Skip to main content
Menu

Cardiovascular Disease

Vytorin, Holed Under the Waterline

Merck and Schering-Plough have released the data on a study of genetically high-LDL patients taking a statin alone (Zocor, simvastatin) or the combination of the statin and Schering-Plough’s cholesterol absorption inhibitor (Vytorin, simvastatin and ezetimibe). Vytorin has a good share of the market, and has already been shown to lower cholesterol.
And so it did this time: the Vytorin patients showed a 58% decrease in LDL, while the Zocor group showed a 41% reduction. But this trial went further, looking at the growth of atherosclerotic plaques. You’d figure that a greater decrease in LDL would mean a greater decrease in the size and growth of plaques.
You’d be wrong. The Vytorin group’s carotid arteries, measured in a standard way (intima-medial thickness, IMT) came out as 0.0111 mm, while the Zocor group’s came out as 0.0058 mm. This is making the headlines as “twice as bad as Zocor”, but the difference actually isn’t statistically significant (p = 0.29). Steve Nissen of the Cleveland Clinic is quoted as saying that this is “as bad a result for the drug as anybody could have feared”, but that’s not quite right. If that p value had been, say, 0.01, that would be worse. Strictly speaking, you can’t call the two groups different. They don’t seem to have been different in cardiovascular outcomes.
But here’s the real point: that’s bad enough. The whole point of Vytorin is that it’s supposed to be more effective than a statin alone, and what you can say about this trial is that it sure didn’t prove that. But that carotid artery thickness is definitely a concern – the numbers appear to have big error bars on them, but they’re certainly not pointing in a good direction. And it’s going to be difficult, perhaps impossible, to ever know if that effect is real, because it’ll be mighty hard to get another trial of this sort off the ground after results like this. How can you enroll a treatment group for a drug that has been shown to have no benefit?
Well, OK, there’s that LDL reduction. But the downstream clinical data (the artery measurements and outcomes) overrule that. The point of taking a cholesterol medication is not to make your lab test numbers go up and down, the point is to have fewer heart attacks and strokes. We use those blood lipid numbers as a convenient surrogate, but it’s been obvious for a long time now that we have, to put it delicately, an imperfect understanding of their relevance. Data closer to real mortality and morbidity outcomes will win.
Now what? This is clearly terrible news for Merck and (especially) for Schering Plough. The companies already were under pressure for having taken so long to work up the data for this trial, which delay ended up just drawing even more attention to these bad results. Now, how do you go out and sell Vytorin (or Zetia, the cholesterol absorption inhibitor alone)? Why do insurance companies have motivation to pay for it? And when are we ever going to understand the complexities human lipid behavior and cardiology?
More on ezetimibe, written in happier days, here , here, here, and here. .

35 comments on “Vytorin, Holed Under the Waterline”

  1. Petros says:

    While it’s good that these data have finally been released, the long delay of reporting essentially negative results is likely to worsen Merck’s reputation further.

  2. Bryan says:

    One thing of note is that this is a comparison between two cholesterol lowering drugs in people with extremely high LDL-C levels. The National Cholesterol Education Program would still classify these individuals after treatment as having high cholesterol. Many studies have shown the benefits of lipid lowering therapies in reducing coronary heart disease and coronary events. It seemed that Merck/Schering may have also been trying to get a fast track to showing benefit of Vytorin through an orphan disease clinical trial (saving time and money). We shouldn’t throw the baby out with the bathwater over these unsatisfying results.

  3. lipiddoc says:

    An intriguing hypothesis may that the beneficial effects of statins is not only from lowering LDL, but from decreasing other markers such as inflammatory markers. All the statin trials show higher dose statins with greater ldl-c reduction decrease CV events so we have concluded that lower LDL levels are associated with lower risk. However this trial shows that lowering ldl in a very high risk population does not reduce events or CIMT. Just a thought.

  4. Bryan says:

    lipiddoc…I assumed this trial was simply comparing a statin to a statin+ezetimibe. Correct me if I’m wrong but this does not rule out that ldl lowering reduces events; they only compared events between the therapies. (also sorry for the multiple posts)

  5. RKN says:

    The Vytorin group’s carotid arteries, measured in a standard way (intima-medial thickness, IMT) came out as 0.0111 mm, while the Zocor group’s came out as 0.0058 mm. This is making the headlines as “twice as bad as Zocor”, but the difference actually isn’t statistically significant (p = 0.29).
    Another to way to put this would be to say Zocor was twice as effective as Vytorin. Which is an an encouraging finding, especially for people like me who take Zocor, especially since at Costco the generic is ~$20 for 200 20mg tablets, about a year’s worth of pills.

  6. John Johnson says:

    Yeah, and cholesterol levels are the better understood biomarkers.
    This is why clinical trials based on survival rule the day in the more serious indications — at the end of the day nobody cares what lab levels are and they are not as cleanly correlated with “hard” outcomes as we would like.

  7. David Hungerford says:

    Complete layperson here with possibly stupid question: I assume the arterial wall thicknesses were comparable across groups going into the study?

  8. Kirsten says:

    Higher dose statins also mean more hepatic side effects. I seem to recall a recent revelation that Zetia wasn’t as easy on the liver as was once thought either. No point is using a combination if there isn’t a clear benefit.

  9. brian says:

    It would appear that it is the pathway (HMG-CoA reductase) that statins inhibit that mostly contributes to their effectiveness.
    And, here’s an interesting thought: drugs that decrease LDL-D or total cholesterol in other ways may actually make things worse! Why? Because the liver and the rest of the body make cholesterol via the above pathway and the presence of cholesterol (from LDL absorbed via LDL receptors, for example) inhibits the pathway. Most pathways tend to work this way.
    In the case of FH patients, this feedback mechanism is a little bit broken, because the LDL receptors fail to take up their normal quota of cholesterol back into the liver cells for eventual excretion into the bile, leaving it circulating in the blood. So, to the liver, this ironically looks like a shortage of cholesterol, causing it to pump out even more. Statins inhibit the production of cholesterol by jamming the pathway by which it is made at HMG-CoA reductase.
    However, Zetia doesn’t work that way. It blocks the absorption (effectively, it causes it to be pumped right back out.) It doesn’t do anything to block HMG-CoA reductase or the rest of the pathway that makes cholesterol. In fact, as there is never-the-less cholesterol circulating, leaving less for the LDL receptors to collect, perhaps it actually somewhat STIMULATES the pathway, leaving it to do its other dirty deeds, while still making LDL-C appear to be lower in the blood…
    It is just possible that we may be witnessing some new science here, folks. Interesting…

  10. weirdo says:

    “Another way to put this would be to say that Zocor was twice as effective as Vytorin”
    Well, you could put it that way, but you would be 100% wrong. This statement by RKN is a prime example of why, in general, the average layperson cannot interpret the result of scientific experiments in a meaningful way. No offense, but it’s true.
    To compare relative effectiveness, one would have had to compare both treatments to placebo which, of course, would have been unethical given the condition of these patients and the presence of an efficacious treatment (statins alone).
    Too, as Derek points out, the IMT differences were not statistically significant. The trial was insufficiently powered to demonstrate a difference, it turns out.
    And that, in the final analysis, is the most damning result. The trial was SUPPOSED to show a difference. It didn’t. In a perfect world, the results would be used to design another study to follow up on the trends.
    But would that be ethical? Isn’t it enough to know that Zocor alone IN THIS PATIENT POPULATION is at least as good (if not better than) Vytorin?
    Doesn’t mean they shouldn’t be tried head to head in other populations.
    Oh, and to David Hungerford, yes, the press release (on the MRK website, anyway) specifically states the two groups started out with very similar IMT (0.68 vs. 0.69).

  11. shane says:

    Geez…..heaven forbid anyone actually improve their diet when there is a smorgasboard of dubious pills on offer.

  12. Pharmacoepidemiologist says:

    Seems to me the winner (and there IS a winner) is Abbott, which bought Kos a couple of years back. Kos doesn’t play around with combining a statin and Zetia–no, it combines a statin and niacin, which seems to work very nicely at reducing MIs. Niacin’s looking very good these days. Indeed.

  13. SNP says:

    To compare relative effectiveness, one would have had to compare both treatments to placebo which, of course, would have been unethical given the condition of these patients and the presence of an efficacious treatment (statins alone).
    For all the name-calling you’re doing, I’m not sure that’s the right answer either. I’d define relative effectiveness (in the sense he used it) as change from baseline in one treatment versus change from baseline in the other. Unless patients on placebo would be rapidly regressing, you don’t need a placebo.
    Of course the real real answer is that the difference is not statistically significant.

  14. emjeff says:

    Derek;
    This is not the place to go off on a rant about p-values (though I am sorely tempted), but the real issue here is not that the p-value is significant or not. The issue is why the combination is doing worse than the control (by a wide margin). How do we explain this biologically? Does it make sense? What is the variability in the response among the two groups? The p-value is not a mystical number which answers any of these questions. I agree that it is bad news for Vytorin, but not because of the p-value, it’s because one of the means is twice the other.
    WhatI’d really like to know about this study (well, all studies) is the probability that the control is better than the combo, as well as the probability that it is worse. You can’t get those probabilities from standard methods; you have to use Bayesian statistics, which still frightens people for some reason.
    See Goodman SN Arch Int Med (1999) 130:995-1004. for an excellent discussion of why p-values, in general, are not worth the electrons used to create them.

  15. HelicalZz says:

    “The Vytorin group’s carotid arteries, measured in a standard way (intima-medial thickness, IMT) came out as 0.0111 mm, while the Zocor group’s came out as 0.0058 mm. ”
    That is not quite accurate. It is an increase over baseline level. A change level.
    So the Vytorin/Zocor group went from a baseline of 0.68 and increased 0.0111 (0.691), while the Zocor only group increased less, but from a higher baseline of 0.69. 0.69 + 0.0058 for 0.696. Quiblling and the overall point of ‘not better’ is still solidly in place.

  16. qetzal says:

    emjeff’s cite should be:
    Goodman SN, Ann Intern Med (1999) 130:995-1004.

  17. Christiane says:

    Anyone ever look at the effect of Zetia on B12? My dad was taking Zetia, and was suffering the signs of some severe B12 deficiency (even taking into account age, the decline was pretty sharp), and was told he’d have to do IM injections of B12. Of course he did not want to do daily shots, so I asked him to ask his doctor if by any chance the Zetia could be interfering with his absorption of B12, and if he could do an experiement: stop taking Zetia and take oral supplements. He got his doctor’s blessing. After a few weeks of this, his B12 levels went back into the normal range. The next step is for him to start taking Zetia again and see if he again suffers B12 deficiency. Yes, I know B12 is absorbed through the terminal ileum and Zetia mainly works in the jejunum, but it does have some action in the ileum as well, from what I read. I’d welcome the insight of any of the true experts here.

  18. Thom says:

    #6 Points out one of the most important implications here. There has been tremendous hype in the last few years about how biomarkers are going to pave the way to faster cheaper drug development. If we can no longer trust well established markers like LDL (and the Avandia controversy raised doubts about HbA1c) then it seems unlikely that biomarkers are going to solve the pipeline problem as their proponents promise. If anything they may lead drug developers down further blind alleys.

  19. milkshake says:

    Zetia story is quite sad: like with Vioxx the medicinal chemists were asked to solve a difficult problem, they came up with a beautiful solution – a drug that was doig cleanly what it is supposed to do – and now this
    it is frustrating, spending the years of effort on an exciting project only to learn that the underlying biology does not behave according to the plan.

  20. Chris says:

    Can someone fill me in on what the Avandia / HbA1c study found?

  21. TFox says:

    #14: “the real issue here is not that the p-value is significant or not. The issue is why the combination is doing worse than the control…”
    *sigh*. I’d think the first question would be whether the data support the statement that the combination is worse than the control, before we try to understand why. That’s what looking at the p-value tells you. In this case the answer is no, the groups are equivalent, as far as this study can tell. Bayesian methods would tell the same story using slightly different words; they are, after all, looking at the same data.

  22. jonadab says:

    Niacin? Don’t you get that from eating vegetables? Or is Kos talking huge amounts of it or something?
    emjeff, you seem, to an educated layman, to be ignoring or handwaving away basic principles of statistics. Exactly what constitutes a margin that we can be at all confident about (much less a “wide margin”) is very much dependent on things like sample size, deviation from mean within each group, and so forth. P-values may not be a perfect measure of statistical significance, but the whole concept is certainly not as irrelevant as you seem to make it out to be.

  23. RKN says:

    Well, you could put it that way, but you would be 100% wrong. This statement by RKN is a prime example of why, in general, the average layperson cannot interpret the result of scientific experiments in a meaningful way. No offense, but it’s true.
    I may be wrong, but it would not be because I’m a “layperson”, I’m not.
    Let x = 0.0111 and y = 0.0058,
    then x/y~ 2.
    Saying y is ~half of x is equivalent to saying x ~twice y. That’s all I was saying.
    The trial was insufficiently powered to demonstrate a difference, it turns out.
    There’s insufficient information at the link Derek provided to allow you to conclude that. Statistical power estimations require a measure of population variance and expected type II error. How do you know what the statistical power of this study was? I would be surprised if a 720 patient cohort didn’t get them to ~0.8, but I can’t tell w/o knowing all the relevant values.

  24. Ian Musgrave says:

    Emjeff wrote:

    I agree that it is bad news for Vytorin, but not because of the p-value, it’s because one of the means is twice the other.

    Actually, whether you use P values or Bayesean factors (which are a right pain, let me tell you), the conclusion is that the mean increases are not different (ie one is not twice the other). Bayesean analysis tends to be less forgiving than frequentist P analysis, and when you factor in prior knowledge about the role of LDL levels and intimal thickness, it is even worse.
    Both Baysean analysis and Fequentist analysis agree, there is no difference between simvastatin combined with ezetimibe and simvastatin alone.
    Then again, most trials of statins produce a decrease in IMT, where in this one, admittedly in a difficult and aggressive disease compared to most of the situations statins are prescribed in, simvistatin had no effect on IMT (the press release does not give error values or 95% confidence intervals, but based on previous studies the “increases” in IMT are on the same order of magnitude as the SEM’s of other studies[typically 4-8 µm]).
    Given that you need a high dose of simvistatin just to hold back the progression of IMT in these patients, where in any other patient groups much lower doses could be expected to produce falls in IMT of 24 µm (compared to the “increase” of 5 µm seen in this study see Current Controlled Trials in Cardiovascular Medicine 2005, 6:3) the very modest lowering of LDL found in the simvastatin combined with ezetimibe group compared to simvistatin alone would not really be expected to produce a dramatic difference in IMT, certainly not one that could be picked up in a trial this size.
    For other patient groups, simvastatin combined with ezetimibe may possibly be more effective, but then there have been surprises before, such as the ACAPS study, where the addition of warfarin to lovistatin markedly reduced the efficacy of lovistatin on IMT.

  25. Malcolm says:

    On the other hand, IMT is just another surrogate outcome (I don’t know — perhaps even a worse one than LDL-C).
    AFAIK, noone has yet done a trial with either ezetimibe or ezetimibe+simvastatin to demonstrate a reduction in mortality and/or cardiovascular events relative to placebo or a statin alone. ENHANCE definitely wasn’t powered to do this.
    In my eyes there’s never been a case for ezetimibe to be considered a first-line drug when the alternatives (statins) have oodles of outcome data on hard endpoints, not to mention a long clinical track record.

  26. Barry says:

    14. emjeff on January 15, 2008 9:28 AM writes…
    ….. See Goodman SN Arch Int Med (1999) 130:995-1004. for an excellent discussion of why p-values, in general, are not worth the electrons used to create them.
    It should be Ann Int Med (1999) 130:995-1004., and it is an excellent reference

  27. Clotted Wretch says:

    The NY Times journal of sighence weighs in…
    http://www.nytimes.com/2008/01/16/opinion/16wed2.html

  28. mad chemist says:

    This is simultaneously troubling and amusing. I have observed (over several years) individuals with low cholesterol (and LDL levels) dropping like flies from totally unexpected MI’s, yet others with high serum cholesterol lasting well into old age.
    So, where did all this cholesterol go, and what impact did lowering it have on the cardiovascular conditions of these patients? Did we medicinal chemists unwittingly effect reduction in the serum concentrations by simply depositing it on arterial walls? Excuse the facetious/snotty concept here, but the dissonance between heart disease prevalence and drug therapies has been rather confounding.

  29. Ian Musgrave says:

    Barry wrote:

    14. emjeff on January 15, 2008 9:28 AM writes…

    ….. See Goodman SN Arch Int Med (1999) 130:995-1004. for an excellent discussion of why p-values, in general, are not worth the electrons used to create them.
    It should be Ann Int Med (1999) 130:995-1004., and it is an excellent reference
    Although when you go through it in detail, especially the companion paper (Ann Int Med (1999) 130:1005-1013) where Bayesian and Frequentist approaches are contrasted, there is a fair bit of smoke and mirrors and vigorous beating of straw men (eg page 1002 where he mistakes fallacious habits of the statistically ignorant for a tenet of Frequentism. If it’s not significant it isn’t a trend). Especially hypothetical statement 2 (pg 1011), where a straw man is beaten to death. It is clear under frequentist rules that the result in hypothetical statement 2 is a false positive, and the Bayesian analysis is the same as a frequentist approach when confronted with such data.
    Hypothetical statement 1 is somewhat more complex. A frequentist given the same result and the same background data would argue along similar lines, but the Bayesian analysis is critically sensitive to the prior probabilities, which are plucked from thin air in this case. Also, as a pharmacologist I cringed at the “it works in blood vessels so this should be added to the positive evidence” statement. Anyone with even a vague understanding of pharmacokinetics can understand why something may work well in vitro and fail completely in vivo.
    In explicit head to head comparisons Bayesian and frequentist approaches yield fairly similar results (though as I have said, Bayesian analysis tends to be more conservative (see Am Coll Cardiol, 2004; 43:1929-1939, eg Fig 6). Again, the choice of priors critically influences the results, and there is no clear method for rigorously evaluation what your prior is on background knowledge.
    Far from being “not worth the electrons”, P values can be very helpful indeed. Things that are very non-significant in Frequentist approaches are very poor evidence in Bayesian approaches, and things that are very significant in Frequentist approaches are very good evidence in Bayesian analysis. It is the borderline area where things are problematic, but they are problematic for both approaches, and careful weighing of data and the logic of inference should be done no matter what statistical approach is taken. No approach, Frequentist or Bayesian, will be helpful if biological reality and good experimental deign are overlooked.
    Again, in this particular case, both Frequentist and Bayesian analysis agree that the combined treatment and simvistatin alone were not different.

  30. Ian Musgrave says:

    The Mad Chemist writes:

    This is simultaneously troubling and amusing. I have observed (over several years) individuals with low cholesterol (and LDL levels) dropping like flies from totally unexpected MI’s, yet others with high serum cholesterol lasting well into old age.

    It’s like smoking and lung cancer. Smoking greatly increases the risk of lung cancer, but there will be individuals who due to genetic or environmental influences (or a combination of both) will develop lung cancer without smoking, and those who due to genetic or environmental influences (or a combination of both) can smoke like chimneys and never get cancer.
    Cholesterol and LDL levels are very significant risk factors, but some folks will be predisposed to MI even with low serum cholesterol or LDL due to other genetic and environmental factors (eg air pollution, mutations in various collagen genes).
    In the case of statins, such as simvistatin, the cholesterol doesn’t go anywhere, the actual synthesis of cholesterol is reduced. Cholesterol levels go down because it isn’t being synthesised in the first place (and what little is synthesised is shuttled off to critical physiological functions).
    So the cholesterol is not going to the artery wall. Indeed in patients treated with statins, on average existing plaques are reduced in size, and new ones are not formed.
    Again, due to genetic and environmental factors, not all patients will respond in the same way, but statins substantially lower the incidence of myocardial infarction (and stroke etc.).

  31. Patrick says:

    “…but statins substantially lower the incidence of myocardial infarction (and stroke etc.).” Ian
    Where is the evidence that statins have any effect on heart attacks?
    A few months ago Pfizer ran an expensive ad campaign touting that Lipitor reduced heart attacks by 67%. When I checked their data, it was clear to me that the 67% claim was just statistical noise. The Pfizer study used patients with multiple risk factors: Half the patients used a placebo, and 3% of them had heart attacks, while 2% of the half who used Lipitor had heart attacks.
    This “improvement” is utter nonsense, and all it really shows that the correlation between LDL levels in the blood and heart disease is very weak indeed.

  32. Anonymous BMS Researcher says:

    I’m surprised nobody here has mentioned the PROVE-IT study, in which my company, BMS ended up proving that in the case of Lipitor versus our Pravachol lower bad cholesterol really was better. Since the study went directly against our economic interests (we had obviously been hoping Lipitor would not turn out superior in preventing events), it had considerable credibility in the medical community.
    BMS has run a large number of trials giving very strong evidence that statins save lives, more I think than most statin makers. Before the results came out, some of those who like to criticize our industry claimed PROVE-IT was deliberately underpowered in the hope of failing to find a difference — but we ended up publishing strong evidence that Lipitor really was better than our product. Now of course Pravochol has gone generic, but Pfizer and patients both still benefit from our science!
    Is GM ever expected to run head-to-head scientific comparisons of their products against Toyotas, and then publish proof that Toyota makes a better product should that turn out to be the case? We pharma companies routinely do such things and accept the consequences when the science turns out differently than we expect.
    We at BMS honestly thought Pravachol and Lipitor had similar efficacy despite Lipitor getting better cholesterol numbers, we would not have done the study had we not thought so — and when the science did not go our way, we still published the results.

  33. Ned says:

    See the following discussion (http://www.theheart.org/article/837867.do Signup is free) for some perspective on this issue, including the three large well powered outcomes studies on ezetimibe currently going on (e.g.; IMPROVE-IT). In particular, highlight the fact that this study uses imaging as surrogate endpoint. To quote Dr Sanjay Kaul (UCLA), “Embracing a therapeutic intervention as a resounding success or rejecting it as a disappointing failure simply on the basis of an imaging outcome study is rather short-sighted and betrays a higher confidence in surrogacy than is otherwise warranted by the existing body of evidence. Bottom line, surrogate end points are a slippery slope. We need to await the results of clinical-outcome studies before drawing any firm conclusions.”

  34. SNP says:

    When I checked their data, it was clear to me that the 67% claim was just statistical noise. The Pfizer study used patients with multiple risk factors: Half the patients used a placebo, and 3% of them had heart attacks, while 2% of the half who used Lipitor had heart attacks.
    Patrick, that’s not “statistical noise”, or at least you can’t identify it as such from anything you mentioned. I think what you mean is that the Number Needed to Treat is a lot less impressive than the risk reduction figure by itself.

  35. BCP says:

    Well said, anonymous BMS researcher.
    Anything about “%reduction in LDL” is meaningless without understanding the absolute conc’s of LDL-C. There has historically been a good correlation with lower LDL-C concentration and reduced mortality (although there is always a residual subset of patients who have low LDL and still suffer MIs).

Comments are closed.