03: ECASS Trials

The ECASS trials


Pretty similar in aims as the ATLANTIS study we talked about last time.


  • 0-6 hrs
  • they excluded severe strokes and people with more than minor early signs of infarction on CT (something NINDS didn’t differentiate on)
  • they also used a slightly bigger dose of tPA (1.1 vs 0.9mg/kg)


  • barhel index and modified rankin score


  • 300 in each group
  • median NIHSS slightly higher than more recent trials at 12/13
  • they excluded over 100 on the basis of CT signs- remember this the next time your general radiologist has given you a report simply saying no bleed. note that these 100 they excluded were excluded after inclusion in the trial after review of CT scan by a study radiologist. That’s a bit naughty as you might suspect.
  • mortality was 22% tPA v 15% placebo.
  • it’s difficult to tell about symptomatic intra-cerebral haemorrhage as they didn’t define this clearly. Overall ICH was 43% tPA vs 37% in the placebo group
  • tPA did make a 6% absolute improvement in MRS <2 (35%v29%)

If the treatment increases mortality then you can see why I consider this a -ve trial.


Not to be deterred they tried again


  • this time they had more stringent CT criteria and tight BP control and hoped this would be better
  • still 0-6hrs
  • randomised in blocks so as to have some in the 0-3 hr group and some in the 3-6 hour group.


  • an MRS less than 2


  • 400 in each group,
  • most at about 4 hrs or so
  • again they excluded 10% in each group retrospectively based on CT findings. So despite a strict definition and extensive training of all the radiologists for the study, they still called it wrong 10% of the time.
  • 40% (tPA) had MRS<2 vs 37% in the TPA group. This was not of any statistical significance. This difference was the same whether you were in the 0-3 hr or 3-6 hr group
  • mortality was equal at 10%
  • in terms of symptomatic intracerebral haemorrhage it’s stil difficult to tell as despite giving a nice clear definition (any new blood on CT associated with clinical deteroration) they don’t give the numbers in the manuscript that I can find. They say it was 2.5 times higher, they just don’t give the numbers

This trial is negative by their primary outcome but mortality at least remained equal. In addition they more than doubled the rate of SICH.


the one that got all the attention a few years back and got the guideline reccomendations extended out to 4.5 hrs


  • the usual stroke patients with a CT and again there were CT exclusion criteria if your stroke was too big
  • 3-4.5hrs only
  • they also excluded NIHSS >25 which is new for this study


  • the now ubiquitous MRS of <2


  • 400 each group
  • took 4 years to recruit 800 patients from 130 centres (that’s 1.5 patient per year per centre)
  • 52% in tPA group and 45% in the placebo group had an MRS<2
  • mortality was 7.7% in the tPA group and 8.4% in the placebo group
  • Symptomatic intracerebral hameorrahge was bizzarely low at 2.4% (still 10 times mor than placebo) despite the fact that overall intracerebral bleeding was 27% in the tPA group and 17% in the placebo group. This is probably explained by the following statement:

In our study, we modified the ECASS definition of symptomatic intracranial hemorrhage by specifying that the hemorrhage had to have been identified as the predominant cause of the neurologic deterioration.

Remember that the investigators get to decide this, and this allows them to define SICH pretty much as they want.

  • Median NIHSS of 9 in tPA group and 10 in the placebo group. (p-value of 0.03 for this). People had more severe strokes in the placebo group.
  • 7% had previous stroke in tPA group and 14% in the placebo group (p-value of 0.03 for this). More patients in the placebo group would have had an MRS >0 before they even had their second stroke
  • they try to adjust for the significant baseline variations in the groups with logistical regression – they tell us that there was still a significant difference after adjustment but you have to be a believer in logistical regression; that it can accurately adjust for not just the stated confounding variables but all the ones that you don’t know about.

By their trial design and definitions this is a positive trial, you just have to decide if the problems outlined above are enough to call it into question. 

Note that this trial enabled recommendations that tPA is safe for 3-4.5 hrs. This despite that all prior trials treating patients at the same time periods were killing patients and were terminated early.