Thirty-seven double-blind RCTs (2,757 patients), 15 non double-blind RCTs (405 patients), 9 non-randomised controlled clinical trials (CCTs; 281 patients), 33 cohort studies, 4 case-control studies and 10 case series were included. The double-blind RCTs included 8 double-blind population or clinic-based RCTs (1,734 patients), 10 double-blind laboratory-based RCTs (149 patients) that used hypoglycaemia as the primary outcome, and 19 double-blind RCTs with hypoglycaemia as a secondary outcome (874 patients). Many of the trials used a crossover design.
Double-blind RCTs.
Seven of the 37 double-blind RCTs (5 studies reported in tables and 2 further studies described in the text) showed that human insulin significantly increased the frequency of hypoglycaemia and/or reduced the symptoms of hypoglycaemia. Two studies were population-based, three were laboratory-based in patients who had not previously reported impaired awareness, and two assessed hypoglycaemia as a secondary outcome. The participants had long-standing diabetes or were on intensive therapy. The remaining 30 RCTs showed no significant difference between insulins in hypoglycaemic measures, although two of these RCTs reported high drop-out rates for serious hypoglycaemia during the study period for human insulin.
None of the RCTs (either population- or laboratory-based) in patients previously reporting impaired awareness of hypoglycaemia showed any significant difference between human and animal insulin.
Non double-blind RCTs.
Four (27%) of the 15 studies reporting hypoglycaemia as a secondary outcome showed that human insulin increased the frequency of hypoglycaemia compared with animal insulin, although the authors described the difference as small and not statistically significant. All four positive studies were set in laboratories and hypoglycaemia was induced experimentally.
Non-randomised CCTs.
Two (22%) of the 9 CCTs reporting hypoglycaemia as a secondary outcome showed that human insulin increased the frequency of hypoglycaemia compared with animal insulin, although this was not statistically significant. Four studies found no difference, while one study did not report any hypoglycaemic event. Two studies investigating hypoglycaemia as a primary outcome reported no difference.
Cohort studies.
Eight (24%) of the 33 studies showed that human insulin increased frequency, severity and/or decreased awareness of hypoglycaemia compared with animal insulin. Nine studies reported no difference between the two treatments, while 3 studies found fewer hypoglycaemia episodes with human insulin and/or a greater awareness of hypoglycaemia.
Case-control studies.
One (25%) of the 4 studies showed a statistically significant association between human insulin and hypoglycaemia.
Case series.
Nine (90%) of the 10 studies showed that human insulin increased the frequency, severity and/or decreased awareness of hypoglycaemia compared with animal insulin. The largest case series found no evidence that hypoglycaemia from human insulin contributed to any of the 50 deaths investigated.