RÉSUMÉ
BACKGROUND: Diabetes mellitus (DM) is common among tuberculosis (TB) patients and often undiagnosed or poorly controlled. We compared point of care (POC) with laboratory glycated haemoglobin (HbA1c) testing among newly diagnosed TB patients to assess POC test accuracy, safety and acceptability in settings in which immediate access to DM services may be difficult. METHODS: We measured POC and accredited laboratory HbA1c (using high-performance liquid chromatography) in 1942 TB patients aged 18 years recruited from Peru, Romania, Indonesia and South Africa. We calculated overall agreement and individual variation (mean ± 2 standard deviations) stratified by country, age, sex, body mass index (BMI), HbA1c level and comorbidities (anaemia, human immunodeficiency virus [HIV]). We used an error grid approach to identify disagreement that could raise significant concerns. RESULTS: Overall mean POC HbA1c values were modestly higher than laboratory HbA1c levels by 0.1% units (95%CI 0.1-0.2); however, there was a substantial discrepancy for those with severe anaemia (1.1% HbA1c, 95%CI 0.7-1.5). For 89.6% of 1942 patients, both values indicated the same DM status (no DM, HbA1c <6.5%) or had acceptable deviation (relative difference <6%). Individual agreement was variable, with POC values up to 1.8% units higher or 1.6% lower. For a minority, use of POC HbA1c alone could result in error leading to potential overtreatment (n = 40, 2.1%) or undertreatment (n = 1, 0.1%). The remainder had moderate disagreement, which was less likely to influence clinical decisions. CONCLUSION: POC HbA1c is pragmatic and sufficiently accurate to screen for hyperglycaemia and DM risk among TB patients.