Prominent cardiologist: Schering-Plough could have known the result of Vytorin trial before unblinding data.

One important aspect of the "who knew what and when" story related to Schering-Plough executives selling massive amounts of SP stock in 2007 is that while the treatment arms of the Vytorin trial were blinded, the treatment effect on the primary end point—change in mean carotid intima-media thickness (CIMT) after two years—was similar in the Zocor (simvastatin-alone) and Vytorin (ezetimibe/simvastatin-combination) groups.

Some cardiologists believe that unblinding the results would have been unnecessary to know the treatment effect was neutral:

Chart courtesy of PharmaGossip.

"Somebody had looked at the end-point examination, the IMT results, and, irrespective of group assignment, could know that a groupwise comparison of CIMT changes showed no statistically significant difference," Dr Allen Taylor (Walter Reed Army Medical Center, Washington, DC) told heartwire. "In my view, once that is known, the trial is functionally unblinded."

Taylor, a researcher involved in numerous clinical trials, including his role as the lead investigator of the Arterial Biology for the Investigation of the Treatment Effects of Reducing Cholesterol (ARBITER-2) CIMT study, stressed that he has no inside knowledge of when the end-point analysis was completed and does not know whether company officials looked at the blinded data, although he believes it that remains open to speculation.

He said it would be difficult to look across all the data without assigning patients to treatment groups and make any sense of presumed outliers. He added that the 18-month gap after the trial was completed and the reporting of the results leaves the companies wide open to question, especially Merck/Schering Plough's contention that they sought to address quality issues with the images.

"Their sequence of actions to rehabilitate the data is in many ways a signal of distress," said Taylor. "What was fueling that? I can see some may readily speculate that it seems highly implausible that 18 months were spent working on the data without there being a reason related to the overall trial results being known to be negative."

Comments

[Adam]

I'm not at all sure I believe that. I've just tried simulating some data to see how obvious a group difference would be on unblinded data.

The difference between treatments would have to be pretty big before you could spot it on unblinded data, and probably much bigger than anyone would reasonably expect in a trial. I simulated data for 1000 patients with a between group difference in a continuous outcome variable equal to the standard deviation of the variable. That's a bigger difference than you would normally expect to see in a clinical trial (although I don't know what difference they were looking for here). This histogram of all the combined data was visually indistinguishable from a normal distribution. The data even passed the Shapiro-Wilk test for the normality of the distribution, even though it was technically bimodal.

So I don't believe that you could look at a distribution on blinded data and rule out a significant difference, just because the distribution looks like the topmost of the 3 curves above.

[Alan]

Adam is correct. You are looking here at the distribution of individual patient results. The treatment-group means are more tightly distributed, by roughly the square root of the number of patients in the trial. (That's about 55x for a 3,000-patient trial) So it is possible to have a very robust treatment effect and still have a unimodal curve.

On the other hand, seeing a bimodal curve in the blinded data would indicate that there's a treatment effect, unless there's a reason for the underlying data to be bimodal. So "blind" isn't really blind, but only if you see two humps.

[cheap darkfall gold]

Hey, great post, very well written. You should post more about this.

[darkfall gold]

thank you for sharing,i like.

[darkfall gold]

We can put forward a contract that says we want something for nothing, but at the end of the day that’s just not going to wor

[dungeon fighter gold]

We can put forward a contract that says we want something for nothing, but at the end of the day that’s just not going to wor