BackgroundDiabetes mellitus (DM) leads to multiple complications, including severe hypoglycaemia events (SHEs). SHEs can impact a patient’s quality of life and compliance and may directly result in additional costs to the health care system. The aim of this review was to evaluate the risk of severe hypoglycaemia in patients with type 1 (T1) and 2 (T2) DM as observed in everyday clinical practice for various drug regimens.MethodsWe conducted a systematic review of observational (retrospective or prospective) studies in the MEDLINE, Embase, and Cochrane Library databases that covered at least 100 children or adults with T1/T2 DM. In T1 DM, basal-bolus/pre-mix insulin (human or analogue) and insulin pump were reviewed, and in T2 DM, basal-bolus/pre-mix insulin (human or analogue), oral antidiabetic drugs supported with basal insulin (human or analogue), sulfonylureas in monotherapy, and combined oral treatment were reviewed. In order to estimate SHE rates, we extracted data on the time horizon of the study, number of patients, number of SHEs, and number of patients experiencing at least one SHE. We used a random effects model to estimate the annual SHE rate. We considered the risk for other antidiabetic medications in T2 DM to be negligible and the results of our main review yielded no observational data for premixes in T1 DM so they were assessed based on relative rates taken from additional systematic reviews. The study, being a desk research, did not involve any human subjects (including human material or human data) and no ethical committee approval was asked for. For the same reason there was no need to collect informed consent for participation in the study.ResultsWe identified 76 observational studies encompassing 707,722.30 patient-years. The estimated annual SHE rate varied from 0.168 (95 % CI 0.123–0.237) for insulin pump up to 1.628 for biphasic human insulin in T1 DM patients, and from 0.0035 for oral antidiabetic drugs up to 0.554 (95 % CI 0.157–7.534) for basal-bolus with human insulin in T2 DM patients.ConclusionsOur review indicates that SHE rates differ between patients depending on treatment regimen. However, SHEs are also driven by other factors. Proper modelling techniques are needed to use various types of information in published studies.Electronic supplementary materialThe online version of this article (doi:10.1186/s12902-015-0052-z) contains supplementary material, which is available to authorized users.