Diagnostic performance of using one- or two-HbA1c cut-point strategies to detect undiagnosed type 2 diabetes and impaired glucose regulation within a multi-ethnic population

Abstract

We compared test performance and cost per case for strategies detecting diabetes on the oral glucose tolerance test (OGTT) using either (a) glycated haemoglobin (HbA1c) ≥ 6.5% (48 mmol/mol) or (b) two HbA1c thresholds where the first cut-point 'rules out' and the second 'rules in' diabetes. HbA1c values in between the thresholds require confirmatory glucose testing for diagnosis. We conducted an analysis of adults aged 40-75 years from the Leicester Ethnic Atherosclerosis and Diabetes Risk (LEADER) cohort (Leicester, UK), from 2002 to 2008, who underwent oral glucose tolerance testing (OGTT) and HbA1c testing. From 8696 individuals (mean age 57.3 years, 73% white Europeans (WE) and 27% South Asians (SA)), HbA1c ≥ 6.5% produced sensitivity of 62.1% for detecting diabetes in WE and 78.9% in SA. Using two selected thresholds, HbA1c ≤ 5.8% (rule-in, 40 mmol/mol) and HbA1c ≥ 6.8% (rule-out, 51 mmol/mol) produced high sensitivity/specificity (> 91.0%) for detecting diabetes, however, 28.8% of the cohort with HbA1c 5.9%-6.7% required a subsequent glucose test. The two cut-point threshold produced a lower cost per case of diabetes detected in WE, compared to HbA1c ≥ 6.5% of £38.53 (1.89 to 86.81) per case, but was more expensive in SA by £84.50 (69.72 to 100.92) per case. Using a risk score to determine HbA1c testing, the same costs per case became £63.33 (23.33 to 113.26) in WE and £69.21 (55.60 to 82.41) in SA. Using a two-threshold strategy may have some benefits over a single cut-point; however, 28.8% of individuals required two blood tests.