5 different ways of spinning RECORD

The FDA Avandia briefing document (see previous post) contains 5 separate reviews of the RECORD trial. At an FDA press briefing Janet Woodcock predicted, with perfect understatement, “there will be a long discussion of RECORD.” Thomas Marciniak, an FDA reviewer who is well-known for his skepticism, was the most critical of the trial. The other reviewers were more generous.

Here are excerpts from the summaries of the 5 reviews of RECORD:

Thomas Marciniak, medical team leader of the division of cardiovascular and renal products, performed the review of cardiovascular events in the trial:

Our summary is that RECORD was inadequately designed and conducted to provide any reassurance about the CV safety of rosiglitazone. The results do confirm and extend the recognized concerns regarding increased heart failure (HF) and HF deaths with rosiglitazone. They also suggest that rosiglitazone increases the risk for myocardial infarction (MI), although the confidence intervals for estimated MI hazard ratios are wide and include no risk while biases in the study suggest that the true risk could be higher.

Norman Stockbridge, director of the division of cardiovascular and renal products, commented on Marciniak’s report:

Dr. Marciniak identified numerous valid issues with the design and conduct of RECORD. These issues leave much in doubt and thus undermine one’s ability to take much comfort from the safety findings. It is less clear to me whether, despite its shortcomings, one can interpret RECORD as showing new evidence of the harm of rosiglitizone.

Ellis Unger, deputy director of the Office of Drug Evaluation, wrote:

…the results on the primary endpoint (time to first cardiovascular death or cardiovascular hospitalization) are not entirely conclusive. Conversely, the findings on all-cause mortality seem readily interpretable and important in RECORD, and they favor rosiglitazone – almost reaching statistical significance (hazard ratio 0.86; 95% CI 0.68 to 1.08, p=0.19). If the almost significant excess in cardiovascular mortality for rosiglitazone found in the meta-analysis of Nissen and Wolski is viewed as a hypothesis for future study, that hypothesis is not substantiated by the results of RECORD. With respect to cardiovascular safety, the strength of reassurance provided by RECORD can (and will) be debated, but, aside from the known risk of heart failure, the study does not appear to demonstrate harm.

Karen Mahoney performed the endocrine review of RECORD:

Although the results of RECORD, on their face, would have met the Agency’s criteria for evidence that rosiglitazone is not associated with an unacceptably increased risk for major adverse cardiovascular events, there continue to be differences of opinion in this regard. The ongoing TIDE study is a large, randomized, double-blind, placebo- and active- controlled cardiovascular outcomes trial, the design of which does not contain many of the limitations of the RECORD study. It is double-blinded and compares the risk of add-on rosiglitazone not only to placebo, but also to pioglitazone, the only other approved thiazolidinedione. The study population has higher cardiovascular risk and is expected to have a higher event rate. The study is conservatively powered. This trial, if it continues to completion, has the potential to address the question of the cardiovascular safety of rosiglitazone more definitively. The Division of Metabolism and Endocrinology Products looks forward to discussion of the advisability of continuation of TIDE.

David Hoberman performed the statistical review and evaluation:

This review has found the sponsor’s analytic methods to be essentially non-controversial, and that any modifications on the margin do not materially affect the results. And that is the important issue: to what extent do either alternative statistical methods or the conduct of the trial, either designed or unanticipated, affect the results in such a way as to lead to different inferences from the ones which would be drawn from the data’s face value. The review addresses the issue of statins and finds that the evidence that differential statin use in the two arms substantially affects the results is weak. With regard to possible under reporting of events, as long as it occurs at random between the groups, the effect would be to make it harder to demonstrate non-inferiority if the groups’ risks were the same. If “inferiority” is the focus of attention, then fewer events would depress the power to show a statistically significant difference between the groups if there were one. Medication added on to dual randomized therapy, depending upon the arm in question, can confound the intent-to-treat analysis. However, there are two mitigating factors. The first is the option to analyze only the time spent on the randomized therapy. This shows no increased risk of ischaemic events on Avandia. The second is that, in both groups, the vast majority of time was spent on the randomized therapy in both groups: 83% in the control group and 75% in the Avandia group. The influence of the add-on medications after going off dual randomized therapy is not calculable, but may be small in the intent-to-treat analysis. Finally, the review considers the possibility that the trial could have reasonably demonstrated the putative inferiority of Avandia and finds that there is evidence that RECORD could have done so.

Finally, Susan Leibenhaut reported the findings of FDA inspections of 3 clinical sites, GSK, and Quintiles:

V. Summary of Inspection Results and Recommendations The findings of the inspections of clinical sites, the sponsor, and the CRO included the following:

  • Procedures, resources, and training for the conduct of the trial were adequate
  • No evidence of tampering with or falsification of data
  • No evidence of serious violations
  • No evidence that publication of the interim CV analysis in June 2007 influenced the conduct of the trial.
  • The endorsement to conduct an interim CV analysis by the 2 members of the Steering Committee and 2 members of the DSMB did not fulfill the requirements of the charters of these committees.


  1. […] J. Graham from the Office of Surveillance and Epidemiology. (See our previous posts here and here for additional coverage of the briefing […]

  2. […] FDA accepted contrarian viewpoints: Depending on your math, FDA officials had at least 5 different views on Avandia that were presented at the meeting. This kind of cacophony is counter to the […]

  3. […] 5 different ways of spinning RECORD (July 9, 2010) […]

Speak Your Mind