Sensitivity and specificity of three diabetes diagnostic criteria in people with non-alcoholic fatty liver disease (NAFLD) and otherwise healthy people: Analysis of NHANES III.
Level 4 - case-series / case-control
Cross-sectional diagnostic accuracy analysis without an independent reference standard
PubMed 37321876 · doi:10.1016/j.pcd.2023.06.003
What was done
Authors conducted a cross-sectional diagnostic accuracy analysis using data from 3,652 participants (mean age 55.6 years, 49.4% men) from NHANES III (1989–1994), of whom 673 (18.4%) had non-alcoholic fatty liver disease (NAFLD). Type 2 diabetes mellitus (T2DM) was defined using three criteria: postprandial glucose (PPG) ≥ 200 mg/dL, fasting plasma glucose (FPG) ≥ 126 mg/dL, and HbA1c ≥ 6.5%. Pairwise sensitivity and specificity across the six combinations of these definitions were estimated in individuals with and without NAFLD. Poisson regressions evaluated whether individuals with NAFLD were more likely to meet two diagnostic criteria while missing the third.
What was found
Compared to individuals without NAFLD, those with NAFLD had lower specificity across pairwise comparisons except when PPG was evaluated against HbA1c (98.28% [95% CI: 97.73%–98.72%] in non-NAFLD vs 96.15% [95% CI: 94.28%–97.54%] in NAFLD). In people without NAFLD, FPG sensitivity was slightly higher than HbA1c (64.62% [95% CI: 55.75%–72.80%] for FPG vs 56.58% [95% CI: 44.71%–67.92%] for HbA1c). Individuals with NAFLD were significantly more likely to be diagnosed by FPG and PPG yet not by HbA1c (prevalence ratio = 2.15; p = 0.020).
Why it matters
Screening for type 2 diabetes in patients with NAFLD using HbA1c alone may underdiagnose cases that are detected by fasting or postprandial blood glucose criteria.
Limits
The analysis relies on cross-sectional data collected between 1989 and 1994, which may not reflect contemporary diagnostic assay performance. There was no independent gold standard for diabetes, forcing pairwise comparisons among the index tests. Diagnoses were based on single-timepoint measurements rather than the repeat testing required in clinical practice.
Cited by
- supports A fasting blood glucose level over 125 mg/dL is one of the diagnostic criteria for diabetes.