Aims: To investigate concordance with medication, as assessed at baseline and at 1- and 2-year follow-up, and to examine factors associated with non-concordance in a UK-resident South-Asian population. Methods: Data from the UK Asian Diabetes Study were analysed. Concordance with medications was assessed and recorded at three time points during the study. Multiple logistic regression was used to investigate the factors associated with non-concordance; the associations of baseline factors with year 1 concordance and baseline plus year 1 factors with year 2 concordance. Results: Data for 403 patients from seven practices participating in the UK Asian Diabetes Study were analysed. The numbers of patients who were non-concordant were: 63 (16%) at baseline 101 (25%) at year 1; and 122 (30%) at year 2. The baseline-measured variables that were significantly associated with year 1 non-concordance included diabetes duration, history of cardiovascular disease, components of the EuroQol quality of life questionnaire, the EQ-5D score, and number of medications prescribed. In multivariable analyses, the most important determinant of year 1 non-concordance was baseline non-concordance: odds ratio 13.6 (95% confidence limits 4.7, 39.9). Number of medications prescribed for blood pressure control was also significant: odds ratio 1.8 (95% confidence limits 1.4, 2.4). Similar results were observed for year 2 non-concordance. Conclusions: Non-concordance with medications was common and more likely in people prescribed more medications. The current target-driven management of risk factor levels may lead to increasing numbers and doses of medications. Considering the high cost of medications and the implications of poor health behaviours on morbidity and mortality, further investigation of prescribing behaviours and the factors affecting patient concordance are required.