (≤3.84 mg/dL, 3.85-4.01 mg/dL, 4.02-4.24 mg/dL, ≥4.25 mg/dL), proteinuria, race, BMI, and log transformed CRP, protein intake and blood pressure group assignment. We tested for interactions between age greater than 60 and GFR and for interactions between sex and GFR in the adjusted models. A sensitivity analysis was conducted to determine whether associations were independent of study group assignment for blood pressure and protein. In addition, to examine the robustness of our findings, we examined the potential for non-linearity of continuous GFR predictors by including quadratic terms in the models and by examining the cubic splines. We tested whether the association of GFR with self-reported frailty differed according to method of GFR assessment by comparing the c statistics corresponding to areas under the ROC curves in adjusted models. Cox proportional hazards models were used to assess the association of frailty with death; mGFR, eGFRcr and eGFRcys were included as covariates in separate models. A p-value of less than 0.05 was considered statistically significant. Analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary NC).

Participant characteristics
Of the 840 MDRD participants, 812 had complete data for the assessment of self-reported frailty and were included in our analyses. The majority was male (60.5 %), the median age was 52 (interquartile range [IQR] 42-61), and a history of hypertension was common (84 %) ( Table 1). As expected based on the MDRD inclusion criteria, very few participants had a history of diabetes (5 %). Approximately 21 % of participants had an mGFR greater than or equal to 45 ml/min/1.73 m 2 , 35 % had an mGFR between 30-44 ml/min/1.73 m 2 and 44 % had an mGFR less than or equal to 29 ml/min/1.73 m 2 .

Prevalence of frailty
The most common frailty components were low physical activity (47 %) and poor physical function (23 %), whereas a smaller proportion of individuals met the criteria for underweight/weight loss (17 %) and exhaustion (13 %). There was a graded association of the prevalence of the components of self-reported frailty with GFR category such that individuals in the lowest GFR category had a higher prevalence of each of the components of self-reported frailty (Fig. 1). Sixteen percent of patients were frail by self-report, 53 % were intermediate frail by self-report, and 31 % were not frail (Table 1). Self-reported frail individuals were slightly older than non-frail individuals. Selfreported frail patients were less likely to be male and less likely to be Caucasian. There were no statistically significant differences in the prevalence of hypertension or diabetes among self-reported frail and nonfrail participants.

Association of mGFR with self-reported frailty
In univariate analysis, higher mGFR was associated with lower odds of self-reported frailty ( Table 2). Creactive protein was not associated with frailty (OR 1.07, 95 % CI 0.95-1.2 per one log unit, data not shown) and was therefore not included in our final models. After adjusting for covariates, the association  Association of estimated GFR (eGFRcr and eGFRcys) with self-reported frailty Associations of eGFRcr with self-reported frailty were similar to those observed with mGFR. Higher eGFRcr Associations of eGFRcys and covariates with selfreported frailty were slightly lower compared to those observed with both mGFR and eGFRcr. Higher eGFRcys was associated with lower odds of selfreported frailty in univariate and multivariable analysis (Table 2). When eGFRcys was modeled using categories, individuals with eGFRcys ≤ 29 ml/min/ 1.73 m 2 were more likely to be frail by self-report (OR 2.30 95 % CI 1.11-4.70).
The association between GFR and frailty did not differ significantly according to age or sex for any method of assessing GFR (p-values for interactions >0.05). The potential for non-linearity of the association of GFR with frailty was examined using a quadratic term in the above models and by examining cubic splines, neither of which differed from the linear models presented. The concordance c statistic (area under the receiver operating characteristics [ROC] curve) for predicting frailty by self-report for the base model was 0. 66 (95 % CI, 0.60 to 0.71) (Fig. 2, Table 4). Adding mGFR (0.68, 95 % CI, 0.62 to 0.73), eGFRcr (0.67, 95 % CI 0.61 to 0.72) or eGFRcys (0.67, 95 % CI 0.62 to 0.72) to the model (as a categorical variables) did not significantly change the c statistic.

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 selfreport 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 creatininebased 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 selfreported frailty and kidney function may not be particularly sensitive to method of measurement or  Abbreviations: mGFR measured glomerular filtration rate by Iothalamate, eGFRcr creatinine estimated glomerular filtration rate, eGFRcys by CKD-EPI equation. All models are adjusted for urine protein, sex, age, race, BMI; body mass index and albumin quartiles and group randomization by diet and blood pressure assignment. All GFR measures are per 10 ml/min/1.73 m2 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 selfreported 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 selfreported 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 selfreport 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][19][20][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][4][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 creatininebased 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.

Conclusions
In summary, we found that worse kidney function was associated with higher prevalence of self-reported frailty in a cohort with a smaller burden of comorbidities than is generally present among patients with advanced CKD. Furthermore, the inverse association between kidney function and self-reported frailty was similar when kidney function was measured or estimated. Self-reported frailty was associated with higher risk of death even after adjusting for kidney function. There is a need for longitudinal and interventional studies to determine whether intervening on frailty can improve outcomes among patients with CKD.