Estimating albumin to creatinine ratio from protein to creatinine ratio using same day measurement: validation of equation

Abstract Background and Aims According to the KDIGO, severity of chronic kidney disease is defined by glomerular filtration rate and albuminuria. Furthermore, these variables are used for estimating both cardiovascular and end-stage renal disease risk. However, albuminuria is not always available, whereas proteinuria is, notably because of the lower costs of proteinuria. Recently, Weaver & al. developed an equation that estimates urine albumin/creatinine ratio (ACR) from protein/creatinine ratio (PCR). For retrospective analyses in clinical research, it might be interesting to be able to estimate ACR from PCR. The objective of this paper is to assess the performance of Weaver & al. ‘s equation in our population. Method In a University hospital, we retrospectively analysed measurement of ACR and PCR obtained on the same day between May 2018 and March 2020. Different assays were considered to measure urine albumin, creatinine and protein (Roche Cobas from May 2018 to May 2019 and Abbott Alinity from May 2019 to March 2020). Only one (the first available) sample per patient was considered. Patients were then categorized according to the KDIGO classification (A1-A2-A3). We then compared categorization of patients according to KDIGO between measured and estimated ACR. Sensitivity, specificity, positive predictive value and negative predictive value of estimated ACR versus measured ACR were calculated considering two different thresholds: A1 versus A2/3 (ACR 30mg/g) or A1/A2 versus A3 (ACR 300mg/g). Results Comparison was done in 2633 and 2386 patients, with Cobas and Alinity, respectively. Median age was 63 (IQR 19) and 64 (IQR 19) years old, 43 and 41% were women and 74 and 78% were diabetic (albuminuria measurement is refunded for diabetic patients in Belgium) with Cobas and Alinity, respectively. Considering measured ACR, patients were categorized in A1, A2, A3 in 65,6%, 25,5%, and 8,8%, respectively with the Cobas assay. With the Alinity assay, the results were 64,2%, 25,5%; and 10,3%, respectively. Considering estimated ACR, patients were categorized in A1, A2, A3 in 64,7%, 25,7%, and 9,6%, respectively with the Cobas Assay. With the Alinity assay, the results were 62,5%, 25,8% and 11,7%, respectively. Regarding A1-A2 (30mg/g) threshold, sensitivity was 85% and 89%, specificity was 91% and 92%, positive predictive value was 83% and 85%, negative predictive value was 92% and 94% for Cobas and Alinity respectively. Regarding A2-A3 (300mg/g) threshold, sensitivity was 93% and 98%, specificity was 98% for both, positive predictive value was 86% and 87%, negative predictive value was 99% and 99,8% for Cobas and Alinity respectively. Values are presented in table 1. Conclusion We observed a good concordance between estimated and measured ACR. Particularly, no patient in category A3 with measured ACR was categorized A1 with the estimating equations. We also observed that concordance was good regardless of our available assays. Thus, negative predictive value was excellent. In conclusion, ACR should be measured when clinically needed, but an estimated ACR can reasonably be obtained from Weaver’s equation and may provide an accurate estimation for retrospective clinical research.