Progress with the study and modifications to the design

Planning for the third international stroke trial (IST-3) of thrombolysis for acute ischaemic stroke with intravenous recombinant tissue plasminogen activator (rtPA) began in 1998. At its outset in 2000, the trial was planned as a pragmatic trial which sought to improve the external validity and precision of the estimates of the overall treatment effects (efficacy and safety) of rtPA in acute ischaemic stroke and to determine whether a wider variety of patients than previously thought might benefit from this treatment. After the drug was approved for use in Europe in 2003, the flexible trial eligibility criteria enabled us to continue recruiting patients who did not clearly meet all the criteria set out in the provisional European licence for the drug. With these changes and the need to comply with the increasingly burdensome requirements of research regulation in Europe, some aspects of the trial protocol and its practical procedures have inevitably had to change over the 12 year course of the study. The initial pilot phase of the study, funded by the UK Stroke Association, had a double-blind design, for which Boehringer Ingelheim provided a supply of drug and matching placebo. However, Boehringer Ingelheim did not wish to provide further drug supplies for the expansion phase, funded by The UK Health Foundation, and hence the design of this phase of the study was modified. With appropriate ethical approval, the expansion phase was designed as a prospective, randomised, open blinded endpoint (PROBE) trial and the first patient was recruited in this phase on 1st August 2003. The main phase of the trial was awarded funding by the UK Medical Research Council and began on 1st April 2005. The protocol for the MRC phase of the trial was published in Trials[1].. At each stage, relevant amendments were made to the protocol, and consolidated in an updated protocol (version 1.93) which was approved by appropriate authorities and published on the trial website http://www.ist3.com (additional data file 1). When advanced imaging techniques became more widely available in the participating centres, the study group developed a nested imaging study evaluating the role of CT and MR perfusion and angiography (CTP/CTA/MRP/MRA) in patients with acute ischaemic stroke. The cost-effectiveness of thrombolysis is being investigated in a sub-study in the Scandinavian region cohort.

Sample size/recruitment target

In our 2004 application to the MRC for funding of the main phase of the trial we stated: 'Assuming the same event rate as observed in the feasibility phases, the trial could very reliably detect an absolute difference of 10% in the proportion of patients dead or dependent at 6 months after treatment. With a sample size of 6000 patients, the trial would have > 99.9% power to detect such a difference (alpha = 0.001) and sufficient power to permit reliable analyses of the pre-specified subgroups. A trial of this size could detect a smaller, but still clinically worthwhile net benefit of as little as a 3% absolute difference in the primary outcome (80% power, at alpha = 0.05).' The application to the MRC for funding was successful, but by the time the grant was activated, two major changes to research regulation came into force: the European Clinical Trials Directive, the UK Clinical Trials Regulations 2004, and, shortly afterwards, the UK Research Governance Framework. These changes led to a large increase in the time taken to recruit and activate new centres and hence significantly reduced the rate of increase in patient recruitment. By 2007, it was clear that the various factors that had delayed recruitment meant that the original target of 6000 was no longer feasible, and we applied for approval to amend the recruitment target and for additional funding to extend the period of recruitment to mid 2011. These changes were summarised in version 1.93 of the protocol (additional document 1) as follows: "If 3500 patients were recruited, the trial could detect a 4% absolute difference in the primary outcome. A sample size of 1000 patients could detect a 7% absolute difference in the primary outcome, which is consistent with the effect size among patients randomised within 3 hours of stroke in the Cochrane review. Amendment in the light of recruitment by 2007: . The Medical Research Council... awarded an extension to funding to permit recruitment to continue until mid 2011, with a revised recruitment target of 3100 which would yield 80% power to detect an absolute difference of 4.7% in the primary outcome.'

Recruitment

The pilot (double blind) phase of the trial began when the first patient was randomised on 5th May 2000; sixteen centres enrolled patients. Recruitment continued without interruption into the open expansion phase (first patient in this phase was randomised on 1st August 2003). In April 2005, recruitment continued with the start of the MRC funded main phase. The independent Data Monitoring Committee (DMC) met regularly throughout the trial, in strict confidence, to review the accumulating trial data and any other relevant external data. Their terms of reference were set out in the protocol, and in the light of the evidence available, the DMC did not recommend any change to the protocol or to recruitment into the study. The trial closed to recruitment on 31st July 2011, with a total of 3035 patients from 156 hospitals in 12 countries (Table 1).

Table 1 Recruitment: number of centres and patients in each country
Full size table

Eligibility

The eligibility criteria are set out in full in the protocol, but can be summarised in terms of the 'uncertainty principle'[2–4]. As this was a pragmatic trial, conducted in the ';real world'[5], in order to ensure a wide variety of patients were included, inclusion and exclusion criteria were kept simple [6]. Patients were eligible if: they had symptoms and signs of clinically definite acute stroke; the time of stroke onset was known; treatment could be started within six hours of this onset; and, CT or MRI brain scanning had reliably excluded both intracranial haemorrhage and structural brain lesions which could mimic stroke (e.g cerebral tumour). In addition, if the patient had a clear indication for intravenous thrombolysis with rtPA, they were to be treated in accordance with local guidelines. Equally, if the patient had a clear contraindication to treatment they were not to be entered in the trial. Only if both the clinician and the patient (or a relevant proxy for the patient) felt that the treatment was 'promising but unproven', could the patient be included in the trial after appropriate informed consent. In order to comply with local regulatory and ethical requirements, the form of the consent procedure applied varied somewhat between countries and between centres within each country. Details of the type of consent process applied will be reported separately.

Baseline data collection and method of treatment allocation

After informed consent had been obtained, the clinician used either a voice-activated telephone or a secure web-based randomisation system to enter a patient in the study. The systems recorded and checked the baseline characteristics of the patients and allocated the treatment only after all baseline data had been entered and passed range and consistency checks. We used a minimisation program to achieve optimum balance for key prognostic factors (see Table 2a), and from January 2006, the minimisation algorithm was additionally stratified by world region and minimised on everything else within world region. Because simple minimisation within centres can lead to alternation of treatment allocation and thus potential loss of allocation concealment, our system also incorporated a degree of random allocation. Patients were allocated to the treatment group that would minimise the difference between the groups with a probability of 0.80. For 244 patients in the initial phase of the trial, the National Institutes of Health Stroke Score (NIHSS) was not measured directly, but was retrospectively estimated by an algorithm (derived from subsequent patients in the trial based on baseline clinical characteristics). In the initial double-blind phase, the clinician was informed of the treatment allocation in the form of a treatment pack number (which might contain active drug or matching placebo). In the open phase, the investigator was informed of the patient's treatment allocation (either thrombolysis with 0.9 mg/kg i.v. rtPA or control). In both phases, the randomisation system informed the local clinician of the dose (of drug or placebo in the double blind phase, or of the drug dose amongst those allocated thrombolysis in the open phase) to be given as a bolus and as infusion, and the patient's unique trial ID number.

Table 2a Clinical characteristics at randomisation used in the minimisation algorithm
Full size table

Baseline characteristics of patients

The characteristics of the patients recorded at baseline are shown, for both treatment groups combined, in Tables 2a and 3 (2b). Patients recruited within 1-2 hours of onset were statistically significantly (p < 0.0001) more likely to have a more severe neurological deficit than those recruited at later time points after onset (Figure 1). Similarly, patients recruited at earlier time points were statistically significantly older than those recruited later (p < 0.0001) (Figure 2). Additionally, the proportion of patients with a visible ischaemic lesion on the baseline imaging rose with time (data not shown).

Table 2b Clinical characteristics at randomisation not used in minimisation algorithm, and recorded at randomisation
Full size table
Figure 1
figure 1

Baseline stroke severity as assessed by NIHSS among patients randomised at different times after stroke onset. Plot displays median and interquartile range. p value for linear regression on exact time since onset p < 0.0001.

Full size image
Figure 2
figure 2

Age among patients randomised at different times after stroke onset. Plot displays median and interquartile range. p value for linear regression on exact time since onset p < 0.0001.

Full size image

Baseline characteristics of the patients in the context of the prevailing evidence

The baseline characteristics of the patients recruited into the study changed over the course of the study as clinician's uncertainties about which types of patients should be treated changed. These changes arose from appreciation of the accumulating evidence from randomised controlled trials and consequent changes in the European Licence and clinical guidelines. At the start of the trial in 2000, rtPA was approved for use in the USA and Canada, but was not approved for use in Europe. In 2003, the European Medicines Agency granted provisional approval for the use of rtPA for patients aged less than 80 years who could be treated within three hours (and who also met a number of other criteria). In 2008, when the ECASS-3 study reported evidence of benefit for treatment between 3-4.5 hours [7], guidelines in various European countries suggested that patients aged less than 80 years could benefit even if treated within 3-4.5 hours. We estimated the number of patients in the trial that would have met either the original 2003 approval criteria or the anticipated new 2011 criteria (modified in the light of the ECASS-3 data). If we defined 'meeting the 2003 approval' as: age < 80, delay from onset to randomisation < 3 hoyrs, NIHSS 6 - 25, SBP < = 185, DBP < = 110, 2.7 < = glucose < = 22 and no history of the combination of prior stroke and diabetes mellitus, 61 patients (2%) met the criteria, and if all the same criteria were applied and the time window was extended to 4.5 hours, 350 (12.7%) met the criteria. This clearly demonstrates that the great majority of the patients in IST-3 fall outside the European approval criteria for rtPA., and hence the study will provide valuable new randomised information in many categories of patient - as well the elderly - who are currently not eligible for treatment.

Development of the statistical analysis plan

The statistical analysis plan is being finalised at present, without reference to the unblinded data, and will be published before the code is broken in early 2012. It will take account of external evidence on factors influencing prognosis and response to treatment, the baseline characteristics of patients entering the study and advances in statistical methodology [8, 9]. The SAP will define a small number of 'primary' subgroup analyses, and a larger number of exploratory secondary analyses [8]. During the course of the trial, a number of novel approaches to the analysis of ordered functional outcome data in clinical trials of stroke treatment have been suggested to improve statistical power Statistical tests, which make use of the ordered nature of outcome data, such as ordinal logistic regression and 'shift analysis', may give more reliable results [9]. However, the technique is not universally applicable and it is important that key underlying assumptions are met for the analyses to be valid [10]. The Trial Steering Committee sought statistical advice, and because the underlying assumptions would not necessarily be met in the IST3 dataset, decided an ordinal analysis was not appropriate for the primary analysis of outcome. This decision was also consistent with all the previous trials of intravenous rtPA, none of which had adopted an ordinal approach for the primary analysis. However, an ordinal analysis has been included as one of the secondary exploratory analyses. The trends in Figures 1 & 2 demonstrate that the statistical analysis plan will also need to take account of these potentially confounding effects on the assessment of treatment effects overall and in subgroups.

Reporting of results by allocated treatment

The trial will be unblinded in early 2012 and the main trial results presented at the European Stroke Conference in Lisbon in May 2012 with the aim of publishing simultaneously in a peer-reviewed journal. The trial primary results will be presented in the context of an updated Cochrane systematic review of all the relevant randomised trials. The imaging sub-study CTP/CTA/MRP/MRA results are due in early 2013. An individual patient data meta-analysis of all trials testing intravenous rtPA, including IST-3, is also being considered.

Discussion

With 3035 patients, IST-3 is the largest ever randomised trial of intravenous thrombolysis for the treatment of acute ischaemic stroke (the two largest previous trials each recruited exactly 800 patients [7, 11]. At its outset in 2000, the trial sought to improve the external validity and precision of the estimates of the overall treatment effects (efficacy and safety) of rtPA in acute ischaemic stroke and to determine whether a wider variety of patients than previously thought might benefit from this treatment. With the approval of the treatment for specific types of patients, the trial was able to evaluate the effects of treatment in patients who did not precisely meet the existing European approval for the treatment; in other words, also sought to extend the criteria for treatment. At the start of the trial, when rtPA had not yet received approval from the European regulatory authorities, patients of all ages were recruited within 3 hours. After the provisional approval for the use of rtPA in Europe in 2003 for those patients aged less than 80 years who could be treated within three hours, participating clinicians generally opted to treat patients who met the conditions of the licence, and hence only recruited only those patients in IST3 that did not strictly meet the licence criteria. The elderly have been under-represented in trials of treatments for acute stroke [12], especially thrombolysis with rtPA, where fewer than 100 patients aged over 80 years have been included in randomised trials of this treatment [13]. IST-3 has included 1617 patients aged over 80, so will provide valuable randomised evidence on the balance of risk and benefit in this much under-researched, but increasingly important age-group [12]. IST3 has shown that pragmatic trial designs with broad and flexible entry criteria can recruit large numbers of patients, even in the face of changing and burdensome regulatory regimes; the emergence of new data from other relevant trials and rapidly changing clinical practice.