Fasting plasma glucose shows higher diagnostic accuracy than HbA1c for undiagnosed diabetes in adults

This systematic review and meta-analysis assessed the diagnostic performance of glycemic markers for identifying undiagnosed diabetes. The analysis included data from population-based studies conducted across continents involving 276,203 adults without previously diagnosed diabetes. The primary exposure metrics were HbA1c, fasting plasma glucose (FPG), and 1-h plasma glucose (1hPG). The comparator was the 2-h plasma glucose standard, defined as a value greater than or equal to 11.1 mmol/L. The primary outcome measured diagnostic performance, specifically sensitivity, specificity, and optimal cutoff values.

The analysis found that FPG demonstrated higher diagnostic accuracy than HbA1c. For FPG, the optimal cutoff was 6.3 mmol/L. At this threshold, the sensitivity was 73% with a confidence interval of 68% to 78%, and the specificity was 91% with a confidence interval of 89% to 94%. In contrast, HbA1c showed lower diagnostic accuracy than FPG. The optimal cutoff for HbA1c was 47 mmol/mol, which corresponds to 6.5% with a range of 6.0% to 6.6%. At this level, the sensitivity was 62% with a confidence interval of 51% to 80%, and the specificity was 92% with a confidence interval of 77% to 96%.

Regional variations in optimal cutoff values were observed. FPG cutoff values ranged from 5.3 mmol/L in North America to 6.5 mmol/L in Asian countries. Similarly, HbA1c cutoff values ranged from 41 mmol/mol, equivalent to 5.9%, in Asian countries to 45 mmol/mol, equivalent to 6.3%, in North America. These findings suggest that uniform global thresholds may not fully reflect population-specific differences in diagnostic performance.

Safety and tolerability findings were not reported in the included studies. Adverse events, serious adverse events, discontinuations, and overall tolerability were not reported. The certainty of evidence (CoE) was low. This low certainty indicates that further high-quality studies are needed to refine diagnostic performance and examine factors influencing accuracy. The funding or conflicts of interest were not reported.

These results suggest that FPG is a more reliable diagnostic marker than HbA1c for detecting undiagnosed diabetes in this population. Clinicians should consider that uniform global thresholds may not fully reflect population-specific differences. The certainty of evidence was low, and the reliability of HbA1c compared to FPG should not be overstated. Questions remain regarding the factors influencing accuracy and the need for further high-quality studies to refine diagnostic performance.