Visual Abstract
Keywords: uremia, quality of life, depression, clinical epidemiology, chronic kidney disease
Abstract
Background and objectives
Uremic symptoms, including fatigue, anorexia, pruritus, nausea, paresthesia, and pain, are attributed to the accumulation of organic waste products normally cleared by the kidneys, but whether kidney function is the primary driver of changes in symptom severity over time is not known. The goal of our study was to evaluate the association between eGFR and uremic symptom severity score in patients with CKD.
Design, setting, participants, and measurements
We identified 3685 participants with CKD not on dialysis in the prospective, observational Chronic Renal Insufficiency Cohort (CRIC) Study with baseline assessment of eGFR and uremic symptom severity. Symptoms were assessed by separate questions on the Kidney Disease Quality of Life-36 instrument (zero- to 100-point scale). The longitudinal association between eGFR and uremic symptom severity score was examined with multivariable adjusted linear mixed-effects models with random intercepts and random slopes.
Results
The mean±SD eGFR at baseline was 44±15 ml/min per 1.73 m2, and participants had a median of six (interquartile range 3–11) simultaneous assessments of eGFR and uremic symptoms over the duration of follow-up. The most prevalent symptoms at baseline were pain (57%), fatigue (52%), paresthesia (45%), and pruritus (42%). In adjusted models, a decrease in eGFR of 5 ml/min per 1.73 m2 was associated with a worsening of the symptom severity score by two points or less for each uremic symptom (P<0.01; zero- to 100-point scale). The association between eGFR and uremic symptom severity score was nonlinear. When starting from a lower initial eGFR, a 5 ml/min per 1.73 m2 decrease in eGFR was associated with a greater magnitude of uremic symptom worsening.
Conclusions
The prevalence of uremic symptoms in CKD is high, with significant variability in patient symptom change over time. Declines in eGFR were associated with worsening of uremic symptom severity, but the magnitude of these changes is small and of uncertain clinical significance.
Introduction
Uremic symptoms, including fatigue, nausea, anorexia, pruritus, paresthesia, and pain, contribute significantly to the burden of disease experienced by patients with CKD. Symptoms contribute to the reduced quality of life encountered by many patients with CKD (1–3), making it unsurprising that patients rank symptoms as one of the most important aspects of their kidney disease and a topic of high priority in CKD-related research (4,5). Furthermore, although certain symptoms such as pain, pruritus, poor appetite, and weakness exceed a prevalence of 50% in patients with advanced CKD (6), even patients with more modest reductions in the eGFR, to levels at or just below 60 ml/min per 1.73 m2, report symptoms (7,8), highlighting how uremic symptoms can affect patients across the spectrum of kidney function. Despite such prevalence, there remains a limited understanding of how uremic symptoms evolve over time in patients with different stages of CKD and the central drivers of change for symptom severity.
Although uremic symptoms have long been primarily attributed to the accumulation of retained uremic toxins with declining kidney function, conclusive evidence for an association between kidney function and uremic symptoms is lacking (9–12). Additionally, there has been no comprehensive study of the relationship between changes in kidney function, changes in uremic symptoms, and other clinical factors that may play a role. Clearly defining the association between uremic symptoms and kidney function has broad implications for the counseling and management of patients with CKD, particularly considering that many providers base the decision to initiate dialysis on the hypothesis that supplementing low native kidney function with dialysis can reduce symptom severity.
The primary purposes of this investigation were first to characterize the burden of uremic symptoms and patterns of change over time in patients with CKD and then to examine the longitudinal association between kidney function, specifically eGFR, and uremic symptom severity. We hypothesized that decreases in eGFR would not be the only factor associated with worsening uremic symptoms and that an exploration of risk factors might identify opportunities for future investigations.
Materials and Methods
Study Design
The Chronic Renal Insufficiency Cohort (CRIC) Study is a prospective, multicenter observational study of men and women with CKD in the United States. Between 2003 and 2008, a total of 3939 participants not on dialysis, aged 21–74 years, and with an eGFR between 20 and 70 ml/min per 1.73 m2 were enrolled. The CRIC Study inclusion and exclusion criteria have been published previously (13). CRIC participants are followed annually until death or withdrawal of informed consent, with data for the current study available through December 2018. Censoring for the current study was defined as start of dialysis or receiving a kidney transplant (n=1204), death (n=800), refusal of further study participation or loss to follow-up (n=248), or December 2018, whichever came first. All participants provided written consent for the CRIC Study. The research was approved by the Institutional Review Boards of the participating centers and was conducted in accordance with the Declaration of Helsinki.
For this analysis, we excluded participants with missing data on uremic symptoms (n=57) or other covariates (n=197) at the baseline visit, resulting in a final analytic population of 3685 participants.
Assessment of Kidney Function
eGFR was calculated from serum creatinine using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) 2009 equation (14), and proteinuria was measured from 24-hour urine collections as previously described (13). For the purpose of descriptive statistics, CKD was categorized into eGFR stage (G stage) according to the Kidney Disease Improving Global Outcomes (KDIGO) G staging system: G1, GFR ≥90 ml/min per 1.73 m2; G2, 60–89 ml/min per 1.73 m2; G3a, 45–49 ml/min per 1.73 m2; G3b, 30–44 ml/min per 1.73 m2; G4, 15–29 ml/min per 1.73 m2; and G5, <15 ml/min per 1.73 m2. Stages G1 and G2 and stages G4 and G5 were combined due to low numbers in the G1 (n=22) and G5 (n=4) groups at the baseline evaluation.
Symptoms of Kidney Disease
The uremic symptoms fatigue, anorexia, pruritus, nausea, paresthesia, and pain were coprimary outcomes for this study. Uremic symptoms were captured annually by the Kidney Disease Quality of Life-36 (KDQOL-36) instrument (15). Individual uremic symptom severity scores were determined by the response to single items on the symptom subscale of the KDQOL-36. Participants reported to what extent they were bothered by each symptom during the past 4 weeks on a Likert scale with five response options ranging from “not at all bothered” to “extremely bothered.” Using the standardized scoring technique, responses to the KDQOL-36 questions were transformed to a scale ranging from zero to 100, with higher scores representing higher symptom severity. Using an averaged score of the six uremic symptoms, symptom severity was categorized as low [score 0–8.9], medium [9.0–18.9], high [19.0–28.9], and very high [29.0–100], based on quartiles of scores in US dialysis-dependent CKD patients (16). We included the KDQOL symptoms domain, which is an averaged score for the symptoms of kidney disease subscale of the KDQOL-36, as a secondary outcome.
Statistical Analyses
Baseline participant characteristics were summarized across G stage using descriptive statistics, with P values for trend determined by linear regression with G stage as an ordinal variable. To estimate the uremic symptom trajectory for each participant, we used linear regression of uremic symptom score on time. Because a minimum of two observations of uremic symptom score were required for this descriptive calculation, the trajectory was estimated for 3297 participants. An absolute value of five points or more per year for the slope of symptom score was considered clinically significant worsening or improvement on the basis of the finding that a ten-point change in a symptom score over 2 years is associated with higher risk of mortality in a kidney transplant population (17).
The association between eGFR and uremic symptom severity, both modeled as continuous, time-updated variables, was estimated using linear mixed-effects models with random intercepts and random slopes, accounting for repeated observations within each participant (18,19). Regression coefficients for the change in symptom severity associated with a 5 ml/min per 1.73 m2 change in eGFR were obtained from multivariable adjusted models (20). To account for nonlinearities of the association of eGFR with symptoms, we tested higher-order polynomials (quadratic and cubic terms) in the models, and the final model was chosen on the basis of the comparison of nested models using likelihood ratio tests and the Akaike information criteria (21).
Linear mixed-effects models included fixed effects of time and were adjusted for clinical center site, age, sex, race, ethnicity, employment, education, marital status, body mass index (BMI), proteinuria, coronary artery disease, diabetes, hypertension, congestive heart failure, peripheral vascular disease, history of malignancy, cerebrovascular disease, depressive symptoms, and smoking. BMI, proteinuria, comorbid medical conditions, and health-related behaviors were included as time-updated covariates. Because depressive symptoms were not measured annually in CRIC participants, the average value across all assessments for each participant was used as a fixed covariate in multivariable models. In a secondary analysis, hemoglobin, parathyroid hormone, serum albumin, and bicarbonate were added as covariates to multivariable models. Additional detail regarding covariates and CRIC data collection is provided in Supplemental Item 1. Partial R2 (22) and SD for the distributions of random intercepts and slopes for final models is provided in Supplemental Table 1.
Missing data are described in Supplemental Table 2. In the final analytic sample (n=3685), there were no missing data at baseline. During follow-up, ≤2% of data were missing for BMI, hypertension, history of malignancy, and laboratory values. For these variables, the value from the previous visit was carried forward. Missing values of urine protein-to-creatinine ratio (ranging from 4% to 19%) were categorized as missing for analysis.
We performed several sensitivity analyses to assess the robustness of our results. Descriptive calculations were repeated using an annual change of 0.3 standard deviations of the symptom score at baseline to define clinically significant change in uremic symptom severity over time instead of an absolute value of five points per year (Supplemental Table 3). Using the final analytic sample (n=3685), multivariable adjusted joint models combining linear mixed-effects models and Weibull proportional hazards models were used to examine the association between eGFR and uremic symptom severity (23). A composite outcome defined as time to death, progression to dialysis-dependent CKD, or receipt of a kidney transplant was used for joint models, and participants were censored for loss to follow-up. Additionally, to examine the possibility that participants with anorexia may have changes in serum creatinine independent of GFR, we repeated the analysis for the association between eGFR and anorexia using the CKD-EPI cystatin C eGFR estimating equation (Supplemental Tables 4 and 5). Finally, given the possibility that using an averaged value across all observations for a participant may result in biased estimates (24), we conducted a sensitivity analysis in which depressive symptoms were modeled as a time-updated covariate, using the last observed value carried forward to fill in missing values (Supplemental Table 6).
In exploratory analyses, a backward elimination algorithm was used to identify significant predictors of uremic symptom severity. All covariates with P values ≤0.2 from an unadjusted model with uremic symptom severity as the outcome were included in a saturated model, and then nonsignificant terms were eliminated in a stepwise manner using the criterion P>0.05. Only time and clinical center site were forced into models. Because existing data on factors that influence uremic symptoms in CKD are lacking, for this exploratory analysis, we considered additional demographic and clinical factors to those that were included in our primary analysis (as listed in Supplemental Table 7).
All analyses were performed using Stata SE v17 (StataCorp, College Station, TX).
Results
Baseline Characteristics
The mean±SD age of study participants was 58±11 years; 45% identified as women and 41% as non-Hispanic Black (Table 1). The mean±SD eGFR at baseline was 44±15 ml/min per 1.73 m2. Although only four participants had an eGFR ≤15 ml/min per 1.73 m2 at the baseline assessment, 537 participants progressed to an eGFR ≤15 ml/min per 1.73 m2 over the duration of follow-up. Compared with participants included in the current analysis, CRIC Study participants excluded from the present study (n=254) had generally the same mean eGFR at baseline and similar mean uremic symptom severity scores (Supplemental Table 8).
Table 1.
Baseline characteristics of CRIC Study participants included in the current study, classified by baseline CKD stage
| Characteristics | Overall | Stage of CKD | |||
|---|---|---|---|---|---|
| G1 and G2 | G3a | G3b | G4 and G5 | ||
| N (%) | 3685 | 545 (15) | 1117 (30) | 1338 (36) | 685 (19) |
| Age, yr, mean±SD | 58±11 | 51±11 | 58±10 | 60±11 | 59±11 |
| Women, n (%) | 1653 (45) | 248 (46) | 452 (40) | 629 (47) | 324 (47) |
| Socioeconomic factors, n (%) | |||||
| Race and ethnicity | |||||
| Black | 1504 (41) | 203 (37) | 452 (40) | 557 (42) | 292 (43) |
| Hispanic | 478 (13) | 42 (8) | 109 (10) | 196 (15) | 131 (19) |
| Employed | 1457 (40) | 333 (61) | 489 (44) | 430 (32) | 205 (30) |
| Completed high school or above | 2924 (79) | 496 (91) | 945 (85) | 1,002 (75) | 481 (70) |
| Currently married | 2033 (55) | 318 (58) | 631 (56) | 724 (54) | 360 (53) |
| Clinical parameters | |||||
| eGFR, ml/min per 1.73 m2, mean±SD | 44±15 | 69±9 | 52±4 | 38±4 | 25±4 |
| Body mass index, kg/m2, n (%) | |||||
| ≤25 | 581 (16) | 106 (19) | 168 (15) | 186 (14) | 121 (18) |
| 25–30 | 1055 (29) | 145 (27) | 338 (30) | 380 (28) | 192 (28) |
| >30 | 2049 (56) | 294 (54) | 611 (55) | 772 (58) | 372 (54) |
| Urine protein-to-creatinine ratio, mg/gCr, n (%) | |||||
| <200 | 1913 (52) | 392 (72) | 710 (64) | 611 (46) | 200 (29) |
| 200–1000 | 844 (23) | 90 (17) | 223 (20) | 344 (26) | 187 (27) |
| ≥1000 | 780 (21) | 40 (7) | 145 (13) | 331 (25) | 264 (39) |
| Missing | 148 (4) | 23 (4) | 39 (3) | 52 (4) | 34 (5) |
| Clinical characteristics, n (%) | |||||
| Coronary artery disease | 811 (22) | 68 (12) | 236 (21) | 334 (25) | 173 (25) |
| Diabetes | 1788 (49) | 167 (31) | 503 (45) | 733 (55) | 385 (56) |
| Hypertension | 3172 (86) | 362 (66) | 957 (86) | 1,217 (91) | 636 (93) |
| Congestive heart failure | 356 (10) | 29 (5) | 85 (8) | 151 (11) | 91 (13) |
| Peripheral vascular disease | 247 (7) | 16 (3) | 52 (5) | 102 (8) | 77 (11) |
| Cerebrovascular disease | 363 (10) | 23 (4) | 115 (10) | 150 (11) | 75 (11) |
| Malignancy within past 5 yr | 181 (5) | 21 (4) | 66 (6) | 66 (5) | 28 (4) |
| Depressive symptoms | 620 (17) | 93 (17) | 155 (14) | 231 (17) | 141 (21) |
| Active smoking | 463 (13) | 65 (12) | 125 (11) | 174 (13) | 99 (14) |
| Laboratory values, n (%) | |||||
| Serum albumin, g/dl | |||||
| ≥3.4 | 3308 (90) | 512 (94) | 1035 (93) | 1181 (88) | 580 (85) |
| <3.4 | 377 (10) | 33 (6) | 82 (7) | 157 (12) | 105 (15) |
| Hemoglobin, g/dl | |||||
| M: <12; F: <11 | 950 (26) | 57 (10) | 191 (17) | 404 (30) | 298 (44) |
| M: 12–13.9; F: 11–12.9 | 1640 (45) | 207 (38) | 504 (45) | 622 (46) | 307 (45) |
| M: ≥14; F: ≥13 | 1095 (30) | 281 (52) | 422 (38) | 312 (23) | 80 (12) |
| Serum bicarbonate, mmol/L | |||||
| ≥22 | 3043 (83) | 503 (92) | 1009 (90) | 1096 (82) | 435 (64) |
| <22 | 642 (17) | 42 (8) | 108 (10) | 242 (18) | 250 (36) |
| Baseline PTH, pg/ml | |||||
| 0–65 | 2219 (60) | 477 (88) | 828 (74) | 737 (55) | 177 (26) |
| >65 | 1466 (40) | 68 (12) | 289 (26) | 601 (45) | 508 (74) |
| Symptomsa median score [IQR], 0–100 scaleb | |||||
| KDQOL, symptom domain | 14 [5, 25] | 9 [5, 20] | 11 [5, 20] | 14 [7, 27] | 16 [7, 25] |
| Fatigue | 25 [0, 25] | 0 [0, 25] | 0 [0, 25] | 25 [0, 50] | 25 [0, 50] |
| Anorexia | 0 [0, 0] | 0 [0, 0] | 0 [0, 0] | 0 [0, 0] | 0 [0, 25] |
| Pruritus | 0 [0, 25] | 0 [0, 25] | 0 [0, 25] | 0 [0, 25] | 0 [0, 25] |
| Nausea | 0 [0, 25] | 0 [0, 25] | 0 [0, 0] | 0 [0, 25] | 0 [0, 25] |
| Paresthesia | 0 [0, 25] | 0 [0, 25] | 0 [0, 25] | 25 [0, 50] | 0 [0, 25] |
| Pain | 25 [0, 50] | 25 [0, 50] | 25 [0, 50] | 25 [0, 50] | 25 [0, 50] |
| Median number of symptoms [IQR]c | 2 [1, 4] | 2 [1, 3] | 2 [1, 3] | 3 [1, 4] | 3 [1, 4] |
CRIC, Chronic Renal Insufficiency Cohort; PTH, parathyroid hormone; IQR, interquartile range; KDQOL, Kidney Disease Quality of Life-36.
All P for trend <0.01, with the exception of nausea (P=0.5).
0 denotes absence of symptom; 100 denotes extreme symptom severity.
Out of six possible uremic symptoms.
Participants reported a mean±SD of 2.4±1.7 possible uremic symptoms out of six at baseline, and there was low correlation among symptom scores at baseline (Supplemental Tables 9 and 10). The symptoms pain (57%), fatigue (52%), paresthesia (45%), and pruritus (42%) were most prevalent at the baseline assessment (Figure 1). Anorexia (21%) and nausea (28%) were less commonly reported. Participants with more advanced CKD reported greater severity for all uremic symptoms and the KDQOL symptom domain (P value for trend <0.01), with the exception of nausea (P value for trend=0.5). Symptom severity scores for participants with stage G4 or 5 CKD was similar regardless of whether they met stage G4 or 5 criteria at enrollment or progressed to advanced CKD during follow-up (Supplemental Table 11). Only 16% of participants reported no uremic symptoms at the first study visit.
Figure 1.
Severity of uremic symptoms at the baseline evaluation, stratified by baseline CKD stage. The severity of symptoms experienced by patients with CKD enrolled in the Chronic Renal Insufficiency Cohort (CRIC) Study was assessed in the preceeding 4 weeks by the Kidney Disease Quality of Life-36 instrument.
Change in Uremic Symptoms over Time
Participants had a mean±SD of 6.9±4.2 simultaneous assessments of eGFR and uremic symptoms over a median 7 years of follow-up (interquartile range 3–11 years). On average, symptom severity scores and the KDQOL symptom domain score worsened by less than one point per year (Table 2; zero to 100 scale). Similarly, when stratified by symptom severity at the baseline visit, average annual change in the symptom severity scores remained low (less than three points per year for all symptoms; Supplemental Table 12). Despite the overall stability of uremic symptom severity scores, numerous participants experienced clinically significant annual worsening or improvement. During follow-up, 34% (n=1248) of participants reported clinically significant worsening of at least one out of six uremic symptoms (defined as mean annual change of five points or more), and 8% (n=296) experienced worsening of three or more symptoms (Supplemental Table 13). Conversely, 27% (n=979) of participants reported significant improvement in one or more symptoms.
Table 2.
Population average annual change in symptom score and percentage of patients reporting marked symptom worsening or improvement over follow-up for CRIC Study participants
| Symptom | Mean Annual Change in Symptom Score (95% CI),a 0–100 Scale | % Worsening,b ≥5 Points/Year | % Stableb | % Improved,b ≥5 Points/Year |
|---|---|---|---|---|
| Fatigue | 0.2 (0.2 to 0.3) | 13 | 77 | 10 |
| Anorexia | 0.3 (0.2 to 0.3) | 9 | 86 | 5 |
| Pruritus | 0.6 (0.5 to 0.7) | 14 | 77 | 8 |
| Nausea | 0.01 (–0.1 to 0.8) | 9 | 85 | 6 |
| Paresthesia | 0.1 (0 to 0.2) | 11 | 79 | 10 |
| Pain | 0.4 (0.3 to 0.5) | 15 | 74 | 11 |
| KDQOL, symptoms domain | 0.3 (0.2 to 0.3) | 8 | 87 | 5 |
CRIC, Chronic Renal Insufficiency Cohort; 95% CI, 95% confidence interval; KDQOL, Kidney Disease Quality of Life-36.
Mean annual change in symptom score for the CRIC Study population (n=3685) was determined by linear mixed-effects model of symptom score on time with random intercept and random slope. A positive value indicates worsening symptom severity; a negative value indicates improving symptom severity.
To estimate the uremic symptom trajectory for each participant, individual slopes were calculated using linear regression of symptom severity score on time. An absolute mean change in the symptom slope of five points per year or more was considered a significant worsening or improvement in the symptom severity score (18). Because a minimum of two observations of symptom score were necessary to calculate individual slopes using linear regression, the number of participants included in this descriptive analysis is 3297.
Longitudinal Association of eGFR and Uremic Symptom Severity
The unadjusted mean change in eGFR for the study cohort was –1.34 (95% confidence interval [95% CI], –1.41 to –1.26) ml/min per 1.73 m2 per year. There was a nonlinear association between eGFR and uremic symptom severity and KDQOL symptom domain score (Figure 2). In multivariable adjusted models, when starting with an eGFR of ≤50 ml/min per 1.73 m2, a 5 ml/min per 1.73 m2 decrease in eGFR was associated with a statistically significant higher uremic symptom severity score (Table 3). The magnitude of change in the uremic symptom severity score was greater at lower initial eGFR values. For example, a decrease in eGFR from 50 to 45 ml/min per 1.73 m2 was associated with a 0.46-point worsening of fatigue severity (95% CI, 0.35 to 0.58), whereas a decrease in eGFR from 20 to 15 ml/min per 1.73 m2 was associated with an increase in fatigue severity of 1.3 points (95% CI, 1.04 to 1.56).
Figure 2.
Longitudinal relationship of eGFR and uremic symptom severity score over a median of 7 years of follow-up in 3685 CRIC Study participants. (A) Fatigue. (B) Anorexia. (C) Pruritus. (D) Nausea. (E) Paresthesia. (F) Pain. (G) KDQOL symptom domain. Each point and associated confidence interval represent the average adjusted uremic symptom severity at a given eGFR. The line represents the slope of symptom change as eGFR decreases over time. Higher symptom scores represent greater symptom severity.
Table 3.
Predicted change in uremic symptom severity score associated with a 5 ml/min per 1.73 m2 decrease in eGFR
| Baseline eGFR (ml/min per 1.73 m2) | Predicted Change in Symptom Score (0–100 Scale)a | ||||||
|---|---|---|---|---|---|---|---|
| Fatigue | Anorexia | Pruritus | Nausea | Paresthesia | Pain | KDQOL Symptom Domain | |
| 20 | 1.3 (1.04 to 1.56) | 0.79 (0.6 to 0.99) | 0.78 (0.52 to 1.04) | 0.73 (0.52 to 0.94) | 0.36 (0.11 to 0.62) | 0.57 (0.28 to 0.85) | 0.67 (0.55 to 0.8) |
| 30 | 1.02 (0.83 to 1.21) | 0.6 (0.45 to 0.74) | 0.62 (0.42 to 0.81) | 0.54 (0.39 to 0.69) | 0.3 (0.11 to 0.49) | 0.44 (0.23 to 0.65) | 0.54 (0.45 to 0.63) |
| 40 | 0.74 (0.61 to 0.88) | 0.41 (0.31 to 0.51) | 0.45 (0.32 to 0.59) | 0.35 (0.24 to 0.46) | 0.24 (0.11 to 0.37) | 0.31 (0.16 to 0.46) | 0.41 (0.34 to 0.47) |
| 50 | 0.46 (0.35 to 0.58) | 0.22 (0.13 to 0.31) | 0.29 (0.17 to 0.41) | 0.16 (0.07 to 0.25) | 0.18 (0.07 to 0.29) | 0.18 (0.06 to 0.31) | 0.28 (0.22 to 0.33) |
| 60 | 0.18 (0.04 to 0.33) | 0.03 (–0.08 to 0.14) | 0.13 (–0.02 to 0.27) | 0.03 (–0.14 to 0.09) | 0.12 (–0.02 to 0.26) | 0.05 (–0.11 to 0.21) | 0.14 (0.07 to 0.21) |
| 70 | –0.09 (–0.29 to 0.11) | –0.16 (–0.31 to –0.01) | 0.04 (–0.25 to 0.17) | –0.22 (–0.38 to –0.05) | 0.06 (–0.14 to 0.26) | –0.08 (–0.3 to 0.15) | 0.01 (–0.09 to 0.11) |
Values reported as means (95% confidence intervals). Predicted symptom severity scores for different values of eGFR were calculated after use of multivariable adjusted linear mixed-effects models with random intercepts and random slopes. Models adjusted for time, clinical site, age, sex, race, ethnicity, employment, education, marital status, body mass index, proteinuria, coronary artery disease, diabetes, hypertension, congestive heart failure, peripheral vascular disease, history of malignancy, cerebrovascular disease, depressive symptoms, and smoking. Mean values for all covariates were used. Bold indicates statistically significant results (P<0.05). KDQOL, Kidney Disease Quality of Life-36.
0 denotes symptom not present; 100 denotes extreme symptom severity. A positive value indicates worsening symptom severity; a negative value indicates improving symptom severity.
In secondary analyses, the addition of hemoglobin, parathyroid hormone, serum albumin, and bicarbonate to multivariable models resulted in a modest attenuation of the association between eGFR and uremic symptom severity, without changes to overall statistical significance (Supplemental Table 14). Use of multivariable adjusted joint models to model the association between eGFR and uremic symptom severity resulted in near identical effect estimates for eGFR (Supplemental Table 15).
Exploratory Analysis of Factors Associated with Uremic Symptom Severity
When an automated process for covariate selection was used to identify demographic, clinical, and laboratory factors significantly associated with uremic symptom severity, eGFR was included in final models for all uremic symptoms and the KDQOL symptom domain (Supplemental Table 16). A higher symptom severity score at the baseline visit, depressive symptoms, and polypharmacy (>15 prescribed medications) were associated with higher uremic symptom severity for all six symptoms and the KDQOL symptom domain. The magnitude of the association of these three variables with uremic symptom severity was greater than or equal to that of eGFR.
Discussion
In this prospective cohort study of 3685 adults with CKD followed for a median of 7 years (interquartile range 3–11 years), the prevalence of uremic symptoms was high, with >50% of patients reporting at least one symptom, regardless of stage of CKD. There was heterogeneity in the change of uremic symptom severity over time, with some individuals experiencing clinically significant symptomatic worsening (34%) and others improvement (27%). We found a nonlinear association between eGFR and uremic symptom severity score such that a greater magnitude of symptomatic worsening occurred with decreases in eGFR at lower initial eGFR values.
The importance of quality of life and patients’ experiences is being increasingly recognized in the care of patients with chronic illness. A better understanding of factors associated with uremic symptom severity is necessary to improve these metrics for patients with CKD (1). To the best of our knowledge, ours is the first study to quantify changes in uremic symptom severity associated with changes in eGFR across multiple stages of CKD. We found that a 5 ml/min per 1.73 m2 decrease in eGFR per year, equivalent to the KDIGO definition of rapidly progressive CKD (20), results in changes of relatively small magnitude to uremic symptom severity score (less than two points for all symptoms on a scale from zero to 100). In a recent study of European older adults with an eGFR <20 ml/min per 1.73 m2 (25), declines in eGFR over 1 year were associated with higher symptom severity. We expand on these findings by demonstrating a graded, nonlinear association between eGFR and symptoms across multiple stages of CKD. Contextualizing the significance of the magnitude of our findings is challenging given that the magnitude of reduction in symptom severity associated with improvement in quality of life has yet to be determined. Considering the SD of symptom severity scores in our cohort ranges from 19 to 31 points, a two-point change represents a very small effect size (0.1–0.07 SD). However, the average change in symptom severity in our cohort was less than one point per year, and so, a 5 ml/min per 1.73 m2 decrease in eGFR over 1 year resulting in an additional change to symptom severity of one or two points could indicate a significant albeit gradual change to symptom trajectory over time.
The current study builds on prior cross-sectional studies by describing the prevalence of co-occurring uremic symptoms across multiple stages of CKD and characterizing longitudinal trends in symptom severity. At the initial evaluation, pain (57%), fatigue (52%), paresthesia (45%), and pruritus (42%) were the most prevalent symptoms. This symptom prevalence is similar to that reported by patients with advanced CKD (26,27) or on dialysis (28,29), which is surprising given that only 19% of our cohort had an eGFR ≤30 ml/min per 1.73 m2 at enrollment. We also find that symptom severity is relatively stable over time, with 54% (n=1994) of participants displaying average rates of change for all symptom scores of less than five points per year. This is similar to the stability in composite measures of physical and mental health in patients with advanced CKD demonstrated by other studies (30,31). However, of particular interest is the 34% of participants who report a marked worsening of at least one symptom over follow-up. Considering this subset, work by Grams et al. (32) suggests that trajectories of patient-reported outcome measures may provide a novel metric for CKD progression, risk of cardiovascular disease, and death. Whether uremic symptom trajectories similarly predict outcomes in CKD warrants further investigation.
In exploratory analyses, the baseline symptom score, depressive symptoms, and polypharmacy (prescription of more than 15 medications) were associated with higher symptom severity for all six uremic symptoms. Depression has been associated with symptom burden and pruritus severity in other studies (27,33) and may affect symptoms via biologic (e.g., increased systemic inflammation) and/or behavioral (e.g., decreased coping ability) mechanisms. Polypharmacy may contribute to symptoms via adverse medication side effects or, conversely, reduced adherence to medications (34,35). Because our analytic models were not designed to test the association specifically between these other clinical factors and uremic symptoms, the results of these exploratory analyses should be interpreted as hypothesis generating only, highlighting the question of the effect of depression and polypharmacy on patient-reported symptoms as an area for further investigation.
The CRIC Study is a large, multicenter, prospective cohort study of adult patients with CKD, capturing repeated assessments of kidney function and uremic symptoms over time. This strengthens our findings by enabling use of statistical methods to account for participant dropout and control for fixed differences between individuals that may influence reporting of symptom severity. Although survivor bias may result in healthier participants contributing more data to our analysis, we did not find a significant difference in our results when joint models were specifically used to investigate this possibility. This study used a widely available measure, the KDQOL-36, to capture patient-reported uremic symptoms. It should be acknowledged that this is not an objective measure of symptoms and that the subjectivity of patient-reported symptoms is influenced by many factors, including the severity of the symptom and its perception by an individual. Furthermore, we have selected as our primary outcomes those symptoms we believe most likely to be attributed to uremia on the basis of clinical observation and biologic plausibility, but we recognize as a limitation of this study the lack of an objective approach for symptom selection. Consistent with other studies in the field (20,28), we used individual questions on the KDQOL-36 to assess uremic symptoms, although we acknowledge more work is required to validate this approach. Specifically, the lack of an established minimal clinically important change for these scores makes contextualization of our results challenging. However, given our findings that trajectories of different symptoms are likely to vary within an individual, analyzing symptom questions separately avoids the potential pitfall that summary metrics of symptom burden fail to capture a scenario in which one symptom improves but another worsens within the same participant. Finally, eGFR captures only one aspect of kidney function, and changes in eGFR represent only one factor contributing to uremic solute level. Uremic solute levels are determined by production, via endogenous pathways, the gut microbiome, or ingested foods, and removal, both renal and nonrenal (36). These inherent limitations of eGFR as a single metric of kidney function may, in part, account for the relatively small changes in uremic symptom severity score that we estimated with changes to eGFR.
In conclusion, this study describes a high prevalence of uremic symptoms in patients with CKD and heterogeneity between patients in symptom severity over time. We demonstrate an association between eGFR and uremic symptom severity, although the magnitude of change in symptom severity associated with relatively large eGFR decline is very small. Although eGFR monitoring will remain an important aspect of care for patients with CKD, when managing uremic symptoms, providers should consider factors in addition to eGFR that may be driving symptom burden.
Disclosures
T. Shafi reports grants from the National Institutes of Health during the conduct of the study; consultancy agreements with Siemens; research funding from Baxter (clinical trial site investigator), CVS (clinical trial site investigator), and Natera (clinical trial site investigator); honoraria from Cara Therapeutics, the National Institutes of Health, Siemens, University of Virginia, and State University of New York, Downstate; personal consulting fees from Cara Therapeutics and Siemens; and serving in an advisory or leadership role for the American Journal of Kidney Diseases, American Journal of Medicine, CJASN, and Kidney360. All remaining authors have nothing to disclose.
Funding
S. Kalim is supported by National Institutes of Health (NIH) grant R01DK124453. E.P. Rhee and T. Shafi are supported by NIH grant R01-NR017399. K.E. Wulczyn is supported by NIH grant T32DK007540-34.
Supplementary Material
Acknowledgments
A condensed version of the primary results of this paper were presented as a poster at the 2021 American Society of Nephrology Annual Kidney Week. The CRIC Study was conducted by the CRIC Study Investigators and supported by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). The data from the CRIC Study reported here were supplied by the NIDDK Central Repository. This manuscript was not prepared in collaboration with Investigators of the CRIC Study and does not necessarily reflect the opinions or views of the CRIC Study, the NIDDK Central Repository, or the NIDDK.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
See related editorial, “Symptoms with or because of Kidney Failure?,” on pages 475–477.
Author Contributions
S. Kalim was responsible for funding acquisition; S. Kalim and E.P. Rhee were responsible for resources; S. Kalim, E.P. Rhee, and T. Shafi were responsible for supervision; S. Kalim, E.P. Rhee, T. Shafi, and K.E. Wulczyn were responsible for conceptualization; S. Kalim, E.P. Rhee, T. Shafi, and S.H. Zhao reviewed and edited the manuscript; S. Kalim, K.E. Wulczyn, and S.H. Zhao were responsible for the formal analysis; T. Shafi, K.E. Wulczyn, and S.H. Zhao were responsible for the methodology and curated the data; and K.E. Wulczyn was responsible for the investigation and wrote the original draft of the manuscript.
Supplemental Material
This article contains supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.13010921/-/DCSupplemental.
Supplemental Item 1. Supplemental methods on covariate and data collection in the Chronic Renal Insufficiency Cohort Study.
Supplemental Table 1. Number of observations included in present analysis stratified by visit year and percentage of missing values for variables among Chronic Renal Insufficiency Cohort Study participants included in the current analysis.
Supplemental Table 2. Additional clinical and laboratory characteristics of Chronic Renal Insufficiency Cohort Study participants included in the current study, classified by baseline CKD stage.
Supplemental Table 3. Baseline characteristics of Chronic Renal Insufficiency Cohort Study participants, stratified by inclusion or exclusion in the current study.
Supplemental Table 4. Population average annual change in symptom score and percentage of patients reporting marked symptom worsening or improvement over follow-up for 3685 Chronic Renal Insufficiency Cohort Study participants, stratified by symptom severity at the baseline assessment.
Supplemental Table 5. Number of uremic symptoms significantly worsening, improving, or remaining stable for each Chronic Renal Insufficiency Cohort Study participant.
Supplemental Table 6. Additional metrics of model fit and model characteristics from multivariable adjusted linear mixed-effects models of eGFR on uremic symptom severity.
Supplemental Table 7. Predicted change in uremic symptom severity score associated with a 5 ml/min per 1.73 m2 decrease in eGFR calculated from multivariable models, including laboratory parameters.
Supplemental Table 8. Comparison of effect estimate for the association between eGFR and uremic symptom severity score between multivariable adjusted linear mixed-effects models and joint models.
Supplemental Table 9. Final multivariable-adjusted models of symptom severity among Chronic Renal Insufficiency Cohort Study participants using automated algorithm for covariate selection.
Supplemental Table 10. Percentage of patients reporting marked symptom worsening or improvement over follow-up for 3297 Chronic Renal Insufficiency Cohort Study participants using 0.3*SD to define marked symptom change.
Supplemental Table 11. Spearman correlation between individual symptom scores at baseline in 3685 CRIC Study participants.
Supplemental Table 12. Spearman correlation between mean annual rates of change in symptom scores in 3685 Chronic Renal Insufficiency Cohort Study participants.
Supplemental Table 13. Comparison of effect estimate for the association between eGFR and anorexia severity score using different eGFR estimating equations in multivariable adjusted linear mixed-effects models.
Supplemental Table 14. Predicted change in anorexia symptom severity score associated with a 5 ml/min per 1.73 m2 decrease in eGFR using different eGFR estimating equations.
Supplemental Table 15. Predicted change in uremic symptom severity score associated with a 5 ml/min per 1.73 m2 decrease in eGFR, with depressive symptoms modeled as a time-updated covariate using last observation carried forward to account for missing values.
Supplemental Table 16. Uremic symptom severity of Chronic Renal Insufficiency Cohort Study participants, stratified by enrollment G stage of 4 or 5 versus progression to G stage 4 or 5 during follow-up.
References
- 1.Porter AC, Lash JP, Xie D, Pan Q, DeLuca J, Kanthety R, Kusek JW, Lora CM, Nessel L, Ricardo AC, Wright Nunes J, Fischer MJ; CRIC Study Investigators : Predictors and outcomes of health-related quality of life in adults with CKD. Clin J Am Soc Nephrol 11: 1154–1162, 2016 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Amro A, Waldum B, Dammen T, Miaskowski C, Os I: Symptom clusters in patients on dialysis and their association with quality-of-life outcomes. J Ren Care 40: 23–33, 2014 [DOI] [PubMed] [Google Scholar]
- 3.Voskamp PWM, van Diepen M, Evans M, Caskey FJ, Torino C, Postorino M, Szymczak M, Klinger M, Wallquist C, van de Luijtgaarden MWM, Chesnaye NC, Wanner C, Jager KJ, Dekker FW: The impact of symptoms on health-related quality of life in elderly pre-dialysis patients: Effect and importance in the EQUAL study. Nephrol Dial Transplant 34: 1707–1715, 2019 [DOI] [PubMed] [Google Scholar]
- 4.Davison SN, Levin A, Moss AH, Jha V, Brown EA, Brennan F, Murtagh FEM, Naicker S, Germain MJ, O’Donoghue DJ, Morton RL, Obrador GT; Kidney Disease: Improving Global Outcomes : Executive summary of the KDIGO Controversies Conference on Supportive Care in Chronic Kidney Disease: Developing a roadmap to improving quality care. Kidney Int 88: 447–459, 2015 [DOI] [PubMed] [Google Scholar]
- 5.González AM, Gutman T, Lopez-Vargas P, Anumudu S, Arce CM, Craig JC, Dunn L, Eckardt KU, Harris T, Levey AS, Lightstone L, Scholes-Robertson N, Shen JI, Teixeira-Pinto A, Wheeler DC, White D, Wilkie M, Jadoul M, Winkelmayer WC, Tong A: Patient and caregiver priorities for outcomes in CKD: A multinational nominal group technique study. Am J Kidney Dis 76: 679–689, 2020 [DOI] [PubMed] [Google Scholar]
- 6.Murphy EL, Murtagh FEM, Carey I, Sheerin NS: Understanding symptoms in patients with advanced chronic kidney disease managed without dialysis: Use of a short patient-completed assessment tool. Nephron Clin Pract 111: c74–c80, 2009 [DOI] [PubMed] [Google Scholar]
- 7.Rocco MV, Gassman JJ, Wang S-R, Kaplan RM: Cross-sectional study of quality of life and symptoms in chronic renal disease patients: The modification of diet in renal disease study. Am J Kidney Dis 29: 888–896, 1997 [DOI] [PubMed] [Google Scholar]
- 8.Hailpern SM, Melamed ML, Cohen HW, Hostetter TH: Moderate chronic kidney disease and cognitive function in adults 20 to 59 years of age: Third National Health and Nutrition Examination Survey (NHANES III). J Am Soc Nephrol 18: 2205–2213, 2007 [DOI] [PubMed] [Google Scholar]
- 9.Shlipak MG, Lash JP, Yang W, Teal V, Keane M, Cappola T, Keller C, Jamerson K, Kusek J, Delafontaine P, He J, Miller ER 3rd, Schreiber M, Go AS; CRIC Investigators : Symptoms characteristic of heart failure among CKD patients without diagnosed heart failure. J Card Fail 17: 17–23, 2011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Zhang X, Bansal N, Go AS, Hsu C-Y: Gastrointestinal symptoms, inflammation and hypoalbuminemia in chronic kidney disease patients: A cross-sectional study. BMC Nephrol 16: 211, 2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.de Goeij MCM, Ocak G, Rotmans JI, Eijgenraam J-W, Dekker FW, Halbesma N: Course of symptoms and health-related quality of life during specialized pre-dialysis care. PLoS One 9: e93069, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Brown SA, Tyrer FC, Clarke AL, Lloyd-Davies LH, Stein AG, Tarrant C, Burton JO, Smith AC: Symptom burden in patients with chronic kidney disease not requiring renal replacement therapy. Clin Kidney J 10: 788–796, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Feldman HI, Appel LJ, Chertow GM, Cifelli D, Cizman B, Daugirdas J, Fink JC, Franklin-Becker ED, Go AS, Hamm LL, He J, Hostetter T, Hsu CY, Jamerson K, Joffe M, Kusek JW, Landis JR, Lash JP, Miller ER, Mohler ER 3rd, Muntner P, Ojo AO, Rahman M, Townsend RR, Wright JT; Chronic Renal Insufficiency Cohort (CRIC) Study Investigators : The Chronic Renal Insufficiency Cohort (CRIC) Study: Design and methods. J Am Soc Nephrol 14: S148–S153, 2003 [DOI] [PubMed] [Google Scholar]
- 14.Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF 3rd, Feldman HI, Kusek JW, Eggers P, Van Lente F, Greene T, Coresh J; CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) : A new equation to estimate glomerular filtration rate. Ann Intern Med 150: 604–612, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Hays RD, Kallich JD, Mapes DL, Coons SJ, Carter WB: Development of the kidney disease quality of life (KDQOL) instrument. Qual Life Res 3: 329–338, 1994 [DOI] [PubMed] [Google Scholar]
- 16.Peipert JD, Bentler PM, Klicko K, Hays RD: Psychometric properties of the Kidney Disease Quality of Life 36-Item Short-Form Survey (KDQOL-36) in the United States. Am J Kidney Dis 71: 461–468, 2018 [DOI] [PubMed] [Google Scholar]
- 17.Taylor K, Chu NM, Chen X, Shi Z, Rosello E, Kunwar S, Butz P, Norman SP, Crews DC, Greenberg KI, Mathur A, Segev DL, Shafi T, McAdams-DeMarco MA: Kidney disease symptoms before and after kidney transplantation. Clin J Am Soc Nephrol 16: 1083–1093, 2021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Littell RC, Pendergast J, Natarajan R: Modelling covariance structure in the analysis of repeated measures data. Stat Med 19: 1793–1819, 2000 [DOI] [PubMed] [Google Scholar]
- 19.Laird NM, Ware JH: Random-effects models for longitudinal data. Biometrics 38: 963–974, 1982 [PubMed] [Google Scholar]
- 20.Kidney Disease: Improving Global Outcomes (KDIGO) CKD Work Group: KDIGO 2012 Clinical Practice Guideline for the Evaluation and Management of Chronic Kidney Disease. Kidney Int Suppl 3 1–150, 2013 [Google Scholar]
- 21.Akaike H: Proceeding of the Second International Symposium on Information Theory, Budapest, Akademiai Kiado, 1973, pp 267–281 [Google Scholar]
- 22.Edwards LJ, Muller KE, Wolfinger RD, Qaqish BF, Schabenberger O: An R2 statistic for fixed effects in the linear mixed model. Stat Med 27: 6137–6157, 2008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Crowther MJ, Abrams KR, Lambert PC: Joint modeling of longitudinal and survival data. Stata J 13: 165–184, 2013 [Google Scholar]
- 24.Liu Q, Smith AR, Mariani LH, Nair V, Zee J: Methods for assessing longitudinal biomarkers of time-to-event outcomes in CKD: A simulation study. Clin J Am Soc Nephrol 14: 1315–1323, 2019 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Janmaat CJ, van Diepen M, Meuleman Y, Chesnaye NC, Drechsler C, Torino C, Wanner C, Postorino M, Szymczak M, Evans M, Caskey FJ, Jager KJ, Dekker FW; EQUAL Study Investigators : Kidney function and symptom development over time in elderly patients with advanced chronic kidney disease: Results of the EQUAL cohort study. Nephrol Dial Transplant 36: 862–870, 2021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.van de Luijtgaarden MWM, Caskey FJ, Wanner C, Chesnaye NC, Postorino M, Janmaat CJ, Rao A, Torino C, Klinger M, Drechsler C, Heimburger O, Szymczak M, Evans M, Dekker FW, Jager KJ; EQUAL study investigators : Uraemic symptom burden and clinical condition in women and men of ≥65 years of age with advanced chronic kidney disease: Results from the EQUAL study. Nephrol Dial Transplant 34: 1189–1196, 2019 [DOI] [PubMed] [Google Scholar]
- 27.Sukul N, Speyer E, Tu C, Bieber BA, Li Y, Lopes AA, Asahi K, Mariani L, Laville M, Rayner HC, Stengel B, Robinson BM, Pisoni RL; CKDopps and CKD-REIN investigators : Pruritus and patient reported outcomes in non-dialysis CKD. Clin J Am Soc Nephrol 14: 673–681, 2019 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Rhee EP, Guallar E, Hwang S, Kim N, Tonelli M, Moe SM, Himmelfarb J, Thadhani RI, Powe NR, Shafi T: Prevalence and persistence of uremic symptoms in incident dialysis patients. Kidney360 1: 86–92, 2020 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Murtagh FEM, Addington-Hall J, Higginson IJ: The prevalence of symptoms in end-stage renal disease: A systematic review. Adv Chronic Kidney Dis 14: 82–99, 2007 [DOI] [PubMed] [Google Scholar]
- 30.Meuleman Y, Chilcot J, Dekker FW, Halbesma N, van Dijk S: Health-related quality of life trajectories during predialysis care and associated illness perceptions. Health Psychol 36: 1083–1091, 2017 [DOI] [PubMed] [Google Scholar]
- 31.Da Silva-Gane M, Wellsted D, Greenshields H, Norton S, Chandna SM, Farrington K: Quality of life and survival in patients with advanced kidney failure managed conservatively or by dialysis. Clin J Am Soc Nephrol 7: 2002–2009, 2012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Grams ME, Surapaneni A, Appel LJ, Lash JP, Hsu J, Diamantidis CJ, Rosas SE, Fink JC, Scialla JJ, Sondheimer J, Hsu C, Cheung AK, Jaar BG, Navaneethan S, Cohen DL, Schrauben S, Xie D, Rao P, Feldman HI; CRIC Study Investigators : Clinical events and patient-reported outcome measures during CKD progression: Findings from the Chronic Renal Insufficiency Cohort Study. Nephrol Dial Transplant 36: 1685–1693, 2021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Lockwood MB, Lash JP, Pauls H, Chung SY, Samra M, Ryan C, Park C, DeVon H, Bronas UG: Physical symptom cluster subgroups in chronic kidney disease. Nurs Res 69: 100–108, 2020 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Laville SM, Gras-Champel V, Moragny J, Metzger M, Jacquelinet C, Combe C, Fouque D, Laville M, Frimat L, Robinson BM, Stengel B, Massy ZA, Liabeuf S; Chronic Kidney Disease-Renal Epidemiology and Information Network (CKD-REIN) Study Group : Adverse drug reactions in patients with CKD. Clin J Am Soc Nephrol 15: 1090–1102, 2020 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.McKillop G, Joy J: Patients’ experience and perceptions of polypharmacy in chronic kidney disease and its impact on adherent behaviour. J Ren Care 39: 200–207, 2013 [DOI] [PubMed] [Google Scholar]
- 36.Meyer TW, Hostetter TH: Uremia. N Engl J Med 357: 1316–1325, 2007 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.