西村 基

BACKGROUND & AIMS: Serum retinol (ROH) is commonly used for population level assessment of vitamin A status. High-performance liquid chromatography (HPLC) is considered most accurate method for measuring ROH. However, with the technical difficulty of using HPLC for routine assays, serum retinol-binding protein (RBP) measured by immunological assays is expected to be a surrogate marker for ROH, with reports of a close correlation between serum RBP and ROH. Nevertheless, RBP is not commonly tested to assess vitamin A status with concerns over RBP alterations under various physiopathological conditions. Thus, we reappraised the extent to which RBP could be used as a surrogate marker in representative disorders that alter serum RBP levels. As a related marker, diagnostic utility of transthyretin (TTR) was also evaluated. METHODS: To evaluate the reliability of ROH and RBP assays, specimen stability was assessed in terms of (1) storage at 25, 4, -20, and -80 °C for 1-28 days, (2) five-cycle freeze-thawing, and (3) fluorescent light exposure for 1-14 days. Sources of variation (sex, age, body mass index [BMI], and drinking habits) and reference intervals for ROH, RBP, and TTR were determined in 617 well-defined healthy individuals. To investigate the influence of disorders that affect serum RBP, patients with five diagnostic groups were enrolled: 26 with chronic kidney disease (CKD); 13 with various malignancies in advanced stages (AdM), 12 with acute bacterial infections (ABI), 6 with liver cirrhosis (LC), and 26 with simple obesity (BMI ≥ 27 kg/m2). RESULTS: The stability of RBP and ROH in serum was confirmed under all conditions. In healthy individuals, serum ROH, RBP, and TTR were appreciably high in males with a slight increase in proportion to age and BMI. The major-axis regression line between RBP (x) and ROH (y) in healthy individuals was y = x, with a correlation coefficient of 0.986. In the LC, AdM, and ABI groups, similar strong correlations were observed; however, the regression lines were shifted slightly rightward from the healthy group line, indicating a positive bias in estimating ROH. Interestingly, the same analyses between TTR and ROH revealed similar strong linear relationships in all groups; however, the regression line of each group showed a leftward (opposite) shift from the healthy group line. Based on these observations, we developed a novel regression model composed of RBP and TTR, which gave much improved accuracy in estimating ROH, even under these pathological conditions. CONCLUSIONS: The perfect RBP-ROH correlation in healthy individuals indicates the utility of RPB as a surrogate marker for ROH. Nevertheless, under RBP-altered conditions, a slight overestimation of ROH is inevitable. However, when the TTR was tested together, the bias can be corrected almost perfectly using the novel ROH estimation formula comprising RBP and TTR.