Predictive value of nephelometric and high-performance liquid chromatography assays of urine albumin for mortality in a high-risk Aboriginal population.

BACKGROUND Urine albumin assays by high-performance liquid chromatography (HPLC) yield greater values than immunoassays at lower albumin levels. We compared predictive values of albumin-creatinine ratios (ACRs) by these 2 techniques for mortality in Aboriginal people. STUDY DESIGN & SETTING This was a longitudinal study of 741 adults in a remote Aboriginal community who participated in a baseline health survey between 1992 and 1998 at ages ranging from 18 to 84 years (mean, 34 years). All natural deaths were documented on follow-up until 2006. Urine albumin concentrations were measured simultaneously by using both nephelometric and HPLC techniques on baseline urine samples retrieved from -70 degrees C storage, as well as creatinine concentrations, and ACRs were derived. Age- and sex-specific tertiles of ACR were compiled. Cox regression analyses were used to evaluate the predictive value of ACR for natural deaths by ACR tertiles and again by z score changes in ACRs as continuous variables. RESULTS Participants were followed up for a median of 11 years, during which a total of 119 natural deaths were documented. ACRs on baseline urine samples were greater by HPLC than immunoassay at lower ACR ranges, but were fairly concordant at levels greater than 100 mg/mmol. Levels of ACR by both techniques were strong predictors of death, but correlations of death with ACR tertiles and with ACR levels on a continuum were similar for the 2 techniques. LIMITATIONS The age- and sex-specific tertiles used might introduce some risk of bias in the assessment of predictive value. In addition, assays were performed on urine after more than 10 years of cold storage. CONCLUSION Despite different absolute values, this study did not show that ACR level by either technique was superior in predicting deaths.