The Thiazolidinedione Controversy in Cardiovascular Risk


Philip R. Liebson, MD, Rush University Medical Center, 1653 West Congress Parkway, Chicago IL 60612

Diabetes is a strong risk factor for cardiovascular (CV) events, and it is essential that antidiabetic agents either improve CV risk or have a neutral effect. The thiazolidinediones (TZDs) have been used to sensitize end organs to insulin to produce a beneficial effect in diabetics by reducing blood glucose and glycated hemoglobin levels. Unfortunately, adverse effects of TZDs on CV outcomes have been demonstrated leading to repeated assessments by the Food and Drug Administration (FDA). Most recently, in July 2010, an advisory committee of the FDA evaluated the safety of rosiglitazone and delivered an ambiguous verdict. It indicated that the drug raised the risk of CV events compared with other diabetes drugs but recommended leaving the drug on the market with various degrees of warnings or restrictions.

Because this controversy affects the scope of preventive cardiology, this clinical trial review will provide a chronologic summary of the findings and recommendations about TZDs.


The TZDs have been on the market for more than 10 years. Their effect on insulin sensitivity is due to activation of the peroxisome proliferation-activated receptor–λ (PPAR-λ). Although there is evidence that PPAR-λ may produce beneficial effects on atherosclerosis, such as slowing of progression of carotid intima-media thickening and coronary restenosis, early studies showed toxic effects of the TZDs, including hepatic and cardiac failure.1 On the favorable side, soon after licensing, the TZDs were demonstrated to have antioxidant and anti-inflammatory effects.2,3

The first TZD, troglitazone, was removed from the market because of hepatotoxicity. The other 2 TZDs, rosiglitazone and pioglitazone, were marketed in 1999. Since then, rosiglitazone has been the subject of several FDA analyses based upon a number of clinical trials and meta-analyses indicating adverse CV effects.

Individual Studies Indicating Beneficial or Neutral CV Effects of TZDs on CV Risk

A summary of benefits of TZDs on atherosclerosis is found in the review of Dandona and associates.4 These include reduction of carotid intima-media thickness (IMT) within a few months of treatment5 and beneficial effects on progression of coronary atherosclerotic lesions after angioplasty by both rosiglitazone6 and pioglitazone.7 However, there is a paucity of studies demonstrating beneficial clinical effects.

In the Pioglitazone Effect on Regression of Intravascular Sonographic Coronary Obstruction Prospective Evaluation (PERISCOPE trial, a study of the progression of atherosclerosis in 543 patients with coronary disease and type 2 diabetes, treatment with pioglitazone resulted in less progression of atherosclerosis than glimepiride based on intravascular ultrasound evaluation of serial determination of atheroma volume.7 Unfortunately, more than one-third of patients did not complete the study endpoint assessment and there was <1% difference in change between groups. This type of study is not as valuable for reflecting benefit in comparison with a placebo-controlled trial.

The only successful CV clinical outcomes trial of TZDs, Prospective Pioglitazone Clinical Trial in Macrovascular Events (PROACTIVE), demonstrated that a high dose of pioglitazone vs placebo in 5238 diabetic patients with macrovascular disease over an average of 34.5 months reduced the rate of fatal and nonfatal myocardial infarction (MI) by 16% and all macrovascular events by 10%.8

In the Carotid Intima-Media Thickness in Atherosclerosis Using Pioglitazone (CHICAGO) trial, comparing pioglitazone with glimepiride in assessing changes in carotid IMT over a period of 18 months, pioglitazone slowed progression of mean carotid IMT and maximum carotid IMT.9

By contrast, a recent study on changes in atheroma volume of the coronary arteries in 672 patients receiving rosiglitazone or glimepiride demonstrated no significant difference in progression.6 A retrospective analysis of 227,571 diabetic Medicare patients, initiated with pioglitazone and rosiglitazone with three-year follow-up, demonstrated an increased risk of rosiglitazone for stroke (27% increase), heart failure (14%), and composite CV events (18%), with more favorable risk effect of pioglitazone.10 A retrospective cohort study of 39,736 patients found a significantly lower risk of heart failure and death with pioglitazone in comparison with rosiglitazone.11

Several recent studies of prospective studies of rosiglitazone have shown no association of rosiglitazone with increased cardiovascular events (A Diabetes Outcome Progression Trial [ADOPT], Rosiglitazone Evaluated for Cardiovascular Outcomes and Regulation of Gylaecemia in Diabetes [RECORD], Bypass Angioplasty Revascularization Investigation-2 Diabetes [BARI-2D], Action to Control Cardiovascular Risk in Diabetes [ACCORD] trial, and Veteran Affairs Diabetes Trial [VADT]).4

Meta-Analyses Evaluating TZDs

Much of the concerns have focused on rosiglitazone. Some of the salient meta-analyses published over the last six years assessing TZD effects on CV risk factors, CV events, and MI follow.

An analysis in 2004 of 23 randomized controlled trials comparing pioglitazone or rosiglitazone with placebo for 12 weeks in regard to CV risk factors demonstrated similar effects on glycemic control (hemoglobin A1c decreased 1.0%–1.5%) and body weight (increase of 3.0 kg).12 Pioglitazone significantly decreased triglyceride levels and increased high-density lipoprotein cholesterol compared with placebo, whereas rosiglitazone, although producing a similar increase in high-density lipoprotein cholesterol, showed no change in triglycerides and an increase in low-density lipoprotein cholesterol. These results suggested a more beneficial effect of pioglitazone on lipids than rosiglitazone compared with placebo.

Four other relevant meta-analyses of CV events comparing TZD vs control groups are reviewed here, three of which were published in 2007 and one in 2010. Control groups primarily included other antiglycemic agents and rarely placebo. Rosiglitazone groups usually included other antiglycemic agents.

A total of 19 trials enrolling 16,390 diabetic patients treated with pioglitazone vs placebo and followed for 4 months to 3.5 years demonstrated a significantly lower risk of combined outcome events (death, MI, stroke) by 18% with pioglitazone.13 However, serious heart failure was increased with pioglitazone (2.3% vs 1.8% of controls). In secondary outcome analysis of individual components, pioglitazone showed a significant benefit on combined death/MI (P=.04) and borderline benefit for MI (P=.08) and stroke (P=.09).

In a comparison of 28,381 Medicare patients treated with pioglitazone or rosiglitazone, a 15% greater mortality was found in the rosiglitazone group, as well as a 13% increase in risk for heart failure.14 There were no differences in MI or stroke.

The other meta-analyses involve rosiglitazone vs controls. In an evaluation of four clinical trials of rosiglitazone vs placebo for prevention or treatment of type 2 diabetes, 14,291 patients were followed for one to four years for individual risk of MI, heart failure, or CV mortality.15 A 42% increase in MI and 109% risk for heart failure were found without a significant increase in CV mortality.

Another meta-analysis of rosiglitazone published the same year (2007) evaluated 42 trials involving 27,847 patients over at least 24 weeks. There was a 43% increase in MI and 64% increase in mortality in the rosiglitazone group compared with controls.16 Concern has been expressed on the inclusions (only 42 of 116 studies, short duration) and statistical treatment of the data.17 Critiques of this evaluation included variable drug inclusion and exclusion criteria, low overall event rates, and absence of reports of stroke and noncardiovascular deaths. However, these criticisms, which may apply generally to meta-analyses, reflect the problems associated with this type of evaluation.

In 2010, the same authors updated their meta-analysis of rosiglitazone and pioglitazone, again evaluating studies of more than 24 weeks, reporting adverse CV events, including 56 trials totaling 25,531 patients, an increase in 30% from the previous study.18 A total of 146 trials were excluded primarily because of short-term phase 1 trials, less than 24 weeks duration, and no control groups. Similar to the previous meta-analysis, there was a significant increase in MI in the rosiglitazone group (by 28%) but not in CV mortality.

These meta-analyses demonstrate an adverse effect for rosiglitazone on CV events in comparison with control agents and pioglitazone.


In 2009, RECORD, an open-label trial of rosiglitazone, was published and is discussed here because of consequences involving a recent FDA decision (see below).19 In this open-label study, 4447 patients on metformin or sulfonylurea were assigned to an addition of rosiglitazone or continuation of the initial therapy alone. This was a non-inferiority study, meaning that there was an evaluation of whether rosiglitazone was not inferior to the risk of the other agents for CV risk consequences. Over a 5.5-year follow-up, there was a 110% increased risk of heart failure in the rosiglitazone addition group but no evidence of inferiority of risk for stroke, MI, or CV death. The concerns about this study analysis are discussed below.

Scientific Advisories

Two scientific advisories on TZD use have been reported, the first in 2003 and a recent one in 2010. The 2003 advisory from the American Heart Association (AHA) and American Diabetes Association was primarily concerned with the increased risk of edema and heart failure with these agents in type 2 diabetes.1 It was recommended that patients in this category with known left ventricular dysfunction in whom heart failure develops while on TZDs should have this agent discontinued and that TZDs should not be initiated in patients with New York Heart Association class III or IV congestive heart failure.

The most recent scientific advisory from the AHA and American College of Cardiology Foundation expressed the opinion that there is “some inconclusive evidence of potential harm of rosiglitazone but not pioglitazone” for macrovascular events.20 The recommendations for heart failure were unchanged from 2003. In regard to intensive vs standard glycemic control in patients with high CV risk, they alluded to the 20% increase in risk of death in the intensive treatment group in the ACCORD trial, with “Preliminary post hoc exploratory analyses ... not suggest(ing) a link between differences in the use of drugs (including rosiglitazone) and the increased deaths in the intensive-treatment groups.”21

The Long Road Travelled

The experience with the TZDs reflects similar experiences with other classes of agents such as the calcium channel blockers and cyclooxygenase inhibitors. As indicated, the first TZD to be marketed, troglitazone, was discontinued because of hepatic and cardiac toxic effects. In May 2007, the FDA released a safety alert about possible increased risk of CV ischemic events with rosiglitazone. This resulted primarily from the Nissen meta-analysis indicating a 43% increased risk of MI and possible increase in CV mortality.16 However, an FDA advisory panel decided later that year not to withdraw rosiglitazone from the market, but a black box warning was provided indicating the potential risk of myocardial ischemia, especially (1) in patients with heart disease taking nitrates and (2) when added to established insulin therapy.20

Fast-forward to 2010. On April 20, the last scientific advisory of the AHA on rosiglitazone was published. This came two months after a 334-page US Senate Finance Committee report was released on an investigation of documents related to rosiglitazone and the activities of the parent drug company GlaxoSmithKline (GSK).22 The report stated that “[t]he totality of evidence suggests that GSK was aware of cardiac risks associated with [rosiglitazone] years before such evidence became public .... GSK executives intimidated independent physicians, focused on strategies to minimize findings that [rosiglitazone] may increase cardiac risk, and sought ways to downplay findings that the rival drug, . . . (pioglitazone), might reduce cardiovascular risk.”23

To summarize the evidence thus far, there is no evidence that rosiglitazone prevents vascular events in diabetics. It appears less safe than pioglitazone. Further clinical trials have been recommended by consensus panels.

The FDA has compelled the parent drug company to undertake an extensive clinical trial. This is the Triazolidinediones Intervention With Vitamin D Evaluation (TIDE), begun in 2009 and expected to be completed in 2015.24 It encompasses a comparison of rosiglitazone and pioglitazone in regard to composite CV events. It involves more than 240 sites throughout the world and more than 16,000 patients. As indicated above, two previous comparison trials demonstrated a greater risk of rosiglitazone for mortality and heart failure14 and all-cause mortality, MI, and stroke,10 respectively. Whether TIDE will add any further information on rosiglitazone is open to question.

Information on some of the communications within the company is presented in a short commentary by Nissen.25 According to the Nissen commentary, the 2007 meta-analysis manuscript was sent by a journal reviewer inappropriately to the drug company, and after an internal review by statisticians at GSK, GSK agreed with the meta-analysis conclusion of increased risk for ischemic events with rosiglitazone. Following this finding, the company decided to unblind the RECORD trial (see above), after which the trial steering committee published an interim analysis, although the data available were underpowered. The published RECORD paper indicated a very low MI rate, one-third that of a similar study with pioglitazone questioning the ascertainment of MIs in the RECORD study. There were also some questions about the adherence rates of patients on rosiglitazone. The point of this is that reported results of clinical trials may be deceptive unless attention is paid by the reader to discrepancies in information on adherence rates and effective power of the study, which may change with early unblinding and closure of the study.

The latest decision at the time of this report was made by the FDA Advisory Committee in July 2010 to keep rosiglitazone on the market (vote was 20 to 12), but with stricter label warnings. A heavy focus was made on the results of the RECORD trial, which the company used as one of the main defenses of the CV safety of rosiglitazone. The FDA now has to act on the advice of the committee, which it usually does.