Reevaluation of CV Outcomes in the RECORD Trial
Reevaluation of CV Outcomes in the RECORD Trial
The reevaluation of RECORD CV outcomes included a blinded systematic review of the original RECORD database and study documents supplemented by an attempt to contact study investigators for additional participant follow-up information. Because of logistical challenges, only a modest number of additional person-years of information were obtained compared with the initial RECORD data. The results of this reevaluation and analyses provide several key findings. First, results using the DCRI CEC data including extensive sensitivity analyses were consistent with the originally published RECORD results and did not show statistically significant differences between the rosiglitazone and the metformin/sulfonylurea groups for the composite end point of CV death, MI, or stroke and the individual components. Second, a comparison of event adjudication by 2 CEC groups using the same end point definitions showed only a small increase in the number of stroke or MI events reported by the DCRI CEC group compared with the original RECORD CEC group. The DCRI CEC group adjudicated more deaths as CV deaths because physicians were more likely to adjudicate the cause of death as unknown rather than non-CV cause when limited information was available and deaths with unknown cause were included in the CV death analyses per protocol. Third and finally, comparative treatment results were similar using a contemporary set of end point definitions being developed by the FDA compared with the historical definitions.
Conclusions
The reevaluation of RECORD CV outcomes included a blinded systematic review of the original RECORD database and study documents supplemented by an attempt to contact study investigators for additional participant follow-up information. Because of logistical challenges, only a modest number of additional person-years of information were obtained compared with the initial RECORD data. The results of this reevaluation and analyses provide several key findings. First, results using the DCRI CEC data including extensive sensitivity analyses were consistent with the originally published RECORD results and did not show statistically significant differences between the rosiglitazone and the metformin/sulfonylurea groups for the composite end point of CV death, MI, or stroke and the individual components. Second, a comparison of event adjudication by 2 CEC groups using the same end point definitions showed only a small increase in the number of stroke or MI events reported by the DCRI CEC group compared with the original RECORD CEC group. The DCRI CEC group adjudicated more deaths as CV deaths because physicians were more likely to adjudicate the cause of death as unknown rather than non-CV cause when limited information was available and deaths with unknown cause were included in the CV death analyses per protocol. Third and finally, comparative treatment results were similar using a contemporary set of end point definitions being developed by the FDA compared with the historical definitions.
