Determining the Accuracy of Diabetes Tests

A meta-analysis on the accuracy, sensitivity, and specificity of the index tests for early detection in type 2 diabetes.  

Early detection of diabetes through screening may provide benefits including leading to the lesser or delayed incidence of microvascular and macrovascular complications of the eyes, kidneys, nerves, feet, and heart. Most diabetes and prediabetes screening across the globe involves questionnaires/risk scoring tools and tests involving fasting plasma glucose, HbA1c, and random blood glucose. The ADA guidelines adopted the HbA1c of 6.5% as the cut off for type 2 diabetes mellitus (T2DM). However, there is little information regarding the accuracy of the diagnostic tests for T2DM and prediabetes.  

The purpose of this meta-analysis study is to determine the accuracy, including the sensitivity and specificity, of diabetes screening in previously undiagnosed individuals. Data from this study was used for the early detection of T2DM and prediabetes. Studies that included diagnostic accuracy of blood glucose tests in patients 18 years and older for T2DM and prediabetes were included. Tests of interest included random plasma glucose, fasting plasma glucose, HbA1c, and post-prandial glucose. The 75-gram oral glucose tolerance test was used as a reference. Studies that were commentary, editorial, included pediatrics populations, pregnant women with T2DM, patients with T1DM, or conducted in animals were excluded. The studies’ sensitivity and specificity, positive and negative likelihood ratios, and 95% confidence interval were determined.  

A total of 37 articles were utilized in the meta-analysis after removal of duplicates and abstract screening. Of the 37 articles, 21 studies assessed only HbA1c tests, nine studies assessed fasting plasma glucose, four studies evaluated both HbA1c and fasting blood glucose, two studies assessed fasting capillary glucose, and 1 study assessed random capillary blood glucose. The 37 studies comprised 107,534 participants for diabetes, 39,846 participants for diabetes and prediabetes, and 667 for prediabetes alone. The vast majority at 44% (n=16) of the studies used WHO guidelines for diagnosing diabetes and prediabetes, while 42% (n=15) used ADA guidelines, and 14% (n=5) used both WHO and ADA guidelines.  

Seventeen studies assessed the accuracy of HbA1c at 6.5%, which showcased a sensitivity of 50% (95% CI: 0.417-0.588), a specificity of 97% (95% CI: 0.953-0.984), a positive likelihood of 18.328, and a negative likelihood of 0.512. Ten studies assessed the use of fasting plasma glucose at 126 mg/dL, which showed a sensitivity of 59.4% (95% CI: 46.6-71), a specificity of 98.8% (95% CI: 96.5-99.6), a positive likelihood of 47.825 (95% CI: 19.19-119.73), and a negative likelihood of 0.411 (95% CI: 0.305-0.555). Based on the sensitivity and specificity profile, the optimal cutoff value for HbA1c for diagnosing purposes was estimated at 6.03%, with the pooled sensitivity at 74% (95% CI: 68-79) and specificity at 87.2% (95% CI: 82-91). This cut off for HbA1c was deemed as moderate quality per GRADE determination. Additionally, the optimal cut-off for fasting plasma glucose for diagnosis was 104 mg/dL, with a pooled sensitivity of 82.3% (95% CI:  74.6-88.1) and specificity of 89.4% (95% CI: 85.2-92.5).  

This meta-analysis provided an overview of the accuracy of the diagnostic tests for diabetes. It concluded that for detection of diabetes in previously undiagnosed populations, the optimal threshold for HbA1c was 6.03% and 104 mg/dL for fasting plasma glucose. This study comprised 37 studies, including a large population size with varying ethnic backgrounds. However, it excluded studies that were non-English language, and no indirect comparisons amongst the different index text were established. Further studies on the relationship between ethnicity, economic background, and diagnostic test cutoffs should be conducted. Although the diabetes cutoffs are defined per guidelines, lowering the threshold for higher sensitivity for diagnosis screening may be considered. 

Practice Pearls: 

• A fasting plasma glucose of 126 mg/dL showed a sensitivity of 59.4% and a specificity of 98.8%. 
• An HbA1c of 6.5% had a sensitivity of 50% and a specificity of 97%. 
• Optimal screening and cutoff for T2DM include HbA1c of 6.03% and fasting plasma glucose of 104 mg/dL. 

Barry, Eleanor, et al.”  “Efficacy and Effectiveness of Screen and Treat Policies in Prevention of Type 2 Diabetes: Systematic Review and Meta-Analysis of Screening Tests and Interventions” BMJ, 2017, p. i6538., doi:10.1136/bmj.i6538. 

Kaur, Gunjeet, et al.”  “Diagnostic Accuracy of Tests for Type 2 Diabetes and Prediabetes: A Systematic Review and Meta-Analysis” Plos One, vol. 15, no. 11, 2020, doi:10.1371/journal.pone.0242415. 

Shana Indawala, PharmD Candidate, University of South Florida Health, Taneja College of Pharmacy