Evaluating non-randomised intervention studies - NIHR Health ...

Health Technology Assessment 2003; Vol. 7: No. 27Appendix 8Descriptions of the IST and ECSTThe International Stroke Trial(IST) 132Description of the trialThe IST investigated the safety and efficacy ofaspirin and heparin on the outcome of ischaemicstroke. Between January 1991 and May 1996,19,435 patients with suspected acute ischaemicstroke entering 467 hospitals in 36 countries werecentrally randomised using a minimisationalgorithm within 48 h of stroke onset. Using a2 × 2 factorial design, half of the patients wereallocated ‘heparin’ and half ‘avoid heparin’, whilstsimultaneously half were allocated ‘aspirin’ andhalf ‘avoid aspirin’. Aspirin was prescribed at300 mg per day for 14 days (or the duration ofhospital stay). At discharge, clinicians wereencouraged to prescribe all patients long-termaspirin. Among patients allocated heparin,subcutaneous heparin was administered for14 days at one of two randomly allocated doses:5000 IU twice daily (low dose) and 12,500 IUtwice daily (medium dose). All analyses presentedbelow combine these two dose groups.The primary outcomes were death within 14 daysand death or dependency at 6 months. Outcomedata were 99.99% complete for 14-day outcomeand 99.2% complete for 6-month outcome. Interms of compliance, low-dose heparin wasreceived throughout the scheduled treatment by90% and medium-dose heparin by 88% of thoseallocated to it; 94% of those allocated to ‘avoidheparin’ did not receive it. Aspirin was takenthroughout the scheduled treatment period by92% allocated to receive it; 93% of those allocatedto avoid it did not receive it.In both aspirin- and heparin-allocated patientsthere were non-significantly fewer deaths within14 days (9.0% heparin versus 9.3% no heparin;9.0% aspirin versus 9.4% no aspirin). At 6 monthsthere was a non-significant trend towards a smallerpercentage of aspirin group being dead ordependent [62 versus 63%, 2p = 0.07; a differenceof 13 (SD 7) per 1000]. After adjustment forbaseline stroke severity, the benefit from aspirinwas significant [14 events prevented per 1000patients (SD 6), 2p = 0.03]. There was nointeraction between aspirin and heparin in themain outcomes.Baseline characteristics in the ISTThe distribution of 15 baseline characteristicsassociated with prognosis following stroke isreported in Table 36. These data were collectedduring the randomisation process and aretherefore available for every randomised patient.The ORs describe the relationship between achange in the baseline variable and the outcomeof death or disability at 6 months.The trial protocol required the use of a computedtomography (CT) scan to rule out intracranialhaemorrhage. However, where obtaining a CTscan was likely to require a long delay, and theclinician regarded it as very likely that the strokewas ischaemic, a non-comatose patient could berandomised before the CT scan. Table 36 showsthat a high proportion of patients did not reportCT scan results at baseline, and that the absenceof a CT scan indicated poor prognosis. Thisobservation arises through there being arelationship between stroke severity and thedecision to obtain a CT scan.For the purpose of our analyses, the last sixvariables were converted to a score (0–6) giving thenumber of presenting neurological characteristics.The distribution of this score is given in Table 37.Deficit score (Table 37), stroke type andconsciousness (Table 36) appear to have the largestranges of event ranges across their categorisations,indicating that they are likely to be three of themost prognostic variables.Adaptations made to the IST for theresampling projectFor the resampling project, only the comparisonof ‘aspirin’ against ‘avoid aspirin’ was consideredfor the outcome of death or disability at 6 months.To use the IST dataset for this project, themulticentre nature of the trial was exploited toconstruct a series of smaller randomised and nonrandomised‘sub-studies’. Centres falling withingeographical regions were grouped together suchthat sufficient participants were accrued in each167© Queen’s Printer and Controller of HMSO 2003. All rights reserved.