Identifying Optimal Sample Types and Decision Thresholds for the Urinary Albumin–Creatinine Ratio

A major aim of the study by Saydah and coworkers (1) described in this issue of Clinical Chemistry was to evaluate the prevalence of albuminuria in adults participating in the National Health and Nutrition Examination Survey, 2009–2010 as an estimate of the prevalence of increased urinary albumin excretion in the US population. As their data show, increased urinary albumin is a common disorder, with >3% of a cross-sectional sample of the US population showing an increase with repeat sampling. In addition, their evaluation of >5000 paired random and first morning urine samples has provided a large data set that bears on how sample type influences test results and how values for the albumin–creatinine ratio should be interpreted. Although clinical testing of albumin excretion in the urine has been applied widely for many years, there are still questions about how best to perform such testing and interpret the results (2). Additional data that characterize more fully the sources of pre- and postanalytical variation and better approaches for standardizing measurements are needed to help resolve these remaining issues. Several previous studies reported that urine samples obtained in the first morning void yield lower values than random samples collected throughout the day (3–5), and the first morning void sample has been observed to correlate more closely with results of a 24-h collection than results for a random urine sample (5). It is logical that random urine samples collected later in the day would have higher values than first-void samples, because urinary excretion of albumin usually is higher during the day. These diurnal variations in excretion rate may be related to changes in posture or exercise. Data from the large number of paired samples analyzed by Saydah et al. (1) indicate that random urine samples yielded values that are about 50% higher, near the decision threshold (30 mg albumin/g creatinine) applied in their study. The 95th percentile for a first-void sample was 28.2 mg albumin/g creatinine, compared with 46.5 mg albumin/g creatinine for a random sample. The one caveat in this quantitative comparison is that shipping and handling differed for the 2 sample types. Although urine albumin is considered stable under the storage and shipping conditions used, the additional transportation time of the first morning samples might have contributed slightly to their lower values. The finding of substantially lower values for the albumin–creatinine ratio for first-void samples provides support for some clinical guidelines, such as the Kidney Disease: Improving Global Outcomes (KDIGO) Clinical Practice Guideline for Evaluation and Management of Chronic Kidney Disease, which currently recommends following up an increased value obtained for a random sample with testing on a first morning void sample (6). A sequential testing strategy that uses a random sample as the initial sample is a logical clinical approach, in that it increases diagnostic sensitivity, with fewer false-negative results, relative to the use of a first-void sample. This approach also has practical value in that obtaining a random sample does not entail a patient making a repeat visit. The lower specificity for testing of random urine samples is addressed by guidelines recommending follow-up testing with a first-void sample. Saydah et al. report that fewer than half of the increased values obtained with random urine samples were confirmed in repeat testing with first-void samples. For a test with a high within-person imprecision of ≥30% (2, 3), repeat testing is important to confirm a persistent increase in the urinary excretion of albumin. This approach is advocated in standards of practice outlined by the American Diabetes Association (7), although these standards do not specify that follow-up testing should be performed on first-void samples. One might argue that the albumin–creatinine ratio obtained with a first-void sample is essentially a different test than the ratio obtained with a random urine sample and that a different decision threshold should be applied, analogous to that of morning and evening cortisol sampling. What is not clear is how that would improve diagnostic performance compared with current guidelines. Having a single decision threshold has the virtue of simplicity, but clinicians and laboratorians should be aware of the influence of sample type on values for albumin excretion. Many guidelines, such as those of the American Diabetes Association (7), recommend interpreting values for the albumin–creatinine ratio by using a single cutoff of 30 μg albumin/mg creatinine (equivalent to 30 mg albumin/g creatinine) for all adults. This cutoff value is a clinical decision threshold that is based on risk of increased progression of kidney disease, rather than on a population-determined reference limit. Use of a single decision threshold does not account for differences due to sex or race. The daily output of creatinine is known to differ not only between men and women but also between races (8–11). The lower creatinine output of women produces a smaller denominator for the albumin–creatinine ratio and thus higher mean values for women. Some studies have proposed different decision thresholds according to sex and race. Saydah and coworkers suggest that further evaluations of the population distributions of values for men and women would be of interest and be relevant to sex-specific decision thresholds. Some studies have observed a higher prevalence of increased albumin excretion for women when a single decision threshold is applied to both sexes, but a statistically significant difference between men and women in microalbuminuria was not apparent in the current study. Of additional concern in establishing reference intervals or decision thresholds for measures of urinary albumin is the lack of reference materials and standardization for the current assays of urinary albumin. There are ongoing efforts to address both of these problems (12, 13). Comparisons of results obtained with different methods generally have not found large relative biases (2, 13), but until better assay standardization is achieved, some uncertainty will remain about how accurately values obtained from different laboratories or from a specific study relate to true values. The entire issue of trying to define a decision threshold value, whether for all adults, for males and females, or for specific racial groups, may be partially misdirected, in that several studies have found that microalbumin is a continuous variable without a well-defined threshold value for defining risk for the progression of renal disease or proteinuria (14–16). Better recognition of albumin excretion as a continuous risk factor might be achieved by developing graded risk categories based on different intervals of albumin excretion rates, as is currently done for assessing cardiovascular risk with measures of C-reactive protein. Albumin excretion rate is firmly established as a tool in the clinical assessment of the onset and progression of renal glomerular disease, and the data of Saydah et al. (1) support current clinical guidelines that recommend repeat testing. Their data, however, also raise questions regarding the effects of sample type on the prevalence of albuminuria and what thresholds to use for interpretation. Until the clinical laboratory community can standardize the measurement of urine albumin and develop appropriate reference materials, some uncertainty will remain about the interpretation of albumin–creatinine ratios generated with different laboratory methods.