An observational study using data from US healthcare databases suggests that in elderly patients with type-2 diabetes, rosiglitazone may be associated with a slightly greater risk of all-cause mortality than pioglitazone.

There is considerable debate over the absolute and relative safety of the two available thiazolidinedione oral anti-diabetic drugs, rosiglitazone and pioglitazone. Various epidemiological studies and meta-analyses have confirmed that both increase the risk of congestive cardiac failure, although it is uncertain whether there are qualitative differences between them. The authors of this study used data from two US healthcare administrative databases to compare outcomes in elderly patients with type-2 diabetes who were treated with one or other of these drugs. They identified patients aged over 65, treated for type-2 diabetes who started one of the study drugs between January 1999 and December 2004 for one database or 2005 for the other. Previous medical history was examined to identify pre-existing relevant medical conditions. Primary outcome for the analysis was all-cause mortality, with myocardial infarction, stroke, and hospitalisation for heart failure as secondary outcomes.

The selection algorithm identified 28,361 patients receiving a first prescription for one of the study drugs, evenly split between pioglitazone (50.3%) and rosiglitazone (49.7%). Users were generally similar at baseline, however rosiglitazone users were slightly more likely to have a history of cardiovascular disease and less likely to be users of statins or beta-blockers; they were also more likely to have been in a nursing home and have had more hospital days.

Median (mean) exposures were 217 (380) and 215 (369) days for pioglitazone and rosiglitazone users, respectively to give a total of 29,060 person-years follow-up. Over this period, 1,869 patients died. Deaths in rosiglitazone users were slightly higher than those in pioglitazone users: 59.7 per 1,000 person-years for pioglitazone and 69.2 for rosiglitazone, to give a crude incidence rate ratio (IRR) of 1.17 (95% CI, 1.06 to 1.28). After adjustment for potential confounding factors, IRR for rosiglitazone users was 1.15 (95% CI, 1.05 to 1.26).

There were no significant differences for MI or stroke, however rosiglitazone was associated with a greater risk of hospitalisation for heart failure (IRR 1.13; 95% CI 1.01 to 1.26): the absolute difference in event rates was 4.0 (42.0 vs. 46.0) per 1,000 person years.

Based on their analysis, the authors conclude that in elderly patients with type-2 diabetes, initiation of rosiglitazone is associated with a higher all-cause mortality than initiation of pioglitazone. They note that their study has limitations, and although the figures were adjusted for potential confounding factors, residual confounding may have remained.

[Editor's comment: adds more data to the debate, but probably little more clarity. The actual difference in events was small - 9.5 per 1,000 patient-years; and even though the study collected data for between 6 and 7 years, most patients didn't take the drugs for long - mean exposures were about a year. The authors have gone to some lengths to reduce confounding factors, nevertheless the differences have to be treated with care due to the study design. Bandolier has commented on the problems with limitations - they conclude (in Bandolier 2007; 14(7): 1-3) that relative risks below 2, and certainly below 1.5, should be treated with caution, especially outside a randomised controlled trial.]