Influence of renal biomarker variability on the design and interpretation of occupational or environmental studies

Abstract

Objectives: To quantify and identify sources of within- and between-subject variability of microalbumin, N-acetyl-beta-D-glucosaminidase (NAG) and alanine aminopeptidase (AAP), three biomarkers used for early detection of renal injury, and to assess the consequences of this variability for the design and power of epidemiological studies.

Methods: Urinary excretion of microalbumin, NAG, AAP and creatinine as well as blood pressure (BP) were measured three times over a 2-year period among 142 healthy male workers. To minimise physiopathological and analytical sources of variation, standardised methods were used for urine collection and assays, and severe exclusion criteria were applied. At the first and third examinations, subjects completed the same questionnaire, providing information about their personal characteristics, tobacco and alcohol consumption, and health. A linear mixed model was used to estimate the within- and between-subject variance components and to analyse the relation between subjects' characteristics and the biomarkers.

Results: No change in the mean value of any of the biomarkers was observed over the 2-year period. Intra-class correlation coefficients between repeated measurements were 0.53, 0.57 and 0.56 for microalbumin, NAG and AAP, respectively; the between-subject variance was slightly higher than the within-subject variance. Subjects' age, BP, body mass index and smoking and drinking habits explained 7.2%, 12.5% and 4.2% of the total variance of microalbumin, NAG and AAP, respectively.

Conclusions: In this healthy population of male workers, day-to-day differences in biomarker values appeared to be nearly as great as differences between subjects. The within-subject variance of these biomarkers is not high enough to justify systematic repeated measurements in epidemiological surveys. But, in some situations where the number of subjects is limited, measuring the subjects twice may improve study power by reducing the total variance by about 25% for each biomarker. Taking the above covariates into account would slightly improve study power and the accuracy of parameter estimates for NAG, but would add little to the analysis of microalbumin and AAP.