In this prospective cohort study of patients with CKD, we set out to compare the prognostic performance of two commonly utilized circulating markers of protein carbamylation. Protein carbamylation has emerged as a novel, mechanistically driven risk factor for adverse clinical outcomes in CKD that is of particular interest as it appears to be modifiable [29,30,31]. Studies have used a variety of assays to assess carbamylation, but their predictive performance has never been compared, limiting the ability to summarize and contrast findings, and to optimally plan future studies. As expected, in the present study we found that C-Alb and HCit were each independently associated with a higher risk for all-cause mortality and CKD progression, corroborating several prior studies [19,20,21, 32]. We now also show that both biomarkers demonstrated remarkably similar risk associations to each other when utilized concomitantly. The direction and the order of magnitude of the point estimates for each of the markers and respective outcomes were consistent though not identical (most notably for the ESKD outcome). Nevertheless, the 95% confidence intervals around each of the point estimates overlapped considerably, highlighting their similarities. The two markers appeared to also provide similar modest incremental predictive information when added to base models that included standard risk factors for death or CKD progression. Such findings indicate that HCit and C-Alb perform similarly when predicting important clinical outcomes.
It is well reported that levels of C-Alb or HCit differ significantly between individuals who experience meaningful clinical outcomes and those who do not (e.g., mortality, CKD progression, cardiovascular events, even anemia and erythropoietin resistance) [1]. Furthermore, the incremental predictive value of these markers in addition to established risk factors has been shown in preliminary reports [20]. The similar performance of these two markers in our study suggests broad-based comparisons across studies using either marker are possible. Clinical outcome studies in CKD reporting on associations with either assay are likely providing similarly useful information on the risks of carbamylation burden. Importantly, free HCit is a commonly reported analyte measured on metabolomic platforms and the numerous metabolomic studies and databases currently containing HCit data could potentially be utilized to answer key questions in carbamylation science and advance the field rapidly. Future research is needed to determine if any of the differences between C-Alb and HCit observed in our study would result in meaningful differences if applied clinically (e.g., if any differences could influence prospective clinical decision-making).
Upon categorizing the study population into quartiles for each of the two different carbamylation biomarkers, the corresponding concomitant clinical laboratory data for the population appeared similar but not identical. While BUN levels were nearly identical, for example, in the highest C-Alb and HCit groups, the eGFR and proteinuria in these same groups differed with the high HCit group showing worse values (i.e., lower eGFR and higher proteinuria). Indeed, the small molecule HCit was more strongly correlated with eGFR than C-Alb; thus, perhaps HCit is a more sensitive marker of eGFR, and this may have driven some of our findings despite statistical adjustments for eGFR made throughout. Regardless, HCit’s predictive and prognostic performance in models, compared to C-Alb, did not appear different. The inconsistent adjusted associations between the carbamylation markers and other variables shown in Table 2 warrant additional study. While competitive glycation may impact carbamylation levels in diabetics [33], age or race and ethnicity differences are less readily explained. As these associations were not adjusted for multiple testing, they should be interpreted cautiously and require further validation.
In a small ancillary study of 45 hemodialysis patients, investigators reported that protein bound as well as “total” (free and protein bound) HCit levels correlated with measures of albumin carbamylation [22]. Our study corroborates such expected positive correlation between the different carbamylation measures yet provides several new insights. The clinical prediction modelling presented herein using both markers from the same time point is the greatest advancement. In addition to a significantly larger sample size from a distinct patient population (non-dialysis CKD), our study also demonstrates that free HCit measures alone correlated robustly to the carbamylation of the most abundant circulating protein, albumin. Free HCit is the form captured by metabolomic platforms such as the one employed in this study, and it does not require steps to cleave peptide bonds (required to measure total HCit), or sample dialysis and protein precipitation (used to remove free HCit and yield only protein bound measures) [34]. Free HCit measures, which are commonly available in existing metabolomic databases, performed remarkably similar to C-Alb. The significantly greater correlation we observed between HCit and C-Alb vs. the prior report may be a function of our use of free HCit alone, sample size, differences in patient populations, or differences in assay techniques.
While statistically correlated (correlation coefficient 0.64), the markers do not appear related to each other in a strict 1:1 manner. Moreover, each marker demonstrated strong correlations to BUN levels, though, again, not with a 1:1 relationship. It is interesting to recall that the correlation reported between serum glucose and glycated hemoglobin (HbA1c), the most widely used surrogate marker for time-averaged glucose to guide diabetes management, is 0.42–0.58 [35]. Even at such correlation strengths, HbA1c is instrumental in determining the long-term clinical risks of fluctuating glucose levels [36]. It is likely that carbamylated proteins with long half-lives similarly better depict time-averaged urea exposure than single blood urea measurements which fluctuate significantly in relation to recent diet, hydration, volume, circulatory status, medications, acute GFR changes, and catabolic state [37, 38]. Furthermore, carbamylation occurs through non-uremic processes (e.g., myeloperoxidase catalyzed oxidation of thiocyanate derived from diet and smoking) [15] and can be exacerbated by amino acid deficiencies from nutritional imbalance, protein-energy wasting, or other means [17]. The half-life of the respective markers’ un-carbamylated substrate (lysine and albumin), differ considerably on the order of hours for lysine and weeks for albumin. Nevertheless, the ongoing release of protein bound HCit to free HCit during protein degradation, may explain why the 2 markers appeared to provide similar information. Presenting HCit as a ratio to total lysine has also been proposed [14, 15], and we provide this analysis in the supplement to show overall results were not significantly changed using either HCit alone or as a ratio to lysine. The incremental prediction value added to the fully adjusted risk models was remarkably similar for either marker, potentially showing the application of these markers could confer similar utility despite any differences observed in their associations to clinical variables and outcomes.
To this end, we must acknowledge that well-conducted epidemiological studies have shown that traditional biomarkers such as eGFR and proteinuria, plus easily available clinical parameters, perform very well in terms of identifying CKD patients at high risk of future adverse outcomes (e.g., the baseline C-statistic using traditional markers to predict CKD progression was ~ 0.9 from a previously published study in CRIC) [39]. As the authors of this prior study conclude, it is unlikely that such a high C-statistic can be significantly improved on or even needs improving. Indeed, we observed only modest changes in C-statistics when carbamylation markers were added, but the key finding for this report is that the magnitude of change was similar for each marker. Beyond C-statistics, which can have several limitations [40], carbamylation markers can provide important pathophysiological information which can lead to novel therapeutic targets as well as monitoring of treatment effectiveness. Further, the NRI performance of the two markers appeared to be more clinically meaningful and similar across both.
Our results need to be interpreted in the context of their limitations. We were only able to compare single time point measurements of the biomarkers at baseline without subsequent longitudinal measurements. External validation ideally employing serial measurements to assess performance characteristics remains a future direction. Also, as HCit can exist in both protein-bound and free forms in biological systems, we could not account for relative differences from our metabolomic data that generated only free HCit results for this study. Some previous studies have used protein bound HCit as a biomarker for carbamylation and it is possible our findings may have differed if doing so in this study [14, 15]. Thus generalizability to other HCit based studies requires cautious interpretation based on the assay methods. Further, the objective of this study was to compare the performance of the two biomarkers, rather than provide an accurate estimate of the hazard ratios associated with each. Nevertheless, strengths of our study include the large sample size and use of the CRIC study, the largest prospective cohort study for CKD in the United States offering patients from diverse racial and ethnic backgrounds, and rigorous outcome ascertainment and covariable data.
- The Renal Warrior Project. Join Now
- Source: https://bmcnephrol.biomedcentral.com/articles/10.1186/s12882-024-03619-6