More on Stroke Thrombolysis (IST3)

Next time you are thinking about thrombolysing a stroke patient you might like to take a look at the largest ever RCT of stoke thrombolysis. Now if you read just the conclusions of The benefits and harms of intravenous thrombolysis with recombinant tissue plasminogen activator within 6 h of acute ischaemic stroke (the third international stroke trial [IST-3]): a randomised controlled trial, The Lancet  379  (9834) pp 2352-2363 you could be forgiven for thinking that this was a positive study that favours thrombyolysis. Read the papoer however and things look less rosy.

The main aim of IST3 was to expand the range of people who could be given rTPA for stroke. If the patient met a clear indication for lysis according to the current licence they got the drug. If they met a clear absolute contraindication they didn't. Where the included patients were supposed to fall was where the clinician wasn't sure if it would help. What this was really trying to get at was the over 80s and the 4.5-6 hour patients.

3100 patients were randomised in an open label study (which favours the intervention), the primary outcome measure was the Oxford Handicap Scale (which is like the modified Rankin scale with the dead category removed leaving 5 possible categories - normal, not normal but independant, not normal but needs help at home, not normal and dependant, terrible), used telephone or postal follow-up to assess the scale at 6 months with a patient or carer ("so how's Grandma doing, knowing that she got the wonder drug?/ Now, your Grandma missed out on the wonder drug didn't she, how is she going? Oh dear") and what difference did they find between the two groups? Lives saved? People kept at home and independant?

The only difference that could be found between the two groups in real terms was the excess of dead patients in the rTPA group at the end of 7 days. The primary outcome measure (functional outcome at 6 months) demonstrated no difference. Deaths at the end of 6 months were the same suggesting that big nasty strokes kill people but rtPA does a great job of speeding up the process. But thanks to the wonders of statistical jiggery-pokery an improvement in the treatment group was able to be constructed (I mean detected) by ordinal logistic regression analysis. Now when a statistical method needs that many words to describe it you should start to be very nervous. Especially when you are looking at a pretty simple question like "does this wonder drug make a difference to the number of people alive and indepenant 6 months after a stroke?" and you are using over 3000 people to find that difference.

The whole thing is just too funny/depressing depending on my mood to write about it more here. The latest issue of Emergeny Medicine Australasia has an editorial on it, the team at Emergency Medicine Abstracts have gone to town on it, Cliff Read of has had a look at it.

Bottom line, if the drug is so good why can't anyone do a study that clearly shows it and why do the studies that proport to show it need to be so distorted between the results and conclusions sections?