We are not supposed to criticize Nissen and group for their potential conflicts of interests as a result of taking fees from Takeda for the JAMA meta-analysis, and acting as principle investigators in other Takeda/Actos trials [e.g.ongoing PERISCOPE trial]. However, the whole tone of the article reporting the latest Nissen study of Actos is completely different than the tone taken by the same authors in their frontal attack on GSK/rosiglitazone in the May 2007 NEJM meta-analysis. For example, they completely slammed GSK for purportedly not publishing results of their clinical trials. This was in fact incorrect as many of the studies used in that review have in fact been published. However, they make not a peep about the fact that of the 19 trials used in the Actos meta-analysis, only 2 have been published. Further, Takeda was highly criticized for this failure to publish trial results in the July 2007 FDA hearings.

In the accompany JAMA rehash of the NEJM meta-analysis of Avandia/rosiglitazone, over and over the numbers are butchered (again) in favor of the controls and against Avandia?!?! For example, in the /211 (Dargie) clinical trial with NYHA Class I&II subjects, Singh & Furberg report 5 MIs in the Rosi group and 0 in control, when the actual numbers in the clinical trial reports on which they claim they are relying reflect 4 and 2 MIs, respectively. Further, the 19 heart failures (HF) in the Rosi group include the 4 MIs, effectively inflating the HF numbers. Similarly, the numbers for the DREAM (Gerstein) study are misreported: 16/9 MIs vs actual 15/11 MI’s. The ADOPT (Kahn) nos. for HF and all-cause mortality also appear to be misreported; although the nos. are in favor of Rosi, that favor has been minimized by incorrect reporting (e.g. all-cause mortality 34/62 reported vs 12/36 actual). The RECORD (Home) nos. include both adjudication and nonadjudicated cases for all categories (when adjudicated nos. are available). This is just not the best science from a publication point of view. Given there are only four trials, with very few adverse events, even the difference in a few misreported events can significantly skew the significance (p) of the results, or the finding on heterogenous. Again, the meta-analysis lumps diabetics with non-diabetics. Only short shift is given to how heterogeneous analysis was done. Duration of diabetes was not considered, which has previously been found to make the groups heterogeneous. Meta-analysis of these same studies has previously been found to be heterogenous by the Cochrane group, whose methods are the ones the authors are supposedly using. This further report does not make the prior numbers “more robust ,” but is just another misleading rehash.