Relationship between glycated hemoglobin and glucose concentrations in the adult Galician population: selection of optimal glycated hemoglobin cut-off points as a diagnostic tool of diabetes mellitus

Abstract

Aims/hypothesis: To analyze the relationship between glucose and glycated hemoglobin (HbA(1c)) in the adult Galician population, evaluate the use of HbA(1c) for the screening and diagnosis of diabetes, and calculate the diagnostic threshold required for this purpose.

Methods: We analyzed data on 2848 subjects (aged 18-85 years) drawn from a study undertaken in 2004 to assess the prevalence of diabetes in Galicia. For study purposes, diabetes was defined using the criteria recommended in 2002. Participants were classified into four glucose-based groups. The relationship between glucose and HbA(1c) was described using linear regression models, generalized additive models and Spearman's correlation. Diagnostic capacity was assessed, and optimal HbA(1c) cut-off points were calculated as a diabetes marker using the receiver operating characteristic curve.

Results: Prevalence of pre-diabetes, unknown diabetes and known diabetes was 20.86, 3.37 and 4.39%, respectively. The correlations between HbA(1c) and fasting glucose were higher than those obtained for HbA(1c) and glycemia at 2h of the oral glucose overload (0.344 and 0.270, respectively). Taking glucose levels as the gold standard, a greater discriminatory capacity was obtained for HbA(1c) (area under de cruve: 0.839, 95% confidence intervals: 0.788-0.890). Based on the study criteria, the optimal minimum and maximum HbA(1c) values were 5.9% and 6.7%, respectively.

Conclusions/interpretation: HbA(1c) did not prove superior to glycemia for diagnosis of diabetes in the adult Galician population, and cannot therefore be used to replace the oral glucose tolerance test for screening and diagnosis purposes. Indeed, determination of glucose is essential to verify the diagnosis in the majority of cases.