BACKGROUND: Both fasting plasma glucose (FPG) and HbA1c are recommended for the diagnosis of diabetes and prediabetes by the American Diabetes Association (ADA), and for diabetes by the World Health Organization. The ADA guidance is influential on clinical practice in many developing countries, including in the Caribbean and Latin America. We aimed to compare the prevalence and characteristics of individuals identified as having diabetes and prediabetes by FPG and HbA1c in a predominantly African ancestry Caribbean population. METHODS: A representative population-based sample of 1234 adults (≥25 years of age) resident in Barbados was recruited. Standard methods with appropriate quality control were used to collect data on height, weight, blood pressure, fasting lipids and history of diagnosed diabetes, and to measure fasting glucose and HbA1c. Those with previously diagnosed diabetes (n = 192) were excluded from the analyses. Diabetes was defined as: FPG ≥7.0 mmol/L or HbA1c ≥6.5%; prediabetes as: FPG ≥5.6 to <7mmol/L or HbA1c ≥5.7 to <6.5%. RESULTS: Complete data were available on 939 participants without previously diagnosed diabetes. The prevalence of undiagnosed diabetes was higher, but not significantly so, by HbA1c (4.9%, 95% CI 3.5, 6.8) vs FPG (3.5%, 2.4, 5.1). Overall 79 individuals had diabetes by either measure, but only 21 on both. The prevalence of prediabetes was higher by HbA1c compared to FPG: 41.7% (37.9, 45.6) vs 15.0% (12.8, 17.5). Overall 558 individuals had prediabetes by either measure, but only 107 on both. HbA1c, but not FPG, was significantly higher in women than men; and FPG, but not HbA1c, was significantly associated with raised triglycerides and low HDL cholesterol. CONCLUSION: The agreement between FPG and HbA1c defined hyperglycaemia is poor. In addition, there are some differences in the phenotype of those identified, and HbA1c gives a much higher prevalence of prediabetes. The routine use of HbA1c for screening and diagnosis in this population would have major implications for clinical and public health policies and resources. Given the lack of robust evidence, particularly for prediabetes, on whether intervention in the individuals identified would improve outcomes, this approach to screening and diagnosis cannot be currently recommended for this population.