OBJECTIVES: Severe (requiring assistance from another individual) hypoglycaemic events (SHEs) are important from both clinical and economic perspective. The aim was to assess SHEs risk related to major drug groups in type 1 and 2 diabetes (T1, T2) in real-life settings. METHODS: We systematically reviewed Medline, EMBASE and Cochrane databases for recent (<10 years), large (n ≥100) observational studies for drugs associated with higher SHEs risk by available guidelines (IDF, ADA, EASD) – i.e. insulins and sulphonylureas. We extracted number of SH events and patients experiencing at least one, number of participants and time horizon. Using a random effects Poisson model within MCMC framework we estimated the annual SHEs rates whenever the heterogeneity and number of studies allowed a reasonably credible estimate. RESULTS: Search strategies yielded 5220 studies, 525 full texts were analysed, and 43 were finally enrolled: 15 regarding long-acting insulin analogues (LAA), 13 – pumps, 21 – other insulins (OI), 9 – sulphonylureas. Full texts were rejected mostly due to: inappropriate SHE definition, patients using a mixture of treatments, no precise treatment definition. We obtained the following ranges for SHE rates in individual studies. For T1: LAA (n=13, 0.10-1.47), pumps (n=17, 0-0.58), OI (n=12, 0.20-1.26). For T2: LAA (n=5, 0-0.98), OI (n=10, 0-0.59), OI + sulphonylureas (n=4, 0-0.59), OI + other OADs (n=8, 0-0.72), sulfonylureas (n=6, 0-0.11). Pooled data suggest the following mean annual SHE rates (95%CI). For T1: LAA 0.54 (0.33-0.88), pumps 0.17 (0.11-0.25), OI 0.67 (0.36-1.26). For T2: OI 0.34 (0.07-0.68), sulfonylureas 0.06 (0.02-0.15). CONCLUSIONS: SHEs risk differs for T1, T2 and available drugs. However SHEs frequency may not seem large, the fear of such an event, its danger and economic consequences may be an important issue in managing diabetes. Limited availability of studies and heterogeneity of data make it difficult to come up with precise rate estimation.