Point-of-care measurements of HbA(1c): simplicity does not mean laxity with controls.

Point-of-care HbA1c measurements (POC-A1Cs) have been adopted by many diabetes clinics to improve the quality of care provided to their patients (1). Herein, we show that reliability of this approach might be questioned. POC-A1Cs routinely used in the ambulatory section of our diabetes clinic was evaluated on 100 diabetic patients (type 1, n = 58; type 2, n = 42) attending the clinic from 1 October 2011 to 30 November 2011. Patients with abnormal hemoglobin traits or shortened erythrocyte life span were excluded. Blood-capillary samples were analyzed by POC-A1C (DCA Vantage; Siemens Medical Solutions Diagnostics, Cergy-Pontoise, France) and venous EDTA-anticoagulated blood specimens by the central laboratory high-performance liquid chromatography measurement (Tosoh HLC-723 GHb G8; BioSciences, Lyon, France). Both methods were certified (NGSP/Diabetes Control and Complications Trial [DCCT] and International Federation of Clinical Chemistry and Laboratory Medicine [IFCC]). Internal quality evaluation showed CVs consistently below 3%.

HbA1c values obtained from POC-A1C were found to be below those given by the central laboratory in 98% of the cases. POC-A1C values differed by a mean of −0.50 ± 0.28%. Central laboratory and the POC-A1C values were correlated, but the regression equation suggested a slight proportional bias (slope: 0.87) and a greater constant bias (intercept with y-axis: 0.37%). Bland-Altman statistics showed a significant correlation between the delta and the mean of HbA1c. The higher the HbA1c value was, the greater the discrepancy between both methods. To evaluate whether these discrepancies in HbA1c values can interfere with decision making, we assessed the possible POC-A1C–induced errors in categorization at the different HbA1c threshold levels used by the clinicians to modify hypoglycemic treatment. If the therapeutic HbA1c objective was ≤6.5%, then 11% of the population was incorrectly considered in the target by POC-A1C. This proportion of misclassification increased to 24% when the therapeutic target was ≤7% and decreased thereafter (≤7.5%, 12%; ≤8.0%, 8%). The higher misclassification rate observed for a 7% threshold is due to the fact that the proportion of patients around this value is especially high in our unselected cohort (HbA1c median: 7.28%). This real-life analysis differed from bench tests, which are usually performed to validate POC-A1C methods (2). Similar tendencies to an underevaluation of HbA1c by POC methods have been noted already by Holmes et al. (3) and by Twomey et al. (4) in the context of the U.K. “pay-for-performance program.” At the time of the current study, no sign of a possible drift in HbA1c determination was given by external quality-control procedures. One cannot minimize the clinical relevance of this transitory drift observed with the POC-A1C device. The solution for maintaining routine POC-A1C use involves every participant in the chain. First, lot-to-lot stability must be improved and controlled by the manufacturer as already suggested by Little et al. (5). External quality-control procedures should be more frequent and reactive. Clinicians should be aware of any discrepancies between POC-A1C and central laboratory values and, if necessary, carry out a local audit as we did. Finally, it should be dangerous to rely only upon POC-A1C to evaluate the quality of long-term glucose control in diabetic patients. Measurement of HbA1c by laboratory method should be performed at least once a year.