Reproducibility of the diagnosis of diabetes by the 1-hour oral glucose tolerance test is substantial Authors: Anne E. Sumner, Anika J. Shah, Ananya Bharadwaj, Grace G. Smith, Kauthrah Ntabadde, Bella S. Izere, Christopher W. DuBose, Ram Jagannathan, David B. Sacks Published: 2026 ABSTRACT Diagnostic reproducibility of diabetes was substantial (ҝ-statistic = 0.74, 95%CI: 0.58–0.90) in 154 African descent individuals with 1 h-OGTTs performed 10 ± 7d (mean ± SD) apart. Diabetes was diagnosed in 34 individuals with 65%(22/34) detected at both visits (non-switchers) and 35%(12/34) at one visit (switchers). With insulin resistance greater in non-switchers, 1 h-OGTT reproducibility may be better if insulin resistance present. Introduction The two-hour oral glucose tolerance test (2 h-OGTT) has been the diagnostic standard since 1979 for diabetes [1]. In 2024 the International Diabetes Federation (IDF) concluded to diagnose diabetes the 1 h-OGTT was less expensive and time-consuming and better detects cardiometabolic risk [2]. While reproducibility of the 2 h-OGTT for the diagnosis of diabetes is excellent [3,4], diagnostic reproducibility of 1 h-OGTT is unknown. The IDF focuses on five types of diabetes: Type 1, Type 2, pancreatogenic, gestational and malnutrition-induced [5]. Additional forms include monogenic and medication-induced diabetes. This heterogeneity represents a challenge in developing a universally applicable diagnostic test. Furthermore, type 2 diabetes (T2D), the most prevalent type, has three physiological subtypes: primarily insulin-resistant, primarily beta-cell insufficient and both types combined [6]. These three subtypes vary in presentation. For example, at diagnosis of T2D, hyperglycemia is more pronounced when the etiology is predominantly insulin resistance rather than beta-cell insufficiency [7]. To diagnose diabetes, the 2 h-OGTT requires a 2 h-glucose concentration of ≥200 mg/dL. However, the diagnostic threshold for the 1-hour OGTT is higher at ≥209 mg/dL [2]. We postulate that the higher threshold required by the 1 h-OGTT may favor diabetes detection in individuals with insulin resistance. Our goals for diabetes detection by the 1 h-OGTT were to determine: (1) reproducibility and (2) whether insulin resistance was associated with reproducibility. Methods Two 1 h-OGTT(Trutol75) were performed 10 ± 7d (mean ± SD) apart in 154 individuals (African-immigrants: 91%(140/154), African-Americans: 9%(14/154), male: 60%(92/154), age:40 ± 12y, range 20–68y, BMI:28.1 ± 4.8, range:19.3–43.6 kg/m2). Glucose, insulin, and C-peptide levels obtained at 0 and 1 h. The protocol was approved by the NIH Institutional Review Board (ClinicalTrials.gov Identifier: NCT00001853). Participants provided written informed consent. IDF criteria were used to diagnose: normal glucose tolerance (NGT) 1 h-glucose <155 mg/dL, intermediate hyperglycemia (IH): 1 h-glucose between ≥155 and ≤208 mg/dL, and diabetes: 1 h-glucose ≥209 mg/dL. Insulin resistance was assessed three ways: HOMA-IR [8] (fastingglucosemgdL×fastinginsulinuUml405), HOMA2-IR (formula is proprietary) [9] and 1 h-Matsuda Insulin Sensitivity Index [10] (10,000fastingglucosemgdL×fastinginsulinuUmL×glucosemgdL@60min×insulinuUmL@60min). Assays Blood for glucose measurement were collected in fluoride and oxalate tubes and immediately placed on ice. Samples for insulin and C-peptide remained at room temperature. After the OGTT ended, samples were taken to the laboratory for immediate processing. Plasma glucose was measured using hexokinase (Abbott Architect). Serum insulin and C-peptide measured by immunoassay (Roche Cobas 6000). Statistics Reproducibility was assessed by kappa-statistic (Excellent: 0.81–1.0, Substantial: 0.61–0.80, Moderate: 0.41–0.60, Fair to Poor: 0.00–0.40) [11]. Absolute, positive and negative agreement were determined. Comparisons were by unpaired t-tests. Data managed with Research Electronic Data Capture (REDCap) [12]. Analyses performed with STATAv19.0. Results At OGTT-1, prevalence of DM, IH and NGT were: 19%(29/154), 27% (42/154) and 54%(83/154), respectively (Fig. 1). At OGTT-2, prevalence of DM, IH and NGT were 18%(27/154), 29%(45/154) and 53% (82/154), respectively (Supplement Table). Out of 34 individuals diagnosed with diabetes, 64%(22/34) were diagnosed at both OGTTs (non-switchers) and 35%(12/34) were diagnosed with DM at one OGTT and either NGT (5/12) or IH (7/12) at the other OGTT (switchers). The overall absolute agreement between the 2 OGTTs was 92.2% (95% CI: 86.9%–95.5%) while the positive agreement for identifying diabetes was 78.6% (95% CI: 66.2%–87.3%), and negative agreement for identifying non-diabetes was 95.2% (95% CI: 91.9%–97.3%). Cohen’s kappa for diabetes classification was 0.74 (95% CI: 0.58–0.90). Non-switchers were more insulin-resistant than switchers, with higher HOMA-IR (4.15 ± 2.50 vs. 2.42 ± 1.52, P = 0.026), HOMA2-IR (2.13 ± 1.06 vs. 1.44 ± 1.83, P = 0.048), lower Matsuda Indexes (2.21 ± 0.93 vs.4.03 ± 2.242, P = 0.006) and higher 1 h-glucose levels (OGTT-1: 252 ± 52 vs. 187 ± 42 mg/dL, P < 0.001, OGTT-2: 262 ± 61 vs. 204 ± 32 mg/dL, P < 0.001). C-peptide levels at 1 h were >600 pmol/L in all participants. Discussion As the first study to assess the reproducibility of the 1 h-OGTT, we found substantial agreement for reproducibility of the diagnosis of diabetes (kappa-statistic: 0.74 (95% CI 0.58–0.90). In addition, the high negative agreement of 95.2% suggests that participants without diabetes were classified very consistently across visits. However, while the positive agreement among participants with diabetes was somewhat lower at 78.6%, it remained strong. Overall, these findings support good reproducibility of diabetes classification. However as non-switchers were more hyperglycemic and insulin-resistant than switchers, we suggest that reproducibility may be better in the presence of insulin resistance. To address this possibility, we analyzed insulin resistance using three different measures of insulin resistance: HOMA-IR, HOMA2-IR and the 1 h-Matsuda Index [8–10]. They provided similar results, but HOMA-IR was the least expensive. The formula for HOMA-IR is published, and available at no cost. In contrast, the formula for HOMA2-IR is proprietary and its use outside of university settings incurs a charge. The 1 h-Matsuda Index is the most expensive. As both HOMA formulas require measuring fasting levels of glucose and insulin, the 1 h-Matsuda Index also requires glucose and insulin concentrations at 1 h. A key strength of this investigation is that it was conducted in African descent cohort with 91% of the participants being African immigrants to the United States. No individuals of African descent were included in the meta-analyses the IDF used to formulate 1 h-OGTT recommendations [13,14]. As T2D characterized by beta-cell insufficiency rather than insulin resistance is especially common in Africans [7,15,16], we could observe the consequences of using the 1 h-OGTT across a wide range of insulin resistance. Our observations are only applicable to T2D. Other types were ruled out. All participants with diabetes had: C-peptides levels >600 pmol/L, no history of pancreatitis and BMI between 22.2–41.4 kg/m2. No participants were pregnant or taking medication known to affect glucose metabolism. However, without genetic testing monogenic diabetes cannot be ruled out. As the 1 h-OGTT is applied across populations, modifications which further increase its value merit studying. With a 1 h-threshold of 209 mg/dL for diabetes diagnosis, agreement between OGTTs was substantial but not excellent. As non-switchers were more hyperglycemic and insulin-resistant than switchers, reproducibility may be greater in populations in whom insulin resistance is the predominant cause of T2D. This hypothesis could be examined in existing cohorts [13]. If validated, there could be value in modifying diagnostic recommendations to the 1 h-glucose threshold of 209 mg/dL to include degree of insulin resistance or body mass index as a proxy of insulin resistance or adding fasting glucose [17]. The goal is to build on the strength of the 1 h-OGTT and establish a universal, cost-effective, reliable, reproducible, and feasible testing paradigm for the early diagnosis of diabetes across populations. Supplementary Material