Agreement and bias of three capillary blood glucose meters compared with venous plasma glucose across clinically relevant glycemic ranges in primary care: a multicenter cross-sectional study

<title>Abstract</title> <p>Background Capillary blood glucose meters remain widely used in diabetes management despite the increasing adoption of continuous glucose monitoring systems. Accurate glucose measurements are essential for therapeutic decision-making, particularly in patients at risk of hypoglycemia. This study aimed to evaluate the agreement and bias of three commonly used capillary blood glucose meters compared with venous plasma glucose and to assess their performance across clinically relevant glycemic ranges in a real-world primary care setting. Methods We conducted a multicenter cross-sectional observational study in seven primary care centers. Patients with diabetes undergoing routine blood testing were opportunistically recruited. Simultaneous glucose measurements were obtained using three capillary blood glucose meters (Contour, Glucomen, and FreeStyle) and venous plasma glucose as the clinical comparator. Agreement was assessed using Pearson’s correlation coefficient, Bland–Altman analysis (mean bias and 95% limits of agreement), and mean absolute difference (MAD). An ISO-inspired tolerance criterion was calculated as ± 15 mg/dL for glucose values &lt; 100 mg/dL and ± 15% for values ≥ 100 mg/dL. Analyses were stratified into three glycemic ranges: &lt;100, 100–180, and ≥ 180 mg/dL. Results A total of 142 measurement episodes were recorded, of which 131 included complete data from all four methods. Correlations between devices and venous plasma glucose were high (r = 0.953–0.970). FreeStyle showed the closest agreement with venous plasma glucose, with a mean bias of − 2.67 mg/dL, a MAD of 8.08 mg/dL, and 96.9% of measurements within the predefined tolerance limits. Contour and Glucomen systematically overestimated glucose values, with mean biases of + 10.73 mg/dL and + 6.85 mg/dL, respectively. Differences were most pronounced in the &lt; 100 mg/dL range, where Contour showed a mean bias of + 14.10 mg/dL and only 55.2% of measurements within tolerance, compared with 96.6% for FreeStyle. Conclusions High correlation between blood glucose meters does not necessarily indicate clinical interchangeability. Systematic biases were identified, particularly at lower glucose concentrations, where inaccuracies may affect hypoglycemia detection and therapeutic decisions. Consistent use of the same device should be encouraged, and caution is warranted when switching between glucose meters in clinical practice. Trial registration Not applicable. This study was an observational diagnostic agreement study and did not involve a healthcare intervention requiring trial registration.</p>