Rosiglitazone goes on the RECORD, but is it a hit?

A big new rosiglitazone trial, RECORD, found no increase in overall cardiovascular morbidity and mortality, the primary endpoint of the trial, but is unlikely to quiet the drug’s critics or end the controversy over the drug. The trial is being presented at the American Diabetes Association (ADA) meeting in New Orleans and simultaneously published online in the Lancet.

The multicenter, open-label trial randomized 4,447 type 2 diabetics on metformin or sulfonylurea monotherapy to receive either rosiglitazone in addition or an active control consisting of a combination of metformin and sulfonylurea. After 5.5 years of followup there were 321 events in the rosiglitazone group and 323 in the control group.  Consistent with previous trials, rosiglitazone caused an increase in heart failure and fractures. In addition, a nonsignificant increase in MI was observed.

An accompanying editorial by Ravi Retnakaran and Bernard Zinman notes that “definitive conclusions about the relation between rosiglitazone and cardiovascular disease remain elusive, owing to study limitations. Specifically, the study was open label in design and cardiovascular disease event rates were much lower than anticipated. Furthermore, the findings are inconclusive for myocardial infarction, for which a non-statistically significant increased risk was noted in the rosiglitazone group (HR 1·14, 0·80–1·63).” The editorialists go on to support the ADA/EASD consensus statement which recommends pioglitazone over rosiglitazone.

Steven Nissen, who helped ignite the Avandia controversy originally, was highly critical of the RECORD trial. Here are the comments he sent to CardioBrief:

“The RECORD trial is a seriously flawed study. At the time the interim analysis was published, approximately 30% of patients were no longer taking Avandia. The authors fail to disclose how many patients had stopped the drug in their Lancet manuscript, but it seems like that this number approached 50%. In a safety study, withdrawal of study drug biases the trial toward showing no difference between therapies thereby concealing hazards.

“Furthermore, in those patients with ischemic heart disease, there was as strong trend toward an excess myocardial infarction rate with Avandia treatment.

“There are important irregularities. More of the Avandia  patients received statins, a class of drugs known to reduce cardiovascular event rates in diabetics.”

GSK sent the following response to Nissen’s remarks:

“RECORD was a robust study that met its primary endpoint and delivered statistically significant results. Long-term clinical trials like this are the most reliable way to evaluate cardiovascular outcomes. The dropout rates in RECORD for all treatment groups did not affect the primary endpoint, which was achieved.

“A significant number of patients in the Avandia group completed the study on treatment. By study’s end, 1344 subjects or 60.5% of the Avandia group were still on drug, while 1131 subjects or 50.8% of the control group remained on drug. These numbers are not surprising, or unusual, for a long term, open-label trial in which patients were followed for more than five years.

“Because a patient may stop taking a medicine early in a trial or just before trial’s end, it’s important to consider the amount of time patients were exposed to Avandia, over the course of the trial. For the patients that were part of the Avandia arm of the study, they remained on Avandia 87.9% of the study time.”

Sanjay Kaul sent the following comment to CardioBrief:

“In contrast to the superiority trial in which if compliance with treatment is low, the chances of detecting treatment differences will diminish, low treatment compliance in noninferiority trial will result in increased likelihood of detecting “no difference”. Therefore, a dual strategy incorporating all patients randomized to treatment (intention to treat) and those actually on treatment (per-protocol) approaches is recommended, and the inference strengthened only if both approaches support noninferiority. This is clearly spelt out in the CONSORT guidelines for noninferiority trials published in JAMA in 2006. I am surprised Lancet reviewers and editors allowed this study to be published without these key analyses. I would like to see the investigators/sponsors shine the spotlight on these data. I would not be surprised to find out if the noninferiority criterion is not met in the per-protocol analysis.

“With regards to the 26% hazard in patients with pre-existing heart disease, even if the differences were statistically significant, the proper statistical approach to help interpret the results is via an interaction analysis. This analysis asks the question whether the drug effect differs in those with pre-existing heart disease from the overall cohort.  The key caveat is that these analyses are underpowered and not prespecified thereby leading to false negative and false positive conclusions. Nonetheless, by my back-of-the-envelope calculation, the interaction P value is not significant, thereby suggesting no significant differences between these two subsets. In such cases, the estimate of a treatment effect in the overall cohort (null effect) is the most reliable effect in any subset.

“Bottom line, there are key flaws in trial design (open label leading to imbalances in disease-modifying therapies such as statins and thiazides) and analysis (failure to report noninferiority in per-protocol analysis) that preclude any definitive conclusions to be drawn from this trial. I am disappointed at this missed opportunity to inform the benefit-risk profile of this drug!”

Update (June 7):

Kaul provided the following clarification of his remarks:

Some may argue that the per-protocol analysis gives the least biased estimate of noninferiority. Hence, it deserved more spotlight than came across in the manuscript. It should have been presented in a Table or a Figure and emphasized in abstract. Instead it was buried in text (which I clearly missed –Mea culpa!). Nonetheless, by my reading, noninferiority criterion was not met in the per-protocol analysis, the upper limit of 95% confidence interval (1.21) exceeding the noninferiority margin of 1.20. It is important to show the drop out rate (or compliance) in order to fully understand the results of the per-protocol analysis. I did not see this data presented in the manuscript. I think Nissen also had issues with this.

Similarly, in contrast to the author’s interpretation, analyses within each treatment stratum (rosiglitazone vs metformin and rosiglitazone vs sulfonylurea, with background sulfonylurea and background metformin respectively did not meet noninferiority criterion because the upper limit of 95% CI exceeded HR of 1.20.

The interaction term by my analysis is 0.72 (0.51-1.02) with unadjusted P value of 0.06 (close to the reported P of 0.055). When adjusted for multiple comparisons (I have assumed 9 pre-specified subgroups shown in Figure 3), the interaction P value is 0.40. Thus, using robust criterion for subgroup analysis that adjusts for multiple comparisons, the interaction is not significant. Furthermore, this is an example of qualitative interaction (difference in direction) and such types of interactions are seldom replicable and often spurious.

Here is the Lancet press release:


Using rosiglitazone (Avandia) in combination with standard diabetes treatments (metformin or a sulfonylurea) to lower blood glucose in type 2 diabetics does not increase the risk of cardiovascular disease or death. However, the study confirms that using rosiglitazone more than doubles the risks of heart failure, and also increases the risk of fractures, mainly in women. The findings of the RECORD study are published in an Article Online First and in an upcoming edition of The Lancet. They are being simultaneously presented at the American Diabetes Association (ADA) meeting in New Orleans, USA.

Rosiglitazone belongs to a class of drugs called thiazolidinediones, and has been proven by numerous studies as an effective agent to control blood glucose. But studies have also suggested an increased risk of heart attacks caused by the rosiglitazone, and concern about adverse effects has reduced its use. In this randomised trial, Professor Philip Home, The Medical School, Newcastle University, UK, and colleagues looked at 4447 patients with type 2 diabetes already on either metformin or a sulfonylurea with a mean haemoglobin A1c concentration* (HbA1c) of 7.9%. Patients were assigned received addition of either rosiglitazone (2220 patients) or to a combination of metformin and sulfonylurea (control group) (2227). The primary endpoint was cardiovascular hospitalisation or cardiovascular death.

The researchers found that 321 people in the rosiglitazone reached the primary endpoint compared with 323 in the control group – ie, no statistically significant difference. They also found that heart failure causing admission to hospital or death occurred in 61 people in the rosiglitazone group and 29 in the control group—thus risk of heart failure was more than doubled for rosiglitazone patients. The risk of arm and lower leg fractures also increased by 57% in patients given rosiglitazone—although the increased risk for women (82%) was much higher than that for men (23%).

The authors say*: “It is good to have robust evidence that this useful medication does perform similarly to other glucose-lowering medications in regard of cardiovascular events. It is also good to see it perform better in controlling blood glucose in the longer term. The data on fractures and heart failure are known class effects, and here the study provides useful data to help clinicians and people with diabetes decide when it is not safe to use rosiglitazone.”

They conclude: “What are the clinical implications for the future use of rosiglitazone? Rosiglitazone is not recommended for people with a history of heart failure or with previous problems that might have led to myocardial dysfunction. Rosiglitazone should be used with caution in women at high risk of fractures. Although our evidence is insufficient to rule out a small increased risk of myocardial infarction caused by rosiglitazone when compared with other glucose-lowering agents, rosiglitazone does not increase overall cardiovascular morbidity or mortality.”

In an accompanying Comment, Dr Ravi Retnakaran and Dr Bernard Zinman, Mount Sinai Hospital, Toronto, and University of Toronto, Canada, say that half-maximum doses of rosiglitazone (or a related drug pioglitazone) in combination therapy could be considered as it is generally accepted that half-doses give better-than-half results, while limiting side-effects. They conclude: “This combination therapy is currently being assessed for the prevention of diabetes in individuals with impaired glucose tolerance. If the efficacy of this strategy is confirmed, we might be able to find the optimal way to use this class of medications in the treatment of type 2 diabetes.”

Here is the GlaxoSmithKline press release:

Large, long-term study shows Avandia has no increased overall cardiovascular risk compared to other commonly used diabetes medicines

RECORD trial also demonstrates durable blood sugar control over time with Avandia
Clinical trial results presented today at the American Diabetes Association annual meeting show that overall rates of cardiovascular hospitalization and cardiovascular death are similar in patients taking Avandia (rosiglitazone) compared to those receiving metformin and sulfonylurea.

Results from the 4,447-patient Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycaemia in Diabetes (RECORD) study, which was sponsored by GlaxoSmithKline, Avandia’s maker, were also published online today in the medical journal The Lancet.

RECORD is a large, prospective, randomized, controlled study that was initiated in 2001, and designed to compare cardiovascular outcomes of patients on Avandia added to metformin or sulfonylurea to those on metformin and sulfonylurea. The study confirmed its primary hypothesis. It showed that cardiovascular hospitalization or cardiovascular death (which includes heart attack, congestive heart failure, and stroke) was not statistically different between the two groups after an average of 5.5 years of therapy.  This translates into 321 events (14.5 percent) among patients receiving Avandia compared to 323 events (also 14.5 percent) for patients receiving control medications (hazard ratio=0.99, 95% confidence interval=0.85 to 1.16) (Data on the individual components are provided below.)

“RECORD provides a robust assessment of rosiglitazone’s cardiovascular safety and achieving the study’s primary endpoint leads us to conclude that rosiglitazone carries no increased risk of overall cardiovascular death or hospitalization compared to the most commonly used diabetes medicines, metformin and sulfonylurea, which have been used for decades,” said Dr. Philip D. Home, chairman of the RECORD Steering Committee and a professor at Newcastle University.  “Millions of patients with diabetes do not achieve good blood sugar control, even with combination therapy, making a wide range of therapeutic options critical for the management of this growing and deadly disease.”

Glycaemic control, as measured by mean hemoglobin A1c, was statistically significantly better in the patient group randomized to Avandia after the average of 5.5 years of treatment in the trial. RECORD’s glycaemic result is consistent with the superior long term glycaemic control (for nearly five years) previously demonstrated by Avandia versus metformin and sulfonylurea in the ADOPT study.

Although RECORD was not designed to study microvascular endpoints, previous studies such as the United Kingdom Prospective Diabetes Study, which studied other medicines, have found that improved glycemic control may reduce complications of diabetes that can lead to amputations, kidney failure and eye damage.

Congestive heart failure is a known effect of Avandia and other drugs in the thiazolidinedione (TZD) class and is prominently described in a boxed warning in current labels of both TZDs. As anticipated, rates of CHF in RECORD were higher in the group randomized to Avandia (61 events or 2.7 percent versus 29 events or 1.3 percent) (hazard ratio=2.10, 95 percent confidence interval=1.35 to 3.27). The difference was statistically significant.

Other adverse events in patients randomized to Avandia were consistent with those reported in the U.S. Prescribing Information for Avandia, including bone fractures, which were higher in the group randomized to Avandia and were seen mostly in the bones of the arm, hands, lower leg and feet, and predominantly in women.

Among patients randomized to Avandia, the results for the following prespecified secondary endpoints also showed:

•    Fewer deaths from any cause (136 events or 6.1 percent versus 157 events or 7 percent for control) (hazard ratio=0.86, 95 percent confidence interval=0.68 to 1.08)
•    Fewer deaths from cardiovascular disease (60 events or 2.7 percent versus 71 events or 3.2 percent) (hazard ratio=0.84, 95 percent confidence interval=0.59 to 1.18). Among these events, there were more CHF-related deaths (10 versus 2), though fewer deaths from heart attack (7 versus 10) and fewer from stroke (0 versus 5).
•    Fewer events of a combination measure known as MACE, which includes cardiovascular death, heart attack or stroke (154 events or 6.9 percent versus 165 events or 7.4 percent) (hazard ratio=0.93, 95 percent confidence interval=0.74 to 1.15).
•    More heart attacks (64 out of 2,220 patients or 2.9 percent versus 56 out of 2,227 patients or 2.5 percent) (hazard ratio=1.14, 95 percent confidence interval=0.80 to 1.63).
•    Fewer strokes (46 events or 2.1 percent versus 63 events or 2.8 percent) (hazard ratio=0.72, 95 percent confidence interval=0.49 to 1.06)

None of the differences for these secondary endpoints (any cause death, cardiovascular death, MACE, heart attack and stroke) was statistically significant.

RECORD showed no evidence of an increase in malignancies. There were fewer events of pancreatic cancer in patients taking Avandia although the numbers were small.

“RECORD provides important and reassuring information about Avandia for physicians fighting diabetes,” said Dr. Ellen Strahlman, GSK’s Chief Medical Officer. “Clinical outcomes trials like this offer the highest standard of evidence when considering the benefits and risks of medicines.  RECORD clearly demonstrates that overall rates of cardiovascular hospitalization and death are similar in patients taking Avandia compared to those receiving metformin and sulfonylurea. We believe that Avandia remains an important diabetes medicine for the appropriate patients.”

Avandia, introduced on the market in 1999, is a member of the TZD class of oral anti-diabetic medications which improve blood sugar control by directly targeting insulin resistance, an underlying cause of type 2 diabetes. This helps the body respond better to its own natural insulin. Avandia is indicated as an adjunct to diet and exercise to improve glycemic control in patients with type 2 diabetes mellitus.

More than 18 million Americans have type 2 diabetes, making it the most common form of diabetes, which accounts for about 90 to 95 percent of diagnosed diabetes cases. Type 2 diabetes has been associated with a variety of serious health problems including heart disease, stroke, eye damage, kidney failure and foot problems that lead to amputations.

GlaxoSmithKline – one of the world’s leading research-based pharmaceutical and healthcare companies – is committed to improving the quality of human life by enabling people to do more, feel better and live longer. For more information please visit:

Important safety information for Avandia® (rosiglitazone maleate)
Prescription AVANDIA, along with diet and exercise, helps improve blood sugar control in adults with type 2 diabetes. Taking AVANDIA with insulin or nitrates is not recommended.

AVANDIA can cause or worsen heart failure.  If you have severe heart failure (very poor pumping ability of the heart) you cannot be started on AVANDIA. AVANDIA is also not recommended if you have heart failure with symptoms (such as shortness of breath or swelling) even if these symptoms are not severe.

AVANDIA may increase your risk of other heart problems that occur when there is reduced blood flow to the heart, such as chest pain (angina) or heart attack (myocardial infarction). This risk appeared higher in patients taking medicines called nitrates or insulin.

If you have chest pain or a feeling of chest pressure, you should seek immediate medical attention, regardless of what diabetes medicines you are taking. If you take AVANDIA, tell your doctor right away if you: have swollen legs or ankles, a rapid increase in weight or difficulty breathing, or unusual tiredness; experience changes in vision; become pregnant.

Before taking AVANDIA, review your medical history and tell your doctor if you:
•    Have heart failure or other heart problems, or are on any medicines for high blood pressure, high cholesterol or heart failure, or for prevention of heart disease or stroke.
•    Take insulin or nitrate medicines.
•    Have a type of diabetic eye disease called macular edema.
•    Have liver problems or had liver problems while taking REZULIN® (troglitazone).
•    Are pregnant or planning to become pregnant.
•    Are breastfeeding or planning to breastfeed.

Women taking AVANDIA should know that AVANDIA may increase the risk of pregnancy. More fractures have been observed in women taking AVANDIA. Other possible side effects of AVANDIA include anemia and hypoglycaemia. Your doctor should do blood tests to check your liver before you start AVANDIA and during treatment as needed.

For more information about AVANDIA, please see Medication Guide or full Prescribing Information at

You are encouraged to report negative side effects of prescription drugs to the FDA.
Visit, or call 1-800-FDA-1088.


Don’t lose touch with CardioBrief. Click here to sign up for a daily email newsletter.



  1. […] Rosiglitazone goes on the RECORD, but is it a hit? […]

  2. […] Rosiglitazone goes on the RECORD, but is it a hit? (June 5, 2009) […]

Speak Your Mind