Aims/hypothesis <p>We aimed to evaluate stages in the development of type 2 diabetes by combining fasting plasma glucose (FPG) with 1 h plasma glucose (PG) or HbA_1c levels.</p> Methods <p>Using data of 1174 middle- and older-aged Brazilian adults from the Rio Grande do Sul centre of the ELSA-Brasil cohort, 2017 to 2024, we developed stages based on previously recommended thresholds of mild (FPG: 5.6–6.0 mmol/l; 1 h PG: 6.7–8.5 mmol/l) and moderate (FPG: 6.1–6.9 mmol/l; 1 h PG: 8.6–11.5 mmol/l) hyperglycaemia. Similarly, we developed stages combining FPG levels with mild (39–41 mmol/mol; 5.7–5.9%) or moderate (42–46 mmol/mol; 6.0–6.4%) HbA_1c levels. We additionally assessed these staging schemas requiring an initial clinical score cutoff of ≥10% probability of developing diabetes in 10 years before proceeding to laboratory testing. We calculated baseline insulin responsiveness (insulin secretion-sensitivity index-2, ISSI-2) and the frequency of an estimated high risk of complications (Whitehall subphenotype clusters) across stages. We estimated relative risks of developing diabetes ascertained by self-report and glucose measurements after a 5.31 (0.44) year follow-up using robust Poisson regression with an offset for person-years. We calculated the prognostic properties of staging schemas in the prediction of diabetes.</p> Results <p>The FPG/1 h PG schema stratified participants into three stages of decreasing ISSI-2 levels and increasing frequency of high risk for complications. The relative risk of incident diabetes increased progressively up to stage 3 in crude and adjusted models (15.4 [95% CI 6.1, 38.8] and 11.4 [4.5, 18.0]), respectively. This schema achieved 89.1% sensitivity and 53.7% specificity in the prediction of incident diabetes. When staging was applied with the clinical score, specificity improved (60.3%), and the need for laboratory testing decreased by 27.1%. The lower frequency of altered HbA_1c allowed for two-stage FPG/HbA_1c schemas, and produced lower relative risks and poorer prognostic properties compared with FPG/1 h PG schemas. Performance improved when combined with the clinical score. FPG/HbA_1c staging schemas were superior to the binary categorisation of prediabetes (intermediate hyperglycaemia).</p> Conclusions/interpretation <p>FPG/1 h PG schemas resulted in a nuanced stratification of intermediate hyperglycaemia, with superior prognostic properties compared with FPG/HbA_1c schemas. Applying a predictive clinical score in staging reduced laboratory testing and false positives.</p> Graphical Abstract <p></p>