Longitudinal Changes in Protein Carbamylation and Mortality Risk after Initiation of Hemodialysis

Abstract

Background and objectives 

Carbamylation describes a post–translational protein modification associated with adverse outcomes in ESRD, but the risk implications of changes in carbamylation over time are not well understood.

Design, setting, participants, & measurements 

We investigated the 1-year natural history of protein carbamylation in patients initiating maintenance hemodialysis and determined the prognostic value of longitudinal carbamylation changes in relation to mortality. In a nested patient-control study, we measured serial carbamylated albumin concentrations in select participants from a large incident dialysis cohort followed from 2004 to 2005 (n=10,044); 122 individuals who survived at least 90 days but died within 1 year of initiating hemodialysis (patients) were randomly selected along with 244 individuals who survived for at least 1 year (controls; matched for demographics). Carbamylated albumin concentration was measured using plasma collected at dialysis initiation and every subsequent 90-day period until 1 year or death.

Results 

Baseline carbamylated albumin concentration was similar between controls and patients (mean±SD; 18.9±0.7 and 19.8±1.1 mmol/mol, respectively; P=0.94). From dialysis initiation to day 90, carbamylated albumin concentration markedly fell in all patients, with controls −9.9±0.8 mmol/mol (P<0.001) and patients −10.0±1.2 mmol/mol (P<0.001). Adjusted repeated measures analysis of carbamylated albumin concentration from dialysis initiation to 1 year or death showed that the mean change (95% confidence interval) in carbamylated albumin concentration from baseline to final measure differed significantly between groups (−9.3; 95% confidence interval, −10.8 to −7.7 for controls and −6.3; 95% confidence interval, −7.7 to −2.8 for patients; P<0.01). There were no such between-group differences in blood urea levels, Kt/V, or normalized protein catabolic rate. Mortality prediction assessed using c statistics showed that carbamylated albumin concentration, when modeled continuously as the difference from baseline to final, improved a fully adjusted model from 0.76 to 0.87 (P=0.03).

Conclusions 

Protein carbamylation decreased with dialysis initiation, and a greater reduction over time was associated with a lower risk for mortality. Carbamylation changes were able to predict individuals’ mortality risk beyond traditional variables, including markers of dialysis adequacy and nutrition.

Introduction

Throughout their lifespan, human proteins are exposed to chemical reactions that can alter their structural and functional properties. Spontaneous post–translational protein modifications occur from the binding of reactive molecules to protein functional groups as seen, for example, in glycation reactions. Because post-translational modifications can permanently change protein structure and function, they provide a mechanistic chemical link to the adverse pathophysiology of certain metabolic diseases. Carbamylation is a protein modification that results from constant exposure to urea and its byproduct, cyanate, which both increase as kidney function declines (¹). Carbamylation reactions occur on not only proteins but also, free amino acids, and these targets compete with each other for binding, such that amino acid deficiency exacerbates protein carbamylation (²). Furthermore, carbamylation is not solely related to urea and amino acid levels; cyanate can also be generated by myeloperoxidase and peroxide-catalyzed oxidation of thiocyanate (derived from diet and smoking) at local sites of inflammation (³).

Carbamylation is capable of changing the charge, structure, and functional properties of proteins, and these modifications, in turn, can trigger inappropriate molecular and cellular responses, resulting in adverse clinical outcomes (⁴,⁵). Systemic measures of carbamylation burden, such as carbamylated albumin (C-Alb), have been associated with adverse outcomes in patients with ESRD, including resistance to recombinant erythropoietin treatment, heart failure, and mortality (²,^6–8). For example, in a cohort of 347 patients on prevalent hemodialysis with 5 years of follow-up, the risk for death among patients with carbamylation values in the highest tertile was more than double the risk among patients with values in the middle or lowest tertile (⁷). Similar findings were evident in a larger prevalent hemodialysis cohort (n=1161) followed for 1 year (²). Protein carbamylation associates with death and cardiovascular outcomes in patients with preserved kidney function as well, highlighting carbamylation’s diverse pathophysiology (³). However, only single–time point measures of carbamylation have been used in association studies; thus, the natural history and prognostic value of carbamylation changes over time are unknown. Understanding the temporal relationship between changes in carbamylation and mortality risk is particularly important, because there is emerging evidence that carbamylation is a modifiable process through interventions, such as amino acid therapy (⁹). Herein, we examined serial carbamylation measures in a cohort of patients on incident hemodialysis, hypothesizing that decreases in C-Alb would reduce mortality risk and that changes in carbamylation could risk stratify patients with ESRD beyond traditional risk factors for death.

Materials and Methods

Study Population

The Accelerated Mortality on Renal Replacement Study is a prospective cohort study of 10,044 patients on incident hemodialysis in all 1056 United States centers operated by Fresenius Medical Care North America between June of 2004 and August of 2005. Full details have been previously published (^10–12). Plasma samples were obtained within 14 days of initiation of outpatient hemodialysis and every subsequent 90 days for 1 year or until death. Clinical data were prospectively collected and included demographic information, coexisting conditions, and results of routine laboratory tests. The study was approved by the institutional review board of the Massachusetts General Hospital, which waived the need for informed consent, because all personal identifiers were deleted from samples and data before transfer to the investigators.

We examined mortality in a nested patient-control sample of this cohort. Patients were defined as individuals who died during the first year after hemodialysis initiation, and controls were defined as survivors through this period. We randomly selected 122 patients from the entire cohort along with 244 controls who were frequency matched for age, sex, and race. Because we were interested in longitudinal changes in carbamylation, we required that each patient had at least two plasma samples to analyze, effectively restricting inclusion to individuals who survived at least 90 days (allowing them to contribute baseline and day 90 samples). The 244 controls contributed data across the entire study duration, whereas the 122 patients contributed complete data at baseline and day 90. By day 180, patients were reduced to n=95 because of deaths, and by day 270, they were reduced to n=60.

Exposure Assay

Plasma C-Alb was measured as the ratio of millimoles of C-Alb to moles of total albumin (analogous to the measurement of percentage glycated hemoglobin) using high–performance liquid chromatography and tandem mass spectrometry. A complete description of the mass spectrometric assay for C-Alb and its analytic validation have been previously described (²). Repeat measurements of a representative patient sample showed a coefficient of variance of 1.9%.

Statistical Analyses

Our primary analysis examined C-Alb changes from baseline to all postbaseline visits (until 1 year or death). We used a maximum likelihood, mixed effects, repeated measures model (¹³,¹⁴) that incorporated all longitudinal observations of C-Alb in patients and controls. The model included the terms of outcome (patients versus control), visit, and outcome × visit interaction with baseline C-Alb and baseline variables that differed between patients and controls as covariates (systolic and diastolic BP, history of congestive heart failure, estimated baseline residual renal function, and initial vascular access). We also added time–averaged total serum albumin as a covariate of interest, because although baseline albumin was similar between groups, over time, albumin differed between patients and controls (Table 1). Standard assessment of residual renal function using timed urine collection (¹⁵) was not available; thus, we used an estimation equation to approximate residual renal function (¹⁶): Average of urinary urea and creatinine clearance (ml/min per 1.73 m²) =2.4× serum urea^0.984× serum creatinine^−1.868.

Table 1. - Baseline characteristics of the study population

Characteristic	Survived 1 yr, n=244	Died within 1 yr, n=122	P Value
Age, yr	68±14	68±14
Women, %	130 (53)	65 (53)
White race, %	148 (61)	75 (62)
Body mass index	26.8±7.2	26.7±8.3	0.85
BP, mmHg
Systolic	149±24	139±25	<0.001
Diastolic	75±14	70±13	<0.001
Cause of renal failure, %			0.09
Diabetes	118 (48)	59 (48)
Hypertension	88 (36)	51 (42)
GN	13 (5)	9 (7)
Other	25 (10)	3 (2)
Initial vascular access, %			0.02
Fistula	85 (36)	28 (23)
Graft	33 (14)	16 (13)
Catheter	119 (50)	74 (61)
Average dialysis treatment time, min	220±23	220±38	0.70
Comorbidities, %
Coronary artery disease	29 (12)	17 (14)	0.58
Congestive heart failure	27 (11)	24 (20)	0.03
Stroke	6 (2)	5 (4)	0.39
COPD	7 (3)	4 (3)	0.82
Malignancy	8 (4)	6 (5)	0.44
Lipid disorders	36 (15)	15 (12)	0.52
Anemia	94 (39)	42 (34)	0.44
Peripheral vascular disease	14 (6)	7 (6)	>0.99
Atrial fibrillation	5 (2)	2 (2)	0.79
Liver disease	10 (4)	3 (2)	0.42
Laboratory tests
Hemoglobin, g/dl	10.4±1.4	10.3±1.4	0.49
Baseline albumin, g/dl	3.5±0.5	3.4±0.5	0.17
Study average albumin, g/dl	3.6±0.3	3.4±0.4	<0.001
Ferritin, ng/ml	195 (101, 369)	208 (93, 391)	0.90
Transferrin saturation, %	19.9±10.0	20.1±9.9	0.89
Phosphorus, mg/dl	4.6±1.4	4.4±1.4	0.26
Parathyroid hormone, pg/ml	200.5 (109.5, 333.6)	209.4 (103, 374.8)	0.41
C-reactive protein, mg/L	3.2 (2.4, 4.8)	4.4 (2.4, 4.0)	0.46
BUN, mg/dl	53.5±18.9	55.2±23.0	0.48
Urea reduction ratio	69.5±9.9	69.1±11.5	0.74
Kt/V a	1.34±0.3	1.33±0.4	0.83
Residual renal CLurea,creatinine, b ml/min per 1.73 m2	5.3±4.3	7.1±6.3	0.002

Values reflect baseline measurements recorded at the time of initiating dialysis and are n (%), means±SD, or medians (25th percentile, 75th percentile) as presented. COPD, chronic obstructive pulmonary disease; CL_{urea,creatinine}, average urinary clearance of urea and creatinine.

^aKt/V reflects the average over the first 90 days of dialysis.

^bCL_{urea,creatinine} is estimated from serum creatinine and urea values (see Materials and Methods).

Changes from baseline to all postbaseline visits for C-Alb and other continuous variables of interest (Kt/V and urea reduction) were analyzed by the same repeated measures model. The overall P value from the maximum likelihood, mixed effects, repeated measures model (as shown in Figure 1) represents the overall significance level for the patient versus control difference in C-Alb change measured over all follow-up times. The P value for the specific change from baseline to final measure (taken at 1 year or death) is shown in Table 2. To assess whether this analysis was biased because of time for follow-up, we conducted a sensitivity analysis using an adjusted logistic regression model comparing patient and control differences in C-Alb only in individuals surviving past day 270 of the study.

Figure 1.: Mean (SD) carbamylated albumin through time. Overall, P=0.02 for the patient-control difference in carbamylated albumin change from baseline to each time point adjusting for the variables that differed between patients and controls (systolic and diastolic BPs, vascular access type, estimated residual renal function, average serum albumin, and history of congestive heart failure).

Table 2. - Repeated measures analysis of changes in carbamylated albumin, urea reduction ratio, or Kt/V from dialysis initiation to study completion (1 year or death)

Measure	Control, n=244	Patient, n=122	P Value
Carbamylated albumin, mmol/mol	−9.3 (−10.8 to −7.7)	−6.3 (−7.7 to −2.8)	<0.01
Urea reduction ratio	4.1 (3.2 to 5.1)	3.1 (1.7 to 4.5)	0.25
Kt/V	0.2 (0.1 to 0.2)	0.1 (0.0 to 0.2)	0.16

Values shown are adjusted least squares means taken from baseline to final measure (1 year or death) and 95% confidence intervals estimated from the models. Models included outcome, time point, and outcome × time point interaction with carbamylated albumin, urea reduction, or Kt/V plus time-averaged albumin, residual renal function, systolic BP, diastolic BP, initial vascular access type, and history of congestive heart failure as covariates in each model.

To quantify the prognostic value of C-Alb measurements using multiple statistic approaches, we used a logistic regression model with patient or control status as the outcome, adjusted it for significant risk factors, and then, evaluated the improvement in model performance introduced by adding a participant’s change in C-Alb from baseline to final measure taken. We used c statistics to compare model discrimination, category–free net reclassification improvement to assess the ability of the model to correctly reclassify an individual’s risk, and integrated discrimination improvement to examine the ability of the model to increase average sensitivity without reducing specificity (¹⁷). Multivariable model 1 included the variables significant in univariate analysis (noted above). Model 2 added additional variables previously reported to influence survival on dialysis: time-averaged Kt/V, BUN, normalized protein catabolic rate, transferrin saturation, phosphorous, hemoglobin, ferritin, parathyroid hormone level, body mass index, history of diabetes mellitus, and history of coronary artery disease (¹¹,^18–20).

Because C-Alb values reflect the effects of multiple risk factors, including urea load and dialysis adequacy along with amino acid balance and nutritional status, and to further quantify and visualize patient-control differences in C-Alb relative to other relevant measures, we plotted patient and control distribution curves for C-Alb, Kt/V, BUN, and normalized protein catabolic rate (Figure 2). The plots represent the average final measures taken for subjects in each group. P values represent comparisons of the frequency distribution for each group (regardless of whether using logistic regression or area under the curve comparisons, the significance of the differences did not change). To further assess the correlation between C-Alb and urea, we used Pearson correlations to correlate average C-Alb values from a given time point to average BUN levels taken over the 90 days that preceded the C-Alb time point. The same approach was used to look at the correlation between C-Alb and total albumin.

Figure 2.: Patient and control distributions of carbamylated albumin, Kt/V, BUN, and normalized protein catabolic rate. (A) Distribution of final carbamylated albumin level measures in the study population by patients (blue; n=122) and controls (green; n=244; P=0.01). (B) Distribution of the final averaged Kt/V measure in the study population by patients (blue) and controls (green; P=0.55). (C) Distribution of the average BUN measure in the study population by patients (blue) and controls (green; P=0.51). (D) Distribution of normalized protein catabolic rate measured in the study population by patients (blue) and controls (green; P=0.15). P values represent the frequency distribution comparisons.

All analyses were performed using SAS software, version 9.4 (SAS Institute Inc., Cary, NC). Two–sided P values <0.05 were considered statistically significant.

Results

The baseline characteristics of the patient and control groups were largely similar (Table 1), although patients had lower systolic (mean±SD; 139±25 versus 149±24 mmHg, respectively; P<0.001) and diastolic (70±14 versus 75±14 mmHg, respectively; P<0.001) BPs, had higher percentages of congestive heart failure (20% versus 11%, respectively; P=0.03), and more frequently initiated dialysis with a catheter (61% versus 50%, respectively; P=0.02). Although precise residual renal function measures were not available, limited estimations made using serum creatinine and urea suggested that patients had slightly higher residual function (Table 1). Other common mortality risk factors (e.g., history of coronary artery disease and stroke) trended in the expected directions but did not show statistically significant differences. There was no significant difference in BUN concentrations between patients and controls (55.2±23.0 versus 53.5±18.9 mg/dl, respectively) and no difference in dialysis adequacy as measured by Kt/V (1.3±0.4 versus 1.3±0.3, respectively). All subjects were dialyzed using polysulfone high–flux dialyzers, with no difference in model type between groups (Supplemental Table 1).

Baseline C-Alb was similar between patients and controls (19.8±1.1 and 18.9±0.7 mmol/mol, respectively; P=0.94) (Figure 1). From baseline to day 90, C-Alb fell markedly in all subjects, with patients decreasing by −10.0±1.2 mmol/mol (P<0.001) and controls decreasing by −9.9±0.8 mmol/mol (P<0.001; between-group P=0.16). From day 90 onward, patients showed a progressively higher average level of carbamylation compared with controls (Figure 1), and this difference was clear when assessing the overall patient versus control difference in C-Alb changes measured over all time points (i.e., every 90 days from baseline to 1 year or death; P=0.02). Notably, there was a significant increase in C-Alb between day 90 and final measurements among the patients compared with in controls (mean change +31%±4% among patients versus −5%±3% in controls; P=0.02).

The complete magnitude of change in carbamylation in each group over the study period (i.e., from baseline to 1 year or death) is seen in Table 2. After adjusting for variables that differed between patients and controls, the least squares mean change in C-Alb (95% confidence interval [95% CI]) for patients was −6.3 (95% CI, −7.7 to −2.8 mmol/mol) and controls was −9.3 (95% CI, −10.8 to −7.7 mmol/mol; P<0.01). Using similar analyses, the mean change in Kt/V showed no significant difference between patients and controls (0.1; 95% CI, −0.0 to 0.2 versus 0.2; 95% CI, −0.1 to 0.2, respectively; P=0.16). Moreover, the analysis showed no significant changes in urea reduction (Table 2). To assess whether this analysis was influenced by time for follow-up, we conducted a sensitivity analysis comparing adjusted patient versus control differences in C-Alb only in individuals surviving past day 270 of the study. Even in this subgroup with the longest available follow-up time, there remained a significant patient-control difference in C-Alb measures (P=0.02).

Next, to further quantify and visualize patient-control differences in C-Alb relative to other measures of urea, dialysis adequacy, and metabolism, we plotted patient and control distribution curves for relevant variables using the final measure taken in the study (Figure 2). Although C-Alb showed patients to have a right-shifted curve (i.e., higher values) relative to controls (P=0.01) (Figure 2A), the plots for Kt/V (P=0.55) (Figure 2B), BUN (P=0.51) (Figure 2C), and normalized protein catabolic rate (P=0.15) (Figure 2D) showed no significant differences between patients and controls. In a related analysis, the correlation between C-Alb and urea levels across varying time points seemed modest by ranging from R² of 0.26 to R² of 0.39 (all P<0.001), further suggesting that C-Alb reflects more than just patients’ blood urea concentrations at the time of blood sampling. The correlation between C-Alb and total serum albumin across each time point showed modest inverse correlations between −0.10 and −0.20 (P value =0.05 to P value <0.001), also showing the relative independence of these variables. In exploratory analyses, we looked at >25 clinical variables as predictors of changes in carbamylation and found that, indeed, average BUN and albumin associated with changes in C-Alb (Supplemental Table 2).

Finally, we quantified how well C-Alb levels could improve mortality risk prediction across the study population. As shown in Table 3, addition of C-Alb to models adjusted for variables differing between patients and controls and adjusted for known ESRD mortality risk factors increased the c statistic from 0.76 to 0.87 (P=0.03). Similarly, addition of C-Alb to the model predicting 1-year mortality led to a significant improvement in the classification accuracy, with a net reclassification improvement of 0.60 (0.12–1.10; P=0.002) and an integrated discrimination index of 0.22 (0.12–0.33; P<0.001).

Table 3. - Added predictive ability for mortality with carbamylated albumin

Variable	Unadjusted Model	Model 1	Model 2
C-Alb OR (95% CI)	2.12 (1.23 to 3.66)	1.83 (1.07 to 3.35)	2.73 (1.06 to 7.40)
c Statistics
Base model	0.69	0.63	0.76
Base model and C-Alb	N/A	0.79	0.87
P value		0.03	0.03
NRI	N/A	0.63 (0.16–1.10)	0.60 (0.12–1.10)
P value		0.01	0.002
IDI	N/A	0.15 (0.06–0.25)	0.22 (0.12–0.33)
P value		0.001	<0.001

Improvement in model performance assessed using change in C-Alb from baseline to final measure. Model 1 is the base model adjusted for variables that differed between patients and controls at baseline (residual renal function, systolic BP, diastolic BP, initial vascular access type, and history of congestive heart failure). Model 2 is model 1 plus body mass index, average serum albumin, transferrin saturation, phosphorous, hemoglobin, ferritin, parathyroid hormone level, Kt/V, BUN, normalized protein catabolic rate, history of diabetes mellitus, and history of coronary artery disease. C-Alb, carbamylated albumin; OR, odds ratio per unit increase in the final averaged C-Alb; 95% CI, 95% confidence interval; N/A, not applicable; NRI, net reclassification index; IDI, integrated discrimination improvement.

Discussion

In this study, we showed that, in patients with ESRD, protein carbamylation is universally reduced on initiation of maintenance hemodialysis. Over time, individuals who experienced the greatest reduction in carbamylation level showed a clear survival advantage. Interestingly, this finding held true independent of blood urea levels and standard dialysis adequacy parameters, such as Kt/V, which did not differ between 1-year survivors and those who died during the first year of dialysis. When examining established ESRD mortality predictors (including standard clinical indicators of uremia and dialysis adequacy), C-Alb significantly improved mortality prediction models. Protein carbamylation could, thus, potentially serve as a therapeutic target to reduce ESRD mortality risk.

There is growing evidence that protein carbamylation mechanistically contributes to the adverse clinical outcomes of uremia. A major chemical effect of carbamylation is neutralization of positively charged lysines, which changes protein-water interactions and alters ionic interactions on the protein surface (²¹). Such changes can alter secondary and tertiary protein structures, leading to functional changes and resulting in molecular and cellular dysfunction (¹,²²,²³). Studies have implicated carbamylation in changes in protein charge (²⁴), conformation (²⁵,²⁶), stability (²⁷), enzyme and hormone activity (^28–32), binding properties (^33–35), receptor-drug interaction (³⁶,³⁷), and cellular expression and responses (^38–43). For example, carbamylated collagen and LDLs accelerate the biochemical events of atherosclerosis (²¹,⁴⁴,⁴⁵), and carbamylated erythropoietin loses its activity (⁸,⁴⁶,⁴⁷). Overall, these observations suggest multiple mechanistic pathways through which carbamylation may be contributing to adverse ESRD outcomes.

Clinical studies have linked carbamylation burden to increased risk for inflammation, arthritis, cataract formation, and most robustly, heart disease (¹). For example, carbamylation independently predicted risk of coronary artery disease, future myocardial infarction, stroke, and death in two distinct cohorts of patients without ESRD (³). Moreover, carbamylation levels in both patients without and with ESRD were shown to be independently correlated with concentrations of brain natriuretic peptide and incident congestive heart failure, possibly invoking an association between carbamylation and progressive ventricular dysfunction (⁶,⁴⁸). We have shown that carbamylation levels strongly associate with erythropoietin resistance in patients on dialysis after controlling for traditional risk factors, including markers of inflammation (⁸). Most significantly, using single–time point measurements of total body carbamylation burden, such as C-Alb and homocitrulline, elevated protein carbamylation associates with ESRD mortality (²,⁷).

Our study adds to this growing body of literature by showing that changes in carbamylation over time predict outcomes. In our sample population, baseline carbamylation levels did not differ between patients and controls. Although all subjects experienced a reduction in carbamylation on initiating dialysis, the net degree to which carbamylation was reduced differed significantly between survivors and those who died. When looking specifically at average C-Alb from day 90 to study end, C-Alb increased among patients, whereas it continued to fall in controls. Thus, a single baseline carbamylation measure was inadequate in risk stratifying these patients on incident hemodialysis, whereas the longitudinal change (from either baseline or day 90) was able to do so. Additional studies are required to determine precisely what degree of change in carbamylation and over what period of time are associated with higher risk.

The mediators driving changes in carbamylation may include urea load, amino acid balance, and other pathways described above (^1–3). The traditional markers of urea clearance, such as Kt/V, urea reduction, and blood urea levels, did not differ between patients and controls in this population, however, suggesting that carbamylation is not simply a urea equivalent (further evidenced by the only modest correlation between C-Alb and urea level). Analogous to measuring glycated hemoglobin to determine long–term glucose control in diabetes mellitus, carbamylated proteins may more clearly depict time–averaged urea load than single blood urea measurements, which can fluctuate 40%–70% around dialysis sessions (²⁰). Moreover, carbamylation can also occur through nonuremic processes (e.g., myeloperoxidase-catalyzed oxidation of thiocyanate) (³) and can be exacerbated by amino acid deficiencies (²,⁹). C-Alb measures predicted outcomes, whereas the normalized protein catabolic rate in our study did not, again suggesting that C-Alb is not merely a surrogate for the established nutritional protein indicator. Carbamylation measures integrate several pathologic pathways and therefore, may serve as unique and powerful risk markers.

Baseline total albumin did not differ between patients and controls, but albumin at subsequent time points was higher in survivors (Table 1). Albumin also inversely tracked with changes in C-Alb (Supplemental Table 2). However, it is important to disentangle total serum albumin from C-Alb levels: albumin and C-Alb remained independent predictors of mortality, tests for collinearity were negative, the correlation between total albumin and C-Alb at any time point was modest (R²≤0.2), and C-Alb significantly improved models predicting mortality using established risk factors in ESRD, including total albumin measures (Table 3). Although some of the presumed mediators of a low serum albumin could also drive increased carbamyaltion (e.g., protein energy wasting), the distinction between total albumin and C-Alb likely reflects carbamylation measures’ unique integration of dialysis adequacy, urea load, and amino acid balance among other pathways. Notably, prior studies using C-Alb and the nonalbumin–based carbamylation measure homocitrulline yielded similar results (²,⁷).

It is conceivable that individuals who show elevated carbamylation levels or fail to adequately reduce their carbamylation levels, despite conventional medical optimization, may benefit from carbamylation-targeted interventions. Recently, we showed that amino acid therapy can reduce protein carbamylation in patients with ESRD without signs of malnutrition (⁹). This is being further investigated in an ongoing randomized clinical trial (NCT02472834). Alternatively, carbamylation could serve as an auxiliary marker of dialysis adequacy, guiding when changes to treatment are necessary. Additional clinical studies must test if carbamylation levels can be modulated through intervention and whether such changes affect clinically meaningful end points.

Our study has notable limitations. First, the sample size was modest (n=366), but our study is the first, to our knowledge, that measures serial carbamylation levels over time in an ESRD cohort. Second, although we were able to control for most of the commonly cited variables relevant to ESRD outcomes, our study design cannot exonerate the possibility of residual confounding, and the associations noted cannot conclude causation. Although we attempted to address the importance of residual renal function using serum filtration markers, gold standard measurements of residual kidney function would have strengthened our findings. Third, the study design required that individuals contribute more than one data point of carbamylation to the study, by definition meaning that all patients survived to at least day 90. In this sense, the patients were not a purely indiscriminate selection. However, excluding patients who died in the first 90 days of the study likely would select for healthier individuals with presumably lower carbamylation levels, ultimately reducing the patient-control differences that we observed.

To conclude, in this report, we describe that changes in carbamylation can strongly predict mortality in patients with ESRD, adding beyond single–time point measurements and traditional risk factors. Uncontrolled carbamylation measures could identify individuals on dialysis at increased risk for uremic complications—individuals who may stand to benefit from interventions targeted at reducing carbamylation, such as modified dialysis prescriptions or amino acid support.

Disclosures

Provisional applications for United States and international patents related to the contents of this manuscript have been filed by S.A.K., R.T., A.H.B., and their affiliated institutions. R.T. is a consultant to Fresenius Medical Care North America (Waltham, MA).