Thursday, November 19, 2009

Merck's Massive "Head-to-Head" IMPROVE-IT Study May Never Reach A Meaningful Endpoint


Given the outstanding quality of my anonymous commenter audience of late, I probably ought to just let my commenters write this blog -- as they are far more adept at pointing these things out, and succinctly so, than I -- very nice work, "Anonymous" [BTW, the image at right is a letter, from the Congressional investigations begun last year into these matters -- click it to get all the background on other arguable cholesterol-study shenanigans, circa Summer of 2008]:

. . . .Mark these words: IMPROVE-IT will be at the very best a neutral study -- it will show no definitive difference at all between Vytorin and generic simvastatin. Given the number of events recorded to date (~2,300) in 15,000 patients enrolled and with enrollment having started in October of 2005, the PIs retained by Merck actually should have enough statistical power to detect a clinically meaningful difference in events. [Editor's Note: Remember here also that the PIs unblinded the IMPROVE-IT data, to an independent safety panel, immediately after the SEAS cancer signal -- and then the IMPROVE-IT study size was increased. Interesting, no? -- But back to the main story, now.]

The problem is, they haven't, and this is mainly due to two arguably errant assumptions by Merck's study designers:

(1) all LDL-C lowering is not equal in terms of its association with event reduction, and

(2) Merck's assumed relationship between LDL-C reduction and cardiovascular event risk is not linear.

On the latter point, Merck believes that for every 1.6 mg/dL decrease in LDL-C, cardiovascular event risk is reduced by a relative 1%. What Whitehouse Station is too blind to see is that this relationship does not hold true at the LDL-C values expected in this study (baseline ~100 mg/dL, and on-Rx values in the 50-65 range. In fact, this relationship is actually curvilinear, and the cardiovascular event reduction gets progressively smaller and smaller as LDL-C declines below 100. The "joke" will be on New Merck. It would be funny if it weren't for the tens of millions of patients that will get Zetia/Vytorin instead of a more beneficial therapy (niacin, anyone?) at an aggregated spend of $4 billion/year until we find that out.

-- Anonymous, November 18, 2009 10:15 PM. . . .

I think that has it just about perfectly right. For its part, New Merck is supposed to release a newly revised/projected end-date for IMPROVE-IT in the coming weeks, according to on-the-record remarks made by the study PIs, at AHA in Orlando, this week. Will the new date be in 2013? 2014? Or later? We'll see -- but my money's on the above analysis, regardless of what Merck "forecasts" (or guess-timates, more pointedly) here, as 2009 draws to a close.

3 comments:

Anonymous said...

Check out the sidewiki.

http://www.saphris.com/saphris/index.jsp

Condor said...

I am still making up my mind about whether I trust Google to reamin true to its motto: "Don't Be Evil". . .

So, I'll likely uninstall the Google Toobar, and with it, the so-called "Sidewiki" capability. . . but for those that use it, here is the link I think the anonymous commenter intended for you to see.


[To be a little bit less opaque, here -- I am just a little skeptical that this -- 'net wide data posting won't also (eventually) become a data-mining operation, from the back-end, inside Google.]

Namaste

PS: So, net-net, I'll stick to blogging about all these pages -- in an arena where I decide how far and wide my identifying details are spread. Though the direct comment feature is intoxicating, for its immediacy, its ubiquity -- and yes, its intimacy.

Anonymous said...

I understand the anticipation of IMPROVE-IT because it alone will answer the question of whether zetia works at all. But what about SHARP? Isn't that the next study to anticipate? Sadly it will not provide any data on zetia- zetia is just attached to the study without its own control group. There is a 1 year lipid arm of simva alone in 1000 patients - who then get randomized to placebo or combination. That would have been the correct design because the present design is effectively worthless when it comes to zetia. What a waste. Will they even report the 1 year event data for simva alone vs. simva + zetia? More can be found at:
http://www.sharpinfo.org/