BMJ authors take back inaccurate statin safety statements

bmjcover514Last October, the BMJ published a paper by a group of researchers from the United States and Canada questioning the use of statins in patients considered at low risk of cardiovascular disease.

The article has been cited eight times since then, according to Thomson Scientific’s Web of Knowledge. It mentioned data from another study that reported a high rate of side effects in patients who used the drugs, between 18% and 20% — suggesting that those who received little or no benefit from the therapy could be more more likely to suffer harm than good.

But that citation turns out to have been flawed — prompting the journal to take the unusual step of removing those “statements” from the article and another it published about the issue that has been cited six times. And in an editorial, BMJ editor Fiona Godlee said she has asked a panel of experts to review the original paper to determine if it ought to be retracted completely:

As editor, I have a vested interest in not retracting the articles unless the case for doing so is completely clear. So I have passed this decision to an independent panel. …

This editorial aims to alert readers, the media, and the public to the withdrawal of these statements so that patients who could benefit from statins are not wrongly deterred from starting or continuing treatment because of exaggerated concerns over side effects.

According to a BMJ release about the matter:

In October last year, The BMJ published an article by John Abramson and colleagues that questioned the evidence behind new proposals to extend the routine use of statins to people at low risk of cardiovascular disease.

The authors re-analysed data from the Cholesterol Treatment Trialists’ (CTT) Collaboration. This showed no mortality benefit associated with treatment of people with a less than 20% risk of developing cardiovascular disease over the next 10 years. This has not been challenged.

However, they also cited data from a separate uncontrolled observational study showing that statin side effects occur in 18-20% of patients. This was repeated in another article published in the same week in The BMJ by Aseem Malhotra – and is the statement the authors have now withdrawn.

This error was due to a misreading of data from one observational study, and was not picked up by the peer reviewers or editors, explains Dr Godlee. “The BMJ and the authors of both these articles have now been made aware that this figure is incorrect, and corrections have been published withdrawing these statements.”

According to Godlee’s editorial, the journal was alerted to the problematic data by Rory Collins, an Oxford University researcher who led the CCT Collaboration which Abramson’s group reanalyzed. Collins, Godlee says, has requested that both the Abramson paper and the Malhotra article be retracted, not corrected.

Godlee, in an artful but appropriate punt, says:

I have decided that the right thing to do is to pass this decision to an independent panel. Iona Heath, former chairwoman of the Royal College of General Practitioners and of The BMJ’s ethics committee, has agreed to chair the panel, whose members will include people with no “dog in this fight,” but with expertise in clinical trial and observational study methodology, and in designing and implementing editorial policies on retraction. Full details of the panel and processes will be published shortly. I have asked that all submissions to the panel be placed in the public domain on bmj.com, and I have committed to implementing the panel’s recommendations in full.

 

9 thoughts on “BMJ authors take back inaccurate statin safety statements”

  1. Editorial by Fiona Godlee of BMJ has just vanished from the BMJ website… page cannot be found… What has happened?

    1. Thanks for flagging that. The BMJ tells us it’s a technical problem with the site and that they’re working on it.

        1. One should also be aware that Fiona Godlee (http://en.wikipedia.org/wiki/Fiona_Godlee) is no stranger to controversy or criticism:
          http://vran.org/in-the-news/bmj-on-the-hot-seat/
          That lawsuit, by Andrew Wakefield against Godlee and BMJ, stated that “The complaint is that the defendants authored, edited and approved articles and made statements which “contained false and defamatory allegations”.
          That lawsuit, however, was thrown out:
          http://www.medscape.com/viewarticle/768700
          A strange criticism of Godlee exists:
          http://pjsaunders.blogspot.jp/2012/06/why-is-editor-of-bmj-making-such-strong.html
          Godlee is also the former Chair of COPE* and has been the EIC of BMJ since 2005**.
          * http://publicationethics.org/annualreport/2005/3
          Maybe COPE’s member publishers, including Elsevier Ltd., the largest paying client/member, should think about adding details about their editors, including COIs, as Godlee has done for her profile:
          ** http://www.bmj.com/about-bmj/editorial-staff/fiona-godlee

  2. These studies are not divine slate tablets and should be scrapped when even a minor flaw is found.

  3. Rory Collins is (if you believe the internet) on record as saying:“Statins should be given to all over-50s, regardless of their health history, because they dramatically cut the risk of heart attacks and strokes in later life.”

    I can’t read the BMJ paper because it is paywalled but I am guessing that there was a misunderstanding of 10% increased risk of diabetes with statin use, somehow morphed into 10% risk of diabetes with statin use. With the diagnosis of diabetes in these groups being around 5% in placebo and 5.5% in treated.

    The Cochrane review on this issue is here:
    http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD004816.pub5/full
    It states that only 2 trials they included for review reported on type 2 diabetes – with the observed increase in diabetes mostly coming from the JUPITER trial – which was also using the higher dose of statin and reporting the most protective effects.

    One of the papers from the JUPITER study is here
    http://www.ahjonline.com/article/S0002-8703%2811%2900262-6/fulltext#s0040
    It reports 254 diagnoses of diabetes in the statin arm and 206 in the placebo arm.
    It also reports 58 MI, stroke, or CVD deaths in the statin arm and 113 in the placebo arm.

    Of more concern is it reports 56 deaths from causes other than MI, stroke or CVD in the statin arm and only 27 in the placebo arm.

    What would concern me is that there might be some hidden bias in the study design whereby someone enrolled in the study might be less likely to seek statin treatment than he/she might otherwise do – as of course they were blinded as to the treatment they were receiving (but their GP ought to have been able to guess based on their blood results). And while I don’t claim to be expert in such things I can’t see a mention of a protocol whereby participants would automatically exit the study if their lipid levels began to rise beyond a certain point:
    “The primary CVD study end point included the development of a first nonfatal MI, nonfatal stroke, hospitalization for unstable angina (UA), arterial revascularization, and confirmed cardiovascular death. All-cause mortality, venous thromboembolism, and physician-diagnosed diabetes were prespecified secondary end points of the trial.” But I welcome correction on this point.

    And if there was an inhibition of normal prescribing procedures due to participation in the study, this would might tend to accentuate the difference in cardiac events. In other words good primary health care and preventative medicine might be just as effective in managing blood lipid level related risk (including statins once a certain level is reached), but the risk of diabetes and the unexplained rise in deaths from other causes associated with unconstrained statin prescription can’t be managed.

    1. In terms of outcome of death I only used the whites group of the JUPITER study. With non-whites included the figures are
      MI, stroke, or CVD deaths statin arm: 83, placebo arm: 157
      Deaths other than MI, stroke and CVD: statin arm: 115, placeb arm: 90
      The diabetes diagnoses already included all ethnicities.

  4. Leaving aside the controversy about the statements under question being valid or not (IMO they can be supported) WHAT IS CLEAR is that the inclusion was not a failing of the Peer reviewers – They never saw this statement as it was added in the peri/post acceptance editorial stage: http://www.bmj.com/content/348/bmj.g3563

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.