Understanding the effect of hyperglycaemia on tuberculosis control in a southern African setting: the impact of HIV and diabetes control. The DARTZ (Diabetes and Risk of Tuberculosis in Zambia) studies

An association between diabetes mellitus and tuberculosis has been established: systematic reviews suggest that individuals with diabetes are three times more likely to develop active tuberculosis disease and have nearly a two-fold increased risk of death from tuberculosis than individuals without diabetes. The context of these associations in southern Africa may differ from the rest of the world due to the high prevalence of HIV and the high prevalence of poorly controlled diabetes mellitus in this region. Most prior studies have measured glycaemia at the time of tuberculosis diagnosis as a proxy for diabetes. Physiological stress from acute TB infection can also cause hyperglycaemia, so we use the term hyperglycaemia to encompass both diagnoses until the two can be differentiated. We conducted a case-control study in Lusaka, Zambia to determine if HIV modifies any association between hyperglycaemia and active tuberculosis. We recruited 3,843 tuberculosis cases and 6,977 controls and found no evidence of an overall association between hyperglycaemia and active tuberculosis, though there was a significant effect modification by HIV: among individuals with HIV there was a positive association. A subset of cases from the case-control study were recruited to a cohort study to determine whether a dose-response relationship exists between hyperglycaemia and tuberculosis treatment outcome. We found no evidence of a relationship. A study of diagnostic accuracy used a further subset of cases from the case-control study to determine the diagnostic accuracy for diabetes mellitus of measures of hyperglycaemia used at the time of tuberculosis diagnosis compared to a reference standard. We found a low proportion of hyperglycaemia measured at the time of tuberculosis diagnosis was due to diabetes mellitus, particularly among individuals co-infected with HIV. Glycated haemoglobin showed no greater test accuracy for diabetes diagnosis at the time of tuberculosis diagnosis than fasting or random blood glucose tests.