DIAS-2 & EPITHET
I’m reviewing these two trials as they both looked at what could be a really interesting and useful idea – using imaging to select patients who would most likely benefit from tPA.
This isn’t particularly important to understand fully but in an ischaemic stroke there is such a thing as an ischaemic penumbra – an area of brain tissue surrounding the central infarct; that if perfusion is quickly reestablished may recover well.
The idea is to use imaging (mainly CT perfusion or diffusion weighted MR) to identify such patients and target treatment at them and not the others.
There is lots or pre-RCT research on this if you’re interested, trying to define what all the above actually looks like in practice, but below are the RCTs involved.
EPITHET
There’s really two things going on here. An RCT for tPA v placebo and an observational imaging study to define the useful penumbra.
PATIENTS
- 3-6 hrs with stroke and a negative baseline CT (or sometimes MRI)
- exclusion criteria fairly extensive and building on the work of the older studies. Kind of like goldilocks and the three bears, not too severe a stroke, not too mild, but just right…
- randomised to tPA v placebo
- before they got the study drug everyone got an MRI looking for the ischaemic penumbra – these results were looked at AFTER treatment and had no impact on deciding who got placebo and who got tPA
OUTCOMES
- primary outcome was a surrogate – infarct growth attenuation (basically did patients who got tPA have a smaller infarct on imaging.) This was to be based on a 90 day scan but there some problems with that as we’ll come to
- secondary outcomes included lots of imaging based composite outcomes including some clinical features.
- this means that the trial will be drastically underpowered to detect anything we care about.
- note the definition of symptomatic ICH is again changed:
ICH with significant clinical deterioration of ≥4 NIHSS points within 36 h of treatment, and parenchymal haemorrhage of grade 2 on CT (blood clots in >30% of the infarcted area with substantial space-occupying effect) adjudicated by a blinded committee
this is worth comment
- the change in definition is stricter and probably more accurate but let’s be clear – it favours tPA
- in all the trials adjudication of SICH is done by a central blinded committee. in some of the trials these are directly paid employees of the company making the drug. in a tPA trial if you’re asked to adjudicate on a case where the patient had deterioration and bleed following the study drug you know full well that they got tPA and not placebo. This means that you aren’t really blinded and there’s potential for bias in classifying patients as SICH. This is a problem for pretty much all the trials.
RESULTS
- 100 patients total (50 tPA, 50 placebo) – and yes you’re right that’s a tiny trial. Whenever’s there’s less than 100 in a group talking about stats in percentages bcomes kind of meaningless.
- in terms of their primary outcome it was a negative trial – they didn’t show a decreased infarct growth by geometric mean. But they reanalyse it with a different technique and guess what it looks much better (and they put it in the abstract too)
- the other thing to mention is that they couldn’t do the 90 day scan (which was part of the primary outcome) in a few patients, so they ended up with 37 in the tPA group (which started with 52) and 43 in the placebo group (which started with 49). The reason for this:
Day 90 imaging was not done for 19 patients in the alteplase group (13 had died and six were lost to follow-up) and eight patients in the placebo group (seven had died and one was lost to follow-up)
note the implications of that – they couldn’t do the 90 day scan because twice as many had died in the tPA group. Nevermind the “6 lost to follow-up” who may well be dead, especially as there was only one lost to follow-up in the placebo group.
They tell us that the mortality difference was non-statiscally significant and in a trial this small it’s always going to be hard to tell, especially when over 10% of the patients in the tPA group are “lost to follow up”.
It really matters what happens to those patients in such a small trial. If those 6 patients are dead then there would be a “statistical difference” in mortality. Whether or not there’s a statistically significant difference (and I doubt it matters here) then you need to decide whether that worries you or not.
- for once there was a baseline difference in NIHSS favouring placebo (14 v 10) which is to be expected when numbers are this small
- in terms of the thing we care about there was an 11% absolute difference in MRS <2 favouring tPA but remember in such a tiny trial this is 18 v 12 patients we’re talking about
- the rest of the paper is about the penumbra stuff and looking at who reperfused and who didn’t. And people who got tPA reperfused more often. Let me say again – this is a surrogate that I don’t really care so much about.
they make this statement in their discussion:
The EPITHET results support the use of reperfusion as a surrogate for outcome in future proof- of-concept stroke trials
having read the details of the trial and looked at the tiny numbers involved I’m not sure I agree.
they also make this statement in the results section of the abstract, the bit that most people will read!
Reperfusion was more common with alteplase than with placebo and was associated with less infarct growth (p=0·001), better neurological outcome (p<0·0001), and better functional outcome (p=0·010) than was no reperfusion.
- When I read that the implication is that tPA was associated with better functional outcome. There’s even a p value beside it. But if you’ve read the results as I’ve presented them or even better, you’ve read the trial then you know that’s not what was found
- What we see here is spin (and this is really important); there is nothing factually incorrect in that statement. It’s just that what they’re describing is 4 separate findings placed in a sentence to link the first few words (“reperfusion is more common with tPA)” with the last few words (“better functional outcome”)
- if you asked the trial to answer this question: is tPA associated with better functional outcome then the answer is NO.
DIAS-2
This is unique in a few respects:
- it used desmotaplase, a variant on alteplase. again there’s lots of basic background research on this and it may turn out that there’s something better or worse about it but in mechanism of action and class it’s similar
- it uses imaging to select the penumbra patients (either CT or MR); this is different form EPITHET in that the penumbra imaging part was only observational there. In DIAS-2 the imaging is used to pick out patients who will hopefully do better
PATIENTS
- 3-9 hr patients
- a similar “goldilocks” selection of patients to EPITHET; with the important addition of MRI signs – MR is much more likely to pick up signs of severe infarct and so allows you to be even more selective, even before you do the perfusion imaging
- a salvageable penumbra needed to be shown on imaging (MR or CT)
- 3 groups, high or low dose desmotaplase and placebo
OUTCOMES
- 90 day clinical outcome; which in this trial is a composite of mRS <3 (it’s usually mRS <2), NIHSS improvement and Barthel index
RESULTS
- 30% had no penumbra by imaging and were excluded. This varies from study to study but it’s a useful figure to keep in your head.
- n 180 patients (60 in each group)
- median NIHSS was 9 (remember in NINDS it was 14)
- in terms of their primary composite no difference was found in outcome but let me break it down a bit:
- mRS<3: placebo – 59%; low dose 54%; high dose 49% – PLACEBO SAVES THE DAY!
- SICH (by the new tPA favouring definition) was 4% in the tPA patients vs 0% in the placebo groups
- Of note if the NINDS definition of SICH is used then placebo jumps to having 15% rate of SICH and bizarrely tPA changes to only 9%. I imagine this is more to do with the imaging used (largely MR) and the size of the trial than it is with tPA being protective against bleeding. (he says tongue-in-cheek…)
- In terms of mortality there were 4 deaths in the placebo group, 3 deaths in the low-dose group and 14 deaths in the high-dose group. They give a table giving causes of death and suggest that the tripling in death rate in the high-dose group has nothing to do with the tPA. Hmmm
Their interpretation in the abstract is illuminating:
The high response rate in the placebo group could be explained by the mild strokes recorded, which possibly reduced the potential to detect any effect of desmoteplase.
This begs the question – if these low grade strokes (selected out by the finest imaging and clinical lessons learned from prior trials) do so well that we can’t see the benefit of this wonder drug then why the hell should we be using it?