Feasibility of hexokinase, glucose oxidase–peroxidase and point-of-care glucometer methods for routine blood glucose estimation: a comparative study from a tertiary care hospital in Nepal

Accurate blood glucose estimation is essential for diabetes management. This study evaluated the analytical performance of a point-of-care glucometer compared with the laboratory-based GOD–POD and reference hexokinase methods in a tertiary care hospital in Nepal. This was a hospital-based cross-sectional study conducted among 205 participants attending Tribhuvan University Teaching Hospital, Nepal. Venous serum glucose was measured using the hexokinase and GOD–POD methods, while capillary glucose was measured using the On Call Extra glucometer. Method agreement was evaluated using Bland–Altman analysis and Passing–Bablok regression. Mean glucose concentrations measured by the hexokinase, GOD–POD, and glucometer were 123.58 ± 53.12, 123.98 ± 51.99, and 117.18 ± 51.45 mg/dL, respectively. The glucometer consistently underestimated glucose concentrations compared with both laboratory methods ( p < 0.01). Bland–Altman analysis showed excellent agreement between the GOD–POD and hexokinase methods (mean bias: 0.40 mg/dL; limits of agreement: −20.1 to 20.9 mg/dL), whereas the glucometer demonstrated greater negative bias. Passing–Bablok regression confirmed no significant proportional bias between the GOD–POD and hexokinase methods but identified a small constant bias for the glucometer. The point-of-care glucometer demonstrated acceptable overall clinical agreement with laboratory reference methods but exhibited clinically significant range-dependent systematic bias, underestimating glucose at higher concentrations and overestimating during hypoglycemia. Consequently, unexpected or extreme capillary glucometer values require confirmatory laboratory testing. Furthermore, the laboratory GOD–POD method showed excellent analytical comparability to the reference hexokinase procedure, offering a reliable, cost-effective alternative for routine testing in resource-limited settings. Not applicable.