Diagnostic accuracy of tests for type 2 diabetes and prediabetes: A systematic review and meta-analysis.
Level 1 - systematic review of randomized trials
Systematic review and meta-analysis of diagnostic accuracy studies with a reference standard
PubMed 33216783 · doi:10.1371/journal.pone.0242415
What was done
This systematic review and meta-analysis (PROSPERO CRD42018102477) evaluated the diagnostic accuracy (sensitivity, specificity, positive and negative likelihood ratios) of screening tests (fasting plasma glucose, random glucose, and HbA1c) against a 75-gram Oral Glucose Tolerance Test (OGTT) or 2-hour post-load glucose reference standard in previously undiagnosed adults. Seven electronic databases were searched through March 9, 2020. Study quality was appraised using the QUADAS-2 tool. Bivariate meta-analysis and optimal cut-off analyses were conducted using Stata and R.
What was found
Across 37 included studies from 7,151 initial records: - For HbA1c at the standard 6.5% threshold (venous sample, n = 17 studies), pooled sensitivity was 50% (95% CI: 42% to 59%), specificity was 97.3% (95% CI: 95.3% to 98.4%), positive likelihood ratio was 18.32 (95% CI: 11.06 to 30.53), and negative likelihood ratio was 0.51 (95% CI: 0.43 to 0.60). - The calculated optimal cut-off for HbA1c in previously undiagnosed adults was 6.03%, yielding a sensitivity of 73.9% (95% CI: 68% to 79.1%) and specificity of 87.2% (95% CI: 82% to 91%). - The optimal cut-off for fasting plasma glucose was 104 mg/dL, with a sensitivity of 82.3% (95% CI: 74.6% to 88.1%) and specificity of 89.4% (95% CI: 85.2% to 92.5%).
Why it matters
Current standard screening with HbA1c at 6.5% misses approximately half of undiagnosed diabetes cases identifiable by OGTT in community settings. Lowering screening cutoffs for HbA1c and fasting plasma glucose substantially improves case detection.
Limits
The total participant sample size across the 37 included studies is not reported in the abstract. Quantitative accuracy metrics for prediabetes and random glucose tests are not provided in the abstract despite being study aims. QUADAS-2 quality assessment results and potential heterogeneity across populations or assay methods are not detailed in the abstract.
Cited by
- supports Hemoglobin A1C testing misses more than half of prediabetes and diabetes cases.