Urinary Creatinine Excretion, Bioelectrical Impedance Analysis, and Clinical Outcomes in Patients with CKD: The CRIC Study

Abstract

Background and objectives

Previous studies in chronic disease states have demonstrated an association between lower urinary creatinine excretion (UCr) and increased mortality, a finding presumed to reflect the effect of low muscle mass on clinical outcomes. Little is known about the relationship between UCr and other measures of body composition in terms of the ability to predict outcomes of interest.

Design, setting, participants, & measurements

Using data from the Chronic Renal Insufficiency Cohort (CRIC), the relationship between UCr, fat free mass (FFM) as estimated by bioelectrical impedance analysis, and (in a subpopulation) whole-body dual-energy x-ray absorptiometry assessment of appendicular lean mass were characterized. The associations of UCr and FFM with mortality and ESRD were compared using Cox proportional hazards models.

Results

A total of 3604 CRIC participants (91% of the full CRIC cohort) with both a baseline UCr and FFM measurement were included; of these, 232 had contemporaneous dual-energy x-ray absorptiometry measurements. Participants were recruited between July 2003 and March 2007. UCr and FFM were modestly correlated (rho=0.50; P<0.001), while FFM and appendicular lean mass were highly correlated (rho=0.91; P<0.001). Higher urinary urea nitrogen, black race, younger age, and lower serum cystatin C level were all significantly associated with higher UCr. Over a median (interquartile range) of 4.2 (3.1–5.0) years of follow-up, 336 (9.3%) participants died and 510 (14.2%) reached ESRD. Lower UCr was associated with death and ESRD even after adjustment for FFM (adjusted hazard ratio for death per 1 SD higher level of UCr, 0.63 [95% confidence interval, 0.56 to 0.72]; adjusted hazard ratio for ESRD per 1 SD higher level of UCr, 0.70 [95% confidence interval, 0.63 to 0.75]).

Conclusions

Among a cohort of individuals with CKD, lower UCr is associated with death and ESRD independent of FFM as assessed by bioelectrical impedance analysis.

Introduction

Despite our growing understanding of multiple risk factors that explain clinical outcomes in the setting of CKD, there remains substantial unexplained variation in the rates of death, ESRD, and other morbidities. Low muscle mass and impaired muscle function are established, powerful predictors of mortality in aging and a variety of chronic conditions other than CKD, including AIDS, cancer, and congestive heart failure (1–4). Muscle may increase insulin sensitivity and may decrease inflammation in addition to allowing for greater activity and resilience in the face of disease (5,6). Cross-sectional studies suggest correlations between low muscle mass and worse renal function in patients with CKD (7–11). We hypothesized that measures of muscle mass would be predictive of mortality and progression to ESRD among those with CKD.

We have previously demonstrated that a low creatinine generation rate is associated with death in patients with severe AKI (12). A large study of dialysis recipients in the United States demonstrated a strong association between lower urinary creatinine (UCr) excretion (before dialysis initiation) and increased mortality (13). A recent study examined UCr in a cohort of Italian patients with CKD and found that lower UCr was strongly associated with death and progression to ESRD (14). The authors of these studies (ourselves included) have hypothesized that the relationship between low UCr and adverse outcomes is due to low muscle mass. However, it remains unclear whether UCr is a reliable marker of muscle mass in patients with CKD because no studies have compared it to a more standard muscle mass measure.

Physiology studies have suggested a strong correlation between muscle mass and UCr (as determined from 24-hour urine collections) in healthy populations (15,16). UCr is influenced by factors other than muscle mass, including creatine and creatinine intake (typically from cooked meat) (17,18) and protein intake (16), as well as the rate of creatine-to-creatinine conversion (19). However, longitudinal studies in healthy participants demonstrate modest within-person variation (20–22). Other commonly used tools for assessing body composition include bioelectrical impedance analysis (BIA) and dual-energy x-ray absorptiometry (DXA), both of which provide estimates of fat free mass (FFM), which is well correlated with skeletal muscle mass (23–26).

We sought to characterize the association of UCr and BIA-derived measures of FFM to evaluate the UCr excretion per kilogram of FFM among individuals with varying degrees of kidney function, and to examine these metrics in relationship to clinical outcomes, including death and the development of ESRD in a large and diverse cohort of individuals with CKD.

Materials and Methods

Participants

The Chronic Renal Insufficiency Cohort (CRIC) is a multicenter cohort study of individuals in the United States with CKD. Criteria for enrollment in CRIC have been previously reported (27). We included all CRIC participants with 24-hour UCr and height measurements, as well as a BIA assessment at the baseline study visit.

24-Hour Urine Collections

All CRIC participants are asked to provide at least one 24-hour urine collection upon study enrollment. Participants are provided with a 3-L specimen container and are given detailed verbal and written instructions in proper collection technique. Participants record the time of first and last micturition in the 24-hour period. Samples with a total collection time of <22 or >26 hours are not processed. In these instances, participants are asked to provide another sample, and the collection instructions are reviewed with them in detail. Valid 24-hour collections are measured for volume, thoroughly mixed, separated into aliquots, and then shipped in chilled containers to the CRIC central lab.

Quality of 24-Hour Urine Collection

We used several methods to assess the quality of urine collections. First, we compared measured UCr to that estimated by a validated equation derived by Ix et al. (which incorporates age, sex, weight, and race), defining as potentially inadequate samples that fell outside of 30% of predicted (28). This criteria would exclude 37.3% of the cohort. Second, we used a study-specific metric, wherein we defined as potentially inadequate urine collections that fell outside of 1 SD of the mean UCr-to-weight ratio in the full cohort; 27.5% of individuals met this criteria. Finally, we derived the expected UCr from the Cockroft–Gault equation (which incorporates age, sex, and weight), again excluding those who fell outside of 30% of predicted values: 40.2% would be excluded on this basis (29).

Determination of UCr

Urinary creatinine concentration was assayed via a spectrophotometric rate reaction using the Jaffe method (Roche Diagnostics, Indianapolis, IN). UCr was calculated as the product of urinary creatinine concentration and 24-hour urinary volume.

BIA

BIA was performed using a Quantum II bioelectrical impedance analyzer (RLJ Systems, Clinton Township, MI) with the participant lying supine. Resistance and reactance were measured with two consecutive readings, and stable values (within 1% of each other) were recorded. CRIC participants with pacemakers or with amputations did not undergo BIA testing. FFM, fat mass, and total body water were calculated using the equations developed by Chumlea et al. (30) Percentage body water was defined as total body water/weight.

DXA

In a subset of 232 participants who had DXA performed at the baseline study visit, whole-body lean mass was assessed by using a Hologic DXA apparatus (Hologic, Inc., Bedford, MA). The measurements were performed in the array mode using standard positioning techniques. Quality control scans were obtained daily. We defined appendicular lean mass as the sum of muscle mass estimates for each limb.

Iothalamate-Measured GFR

In a subset of 1329 participants, GFR was measured using iothalamate-125 clearance, as described previously (31).

Covariate Ascertainment

Serum cystatin C concentration was measured via particle-enhanced immunonephelometry (Dade Behring, Deerfield, IL). Estimated glomerular filtration rate cystatin was estimated using the CKD-Epidemiology Collaboration cystatin equation, and estimated glomerular filtration rate creatinine was estimated by the CRIC creatinine equation (31,32). Caloric and macronutrient intake was assessed via a validated food-frequency questionnaire (33). Covariates were log-transformed if they demonstrated substantial rightward skew on visual inspection of histograms.

Outcomes

Our primary outcomes were all-cause mortality and development of ESRD (as defined by initiation of dialysis or kidney transplantation). Deaths were ascertained from reports of next of kin, death certificates, obituaries, reviews of hospital records, and linkage with the Social Security Death Master File. ESRD was defined as receipt of maintenance dialysis or a kidney transplant and was ascertained primarily through self-report or report from family members of decedents. Information collected on ESRD by study investigators was supplemented by the US Renal Data System.

Statistical Analyses

We used standard descriptive statistics to characterize covariates in the study population. UCr was indexed to height (UCrI), where indicated, in univariable analyses by dividing the UCr by height^3.5. This power was chosen on the basis of the β-coefficient of a log-UCr, log-height regression. Indexing to height was necessary because taller individuals will have higher UCr and FFM despite not being more muscular. We assessed intercorrelations between muscle measures with Spearman correlation coefficients. We tested for differences across tertiles of UCrI using ANOVA, Kruskal–Wallis, and chi-squared testing as appropriate. To create a parsimonious model of UCr, we used linear regression with a reverse stepwise elimination algorithm at a threshold P value of <0.05. The R² values from linear regression models provided the percentage variance accounted for by model covariates. We assessed the strength of the FFM to UCr correlation among participants with low versus high percentage body water via direct comparison of Spearman rho coefficients (34). We evaluated the relationship between the UCr-to-FFM ratio, which we hypothesized may reflect muscle quality, and CKD-stage using t tests.

Primary Analyses

We assessed the relationship between a single baseline UCr and FFM—both indexed to height—and clinical outcomes using Cox proportional hazards models. Proportionality of hazards was assessed using log-log plots and examination of Schoenfeld residuals. Where death was the outcome, censoring was performed at withdrawal or loss to follow-up. Where ESRD was the outcome, follow-up was censored at death, withdrawal, or loss to follow-up. In subhazard competing risk analyses, death was treated as a competing risk for the development of ESRD, and outcomes could be experienced only once. Models are presented unadjusted and then sequentially adjusted for (1) known risk factors for death and ESRD in patients with CKD, (2) factors associated with UCr in univariable analysis, but not plausibly on the causal path from muscle mass to clinical outcomes, and (3) factors associated with UCr in univariable analysis, but plausibly on the causal path from muscle mass to clinical outcomes. Model discrimination was assessed with the Harrell C-index, which is the time-to-event equivalent of a C-statistic in studies with binary outcomes (35). We compared C-index values using linear combination of the C-index estimators. All models were stratified by clinical center. Statistical analysis was performed in Stata software, version 13.1 (Stata Corp., College Station, TX).

Results

Baseline Characteristics

Among an entire study population of 3939, there were 3604 CRIC participants who met inclusion criteria for this study. Table 1 summarizes the baseline characteristics of these participants, stratified by tertile of UCr indexed to height (UCrI). The mean age±SD was 57.8±0.9 years. Roughly half of the population was male, and 42% of the participants identified themselves as black.

Table 1.

Baseline characteristics by urinary creatinine indexed to height

Characteristic	Low UCrI (n=1201)	Middle UCrI (n=1201)	High UCrI (n=1202)	Total (n=3604)	P
UCr (g/24 hr)	0.82 (0.66–1.00)	1.25 (1.07–1.47)	1.80 (1.51–2.15)	1.25 (0.93–1.66)	<0.001
Demographic and clinical
Age (yr)	59.0±10.8	57.7±11.0	56.8±10.9	57.8±10.9	<0.001
Men	574 (47.8)	636 (53.0)	745 (62.0)	1955 (54.2)	<0.001
Black	502 (41.8)	469 (39.1)	542 (45.1)	1513 (42.0)	0.01
Hispanic	140 (11.7)	129 (10.7)	126 (10.5)	395 (11.0)	0.63
Hypertension	1040 (86.6)	1032 (85.9)	1030 (85.7)	3102 (86.1)	0.80
Diabetes	608 (50.6)	564 (47.0)	551 (45.8)	1723 (47.8)	0.05
Cardiovascular disease	457 (38.1)	376 (31.3)	350 (29.1)	1183 (32.8)	<0.001
Current smoking	204 (17.0)	163 (13.6)	101 (8.4)	468 (13.0)	<0.001
Systolic BP (mmHg)	131±24	127±22	126±20	128±22	<0.001
Diastolic BP (mmHg)	72±13	71±13	72±12	71±13	0.07
Education
Less than high school	274 (22.8)	214 (17.8)	225 (18.7)	713 (19.8)	0.001
High school graduate	227 (18.9)	255 (21.2)	202 (16.8)	684 (19.0)
Some college	355 (29.6)	347 (28.9)	349 (29.1)	1051 (29.2)
College graduate or higher	345 (28.7)	385 (32.1)	425 (35.4)	1155 (32.1)
Anthropometry
Height (cm)	169±10	169±10	169±9	169±10	0.96
Weight (kg)	86.1±23.5	91.1±22.8	98.0±22.6	91.7±23.5	<0.001
Body mass index (kg/m2)	30.1±7.5	31.9±7.5	34.5±8.0	32.2±7.9	<0.001
Waist circumference (cm)	103±18	106±18	110±16	106±18	<0.001
Fat-free mass (kg)	57.2±15.6	60.1±15.3	64.3±15.3	60.5±15.7	<0.001
Fat mass (kg)	28.9±14.3	31.0±14.4	33.7±15.3	31.2±14.8	<0.001
Medication use
Steroids	124 (10.4)	121 (10.1)	104 (8.7)	349 (9.7)	0.32
Prednisone	37 (3.1)	28 (2.3)	18 (1.5)	83 (2.3)	0.34
ACE/ARB	777 (65.3)	848 (70.9)	849 (71.1)	2474 (69.1)	0.002
Laboratory results
Creatinine (mg/dl)	1.88±0.69	1.82±0.64	1.78±0.58	1.83±0.64	<0.001
Cystatin C (mg/L)	1.64±0.58	1.50±0.54	1.38±0.46	1.51±0.54	<0.001
eGFR-creatinine (ml/min per 1.73 m2)	41.6±15.7	45.2±16.8	48.6±17.1	45.1±16.8	<0.001
eGFR≥60 ml/min per 1.73 m2	168 (14.0)	205 (17.1)	270 (22.5)	643 (17.8)	<0.001
eGFR 45–59 ml/min per 1.73 m2	274 (22.8)	351 (29.2)	393 (32.7)	1018 (28.2)	<0.001
eGFR 30–44 ml/min per 1.73 m2	442 (36.8)	410 (34.1)	374 (31.1)	1226 (34.0)	0.01
eGFR 15–29 ml/min per 1.73 m2	311 (25.9)	233 (19.4)	164 (13.6)	708 (19.6)	<0.001
eGFR<15 ml/min per 1.73 m2	6 (0.5)	2 (0.2)	1 (0.1)	9 (0.2)	0.10
eGFR-cystatin (ml/min per 1.73 m2)	46.9±21.8	52.5±23.3	58.1±24.3	52.5±23.6	<0.001
iGFRa (ml/min/1.73 m2)	44.9±18.9	47.8±21.2	54.1±21.4	49.2±21.0	<0.001
Urine protein (g/24 hr)	0.23 (0.07–1.12)	0.17 (0.07–0.81)	0.17 (0.08–0.75)	0.18 (0.07–0.90)	0.35
<0.10 g/24 hr	440 (36.8)	441 (36.9)	448 (37.3)	1329 (37.0)	0.02
0.10–<0.50 g/24 hr	313 (26.2)	362 (30.3)	373 (31.1)	1048 (29.2)
0.50–<1.50 g/24 hr	193 (16.1)	182 (15.2)	188 (15.7)	563 (15.7)
≥1.50 g/24 hr	250 (20.9)	209 (17.5)	191 (15.9)	650 (18.1)
Serum albumin (g/dl)	3.9±0.5	4.0±0.5	4.0±0.4	3.9±0.5	<0.001
Hemoglobin (g/dl)	12±2	13±2	13±2	13±2	<0.001
Total cholesterol (mg/dl)	184±46	184±47	183±42	183±45	0.89
LDL cholesterol (mg/dl)	101±36	102±35	104±35	102±35	0.33
HDL cholesterol (mg/dl)	50±16	48±16	45±14	48±15	<0.001
Triglycerides (mg/dl)	152.0±106.0	154.0±111.2	161.8±124.9	155.9±114.4	0.09
Hemoglobin A1c (%)	6.7±1.7	6.6±1.5	6.6±1.4	6.6±1.5	0.003
Serum CO2 (mEq/L)	24.3±3.3	24.3±3.1	24.7±3.2	24.5±3.2	0.13
High-sensitivity CRP (mg/L)	2.37 (0.95–5.70)	2.51 (0.97–6.73)	2.73 (1.19–6.47)	2.53 (1.04–6.30)	0.01
IL-6 (pg/ml)	2.05 (1.23–3.40)	1.87 (1.10–2.97)	1.77 (1.11–2.80)	1.87 (1.14–3.08)	<0.001
NT-Pro BNP (pmol/L)	54.3 (23.5–117.7)	40.3 (17.3–96.3)	27.7 (11.7–65.6)	39.4 (16.7–91.0)	<0.001
FGF-23 (pg/ml)	166 (105–283)	146 (99–224)	121 (84–196)	144 (95–236)	<0.001
Dietary intake					.
Total caloriesb	1784±818	1799±803	1905±833	1830±820	0.002
Proteinb (g)	67±34	69±35	76±37	71±36	<0.001
Protein calories: total	15.2±3.6	15.6±3.9	16.0±3.8	15.6±3.8	<0.001
Caloriesb (%)
Fatb (g)	67±37	68±35	74±38	70±37	<0.001
Carbohydrateb (g)	227±111	227±118	233±113	229±114	0.41
Urine urea nitrogen (g/24 hr)	6.06±2.91	8.36±3.47	11.30±5.01	8.58±4.45	<0.001

Baseline patient characteristics across tertiles of urinary creatinine adjusted for height (UCr). Data are expressed as mean±SD, n (%), or median (interquartile range). P values reflect differences across the three groups. UCrI, urinary creatinine excretion indexed to height; UCr, urinary creatinine excretion; ACE/ARB, angiotensin-converting enzyme inhibitor/angiotensin receptor blocker; iGFR, iothalomate measured GFR; CRP, C-reactive protein; NT-Pro BNP, N-terminal pro-brain natriuretic peptide; FGF-23, fibroblast growth factor-23.

^aAvailable in a subset of 1329 participants.

^bAvailable in 2792 participants.

Association between Muscle Metrics in CKD

FFM was positively correlated with 24-hour UCr (rho=0.50 [P<0.001] in the full cohort; rho=0.78 [P<0.001] after exclusion of persons with potentially inadequate collections via the Ix formulation) (Figure 1A). The mean±SD UCr per kg of FFM was 22.3±8.4 mg/kg per day. The UCr-to-FFM ratio was smaller at lower levels of renal function. For example, among patients with CKD stage 3a, UCr per kg FFM was 22.8±7.9 mg/kg per day, compared with 18.8±7.6 mg/kg per day among those with CKD stages 4 or 5 (P<0.001). Similar interactions were seen when GFR was estimated using iothalamate clearance.

Figure 1.

Correlation between muscle metrics. (A) Fat free mass as measured by bioelectrical impedance versus urinary creatinine (rho=0.50; P<0.001). (B) Appendicular muscle mass as measured by dual-energy x-ray absorptiometry versus urinary creatinine (rho=0.47; P<0.001). (C) Fat-free mass as measured by bioelectrical impedance versus appendicular muscle mass as measured by dual-energy x-ray absorptiometry (rho=0.91; P<0.001).

Subpopulations of the CRIC study underwent specialized measures of body composition. Among 232 participants who had DXA assessment of appendicular lean mass (ALM), ALM was positively correlated with UCr (rho=0.47 [P<0.001] in the full cohort; rho=0.75 [P<0.001] when potentially inadequate collections via the Ix formula were excluded) (Figure 1B) and very strongly correlated with the BIA measure of FFM (rho=0.91; P<0.001) (Figure 1C). Participants who underwent DXA had some differences from those who did not undergo this testing (Supplemental Table 1). The correlation between FFM and UCr remained stable across quintiles of percentage body water (quintile 1 [<42%]: rho=0.39; quintile 5 [>56%]: rho=0.42; P for difference=0.49).

Independent Factors Associated with Urinary Creatinine

FFM, as measured by BIA, accounted for 20% of the variability in UCr appearance in this cohort. When the traditional covariates associated with higher creatinine production (male sex, black race, younger age) were added to the model, 24% of the variance was captured. A stepwise regression model that characterizes the covariates most strongly associated with UCr appears as Table 2. Even after accounting for these factors, 50% of the variance in UCr remained unaccounted. The single measured covariate most strongly associated with UCr was urine urea appearance, which accounted for 34% of the variability of UCr in univariable analyses.

Table 2.

Stepwise multivariable linear regression examining factors associated with urinary creatinine excretion

Covariate	UCr (95% CI) (mg/d)	t-Statistic	P
Urine urea nitrogen, per g/d	59.3 (55.9 to 62.7)	34.0	<0.001
Black race	134.2 (104.0 to 164.4)	8.7	<0.001
Cystatin C, per mg/L	−110.3 (−139.6 to −81.0)	−7.4	<0.001
Diabetes	−89.0 (−119.5 to −58.5)	−5.7	<0.001
Female sex	−142.1 (−191.3 to −93.0)	−5.7	<0.001
Log NT-Pro BNP, per unit	−34.6 (−47.1 to −22.2)	−5.5	<0.001
HDL cholesterol, per mg/dl	−2.6 (−3.6 to −1.6)	−5.1	<0.001
Height, per 10 cm	54.3 (31.4 to 77.2)	4.7	<0.001
Weight, per kg	32.9 (14.2 to 51.6)	3.4	0.001
Fat-free mass, per kg	4.0 (1.6 to 6.5)	3.2	0.001
Age, per 10 yr	−22.8 (−36.9 to −8.7)	−3.2	0.002
Serum albumin, per g/L	53.2 (19.0 to 87.4)	3.1	0.002
Waist circumference, per cm	−2.2 (−4.0 to −0.4)	−2.4	0.02
Active smoker	−48.3 (−91.4 to −5.2)	−2.2	0.03

Covariates are ordered by degree of statistical significance (absolute value of covariate t-statistic). The intercept of this model (the value at the mean of continuous variables and the baseline level of categorical variables) is 1393 (1362–1425) mg/d. Model R²=0.50. Covariates considered in the model but removed via stepwise regression: income, education, proteinuria, systolic BP, diastolic BP, total cholesterol, serum CO₂, log troponin I, log IL-6, log fibroblast growth factor-23, prednisone, angiotensin-converting enzyme/angiotensin receptor blocker use, cardiovascular disease history, hemoglobin A1c, hemoglobin. NT-Pro BNP, N-terminal pro-brain natriuretic peptide; 95% CI, 95% confidence interval.

Association of Muscle Metrics with Clinical Outcomes

Over a median 4.2 years of follow-up, 336 (9.3%) participants died and 510 (14.2%) reached ESRD.

UCrI was strongly and independently associated with both death and ESRD, as illustrated in Figure 2 and Table 3. UCrI remained significantly associated with death and ESRD after adjustment for FFM: adjusted hazard ratio (HR) for death per 1 SD higher UCrI, 0.63 (95% confidence interval [95% CI], 0.56 to 0.72; P<0.001); adjusted HR for ESRD, 0.70 (95% CI, 0.63 to 0.77; P<0.001). These results were consistent across extremes of body mass index (those with body mass index <22 and >30 kg/m²). After exclusion of participants with potentially inadequate urine collections by the Ix formula, the HR for death was 0.58 (95% CI, 0.45 to 0.76; P<0.001) and the HR for ESRD was 0.82 (95% CI, 0.68 to 0.99; P=0.05). Neither of these differed from the HRs in the full cohort (P=0.14 and 0.45, respectively). FFM indexed to height was not strongly associated with clinical outcome in univariable or multivariable analyses. As a hypothesized metric of muscle quality, a higher ratio of UCr to FFM was associated with lower rates of death (HR, 0.60 [95% CI, 0.53 to 0.69; P<0.001]) and ESRD (HR, 0.64 [95% CI, 0.58 to 0.71; P<0.001]). The magnitude of this association was greater after exclusion of collection deemed potentially inadequate by the Ix formula (HR for death, 0.45 [95% CI, 0.33 to 0.62]; HR for ESRD, 0.52 [95% CI, 0.41 to 0.67]).

Figure 2.

Kaplan-Meier graphs of clinical outcomes stratified by urine creatinine indexed to height (UCRI). (A) Overall survival. Log-rank P value for high versus intermediate=0.06; intermediate versus low, P<0.001. (B) Progression to end stage renal disease. Log-rank P value for high versus intermediate=0.11; for intermediate versus low, P<0.001.

Table 3.

Unadjusted and sequentially adjusted hazard ratios for death and ESRD per 1-SD increase in urinary creatinine indexed to height and for for death and ESRD per 1-SD increase in fat-free mass, estimated via bioelectrical impedance analysis, indexed to height

Outcome	Hazard Ratio (95% CI)
	Unadjusted	Model 1	Model 2	Model 3
Per 1-SD change in UCrI
Death	0.65 (0.58 to 0.74)	0.79 (0.70 to 0.89)	0.80 (0.69 to 0.93)	0.83 (0.72 to 0.97)
ESRD (death as censoring event)	0.73 (0.66 to 0.80)	0.76 (0.69 to 0.84)	0.70 (0.62 to 0.79)	0.75 (0.66 to 0.85)
ESRD (death as a competing risk)	0.75 (0.67 to 0.83)	0.81 (0.70 to 0.93)	0.72 (0.61 to 0.85)	0.77 (0.66 to 0.90)
Per 1-SD change in FFM estimated via BIA
Death	1.10 (0.99 to 1.22)	1.01 (0.91 to 1.13)	1.00 (0.79 to 1.27)	0.98 (0.77 to 1.26)
ESRD (death as censoring event)	1.21 (1.12 to 1.31)	0.95 (0.87 to 1.04)	1.39 (1.16 to 1.66)	1.27 (1.06 to 1.53)
ESRD (death as a competing risk)	1.20 (1.10 to 1.29)	0.96 (0.87 to 1.06)	1.32 (1.08 to 1.63)	1.23 (1.00 to 1.50)

Model 1: Adjusted for age, sex, race, serum cystatin C, history of cardiovascular disease, systolic BP, proteinuria, and smoking history. Model 2: Adjusted for covariates in model 1 plus income, education, weight, waist circumference, prednisone usage, angiotensin-converting enzyme/angiotensin receptor blocker use, serum albumin, hemoglobin, serum CO₂, urine urea nitrogen. Model 3: Adjusted for covariates in models 1 and 2 plus diabetes history, LDL cholesterol, HDL cholesterol, log C-reactive protein, log IL-6, log NT-Pro brain natriuretic peptide, log fibroblast growth factor-23, and log troponin I. Model 3, preceding table 3a additionally adjusted for fat-free mass. FFM, fat-free mass; BIA, bioelectrical impedance analysis.

To assess the predictive ability of these metrics in comparison with known risk factors for death and ESRD in the CKD population, we examined and contrasted C-index values (Supplemental Table 2). The addition of UCr to traditional risk factors minimally improved the prediction of mortality and did not significantly improve the prediction of ESRD.

Discussion

Several observational studies have demonstrated associations between creatinine generation rate (often measured via UCr) and clinical outcomes. In a large study of patients with cardiovascular disease, lower UCr was significantly associated with increased all-cause mortality (36). A recent study by Di Micco et al. examined UCr in a large population of Italian patients with CKD and determined that lower UCr was significantly associated with both death and ESRD, even after adjustment for body mass index (14). All of these studies were limited in that there was no additional measure of body composition to confirm the associations seen with UCr. UCr has been reported as a proxy for muscle mass in healthy populations (16), but its utility as a muscle mass marker in CKD has been studied in only a small number of patients (37). We demonstrated that UCr was independently predictive of death and progression to ESRD among individuals with CKD even after adjustment for FFM.

Our multivariable model examining associations with UCr demonstrated that FFM (as measured by BIA) accounts for only a small amount of the variability in UCr. While black race, sex, and age were all significantly associated with UCr, they were not nearly as strongly associated with UCr as protein intake, as assessed by urinary urea appearance. Increased protein intake may augment UCr in several ways. First, meat contains both creatine and creatinine and thus meat ingestion may augment UCr (19). In addition, ingestion of the constituent amino acids of creatine (glycine and arginine) stimulates creatine synthesis (38,39). Higher cystatin C levels—indicating lower GFR—were associated with lower UCr, an effect that has been previously described (40). Other factors, including the presence of diabetes and current smoking, were also associated with lower UCr.

Although DXA data were available in only a small subset of patients, the high correlation between FFM estimates based on DXA and BIA suggests that both modalities are capturing a reliable approximation of FFM, although both may falsely categorize edema as FFM (26,41). Despite BIA estimates of FFM being increased in edematous states, we saw no evidence of stronger correlation of FFM and UCr among those with low versus high percentage body water. Future studies may consider higher-resolution muscle imaging, such as that obtained via magnetic resonance imaging, to account more precisely for edema. Nonetheless, the lack of a strong correlation between BIA and UCr suggests that the latter is affected by factors other than muscle mass alone. Given the relationship between UCr and outcomes, these other factors may be more important risk markers than muscle mass itself.

We could not determine what gives rise to the disparity in prognostic value between UCr and FFM, but our results imply that the prognostic value of UCr may not be due to its status as a marker of muscle mass. A greater degree of extrarenal creatinine clearance is seen as kidney function worsens (40), but our findings are robust after adjustment for kidney function. Despite the strong association with protein intake, UCr remains predictive of outcomes even after adjustment for urine urea nitrogen. The nonenzymatic dehydration of creatine to creatinine appears to occur at a stable rate of roughly 1.6%–1.7% per day in healthy adults, but this rate is altered in disease states (16,42,43). We demonstrated a lower creatinine generation rate per kilogram of FFM among those with lower GFR. This may be due to a lower creatine production rate, a lower muscle concentration of creatine, or a lower rate of creatine to creatinine conversion among these individuals. In this light, UCr and the UCr-to-FFM ratio may reflect muscle quality as well as mass. Prior studies in older adults without CKD have demonstrated that poor muscle quality (as manifested by muscle strength or by the UCr-to-FFM ratio) is more strongly associated with adverse outcomes than is muscle mass alone (4,44–46). Future mechanistic studies should evaluate the creatine synthetic and metabolic pathways to determine whether these processes are fundamentally altered in CKD.

Of principal concern in any study examining UCr is the quality of specimen collection. Indeed, systematic undercollection by the sickest individuals in the cohort could lead to the results we report. In sensitivity analysis, we show that even excluding potentially inadequate specimens, the HR for the outcomes of interest does not change. However, there is no way to determine whether individuals at high risk of death and ESRD have truly lowered daily creatinine generation or greater difficulty performing a 24-hour urine collection. Exclusion of potentially inadequate specimens did increase the correlation coefficient between UCr and the muscle mass measures of FFM and ALM, bringing these numbers more in line with those reported in prior studies (in CKD and non-CKD populations); this finding suggested that UCr was strongly associated with muscle mass (15,16,47). This might be the case if UCr, when properly collected, simply reflects muscle mass. Alternatively, it is reasonable to consider that sicker individuals have non-muscle mass–based perturbations in creatinine generation as discussed above, thus giving rise to the discrepancy. We are encouraged that our results are not purely the result of bias based on the consistent finding of strong, independent associations between UCr and outcome regardless of the method we use to exclude certain specimens. Dedicated physiologic studies of creatinine dysmetabolism are warranted.

Our study should be interpreted in the light of several additional limitations. First, our analysis was based on a baseline sample and thus cannot evaluate the effect of changes in muscle measures over time. The lack of longitudinal exposure data would tend to bias our results toward the null, however. Next, BIA is not the gold standard for muscle mass assessment. However, given its high degree of correlation with DXA measures of ALM, we feel that BIA measures of FFM are reliable in this population. Third, the participants who underwent DXA had a lower body mass index than did those who did not; BIA may not perform as well in a more obese population. Finally, serum cystatin C (used to estimate GFR) may not be fully independent of muscle mass, although its concentration is less dependent on muscle mass than is the case with creatinine (48). Higher cystatin C levels in patients with greater muscle mass would tend to bias our findings toward the null hypothesis, however.

In conclusion, UCr is an independent predictor of death and ESRD in patients with CKD independent of traditional risk factors and FFM as assessed by BIA. In addition, FFM assessed with BIA, although strongly correlated with muscle mass, is not significantly associated with death or ESRD in this population. We hypothesize that UCr may capture information about muscle quality that is independent of muscle mass, such as creatine content. Future studies should evaluate in greater detail the relationship between body composition, UCr, muscle function, muscle metabolism, and clinical outcomes.

Disclosure

None.