Quintin‧速記本

Published 18 August 2009, doi:10.1136/bmj.b3076
Cite this as: BMJ 2009;339:b3076

Editorials

Rosiglitazone or pioglitazone in type 2 diabetes?

Longer term safety data are needed before a change in practice is warranted

The management of diabetes aims to control blood glucose concentrations and lower cardiovascular risk, thus limiting macrovascular and microvascular complications. Despite concerns about using thiazolidinediones in people with heart failure, these drugs remain at the centre of most prescribing algorithms for type 2 diabetes, including those recently published by the National Institute for Health and Clinical Excellence.¹ In the linked cohort study (doi:10.1136/bmj.b2942), Juurlink and colleagues compare the risk of acute myocardial infarction, heart failure, and death in patients with type 2 diabetes treated with rosiglitazone and pioglitazone.² Long term trials show that thiazolidinediones increase insulin sensitivity, lower glycated haemoglobin, and delay the progression of disease as measured by treatment failure with monotherapy.³ Both rosiglitazone and pioglitazone (the only thiazolidinediones currently on the market) have favourable effects on the cardiovascular risk profile over and above those achieved through glycaemic control alone.⁴ In addition, thiazolidinediones may have a place in the treatment of non-alcoholic steatohepatitis,⁵ and they may prevent progression from the metabolic syndrome to diabetes.⁶ However, various studies have shown that thiazolidinediones are associated with an increased risk of fractures,⁷ liver disorders, and heart failure.⁸ Heart failure is thought to be a consequence of thiazolidinedione associated oedema; controversy exists around possible increases in the risk of myocardial infarction and death from cardiovascular disease.⁹

In this context, Juurlink and colleagues used an insurance claims database to investigate differences in cardiac risks and overall mortality between rosiglitazone and pioglitazone.² They found that pioglitazone was associated with a reduced risk of heart failure (adjusted hazard ratio 0.77, 95% confidence interval 0.69 to 0.87) and overall mortality (0.86, 0.75 to 0.98), but not myocardial infarction (0.95, 0.81 to 1.11) compared with rosiglitazone. A similar observation has been made by others.^{10 11}

The study raises two questions: do the findings reflect a true difference in risk between the two drugs, and if so, should this lead to changes in clinical practice? To answer the first question, we need to consider the possibility of the results arising by chance—owing to a systematic error in the study design—or confounding, including confounding by disease severity.

Given the precision of the results and the rigorous study design we can exclude chance and bias as explanations for the results. In addition, given the minimal differences between the unadjusted and adjusted hazard ratios despite adjusting for multiple confounders, and the fact that various sensitivity analyses made no material difference to the results, at first impression the likelihood of residual confounding seems negligible, even if results were not adjusted for smoking, alcohol intake, and body mass index. We have no reason to believe that these risk factors would have differed between people who took either drug, and the risk of myocardial infarction did not differ between the two drugs, which supports this assumption.

However, evidence to date indicates that the risk of oedema does not differ between the two drugs,¹² and the question of residual confounding by disease severity remains. People who used rosiglitazone had had diabetes for longer at entry to the study, and the investigators could not adjust for differences in duration of disease beyond five years. Such residual confounding could have resulted in a difference in the risk of myocardial infarction, which was not seen in this study. Although the study may have lacked statistical power to identify such a difference, the point estimate (0.95) is insufficiently different from 1.0 to support this notion. If the difference in all cause mortality is not entirely attributable to a difference in heart failure—a matter that the investigators could not explore and that has not been confirmed by others⁸—people who used pioglitazone may have been healthier at baseline and this could not be adjusted for because it was not identifiable from the claims data.

This leaves clinicians unsure about the implications for prescribing. Both drugs carry an increased risk of heart failure. This is thought to reflect an unmasking of pre-existing cardiac malfunction resulting from increased fluid and sodium retention—rather than a direct effect on heart muscle—and therefore to be reversible and manageable with spironolactone or amiloride, which would explain the absence of an effect on mortality.⁸ The study reinforces the message that thiazolidinediones should be avoided in patients with heart failure.¹

It may be tempting to move away from prescribing thiazolidinediones altogether in favour of the more recently introduced incretin based treatments. However, long term follow-up data for these products are not yet available. Experience with thiazolidinediones, for which the associated risks of cardiac disease and fractures were identified only years after they were introduced on to the market, reinforces the notion that long term studies of effectiveness and safety in clinical practice are essential to complement knowledge on efficacy gained from randomised controlled trials. Such trials are unlikely ever to provide the full picture. Healthcare databases used in observational studies can be limited because they do not always give us all the information we need, such as cause of death and duration of diabetes. Therefore, enhancements to healthcare databases coupled with well designed studies such as that by Juurlink and colleagues are essential for determining the full risk-benefit profile of medicines.

看來上市的藥品，還是很多要觀望的阿！Rosiglitazone或Pioglitazone誰比較安全呢？這場戰爭還有得打~