Heart risks of diabetes drugs compared

BBC News reports that researchers have said that “a drug to treat diabetes, Actos, would be a 'sensible alternative' to one which was banned last year”. The BBC said that Avandia, also known as rosiglitazone, was suspended in Europe but is still available in the US and Canada.

Avandia is the brand name of the diabetic drug rosiglitazone, which was banned in 2010 after the European Medicines Agency concluded that its benefits did not outweigh the cardiovascular risks. Actos is the brand name of pioglitazone, a drug in the same group, which remains licensed for use in some people whose diabetes cannot be controlled by other drugs. Actos is also known to increase the risk of heart failure, and doctors must closely monitor patients.

This systematic review looked at 16 observational studies of 810,000 people, comparing the cardiovascular risk of the two drugs. It found that rosiglitazone has a higher risk of heart attack, heart failure and death compared with pioglitazone.

Though there are some minor limitations to the review, the findings confirm the increased risk of rosiglitazone, and they support last year’s decision to withdraw marketing authorisation for that drug. Pioglitazone remains licensed for use in type 2 diabetes, provided that patients meet certain eligibility criteria and that any adverse effects of the drug are monitored.

Where did the story come from?

This study was authored by researchers from the University of East Anglia and Johns Hopkins University School of Medicine. Funding was provided by the National Institutes of Health (NIH) and NIH Roadmap for Medical Research. The study was published in the peer-reviewed British Medical Journal .

The news reports accurately reflected the findings of this review.

What kind of research was this?

This was a systematic review and meta-analysis of observational studies. The aim of the research was to compare the cardiovascular risk of rosiglitazone and pioglitazone when used to treat people with type 2 diabetes. These drugs belong to a class of drugs called thiazolidinediones which treat type 2 diabetes by increasing the body’s sensitivity to insulin, and thus lowering blood sugar. Specifically, the study aimed to compare the risks of heart attack, heart failure and overall mortality with the two drugs.

A systematic review is the best way of investigating the relationship between a particular exposure (in this case thiazolidinedione use) and an outcome (cardiovascular adverse effects), using the available evidence. The statistical pooling of results from the identified studies (meta-analysis) can provide an overall estimate of the effect of an exposure or treatment. However, there are inherent limitations due to the differences in the individual studies and their included populations, treatment schedule and duration, and follow-up.

Ideally, a review of the effects of a medication would incorporate randomised controlled trials of the drug, rather than cohort studies. This is because randomisation balances differences between participants that could influence their outcomes. However, if a drug has suspected harms, it would not be ethical to conduct randomised studies. For drugs that are already licensed, observational studies are often used to examine what has happened to people already taking the drug.

Observational studies allow a much larger population to be studied over a longer follow-up period than would be feasible in trials, and to investigate potential harms in a “real world” setting. When this approach is taken, researchers need to consider the possibility that factors other than the exposure/treatment of interest may be affecting the results.

Previous systematic reviews and meta analyses have looked at RCTs of the cardiovascular effects of thiazolidinedione. The researchers report that indirect comparisons of rosiglitazone and pioglitazone (e.g. comparing the results of RCTs of rosiglitazone versus another drug or placebo, and RCTs of pioglitazone versus the same drug or placebo) suggest that rosiglitazone is associated with a greater risk of heart attacks and heart failure than pioglitazone. However, indirect comparisons do have limitations.

The current review assessed whether this increase in risk would also be seen in observational studies that directly compared the two drugs.

What did the research involve?

The researchers searched medical databases to identify observational studies published up to 2010 (cohort or case control studies), which directly compared the risk of cardiovascular outcomes with rosiglitazone and pioglitazone in patients with type 2 diabetes. They also carried out a hand search of reference lists, and identified unpublished studies by looking at the websites of regulatory authorities and drug manufacturers. The main outcome of interest was heart attack. Heart failure and overall mortality were secondary outcomes of interest.

The researchers pooled the results from these studies and calculated the odds of cardiovascular outcomes for the two thiazolidinediones, using statistical methods that took into account possible differences between the studies (heterogeneity). Where possible, they used the results from studies that also took into account factors other than rosiglitazone and pioglitazone, which could affect the risk of cardiovascular outcomes.

What were the basic results?

The researchers identified 16 eligible studies (four case-control and 12 cohort studies), which included 810,000 people taking thiazolidinedione drugs (429,000 people taking rosiglitazone and 381,000 taking pioglitazone). Fifteen studies reported on the outcome of heart attack, eight reported on heart failure, and eight reported on mortality. Follow-up times in the studies ranged from 105 days to seven years. Most participants were over 60 years of age, and 55% were male. Duration of thiazolidinedione use was reported in only four studies, and it ranged between 215 and 450 days.

Compared with pioglitazone, use of rosiglitazone was associated with:

  • a 16% increased odds of heart attack (odds ratio [OR] 1.16, 95% confidence interval [CI] 1.07 to 1.24; 15 studies)
  • a 22% increased odds of heart failure (OR 1.22, 95% CI 1.14 to 1.31; 8 studies)
  • a 14% increased odds of death (OR 1.14, 95% CI 1.09 to 1.20; 8 studies)

From this, the researchers calculated that if 100,000 people were treated with rosiglitazone rather than pioglitazone, there would be 170 excess heart attacks, 649 excess cases of heart failure, and 431 excess deaths.

How did the researchers interpret the results?

The researchers concluded that, in people with type 2 diabetes, use of rosiglitazone rather than pioglitazone is associated with significantly increased odds of heart failure, heart attack and overall mortality.


This large review has compared the cardiovascular risk of the two thiazolidinedione drugs rosiglitazone or pioglitazone in 16 studies including 810,000 people. The findings provide evidence that rosiglitazone has a higher risk of heart attack, heart failure and overall mortality than pioglitazone. The findings support the decision to withdraw this drug from the market.

This is a thorough and well-conducted review. There are a few limitations that should be acknowledged:

  • The included cohorts and case control studies had used broadly similar methods to identify the population of interest (e.g. using pharmacy claims databases to identify people taking rosiglitazone or pioglitazone), and to follow-up cardiovascular outcomes (e.g. searching relevant databases and medical records for the diagnostic codes related to heart attack and heart failure). However, few studies tried to verify the accuracy of any coded cardiovascular outcomes. None provided information on the severity or consequences of these events. Also, few studies checked the validity of the drug prescriptions and checked that these were actually filled and taken by the participants. These things could lead to misclassification of some patients, regarding the drugs that they took and the cardiovascular events that occurred.
  • Ideally, a comparative review of the effectiveness and adverse effects of different drugs would be a review of randomised controlled trials rather than observational studies. Randomisation balances out any differences in other cardiovascular risk factors between people given the different drugs. However, once harms are suspected to be associated with a drug, it would not be ethical to conduct randomised studies. If the drug is already licensed, observational studies are often used to look at what has happened to people already taking the drug. It allows inclusion of a much larger treated population with longer follow-up than in trials. It also assesses the harms in a “real world” setting.
  • This review did not compare the drugs to an inactive placebo drug or to an alternative diabetic drug of another class. Though the calculations inform us that rosiglitazone carries a higher risk than pioglitazone, they cannot tell us about the cardiovascular risk of pioglitazone compared to placebo or to other drugs. This is important, as previous research has shown that pioglitazone also increases the risk of heart failure.

Despite these limitations, there was limited statistical heterogeneity (differences) between the results of the included studies. There was no evidence of publication bias (that studies with particular findings had been selectively published).

The findings confirm the increased risk of rosiglitazone, and they support last year’s decision to withdraw marketing authorisation for that drug. Pioglitazone remains licensed for use in type 2 diabetes provided that patients meet certain eligibility criteria and that any adverse effects of the drug are monitored.

NHS Attribution