So IST-3 is out…
“Thrombolysis is of net benefit in patients with acute ischaemic stroke, who are younger than 80 years of age and are treated within 4·5 h of onset”
- It was meant to have 6000 patients to detect a significant difference in the primary outcome. It enrolled 3000. It admits it isn’t powered to detect the difference it wanted to. They re-jigged the stats once they realised they’d get nowhere near 6000.
- Part way through their data collection, ECASS-III pushed eligibility out to 4.5 hours, so methodology changed at this point
- Patients were enrolled if the treating clinician (and the patient – very weird ) agreed that they felt there was uncertainty regarding benefit or harm for that patient. Patients who met the so-called standard of care indications were treated according to each institution’s own thrombolysis protocol, and those with accepted contraindications were excluded. It is therefore a completely different patient cohort from all previous trials. (This doesn’t stop the same authors lumping it in with the other 11 RCTs in their revision of the Cochrane meta-analysis [which was already fraught with issues])
- It was a multicentre trial in 150 centres. On average each centre provided just 2 patients per year.
- It was sort-of randomised, but not double-blinded. The initial pilot phase (10% of patients) was blinded, but the full trial was open-label. In other words, the treating staff knew what therapy their patient had been given. For some reason, one-and-a-half times as many patients (24% vs. 17%) in the treatment arm received HDU care as in the control arm. This gets glossed over somewhat in the discussion. The relationship of this difference to the fact that clinicians were not blinded is not discussed. It is possible that the rtPA group received better/more intensive care? It is certainly a source of likely bias.
- The “randomisation” process seemed to occur AFTER assessment of patient variables in terms of demographics and stroke severity variables (how this is random is beyond me) and “randomisation” then occurred with these factors/variables taken into the algorithm.
- In the subgroup analyses (which are the only areas to demonstrate statistically significant benefits – see below) there are a lot of complicated statistical manipulations involving logistical regression etc. to attempt to negate confounders and make both arms of each subgroup appear almost identical in demographics and stroke severity (presumably to prevent criticism about the stroke severity and patient discrepancies that plagued most of the other lysis trials)
- Follow up was variable – some was by phone call only.
- It is, in summary, a very muddy and strange methodology and analysis which the authors call a “pragmatic” RCT
- Despite the study being NEGATIVE for it’s primary endpoint (percentage of patients functionally independent at 3 months) the authors somehow ludicrously claim that it demonstrates benefit (the actual numbers are: 2% more patients alive and independant at 6 months [37% vs. 35%], but this did not achieve statistical significance)
- There was a non-statistically significant trend towards improvement in OHS score (a secondary endpoint analysis)
- In patients aged >80, there is a trend towards benefit with rtPA, CI 0.97-1.88. (But in a study showing no overall benefit, this actually means that there is a trend towards harm in ages <80, CI 0.67-1.26 )
- Between 4.5-6 hours, there is a trend towards benefit with rtPA, CI 0.89-1.93. (This unfortunately also raises the problem that between 3 and 4.5 hours, there is a trend towards harm with rtPA, CI 0.50-1.07)
- rtPA is also essentially neutral or trends towards harm up until NIHSS 14, but may show some benefit in more severe strokes (I really hope this is true but again, these are only trends )
- There were more deaths early with rtPA (11% vs. 7%) but that evened up by 6 months
- ICH rate was 7% in the rtPA group vs. 1% in the control group, but for some reason (which makes no first-principle sense if we believe that lysis works), there was more fatal swelling in the rtPA group.
- Interestingly (!) enough, the “blinded” phase of the study trended towards favouring control, CI 0.42-1.98, while the open phase trended towards rtPA, CI 0.89-1.45. Make of that what you will!!!