Background: Sitagliptin, the first of a new class of dipeptidyl peptidase-4 (DPP-4)-inhibitory oral antihyperglycaemic drugs (OHDs), was introduced in Japan in December 2009. In April 2010 a safety alert was issued regarding the risk of serious hypoglycaemic events, and prescribers were recommended to reduce the dose of sulfonylurea (i.e. glimepiride, glibenclamide [glyburide] or gliclazide) in patients receiving a combination of sulfonylurea and sitagliptin. Objective: A propensity score-matched cohort study was performed using Japanese pharmacy prescription receipt data for OHDs in order to confirm reported changes in OHD prescription behaviour for patients receiving sitagliptin before and after the safety alert. Methods: Prescription data from about 6,500 medical institutions throughout Japan during December 2009 to 31 December 2010 were randomly collected from 300 pharmacies, covering 82,064 patients with 629,955 prescriptions for OHDs. Patients who had received a sulfonylurea and sitagliptin (1,788 patients/3,576 prescriptions) before the safety alert were designated as the DPP-4 group. Patients who had received a sulfonylurea but not sitagliptin (30,963 patients/61,926 prescriptions) before the alert were designated as the non-DPP-4 group. Propensity score matching was employed to match baseline characteristics, such as age, sex, type of OHD, metformin use, type of prescribers period for measuring baseline period and type of prescribers' institutions, for 1,783 patients from each group. In the matched cohort, logistic regression analysis was conducted to compare prescription trends before and after the alert. The primary outcome measure of this study was dose of glimepiride, glibenclamide or gliclazide prescribed for DPP-4 and non-DPP-4 patients. Results: In the propensity score-matched cohort, the proportion of glimepiride dose >2 mg of DPP-4 patients was reduced from 45.8 % in Period 1 (before the alert) to 37.5 % in Period 2 (after the alert) (odds ratio [OR] 0.71; 95 % CI 0.579-0.870), whereas in the case of non-DPP-4 patients the proportion was changed from 28.9 % to 29.5 % in the matched cohort (OR 1.03; 95 % CI 0.868-1.215). The mean prescribed glimepiride dose in DPP-4 patients was also reduced from 2.79 ± 1.81 mg in Period 1 (before the alert) to 2.38 ± 1.71 mg in Period 2 (after the alert) [p < 0.0001], whereas the corresponding change in the case of non-DPP-4 patients was from 2.01 ± 1.56 mg to 2.01 ± 1.54 mg (p = 0.94). The difference between the mean prescribed doses in the two groups was statistically significant in both periods. Similar trends of prescription pattern changes were seen for glibenclamide and gliclazide. The reduction of prescribed sulfonylurea dose in DPP-4 patients following the safety alert coincided with a decrease of adverse event reports. Conclusion: Our results indicate that propensity score matching to control for baseline characteristics of individual patients and prescribers is a useful approach to avoid selection bias and confounding effects in evaluating the influence of an event on prescription behaviour. This case-matched study indicated that sulfonylurea prescription behaviour changed significantly after the sitagliptin safety alert. There was a significant reduction in sulfonylurea dose after the alert in DPP-4 patients, but not in non-DPP-4 patients. Our findings should be helpful for assessing and improving the effectiveness of other regulatory safety alerts. © 2013 Springer International Publishing Switzerland.