To our knowledge, this study provides the first assessment of bias, precision and accuracy of ePER determined from PCR in the first and second morning spot urine collections in kidney transplant recipients. Our data showed excellent correlation and uniform agreement below nephrotic-range proteinuria, and moderate bias, precision, and accuracy of ePER in both the first and the second morning spot urine collections at predicting 24-hour proteinuria. Furthermore, measures of agreement improved with an increase in proteinuria. Differences in the measures of agreement were not significantly different between first and second morning urine, except in low-grade proteinuria where the performance of ePER in the second morning urine was slightly better. This finding may have clinical utility given the fact that analysis of the second morning urine sample is more convenient in the outpatient settings. Finally, measures of agreement between ePER in the first and second morning urine and 24-hour proteinuria were similar across different histologic phenotypes of allograft injury.
Proteinuria has been associated with progressive kidney disease, graft loss and mortality in kidney transplant recipients (1-3). There have been several other studies linking proteinuria with allograft failure and death (22-24). In these analyses, the average adjusted relative risk for allograft failure for patients with proteinuria was 2.7, and the average adjusted relative risk of death was 1.98 (25). Moreover, posttransplant proteinuria is highly specific for transplant glomerulopathy, microcirculation inflammation, and de novo/recurrent glomerular disease and the prognosis of these specific disease processes is primarily determined by the associated degree of proteinuria (26). Thus, accurate assessment of proteinuria is necessary for prognostic as well as diagnostic purposes and may be a target for therapy.
Spot sample urine measurements have become a standard of care for the assessment of 24-hour proteinuria in kidney transplant recipients and KDIGO guidelines recommend using the PCR as an alternate to the 24-hour collection method (27). However, the validity of spot urine protein measurement in this population remains unclear. Most diagnostic accuracy studies evaluating data on PCR only reported correlation with 24-hour proteinuria, while several studies also reported on sensitivity and specificity of PCR using various cutoff values (8). High correlation does not imply good agreement between two methods of measurement, because it evaluates only the linear association of two sets of observations. The diagnostic accuracy studies have also examined the sensitivity and specificity of PCR in relation to 24-hour proteinuria. Both are statistical measures of the performance of a binary classification and as such, none of these measures gives the clinician information about quantitative accuracy of the test. To date, only one study evaluated bias, precision and accuracy of PCR measurements and 24-hour proteinuria in kidney transplant recipients (12). However, the authors did not provide information which urine sample was analyzed and weather spot urine and 24-hour urine were collected on the same day. This may have contributed to the suboptimal agreement between PCR and 24-hour protein excretion as previous studies had demonstrated that random spot urine protein measurements show poor correlation and poor agreement with 24-hour collections (13,28,29).
In our study, we used first and second morning void urine collections. It must be emphasized that, although these are spot urine collections, they are not random collections, because they are the first or second voids of the day. Previous data in CKD patients suggested that consistency in the timing of collections may improve the agreement between spot PCR measurements and 24-hour urine collections (29,30). Our data demonstrated excellent correlation between estimated and measured 24-hour proteinuria with sensitivity and specificity values 83% or greater, depending on urine sample and cutoff used. The sensitivity and specificity both increased with greater proteinuria and were similar for the first and second morning urine collections. For example, the optimal cutoff for PCR in the first morning urine was 27 mg/mmol for 24-hour proteinuria >150 mg/day and 26 mg/mmol in the second morning urine. At these cutoff levels the sensitivity was 83%, and the specificity was 86% for the diagnosis of proteinuria in the first and second morning urine. For diagnosing high-grade proteinuria >1g/day, the optimal cutoff values were 107 (sensitivity 93%, specificity 96%) and 105 mg/mmol (sensitivity 89%, specificity 96%) in the first and second morning urine, respectively. These sensitivity and specificity data are consistent with earlier reports (8). The discriminatory ability of PCR in first and second morning urine for different cutoff levels of 24-hour proteinuria remained similar in patients with allograft injury demonstrated in the preceding indication kidney biopsies.
Like for the estimation of GFR, one should know the absolute measures of agreement between ePER and 24-hour proteinuria (i.e., bias, precision and accuracy). Accuracy represents the most useful analysis for the clinician, since it takes into account both bias and precision by expressing how many estimates are dispersed within a given range of their respective measurements (31). Because day-to-day fluctuations in proteinuria have been reported to be as high as 37% (32), accuracy within 30% best provides the proportion of estimates not deviating from measured 24-hour protein excretion. In our study the accuracies within 30% ranged from 56% to 78% and were slightly better than those reported in the study of Akbary et al. (47% to 56%) (12). No significant differences in absolute measures of agreement between first and second morning urine collections were found, except in low-grade proteinuria where the performance of ePER in the second morning urine was better. In this regard, the second morning spot urine may be particularly relevant, because it is easier to collect, and probably represents as uniform and achievable way as possible to collect urine among outpatients. In addition, performance increased with an increase in proteinuria and was better in patients with decreased allograft function. This finding is important given the fact that major diagnostic (e.g., biopsy) or therapeutic (e.g., change in immunosuppression) decisions are most commonly indicated in patients with a decrease in kidney function, new-onset or worsening proteinuria.
Previous diagnostic accuracy studies did not provide information on allograft histology and on the type of urine protein excretion associated with different levels of proteinuria, which may influence the accuracy of PCR measurements. With an increase in proteinuria, urinary levels of individual proteins albumin and α-1 microglobulin were also increased. Nevertheless, α-1 microglobulin increased in parallel with albumin excretion only in patients with low to moderate proteinuria (i.e. < 1g/day), while high-grade proteinuria was mainly associated with an increase in albumin excretion. This is in line with previous observations, which showed that low-grade proteinuria and small increases in urinary albumin may result from proximal tubular damage where urinary albumin often increases in parallel with α-1 microglobulin (33). In those patients with marked glomerular pathology heavy proteinuria composed overwhelmingly of albumin is common, and thus the correlation of total urinary protein and albumin with lower molecular weight tubular proteins may be lost. In our study, the predominance of albuminuria and relatively lower amounts of tubular proteins may explain greater diagnostic accuracy of ePER in patients with high-grade proteinuria. Spot urine protein and albumin excretion were greater in patients with previous diagnosis of AMR and recurrent glomerular disease than in patients with T-cell rejection or other non-rejection findings. In contrast, α-1 microglobulin excretion was not significantly associated with different histologic phenotypes. These associations were similar for first and second morning urine samples. Overall, urinary protein profiles alone could not predict specific histological findings, which is in agreement with previous observations (33,34). Nevertheless, the number of individuals with preceding allograft biopsy and thus different histologic phenotypes of allograft injury was low; hence, these results should be interpreted with caution.
This study has some limitations that should be acknowledged. First, our study only included a Caucasian population and a deceased donor kidney source. This may limit external validity to other more diverse patient populations with a higher proportion of living donor allografts and non-Caucasians. Second, number of patients with nephrotic-range proteinuria was low and correlations between estimated and measured values were more consistent for urines with proteinuria below 3g/day. Therefore, 24-hour urine collection should still be needed for proteinuria quantification in patients with severe proteinuria. Third, surveillance biopsies were not performed and only a small number of patients with histologic data on preceding biopsies were included, making it difficult to comment on the exact performance of the ePER in association with different phenotypes of allograft injury. Finally, we do not have outcome data to determine which measure of proteinuria (i.e., ePER in the first or second morning urine sample, or 24-hour collection) is most strongly associated with transplant outcomes.