We’re a few months late on this one, but Heart, a BMJ title, issued a fascinating retraction notice in August about a meta-analysis on percutaneous coronary intervention (that’s stenting to you and me) after suffering a heart attack, and the drug abciximab, which is used to prevent clotting and additional near-term heart attacks. Abciximab is sold as ReoPro by Eli Lilly.
The article, “Clinical impact of intracoronary abciximab in patients undergoing primary percutaneous coronary intervention: an individual patient-data pooled analysis of randomised studies,” was published last May. But according to the retraction notice, the authors had neglected to include fresh data, already in the literature before their paper went live, that contradicted their overall findings.
Here’s the notice:
Retraction: Notice of unreliable findings
Piccolo R, L Gu Y L, Iversen A Z, Dominguez-Rodriguez A, De Smet B J G L, Mahmoud K D, Eitel I, Abreu-Gonzalez P, Thiele H, Piscione F. Clinical impact of intracoronary abciximab in patients undergoing primary percutaneous coronary intervention: an individual patient-data pooled analysis of randomised studies. Heart 2012 May 24 [Epub ahead of print] doi:10.1136/heartjnl-2011-301101. The findings in this paper are unreliable because it fails to address the data from AIDA STEMI [Lancet 2012;379:923-31], to which the authors had access prior to submission and which contraindicate the paper’s conclusion by showing no advantage for intracoronary administration. The authors were asked to update the paper to include the AIDA STEMI findings but, with the exception of Dr Olivier F Bertrand, they declined. Owing to this difference of opinion, Dr Bertrand asked to be removed from the list of authors, a request to which we acceded. Under these circumstances, the matter was considered by COPE who recommended retraction and this paper has now been withdrawn from Heart.
Our in-house cardiology expert tells us that intracoronary administration means you deliver the drug into the artery directly through a catheter, as opposed to injecting it into a vein and hoping enough of it gets to the desired area by means of the circulation system.
What the notice doesn’t say is that the lead author of the Lancet paper, which was published online in February 2012, is Holger Thiel, co-director of the Heart Center at the University of Leipzig — and the same Thiel H who’s second-to-last on the list of investigators in the Heart paper. AIDA-STEMI was a large trial, with 2,065 patients receiving the drug.
As the Lancet study indicates, route of administration of the drug did not appear to affect outcomes:
The incidence of death (4·5% vs 3·6%; 1·24; 0·78—1·97; p=0·36) and reinfarction (1·8% vs 1·8%; 1·0; 0·51—1·96; p=0·99) did not differ between the treatment groups, whereas less patients in the intracoronary group had new congestive heart failure (2·4% vs 4·1%; 0·57; 0·33—0·97; p=0·04). None of the secondary endpoints or safety measures differed significantly between groups.
As the abstract of the article states:
In patients with STEMI [ST-elevation myocardial infarction — a bad form of heart attack] undergoing primary percutaneous coronary intervention, intracoronary as compared to intravenous abciximab did not result in a difference in the combined endpoint of death, reinfarction, or congestive heart failure. Since intracoronary abciximab bolus administration is safe and might be related to reduced rates of congestive heart failure the intracoronary route might be preferred if abciximab is indicated.
But the Heart paper found that:
In STEMI patients undergoing primary PCI, intracoronary abciximab administration, when compared with the intravenous standard route, can improve short-term clinical outcomes mainly by reducing the risk of death.
Bertrand, the author who asked that his name be removed from the Heart study, co-wrote an editorial accompanying the Lancet study, which Lilly helped to fund, in which he concluded:
[T]he results of primary percutaneous coronary interventions continue to improve owing to a reduction of time between symptom onset and reopening of the relevant vessel, to earlier initiation of oral antiplatelet therapy, to judicious selection of antithrombotic agents, and to the rapid and sustained vessel patency obtained with coronary stenting. The results of the AIDA STEMI trial do not support routine intracoronary delivery of abciximab during such primary procedures.
So why did Thiel and corresponding author Federico Piscione decline to rework their article? Thiel tells us:
First of all the Editor required us to include the AIDA STEMI data after!!! acceptance and also online publication of the individual patient-based meta-analysis.
Second: Including a the AIDA STEMI data would have meant that the first author would have to do a totally new analysis and also a totally new draft of the manuscript
Third and most important: By contract with our clinical research organization we were not allowed to share the individual data before finalization of the overall trial which means last 12-month follow-up. The 12-month follow-up data have very recently been presented at the American Heart Association in November. This means that the Italian group has now got the data but not in the beginning of 2012 where we were not allowed to share the data.
We informed the Editor of these facts, but he constantly ignored these data and retracted the manuscript.
Currently, there is an ongoing complaint to the COPE organization.
But Adam Timmis, editor of Heart, said adding the new data wouldn’t have been a heavy lift editorially:
[W]e had access to documentary evidence (emails) that the authors discussed whether or not they should include AIDA-STEMI and decided not to (against Bertrand’s advice – he subsequently withdrew his name form the paper), perhaps giving them a chance to do a further meta-analysis including AIDA-STEMI data in order to give them another publication. The bottom line here is that the authors submitted for publication data they knew was misleading and refused to take the opportunity to correct this.
Meanwhile, Bertrand tells us that the Heart article
is a sad and complicated story which I hope will never occur to me again.
They tried to split the data to get two publications… but ended up with none. I believe that’s called ‘irony’.
It doesn’t look like it to me. It seems as if the Italians pooled together raw data that had come out of various previous studies. Considering the array of co-authors from the Netherlands, Spain, Denmark, Germany and Canada (Bertrand), these people may have been involved in those previous studies. The AIDA STEMI data was not included though. The conflict with the editor seems to be on whether they could have. And if they could have… they should have.
The editor claims the authors had access to this data (in a sense that’s true, as Thiele is an author). Thiele claims that he could not share the data with the people who actually did the data analysis (Piccolo and Piscione), at the time of submission.
Now for the conflict with Bertrand, I can imagine that there has been some discussion about inclusion of the AIDA STEMI data. The conclusion: it would mean a delay of a year. Bertrand was willing to wait it out. The rest wasn’t. (okay, here’s some room for data splitting accusations)
Of course, this is all speculating…
“by contract… we were not allowed to share the individual data before…”
I call that a cop-out. They were in a rush to publish which over-rode the need to be accurate by including all the available studies(which they claim would have forced them to wait). Sounds like they had the wrong priorities. Especially when including the new data contradicted their original findings.
Doesn’t your heart bleed for the first author if he had to re-do his entire analysis and make a new draft? Researchers: How many times has that happened to you and there was nothing you could do to prevent it, much less complain about it?
And, to promote intracoronary abciximab because it “might reduce” post-procedure coronary decompensation (heart failure) is disingenuous– the p value of that finding was .04, which is marginal at best, and I think is meaningless if you eyeball the confidence interval: 0.33-0.97. 0.97 is trivially different from 1.0, at least in my book.
More likely, in my cynical POV, intracoronary abciximab is more expensive (thus more lucrative) than intravenous abciximab– a critical consideration, to a cardiologist. More expensive is always better (he said sarcastically.)
In fairness to the authors of the Heart article one should note that their findings derived from 30-day follow-up whereas the Lancet paper endpoints were pt-outcomes after 90 days. Indeed looking at the graphs provided in the Lancet paper the gap between arterial and venous drug administration is wider at 30 days and then narrows by 90 days.
This seems like a slightly excessive use of a retraction: meta-analyses aren’t the final word in any scientific question, and they have to be updated. The correspondence quite clearly says that the Heart article included all the data they had access to at the time – it’s not unusual or a cop-out for a trial to have a data embargo like this (arguing about that is a different kettle of worms perhaps, but it happens and is the case here).
Instead of a retraction this meta-analysis should stand, and someone can re-do another adding in the data from the newer Lancet study and publish that as an updated meta-analysis. This is, in fact, exactly how meta-analyses are supposed to work, kind of like a running total for a scientific question as more data is gathered. This seems a really disappointing outcome and one which shows a lack of understanding of the process of evidence-based medicine.
If you go along with their arguments about what they were required to do, this meta-analysis will stand until the next one which includes the Lancet study… give them a break for once… retractions are so messy…
Then the next study comes along saying you need more 90 day studies… mysterious increase in mortality after 30 days in the intracoronary group…
Meanwhile, somebody’s got to treat those emergencies… those not trained for intracoronary will do intravenous, and those trained for intracoronary will do intracoronary and charge more…
If they really thought they’d have to wait a year because of a data embargo, then it was reasonable for them to publish with what they had. But the rest of it doesn’t add up.
I just went on the link to the article: why isn’t a kind on alert “RETRACTED” or similar?