Scrutable

BBC Heart Attack Treatment Story
Ther's the usual saying the same thing early on, hoping people aren't paying attention and repeating it later on stuff throughout: welcome to the BBC News Website!
But Stent Maker does research; research says Stents as good as Open Heart Surgery; presumably Stent Makers say "hey use our stents they're really good".
But Others say " hold on, what you called a heart attack ent what everyone else calls a heart attack matey" so data is skewed to give result you wanted.
Stent Maker says "Fake Fake Fake"

A picture caption says "Stents are used on patients too ill for OHS" so wouldn't there be greater mortality in this group anyway?
Anyway, I gotta dash; thought it might be of some interest to people here.
Happy Days
TB

Cheers mate; yes you're right, twas all done in a it of a rush.

This is going to be all about adjudication. All major trials now have an independent committee of clinical experts who review the data from patients suspected to have a heart attack (or whatever) and decide if it meets the criteria for the endpoint. Importantly, this committee works blinded to treatment.

The published main outcomes were adjudicated. The outcomes from the patients who allegedly had a MI according to the international consensus definition were not published, and so we don't know if and how they were adjudicated. This is crucial to the claims we're seeing here.

The protocol for this trial is here: https://clinicaltrials.gov/ct2/show/record/NCT01205776

It does indeed state "protocol defined MI", which is based it seems on measurements of creatinine kinase. The trial was published in 2016, which is well into the era of measurement of cardiac troponin (cTnI or cTnT, mandatory nowadays for confirmed MI diagnosis). The methods (buried in an online appendix) use CK-MB for "post-event MI" (within 72 hours, but later "spontaneous MI can be based on "rise and fall" of troponin or CK-MB (without any criteria). The main publication does not gove any indication on the diagnostic approach for identifying an MI. It all seems very loose.

The universal consensus definition of MI states that: "cTnI and cTnT are the preferred biomarkers for the evaluation of myocardial injury, and high-sensitivity (hs)-cTn assays are recommended for routine clinical use. Other biomarkers, e.g. creatine kinase MB isoform (CK-MB), are less sensitive and less specific".

In the abstract for the 5-year study, we have: "Death from any cause occurred more frequently in the PCI group than in the CABG group (in 13.0% vs. 9.9%; difference, 3.1 percentage points; 95% CI, 0.2 to 6.1)." Not significant, indeed, but just about every other paper I've ever seen would have reported a hazard ratio. Maybe they are in the full paper, haven't seen that. The HRs would look a bit scary - regulators tend to get itchy about high point estimates for a HR for something bad, wherever the 95%CI are, and this seems to be the reaction of the cardiologists quoted by the BBC.

I'd say we definitely need a post-hoc adjudication of cTn results from that trial, according to the consensus definition. if the results from that are adverse for the stent, the trial is effectively dead.

With regard to they ESC's response, they considered 6 RCTs in their guideline, of which EXCEL was one (although the largest), so their conclusions may well still stand.

Finally, if you're a stentee, lots of other considerations depend whether you get a stent or surgery.

The leaders of that trial have now answered back, rather unimpressively, in the humble opinion of me, that makes their trial look shitter.

From a Medscape news article:

Scientific debate over study findings among informed parties is healthy, the EXCEL leadership wrote in their statement; "to suggest, however, that hundreds of EXCEL investigators, including cardiologists, surgeons, statisticians, and entire academic research organizations conspired to change definitions or withhold important study findings is offensive and without merit."

They say that "reporting procedural MI rates according to Universal Definition was not possible" because it is based on the collection of cardiac troponins, which was optional in the trial and done in only a minority of patients because of cost considerations. Further, it was agreed by all involved, including surgical colleagues, that the Universal Definition "was not suitable because of ascertainment bias, different criteria for PCI and CABG [coronary artery bypass grafting], and lack of demonstrated correlation with prognosis."

David Taggart, MD, PhD, University of Oxford, United Kingdom, who withdrew his authorship from the 5-year EXCEL publication, said the trialists agreed, because of a legitimate concern over ascertainment bias, to use the previously untried and untested Society of Cardiovascular Angiography and Interventions (SCAI) definition of MI in addition to the Universal Definition (UD).

"The importance of the UD was a 'safety check' to allow comparison of these two definitions of MI within the EXCEL trial and against other trials," he told theheart.org | Medscape Cardiology via email. "However, at no time was there any agreement NOT to also present the protocol-specified UD. If so, why was the protocol not updated accordingly?"

This episode of The Life Scientific

https://www.bbc.co.uk/programmes/m0003zth about 19 minutes in discusses the effect of giving an aspirin during a heart attack.

It also has an anecdote about Richard Peto's insistence on not doing subgroup analysis on a study that needs large numbers to see a significant effect. So when the Lancet insisted, he and his collaborators used star signs.

Not sure what the point is here. Platelet blockers are established evidence based therapy for prevention of MI in at risk subjects. And the zodiac based subgroup analysis was good fun, but interpretation of those has moved on in the two decades since then.

If the p for interaction is >0.05, forget it. If it is <0.05, there may be a hypothesis. They prove nothing by themselves and it is well understood that there is no power left in the trial for them to actually prove anything. All major outcomes trials have these, and they are usually reassuring in that the various demographic and disease groups responded to treatment. Or, it is just possible they might reveal something of general importance.