Triple versus guideline antiplatelet therapy to prevent recurrence after acute ischaemic stroke or transient ischaemic attack: the TARDIS RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 10/104/24. The contractual start date was in October 2012. The draft report began editorial review in November 2017 and was accepted for publication in May 2018. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Philip M Bath reports grants from the British Heart Foundation and National Institute for Health Research (NIHR) Health Technology Assessment (HTA) programme during the conduct of the study, and he is a shareholder in Platelet Solutions Ltd (Nottingham, UK). He also notes personal fees from Diamedica (Minneapolis, MN, USA), Nestlé (Vevey, Switzerland), Phagenesis Ltd (Manchester, UK), ReNeuron (Bridgend, UK), Athersys (Cleveland, OH, USA), and Covidien (Zurich, Switzerland) outside the submitted work. Robert A Dineen reports a grant from the British Heart Foundation during the conduct of the study. Hugh S Markus reports grants from the NIHR HTA programme during the conduct of the study. Alan A Montgomery is a member of the NIHR HTA Clinical Evaluation and Trials Funding Board.

Copyright statement

© Queen’s Printer and Controller of HMSO 2018. This work was produced by Bath et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2018 Queen’s Printer and Controller of HMSO

Scientific background

Stroke is devastating to patients, carers and society through high mortality (1 in 4 patients by the first year), morbidity (dependency in 1 in 3 patients by the first year, many of whom need long-term care) and cost (NHS and social care costs of £1.7B a year in England). 1 Both stroke incidence and prevalence will increase as the UK population ages. Following stroke or transient ischaemic attack (TIA), the risk of recurrence is high, especially immediately after the event (approximately 10% over the first few weeks), after which it falls to a total of about 40% by 5 years. Typically, recurrent strokes are more severe than earlier events. 2,3

A TIA (‘mini stroke’) is defined as ‘an acute loss of focal cerebral or ocular function with symptoms lasting less than 24 hours and which is thought to be due to inadequate cerebral or ocular blood supply as a result of low blood flow, thrombosis or embolism associated with diseases of the blood vessels, heart, or blood’. 4 TIAs are important because they are a key risk factor for subsequent stroke. Patients presenting with specific TIA features are at a particularly high risk of a subsequent stroke, as assessed by the ABCD2 score5 (as derived from the ABCD score6), which takes account of age (A), blood pressure (B), clinical symptomology (C), duration of symptoms (D), and presence of diabetes (D). An important caveat is that data for the training databases used to derive and validate the ABCD2 score were collected up to 1998 and 2005, respectively, and so the absolute risk rates of stroke are now lower as enhanced secondary prophylaxis has become standard practice. Although some research groups have validated the ABCD2 scoring system,7,8 its value in diagnosis and prognosis after TIA has more recently been questioned. 9

The risk of recurrence can be reduced, but not abolished, with lifestyle changes; drug interventions comprising antithrombotics, antihypertensives and statins; and carotid endarterectomy (after large artery stroke/TIA). 10–12 Although oral anticoagulants are established for cardioembolic stroke,13,14 other patients with non-cardioembolic ischaemia (the majority) need antiplatelet therapy. 15,16 These interventions are all cost-effective.

The archetypal antiplatelet, aspirin (which is an inhibitor of cyclo-oxygenase), reduces recurrence [i.e. relative risk reduction (RRR)] by 17% in patients with prior stroke or TIA. 17 Clopidogrel (which is an adenosine diphosphate receptor antagonist) was slightly more efficacious than aspirin in the Clopidogrel vs. Aspirin in Patients at Risk of Ischaemic Events (CAPRIE) trial. 18 Importantly, the relative difference in efficacy between aspirin and clopidogrel was highest in patients with prior stroke or myocardial infarction (MI). 19 No trials comparing clopidogrel with control or aspirin have been reported in patients with TIA. Dipyridamole (which inhibits the phosphodiesterase inhibitor-5 and red blood cell uptake of adenosine) reduced recurrence by 16% in comparison with placebo, and was comparable to aspirin, in the European Stroke Prevention Study-2 (ESPS-2) trial. 20 Evidence now suggests that stroke prevention is dependent on the number of antiplatelets [e.g. combined aspirin and dipyridamole reduce events by 23% in comparison to aspirin (or dipyridamole) alone without increasing the risk of bleeding, as seen in the ESPS-2 and European/Australasian Stroke Prevention in Reversible Ischaemia Trial (ESPRIT) trials]. 20,21 As with clopidogrel alone, the difference in efficacy between aspirin and dipyridamole versus aspirin alone was greatest in patients with the highest baseline risk. 22 Similarly, aspirin and clopidogrel was superior to aspirin in cardiac patients. 23,24 However, the superiority of clopidogrel-based dual therapy was not seen in the Clopidogrel for High Atherothrombotic Risk and Ischemic Stabilization, Management, and Avoidance (CHARISMA) trial,25 probably because the apparent benefit in those with prior stroke or MI (who had the highest risk of recurrence) was diluted by lack of efficacy in those with no previous vascular events (who were at lower risk). The risk of bleeding with aspirin and clopidogrel versus aspirin was 30–40% higher in these three trials. The Management of ATherothrombosis with Clopidogrel in High-risk patients (MATCH) trial (aspirin and clopidogrel vs. clopidogrel) found that dual aspirin and clopidogrel also increased bleeding. 26,27

On the basis of these trials and taking account of the prices of branded clopidogrel and dipyridamole-ER [£37 and £10 per month, respectively, British National Formulary (BNF)],28 the National Institute for Health and Care Excellence (NICE) recommended in 2005 that patients should take combined aspirin and dipyridamole after ischaemic stroke or TIA [Technology Appraisal (TA) number 90]. 29 In late 2010, NICE updated its recommendation to aspirin and dipyridamole for TIA, and clopidogrel for ischaemic stroke (TA210);30 these decisions take account of the large drop in the price of clopidogrel, reflecting its generic status (£3.40, BNF 6128) but the lack of significant randomised data and a licence for clopidogrel in patients with TIA. Former and current guidelines have not recommended dual aspirin and clopidogrel because of increased bleeding. 31,32 The preference for combined aspirin and dipyridamole, or clopidogrel alone, over aspirin alone was also recommended by the European Stroke Organisation in its 2008 guidelines. 33 In contrast, the 2011 American Stroke Association secondary prevention guidelines still gave equal recommendations for aspirin (50–325 mg daily) alone, dual aspirin and dipyridamole, and clopidogrel (75 mg daily) alone,34 thereby ignoring the results of recent trials. 18,20,21,35

The above data for stroke reflect long-term prophylaxis, a very different situation from the situation immediately after an event when the risk of recurrence is much higher. Conventional acute antiplatelet therapy is based on aspirin alone for ischaemic stroke reflecting the results of the International Stroke Trial and Chinese Aspirin Stroke Trial (CAST) megatrials. 36,37 However, the effect size is small (absolute risk reduction ≈1.1%) and, until recently, the acute treatment of TIA had not been investigated. As the risk of recurrence falls quickly after stroke or TIA, intensive antiplatelet specific treatment is likely to be needed only for a short period so that the exposure time to hazard (mainly bleeding) is limited. Although clopidogrel-based dual therapy has not proved effective/safe in long-term stroke prophylaxis, early and short-term dual antiplatelet therapy based on clopidogrel or dipyridamole may be useful, at least after TIA/minor stroke, as suggested by several trials [Fast Assessment of Stroke and TIA to prevent Early Recurrence (FASTER),38 EARLY,39 Prevention Regimen for Effectively Avoiding Second Strokes (PRoFESS) early]40 and observational studies [Early use of EXisting Preventive Strategies for Stroke (EXPRESS),41 SOS-TIA42]. In the FASTER trial (n = 392), 90 days of aspirin and clopidogrel (vs. aspirin) showed a trend to reduce stroke by absolute 3.7% (not significant), and increased symptomatic intracranial haemorrhage (sICH) by absolute 1% (not significant) leading to a net absolute benefit of 2.7%. 38 Similarly, the EARLY trial (n = 543, acute ischaemic stroke/TIA) found a tendency to reduced vascular events at day 90 with aspirin and dipyridamole (vs. aspirin, not significant) but no effect on functional outcome. 39 A pattern of observations also seen with aspirin and dipyridamole (vs. clopidogrel) in the PRoFESS early subgroup (n = 1360, mild acute ischaemic stroke). 40

In a meta-analysis of all trials comparing dual with mono antiplatelet therapy in patients with acute stroke or TIA [including ESPS-2, ESPRIT, CHARISMA, MATCH, PRoFESS early, EARLY, FASTER, Clopidogrel and Aspirin for Reduction of Emboli in Symptomatic Carotid Stenosis (CARESS) and Clopidogrel plus Aspirin for Infarct Reduction in acute stroke/TIA patients with large artery stenosis and microembolic signal (CLAIR)20,21,25,26,38–40,43,44], acute dual therapy versus monotherapy within 3 days of ictus significantly reduced stroke recurrence and a composite of vascular events. 45 No significant differences were seen for MI, sICH, major bleeding or death (but there were few events, Table 146). No heterogeneity existed in any analysis, which suggested that the composition of dual and monotherapy was not of primary importance. None of the trials was large enough (each < 1400 patients) to show individual significant differences in stroke or vascular events. Importantly, the magnitude of effect appeared to decline with time from ictus so trials recruiting earlier had greater reductions in their point estimates (albeit non-significant because of small sample size) than those recruiting later: range of odds ratios (ORs) for stroke – early, OR 0.51 to 0.71 (EARLY,39 FASTER,38 PRoFESS early40); later, OR 0.83 to 2.44 (CHARISMA,25 MATCH26). The large Clopidogrel in High-Risk Patients with Acute Nondisabling Cerebrovascular Events (CHANCE) trial47 compared combined aspirin and clopidogrel with aspirin alone in 5170 Chinese patients with acute minor stroke or TIA recruited within 24 hours of the event. The results are similar to the meta-analysis and showed that dual therapy was superior with reduced stroke recurrence [hazard ratio (HR) 0.68, 95% confidence interval (CI) 0.57 to 0.81] and no difference in moderate or severe haemorrhage. 47 The above meta-analysis45 was updated with CHANCE47 and summarises all available data for acute stroke and TIA involving trials comparing dual versus mono antiplatelet therapy;46 a summary of the main findings is given in Table 1. The large POINT trial,48 which also compared combined aspirin and clopidogrel for acute minor stroke and TIA (within 12 hours) in a Western population, reported similar findings to CHANCE.

Current stroke prevention is far from perfect. Stroke is heterogeneous in type (ischaemic vs. haemorrhage; lacunar vs. cardioembolic vs. large artery), severity and outcome, and treatments reduce, but do not abolish, events (‘treatment failure’). In addition, patients may be (relatively) insensitive to treatment (‘treatment resistance’, as identified for aspirin and clopidogrel49); hence, improvements in secondary prevention are still needed.

If combined aspirin and dipyridamole is superior to aspirin for long-term secondary prevention,20,21,50 and aspirin and clopidogrel are probably superior to aspirin in acute minor stroke/TIA,38,41,46,47 then triple antiplatelet therapy (combined aspirin, clopidogrel and dipyridamole) might be better still, providing that the risk of recurrence is high and bleeding does not become excessive. In this respect, the risk of bleeding when adding clopidogrel to aspirin and dipyridamole is likely to be similar to that of when adding clopidogrel to aspirin alone, as dual aspirin and dipyridamole does not increase bleeding over aspirin. 20,21 We have performed a series of ‘proof-of-concept’ laboratory and clinical studies investigating this approach. 51–55 Studies in vitro found that triple therapy was most effective in inhibiting aggregation, platelet–leucocyte conjugation and leucocyte activation. 51,54,55 In multiway crossover Phase I and II trials comparing short-term administration of mono, dual and triple antiplatelet platelet therapies, the combination of aspirin and clopidogrel, with or without dipyridamole, was most potent in inhibiting platelet function ex vivo in both normal volunteers (n = 11) and patients with previous stroke/TIA (n = 11). 52,53

In the only parallel group trial of triple therapy in patients with stroke (Phase II trial, n = 17), was triple therapy feasible to administer for up to 24 months. 56 The comparator was aspirin, chosen as this was the UK standard of care at trial commencement. The trial was stopped early on publication of the ESPRIT trial,21 confirming the superiority of dual aspirin and dipyridamole over aspirin (i.e. it was unethical to continue patients on aspirin alone). Predictably, there was increased bleeding with long-term triple therapy versus aspirin. Although unintended, the participants were at a low risk of recurrence (young age/recruited months after the event/many lacunar strokes), a problem that was also seen in the MATCH and CHARISMA trials. 26,57 A conclusion was that future trials of triple antiplatelet therapy would need to target patients at a high risk of recurrence so that benefit was likely to outweigh hazard. We have also used chronic triple antiplatelet therapy in clinical practice in patients at a very high risk of recurrence, defined as recurrence on dual antiplatelet therapy, and this appeared to be safe. 58

Short-term randomised controlled trials of triple antiplatelet therapy have been reported in patients with acute coronary syndromes (ACSs) or to cover stent insertion (25 studies,59 17,383 patients). In our published meta-analysis, and in comparison with dual antiplatelet therapy, glycoprotein IIb/IIIa-based triple therapy reduced MI in patients with non-ST elevation myocardial infarction (NSTEMI) (OR 0.70, 95% CI 0.56 to 0.88) and ST-elevation myocardial infarction (STEMI) (OR 0.26, 95% CI 0.17 to 0.38) patients, and vascular events in NSTEMI (OR 0.69, 95% CI 0.55 to 0.86) and STEMI (OR 0.39, 95% CI 0.30 to 0.51) patients. 59 Death was also reduced after STEMI; major bleeding and transfusions were non-significantly increased and were few in number such that benefit outweighed hazard in absolute numbers of patients. The number of stroke events were too few to assess any statistical trends, and minimal or no data were available for other antiplatelets (cilostazol, clopidogrel, dipyridamole). 59

Rationale for trial

The Triple Antiplatelets for Reducing Dependency after Ischaemic Stroke (TARDIS) trial was predicated on the following.

• There is a high early risk of recurrent cerebral ischaemic events after stroke and TIA.

• Some patients ‘fail’ on monotherapy with aspirin, clopidogrel or dipyridamole.

• Dual therapy with aspirin and dipyridamole is superior to aspirin alone after stroke.

• Clopidogrel, or dual aspirin and dipyridamole, is the standard of care in the UK (NICE).

• If dual therapy is superior, then triple therapy may be better still.

• Laboratory studies, Phase I/II trials, and routine clinical use support the use of intensive antiplatelet therapy with three agents. 51–55

Hence, triple therapy may be better still in high-risk patients providing the benefits exceed the risk of bleeding.

Objectives

The primary objective was to assess ordinal stroke severity at 90 days after short-term administration (1 month) of triple antiplatelet therapy (aspirin/clopidogrel/dipyridamole) versus standard dual therapy (aspirin/dipyridamole) in patients with very recent ischaemic stroke or TIA.

Design

The TARDIS trial was an international Prospective Randomised Open-label Blinded End point (PROBE) study that recruited from 106 sites in four countries (Denmark, Georgia, New Zealand and the UK). The protocol can be found online. 60

Study settings

The trial was carried out in the stroke services of 106 hospitals in four countries: the UK, Denmark, Georgia and New Zealand. Stroke services were housed in a range of settings including teaching/tertiary and district/secondary hospitals (see list in Appendix 1).

Participants

Adult patients aged ≥ 50 years were eligible for inclusion if they were at risk of a recurrent ischaemic stroke, and had either a non-cardioembolic ischaemic stroke or a non-cardioembolic TIA. Randomisation had to be performed within 48 hours of symptom onset unless the participants had received intravenous thrombolysis, in which case randomisation had to be performed once 24 hours had elapsed after the end of this treatment and post-treatment neuroimaging excluded secondary cerebral bleeding. Patients gave written consent, or written proxy consent was obtained from a relative or carer if the patient lacked capacity. The consent form and patient information sheet are available at www.journalslibrary.nihr.ac.uk/programmes/hta/1010424/#/ (accessed July 2018). The full criteria follow here.

Data collected at baseline

Baseline data collected immediately prior to randomisation included demographic information (age, sex and ethnicity), medical history including vascular risk factors, current antiplatelet therapy (none, aspirin, clopidogrel, dipyridamole, other), and pre-morbid dependency (mRS, scores range from 0 to 6, with a score of 0 indicating no symptoms, 5 indicating severe dependency and 6 indicating death61,62). Clinical information included blood pressure and stroke syndrome [Oxfordshire Community Stroke Project (OCSP)63]. Neurological impairment was recorded in stroke patients using the National Institutes of Health Stroke Scale (NIHSS), for which scores range from 0 to 42 with higher scores indicating a more severe neurological deficit. 64,65 The risk of recurrence after index TIA was assessed using the ABCD2 scale (scores range from 0 to 7 with higher scores indicating a higher risk of recurrence5).

Interventions

Participants randomised to the intervention (intensive, triple antiplatelets) group received combined aspirin (load 300 mg, maintenance 50–150 mg daily, typically 75 mg, given orally, nasogastrically or rectally), clopidogrel (load 300 mg, maintenance 75 mg daily, given orally or nasogastrically) and dipyridamole (200 mg twice daily modified release, given orally; or 100 mg three or four times daily, given orally or nasogastrically). Those randomised to guideline antiplatelet therapy received either combined aspirin and dipyridamole or clopidogrel alone, using the loading and maintenance doses given above.

The two approaches for guideline therapy arose because of a change in clinical guidelines. At the start of the trial, NICE recommended the use of aspirin and dipyridamole for secondary prevention29 and the initial trial protocol defined this as the guideline comparator. Once clopidogrel became generic (and its price fell), NICE updated their guidance in 2010 with a recommendation that clopidogrel should be used first line for secondary prophylaxis after ischaemic stroke (but not TIA because clopidogrel was not licensed for TIA and there was an absence of randomised trial data for this indication)30 and the protocol was updated accordingly.

Randomised antiplatelet drugs were given for 30 days so as to influence the time period of high risk of recurrence without intending to accrue significant bleed. After 30 days, participants were treated in accordance with local/national guidelines, typically with combined aspirin and dipyridamole or clopidogrel alone. Drugs were sourced by each participating hospital with supply from any licensed manufacturer (including generic sources).

Detailed information on compliance with randomised treatment was collected up to 7 days, but not during the subsequent 21 days to end of treatment.

At the start of the trial, clopidogrel was available as a branded drug, Plavix^® (Sanofi Guildford, UK). Subsequently, multiple generic versions became available with the advantage of a much lower cost; as a result, recruiting sites increasingly used generic clopidogrel. A number of studies and a meta-analysis have compared Plavix with generic clopidogrel and these do not report any significant difference in antiplatelet responses or effects on vascular outcomes. 66–69

Randomisation

Demographic and baseline clinical characteristics were entered online into a secure web-based database system,70 which was a bespoke system that also provided randomisation and was based on that used in the Efficacy of Nitric Oxide in Stroke (ENOS) trial and the Prevention of Decline in Cognition After Stroke Trial (PODCAST). 71,72 Baseline data were checked to confirm the patient’s eligibility and the system then assigned the participant to intensive or guideline antiplatelet therapy, with a 1 : 1 allocation.

Patients randomised to the guideline group received either combined aspirin and dipyridamole, or clopidogrel alone, in accordance with local policy and guidelines, and antiplatelets taken prior to randomisation.

• Each site chose what comparator(s) they wished to use for ischaemic stroke and TIA separately. They could elect to use one comparator only, or randomise between the comparators. The principal investigator could change the choice of comparison group(s) via the database at any stage during the trial, but changes took 48 hours to take effect to avoid them being made for individual participants. The elected comparator option of each site (irrespective of whether a participant was eventually randomised to intensive or guideline treatment) has been included as a covariate in all adjusted analyses and investigated as part of the prespecified subgroup analysis.

• As aspirin and clopidogrel have irreversible effects on platelets while platelets circulate for 7–10 days, randomisation also took account of which antiplatelet(s) had been taken shortly before stroke onset or in hospital prior to randomisation. The concern was that the comparator group might amount to triple antiplatelet therapy once pre-randomisation treatment was taken into account. However, because aspirin is widely taken both before stroke and immediately following scanning in hospital, it was allowed in all choices prior to randomisation. For example, a patient on clopidogrel prior to stroke and given aspirin in hospital could not be randomised to aspirin and dipyridamole as this would functionally result in them being exposed to all three drugs for the first week after randomisation. As a result, the following rules were followed at the time of randomisation when determining an appropriate comparator (Table 2).

Randomisation comprised stratification on country and index event (stroke vs. TIA), and minimisation on key prognostic baseline factors [age, sex, pre-morbid function (using the mRS), systolic blood pressure, syndrome (cortical vs. lacunar63), previous antiplatelet therapy (none/mono vs. dual), use of gastroprotection, use of low dose heparin and time to randomisation]. For TIA participants, minimisation also included presence of crescendo TIAs (more than one TIA in the previous week) and ABCD2 score;5 for those with stroke, minimisation also included NIHSS and treatment with alteplase.

Definition of events (updated from the statistical analysis plan)73

Stroke, MI and bleeding (major, moderate) are adjudicated independently by two adjudicators; if they differ in type of vascular event or severity of bleeding, a third adjudicator also assesses the event and a majority view is recorded for analysis. Serious adverse events (SAEs) are assessed by a single adjudicator. All adjudicators are blinded to treatment assignment.

Assessments after randomisation

Participants were seen in clinic at days 7 (on treatment) and 35 (end of treatment plus 5–7 days to allow for wash-out) to ascertain whether any outcome or bleeding events had taken place (this included performing a full blood count) and to determine compliance with treatment. Final follow-up was performed centrally at 90 days by telephone from the co-ordinating centre in each country, with the assessor blinded to treatment allocation. If the participant could not be contacted (following multiple attempts), a questionnaire covering the same outcome measures was sent by post. Identification of recurrent cerebrovascular events used multiple sources of information: assessment by the investigator at days 7 and 35, or via serious adverse event (SAE) reporting, patient reporting at day 90 telephone follow-up, and by the general practitioner following a questionnaire posted to them shortly after day 90. The primary outcome, bleeding and investigator-reported SAEs (including cause-specific case fatality) were validated and categorised by expert adjudicators who were blinded to treatment assignment. Participants who did not receive their assigned treatment or who did not adhere to the protocol were still followed up in full at day 90.

Primary efficacy outcome

The primary efficacy outcome was the incidence and severity of recurrent stroke and TIA occurring by 90 days. The timing of 90 days was chosen to emulate the usual follow-up period in acute stroke trials. Severity of recurrent stroke was assessment of the mRS, a six-level ordered categorical scale:76,77 fatal stroke, non-fatal severe stroke (mRS score of 4–5), moderate stroke (mRS score of 2–3), mild stroke (mRS score of 0–1), TIA and no stroke or TIA. 77 The primary outcome was a balance of ischaemic stroke (potential benefit if reduced) and intracerebral haemorrhage (potential harm if increased).

Originally, the intention was to analyse across nine levels of severity comprising the individual seven levels of mRS, TIA and no event (protocol version 1.1, dated 17 October 2008; protocol version 1.2, dated 20 May 2009). Subsequently, the scale was reduced to five levels: fatal stroke, moderate to severe stroke (mRS score of 2–5), mild stroke (mRS score of 0–1), TIA and no event (protocol version 1.3, dated 20 December 2011; protocol version 1.4, dated 26 February 2013; protocol version 1.5, dated 28 February 2014). Finally, the scale was increased to six levels. 73 The changes resulted from multiple discussions among the statistical and clinical teams on simplifying the scale (from nine levels) but not overmerging mRS levels, in particular, keeping mild, moderate and severe stroke separate. These changes occurred during enrolment when the database was locked. A post hoc analysis comparing these three approaches (5, 6, 9 levels) was performed.

Secondary efficacy outcomes

Prespecified secondary outcomes at day 9080 included activities of daily living [Barthel Index (BI)81], cognition [modified telephone Mini-Mental State Examination (t-MMSE),82 Telephone Interview for Cognition Scale – modified (TICS-M),83 categorical verbal fluency using animal naming84], health-related quality of life [EuroQol-5 Dimensions, three-level version (EQ-5D-3L),85 from which health status utility values (HSUVs) were calculated; EuroQoL Visual Analogue Scale (EQ-VAS)], and mood [short Zung Depression Scale (ZDS)86]. At discharge from initial hospitalisation, duration of hospital stay and discharge disposition (institution or home) were recorded.

Safety outcomes

The main safety outcome was bleeding comprising a five-level ordered categorical scale: fatal, major, moderate, minor and none. 76 The definitions of fatal, major and moderate bleeding were those of the International Society on Thrombosis and Haemostasis and are based on severity, site, fall in haemoglobin and need for transfusion. 74 Additional safety outcomes included all-cause and cause-specific case fatality, early neurological deterioration (defined as an increase from baseline to day 7 of at least 4 points on the NIHSS and/or decrease in the consciousness component of the NIHSS), and SAEs.

Information on recurrent stroke and TIA, MI and bleeding was collected using the SAE reporting forms, with additional information specific to the outcome.

Net balance in efficacy and safety

To assess the net balance between efficacy and hazard, composite end points of any stroke or major (including fatal) bleeding, and death, stroke, MI or major bleeding, were analysed.

Study oversight

The trial was conceived and designed by the grant applicants who wrote the protocol (available at http://tardistrial.org/jevpybki.htm). The study was approved by national and/or local ethics committees in each participating country and site, was registered (ISRCTN47823388) and adopted in the UK by the National Institute for Health Research (NIHR) Stroke Research Network. The trial was overseen by a Trial Steering Committee (TSC) (which included five independent members and a patient–public representative) and an International Advisory Committee (comprising each national co-ordinator). The day-to-day conduct of the trial was run by a Trial Management Committee based at the TARDIS co-ordinating centre in the Stroke Trials Unit in Nottingham. Study data were collected, monitored and analysed in Nottingham. Analysis, interpretation and report writing were performed independent of the funders and sponsor, and no pharmaceutical companies were involved in any part of the trial. The corresponding author wrote the first draft of this report; this and subsequent drafts were edited by the grant applicants, who all approved the decision to submit the manuscript for publication. The corresponding author and two statisticians (LJW, KF) had full access to all the data in the study. In addition, the corresponding author had final responsibility for the decision to submit for publication and is the guarantor for the study.

Missing data

As outcomes such as mRS, EQ-5D-3L/HSUV and BI include scores for death (6, 0 and –5, respectively) and in case treatment was associated with asymmetric effects on death and other outcome measures (e.g. more death and less impairment), an extreme value for death was added to the other outcome scales: EQ-VAS = –1, t-MMSE = –1, SSS = –1, TICS-M = –1, verbal fluency = –1, and ZDS = 102.5. 77,87 Including death in these secondary outcome scores has the additional advantage of increasing statistical power and of anchoring the scales to one another.

Statistical analyses

Statistical analyses were performed according to the published statistical analysis plan73 by Lisa J Woodhouse and Katie Flaherty (with oversight by SP) using SAS^® software (version 9.3) (SAS Institute Inc., Cary, NC, USA; SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. ^® indicates USA registration). Analyses were by intention to treat for all comparisons (i.e. according to the treatment group participants were allocated to and irrespective of the treatment they actually received). Data are shown as number (%), median [interquartile range (IQR)], mean [standard deviation (SD)] and OR (95% CIs).

Analysis of the effect of treatment on the primary efficacy outcome was analysed as a shift in stroke and its severity [fatal stroke, non-fatal severe stroke (mRS 4–5), moderate stroke (mRS 2–3), mild stroke (mRS 0–1), TIA and no stroke or TIA] with adjustment for the factors used in stratification and minimisation at the time of randomisation (index event – ischaemic stroke, TIA; country; guideline randomisation choice – aspirin/dipyridamole, clopidogrel, either; age; sex – female, male; pre-morbid mRS; time from onset to randomisation; number of antiplatelets before index event; stroke syndrome – lacunar syndrome (LACS), posterior circulation syndrome (POCS), partial anterior circulation syndrome (PACS), total anterior circulation syndrome (TACS) systolic blood pressure; gastroprotection – yes, no; use of heparin – yes, no; stroke severity – NIHSS; treated with recombinant tissue plasminogen activator – yes, no; ABCD2 score; number of TIA in last week88) and is reported as an adjusted common odds ratio (acOR) with 95% CIs. The common odds ratio (cOR) represents the odds of a patient on treatment moving categories of outcome as compared with a patient on control; cOR < 1 suggests benefit and cOR > 1 hazard. The OR and significance were calculated using ordinal logistic regression following a check (using the likelihood ratio test) that the assumption of common proportional odds was not violated. For sensitivity purposes, the primary outcome was also analysed without adjustment and as a binary outcome of fatal or major stroke. The heterogeneity of the treatment effect on the primary outcome was assessed in prespecified subgroups by adding an interaction term in an unadjusted ordinal logistic regression model. Similarly, the effect of treatment on the main safety outcome was analysed as a shift in bleeding and its severity (fatal, major, moderate, minor and none) with adjustment for the stratification and minimisation factors. For sensitivity purposes, bleeding was also analysed unadjusted, as a binary outcome of fatal and major bleeding, heterogeneity within subgroups was also assessed. The composite outcomes of stroke or major bleeding, and death, stroke, MI or major bleeding were compared between treatment groups using adjusted Cox regression.

Death was analysed using Kaplan–Meier and Cox regression models. Other outcomes were analysed using adjusted multiple linear regression (BI, ZDS, t-MMSE, TICS-M, verbal fluency, EQ-5D/HSUV and EQ-VAS). All analyses were also performed unadjusted for completeness. The nominal level of significance for all analyses, including interaction testing, was p < 0.05. No adjustment was made for multiplicity of testing for secondary analyses.

Sample size

The TARDIS trial was designed with a vanguard phase to assess safety, feasibility and tolerability [funded by the British Heart Foundation (BHF)] and a main phase to assess safety and efficacy (funded by NIHR Health Technology Assessment).

The null hypothesis (H₀) was that intensive antiplatelets would not alter the frequency and severity of stroke/TIA in participants with previous ischaemic stroke or TIA. The alternative hypothesis was that the frequency and severity of stroke/TIA would differ between those participants randomised to intensive versus guideline antiplatelets. A total sample size89,90 of 4100 (2050 per group) participants with ischaemic stroke or TIA was required, assuming overall type I error rate (significance, alpha) = 0.05 with two-sided significance test. Power (1-beta) = 0.90, OR = 0.68 (equivalent to an OR = 0.57 and RRR = 0.31 for binary stroke), distribution in outcome as in Table 3, treatment crossovers 5% and losses to follow-up 2%, and a reduction of 20% for baseline covariate adjustment. 90,91 No adjustment was made for two interim analyses in view of the stringent stopping rules given to the DMC (see above).

Originally, a five-level stroke/TIA ordered outcome was planned with distribution based on data from the TARDIS trial during the vanguard phase (n = 392) (see Table 3). However, this was changed to a six-level scale to allow discrimination of moderate and severe non-fatal stroke outcomes, and to keep together mRS scores of 2 and 3 (which can be challenging to separate clinically). Distribution of mRS came from blinded data from a TSC report (based on n = 1460 participants).

Protecting against bias, including blinding

Multiple measures were taken to minimise bias: recruitment on the basis of predefined inclusion/exclusion criteria, with exclusion of patients enrolled in other trials; central data registration with real-time data validation and concealment of allocation; blinded central telephone assessment of day 90 outcomes by the National Coordinating Centre (NCC) staff; assessment of patient recall of treatment;92 blinded adjudication of outcomes (stroke/TIA, MI, bleeding), SAEs and CT/MRI scans; research staff trained in trial protocol and processes; analysis by intention to treat; analyses adjusted for baseline prognostic variables, including minimisation factors; adjustment for non-randomised acute treatment (thrombolysis with alteplase).

Neuroimaging scan adjudication

The CT or MRI brain scans were performed in accordance with local site practice at baseline in all patients with ischaemic stroke to confirm the diagnosis. Investigators decided whether or not to perform neuroimaging in patients with TIA depending on clinical need. Sites could also perform follow-up scans at any time point after enrolment according to clinical need (e.g. if a neurovascular outcome or bleeding was suspected). For this publication, scan-derived information is based on radiological reporting from local sites.

Future reports will utilise information based on central adjudication of neuroimages. Neuroimages were submitted to the International Coordinating Centre in Nottingham using one of two methods:

1. Uploaded onto the trial website as uncompressed encrypted non-anonymised digital imaging and communications in medicine (DICOM) files. Once the trial system had validated the files against the expected patient details, the files were then anonymised.

2. Sent by courier on a compact disc read-only memory (CD-ROM) or digital versatile disc (DVD), with files in DICOM format with pseudo-anonymisation of patient details; the patients were identified with their unique study number and initials.

When reviewed, some images were in non-DICOM format [e.g. portable network graphic (PNG), Joint Photographic Experts Group (JPG)] and these were converted to DICOM. The anonymised image files were then presented to a panel of expert adjudicators using a browser-based system driven from the trial database. Adjudicators were trained and assessed using the ACCESS system [https://sirs2.ccbs.ed.ac.uk/sirs2 (accessed August 2018)]93,94 and reviewed scans blinded to treatment assignment. Adjudication parameters were derived from the International Stroke Trial-3 and ENOS trial image adjudication systems71,95 and included information on:

• the presence of an acute stroke lesion – location, mass effect and presence of secondary ischaemia or haemorrhage

• the presence of pre-stroke changes – atrophy, white matter hyperintensities, old stroke.

Information from adjudication was used to inform the final diagnosis for all participants with a received scan. When clinical and radiological information were incongruent, Robert A Direen performed a second adjudication to confirm imaging findings.

Sites, investigators and monitoring

Protocol amendments

Five protocol amendments were made during the trial, these started from version 1.1.

Substudies

Three substudies were funded by the BHF in the TARDIS start-up phase.

Recruitment

The trial was designed in 2007 and initial funding was confirmed in 2008. The trial commenced recruitment on 7 April 2009 and enrolment was halted on the advice of the independent DMC on 18 March 2016, after a total of 3096 participants (out of a planned target of 4100; 75.5%) had been enrolled. Prior to this, the DMC had met on 12 occasions and recommended trial continuation. The advice to stop was based on a combination of three observations:

1. the presence of a significant increase in major (including fatal) bleeding in participants randomised to intensive antiplatelet therapy

2. intensive antiplatelet therapy was associated with a non-significant reduction in the primary outcome, but not enough to numerically outweigh the increase in bleeding

3. a conditional power analysis suggested that the trial was highly unlikely to demonstrate a significant difference in the primary outcome.

Once the decision to halt further recruitment had been made, patients randomised to intensive antiplatelet therapy were telephoned and asked to switch treatment to guideline therapy. The TSC reviewed the same data as well as additional analyses on 12 April 2016 and additionally noted that there was no difference in the net balance between benefit and hazard based on the composite of death, stroke, MI and major bleeding. The TSC stopped recruitment on the basis of futility.

Participant flow

The recruitment curve is shown in Figure 1 and shows three phases of recruitment: an initial start-up phase with very low rate, a vanguard phase funded by British Heart Foundation with an intermediate rate and a main phase with high rate and funded by the NIHR’s HTA programme. The Consolidated Standards of Reporting Trials (CONSORT) flow diagram is shown in Figure 2. By 90 days, 54 (1.7%) patients had died and data were missing from 26 (0.8%) patients. A majority of participants (2955, 95.4%) were recruited in the UK, with others from Denmark and Georgia (134, 4.3%) and New Zealand (7, 0.2%) (Table 4). Four sites recruited a far higher proportion of patients than other sites (‘waterfall plot’; Figure 3).

FIGURE 1.

Recruitment graph by time. Updated from Bath et al. 99 © 2016 World Stroke Organization. This article is distributed under the terms of the Creative Commons Attribution 3.0 License (http://www.creativecommons.org/licenses/by/3.0/) which permits any use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage).

Updated from Bath et al. 99 © 2016 World Stroke Organization. This article is distributed under the terms of the Creative Commons Attribution 3.0 License (http://www.creativecommons.org/licenses/by/3.0/) which permits any use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage).

FIGURE 2.

The CONSORT flow diagram showing the flow of participants in the trial.

FIGURE 3.

Recruitment by hospital sites (‘waterfall plot’). Updated from Bath et al. 99 © 2016 World Stroke Organization. This article is distributed under the terms of the Creative Commons Attribution 3.0 License (http://www.creativecommons.org/licenses/by/3.0/) which permits any use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage).

Updated from Bath et al. 99 © 2016 World Stroke Organization. This article is distributed under the terms of the Creative Commons Attribution 3.0 License (http://www.creativecommons.org/licenses/by/3.0/) which permits any use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage).

Baseline data

Table 5 shows the baseline characteristics of study participants, which were well matched except for a difference in ‘other antiplatelets’ (absolute difference of 0.5%). The mean age was 69.0 (SD 10.1) years and 1945 (62.8%) participants were male. Clinical information at the time of randomisation, and subsequent investigations, judged the qualifying event to be ischaemic stroke in 2220 participants (71.7%), TIA in 838 participants (27.1%) and non-ischaemic stroke/TIA (mimic) in 38 participants (1.2%). The median time from ictus to randomisation was 29.3 (IQR 21.8–39.6) hours; 314 (10.1%) and 651 (21.0%) participants were recruited within 12 and 13–24 hours of onset, respectively. In participants recruited with an ischaemic stroke, the mean NIHSS score was 4.0 (SD 3.8) and 336 participants (15.7%) received thrombolysis. In participants who were thrombolysed, the median times to this commencing from ictus and randomisation were 2.3 (IQR 1.8–3.0) hours and 36.5 (IQR 28.4–42.6) hours, respectively. In TIA participants, the median ABCD2 score was 5.0 (IQR 5.0–6.0), 155 participants (20.1%) presented with a crescendo TIA (defined as more than one TIA over the previous week), and 36 participants (4.3%) were already taking two antiplatelet agents. Among all participants, the mean blood pressure was 143.5 (SD 18.2)/79.5 (SD 11.4) mmHg and the clinical syndrome was cortical (combined total and partial anterior syndromes) in 1593 participants (51.5%) and lacunar in 1288 participants (41.6%). 63

Antiplatelet treatment

Out of the 3096 recruited participants, 1556 were randomised to intensive antiplatelet therapy and 1540 to the guideline group (see Figure 2). More participants were randomised to clopidogrel alone (n = 849, 55.1%) than combined aspirin and dipyridamole (n = 691, 44.9%) in the control group. Adherence to randomised treatment over the first 7 days was fair in both treatment groups: initial treatment was received in 83.5% of patients and 98.4% received at least some randomised treatment during the first week (Table 6). Participants randomised to intensive antiplatelets were more likely to receive all treatment in the first week than those randomised to guideline treatment.

The rate of treatment crossover was low: 6 (0.4%) participants randomised to the intensive group did not receive triple antiplatelets and 30 (1.9%) participants randomised to the guideline group received intensive antiplatelets at some point in the treatment phase. During treatment, 37 (1.2%) participants had carotid endarterectomy leading to temporary cessation of one or more antiplatelets (Table 6). A further 31 (1.0%) participants had antiplatelet therapy replaced with oral anticoagulation following identification of AF. Gastroprotection was recommended in the protocol and was provided in 803 (25.9%) patients: intensive for 405 patients (26.0%) and guideline for 398 patients (25.8%). Following the final DMC review and early closure to recruitment, participants randomised to intensive antiplatelet therapy were immediately asked to switch treatment to guideline therapys.

Outcome assessment

Final follow-up (at day 90) was completed for 3016 (97.4%) participants and vital status was available for all but five (0.1%) (see Figure 2). Outcomes were determined by telephone in 2889 (93.3%) participants and by post in 67 (2.2%) participants.

Primary efficacy outcome

The primary outcome (recurrent stroke or TIA and their severity by day 90) was determined in 3070 (99.2%) participants. Overall, a total of 198 (6.4%) participants had a recurrent stroke or TIA (intensive, n = 93, vs. guideline, n = 105; Table 7), which comprised 118 strokes (ischaemic, n = 96; haemorrhagic, n = 19; unknown type due to absence of neuroimaging, n = 5; both an ischaemic and haemorrhagic stroke, n = 1; and both an ischaemic and unknown type of stroke, n = 1) and 80 TIAs. There was no difference in the incidence and severity of stroke or TIA (acOR 0.90, 95% CI 0.67 to 1.20; p = 0.47) or fatal stroke (aOR 1.95, 95% CI 0.76 to 4.99; p = 0.16) between intensive and guideline antiplatelet therapies (Figure 4 and see Table 7). The likelihood ratio test for the proportional odds assumption was not significant (χ² = 0.178, degrees of freedom = 8; p = 1.00). The primary efficacy outcome by time to recruitment is shown in Table 8.

FIGURE 4.

Primary outcome as distribution of recurrent stroke and TIA, and their severity, at day 90. Comparison by ordinal logistic regression adjusted for baseline factors.

Originally, the trial intended to assess the effect of intervention on a nine-level ordinal outcome for stroke and TIA, and subsequently a five-level outcome. However, these were felt to be over- or under-granular, respectively, and the six-level outcome was chosen for the final analysis. Decisions on these changes were made blinded to treatment assignment. In post hoc analyses, the original outcomes were analysed and gave comparable results (Table 9).

When assessed in prespecified subgroups (Figure 5), there were no significant interactions between the primary outcome and treatment. In the intensive arm, there was a tendency for fewer and less severe strokes/TIAs in participants presenting with a mild stroke, and worse outcomes in participants with a more severe stroke, but this interaction was not significant (p = 0.070). In a sensitivity analysis, the rate of stroke or TIA by day 90 did not differ between participants receiving intensive versus guideline antiplatelet therapy (aOR 0.88, 95% CI 0.65 to 1.18; p = 0.38) (Figure 6 and see Table 7). Similarly, alternative cut-off points were tested for the measures of severity, onset to randomisation and degree of carotid stenosis. These post hoc analyses, and the original results, are shown in Table 10; all the interaction tests remain non-significant.

FIGURE 5.

Primary outcome at day 90 in prespecified subgroups. Unadjusted common odds ratio. a, Depending on local policy and guidelines, sites were given the option to choose which treatment they would give as the comparator, that is, aspirin/dipyridamole, clopidogrel alone or either treatment. NC, not calculable.

FIGURE 6.

Kaplan–Meier curve for recurrent stroke or TIA over 90 days. Comparison by Cox proportional regression adjusted for baseline factors.

Although there was no difference between the treatment groups for stroke alone as an outcome, there was a tendency for more recurrent strokes of haemorrhagic type in the group randomised to intensive antiplatelet therapy (aOR 2.89, 95% CI 1.02 to 8.19; p = 0.046) (see Table 7). The rates of TIA were lower in the intensive treatment group (aOR 0.62, 95% CI 0.39 to 0.96; p = 0.033) (Figure 7 and see Table 7). Analyses based on unadjusted comparisons did not differ qualitatively for the primary outcome.

FIGURE 7.

Kaplan–Meier curve for recurrent TIA over 90 days. Comparison by Cox proportional regression adjusted for baseline factors.

Secondary efficacy outcomes

Central follow-up at day 90

Across all patients, there was no difference between the treatment groups in outcome at day 90 assessed as dependency (mRS, Figure 8), disability (BI), mood (Zung Depression Scale), cognition (t-MMSE, telephone interview cognition status – modified, verbal fluency) or quality of life (EQ-VAS) (see Table 11). Similarly, the rate of MI did not differ between the treatment groups. Analyses based on unadjusted comparisons did not differ qualitatively for any of the secondary outcomes (see Table 11).

FIGURE 8.

Distribution of mRS in all patients at day 90 comparison, by ordinal logistic regression adjusted for baseline factors.

Safety: bleeding

The main safety outcome was the distribution of risk and severity of bleeding (using the ordinal scale of fatal, major, moderate, mild or no bleed) at day 90; this was shifted to more bleeding and bleeding of greater severity in participants randomised to intensive antiplatelet therapy (acOR 2.54, 95% CI 2.05 to 3.16; p < 0.001) (Table 12 and Figure 9). When the distribution of bleeding and its severity was assessed in prespecified subgroups, a statistically significant interaction between bleeding and the site comparator choice was present (Figure 10). Intensive treatment was associated with more bleeding than guideline treatment when compared within sites that chose aspirin and dipyridamole as the comparator arm than when compared within sites that chose clopidogrel alone as the comparator. An interaction was also seen for patients who received thrombolysis; intensive antiplatelets were associated with higher bleeding rates in those who received thrombolysis than those who did not. The effect of treatment on ordinal bleeding by time to recruitment is shown in Table 13.

FIGURE 9.

Distribution of ordinal bleeding at day 90. Comparison by ordinal logistic regression adjusted for baseline factors.

FIGURE 10.

Ordinal bleeding at day 90 in prespecified subgroups. Unadjusted common odds ratio. a, Depending on local policy and guidelines, sites were given the option to choose which treatment they would give as the comparator, that is, aspirin/dipyridamole, clopidogrel alone or either treatment. NC, not calculable.

The rate of severe (fatal or major) bleeding by day 90 was 2.5% in those assigned to intensive antiplatelet therapy and 1.1% among participants receiving guideline antiplatelets [adjusted hazard ratio (aHR) 2.23, 95% CI 1.26 to 3.96; p = 0.006] (Figure 11 and see Table 13). The rates of bleeding increasingly diverged between the treatment groups up to the end of treatment at day 30 but not thereafter. Combined fatal and major intracranial bleeding was increased with intensive antiplatelet therapy (aHR 3.84, 95% CI 1.26 to 11.63; p = 0.018). A non-significant tendency to more fatal or major extracranial bleeding was also seen with intensive antiplatelets (aHR 1.89, 95% CI 0.96 to 3.71; p = 0.064) (see Table 13).

FIGURE 11.

Kaplan–Meier curve for major bleeding over 90 days. Comparison by Cox proportional regression adjusted for baseline factors.

The above differences in ordinal bleeding between intensive and guideline antiplatelets at day 90 were apparent by day 7, and ordinal bleeding, major extracranial bleeding and fatal/major bleeding were different at day 35 (see Table 13). At day 90, intensive antiplatelet was associated with more, and more severe, bleeding when assessed as the worst (see Table 13) or first (see Table 12 and Table 14) bleed. Similarly, intensive antiplatelet therapy was associated, in post hoc analyses, with a higher rate of major bleeding by day 14 of treatment (Table 15).

Serious adverse events

There was no evidence of a mortality difference between the treatment groups (Figure 12 and see Table 7). Excluding primary outcome and bleeding events, the overall incidence of SAEs was similar in the two treatment groups: intensive 21.7% versus guideline 21.4% (acOR 1.02, 95% CI 0.86 to 1.22; p = 0.80) (Table 16 and Figure 13); similarly, the rate of fatal SAEs did not differ between the treatment groups (intensive 0.8%, guideline 1.4%, aHR 0.52, 95% CI 0.25 to 1.05; p = 0.070).

FIGURE 12.

Kaplan–Meier curve for death over 90 days. Comparison by Cox proportional regression adjusted for baseline factors.

FIGURE 13.

Distribution of SAEs at day 90. Comparison by ordinal logistic regression adjusted for baseline factors.

The severity of SAEs, as judged by the reporting investigator, did not differ in their frequency or distribution (see Table 16). When assessed by time in trial, SAEs were more common by day 7 in the intensive group than in the guideline group, although this difference had disappeared by day 35 (Figure 14). When considered by organ, no differences were apparent except that gastrointestinal-based SAEs (excluding reported bleeding events) were more frequent in the intensive than guideline group despite gastroprotection being recommended (2.8% vs. 1.0%; p < 0.001; see Table 16). In a post hoc analysis, gastrointestinal-based SAEs (excluding reported bleeding events) differed for both comparisons of the intensive group with clopidogrel [43 (2.8%) vs. 8 (0.9%); p = 0.005] and with combined aspirin and dipyridamole [43 (2.8%) vs. 7 (1.0%); p = 0.012], suggesting that the presence of dipyridamole was not the only explanation for the increase in SAEs.

FIGURE 14.

Kaplan–Meier curve for SAEs over 90 days. Comparison by Cox proportional regression adjusted for baseline factors.

Composite outcomes

The composite end point of any stroke or fatal and major bleeding occurred in 87 (5.6%) participants in the intensive group and 69 (4.5%) participants in the guideline group (aHR, 1.24, 95% CI 0.90 to 1.70; p = 0.19) (Table 17). Similarly, the composite end point of death, stroke, MI or fatal and major bleeding did not differ between the treatment groups: intensive group 102 (6.6%) vs. guideline group 98 (6.4%) (aHR, 1.02, 95% CI 0.77 to 1.35; p = 0.88). Figure 15 shows a bar chart of the relative frequencies of different measures of net benefit and risk; no differences are apparent.

FIGURE 15.

Bar chart for net benefit: risk at day 90. All comparisons non-significant.

Protocol violations

One or more protocol violations were present in 414 (13.4%) participants, with more present in the intensive (n = 246, 15.8%) than guideline (n = 154, 10.0%) group (p < 0.001) (Table 18). When considered as subtypes of violations, the difference between the treatment groups was maintained for matters of trial practice such as not administering the intervention loading dose or not giving the interventions for at least 16 days.

The TARDIS trial in context of earlier antiplatelet intensity trials

Findings from earlier large trials and meta-analyses suggested that acute dual antiplatelet therapy (either aspirin and dipyridamole, or aspirin and clopidogrel) was superior to one agent in preventing early recurrent cerebral ischaemic events. 45,46 The results of the TARDIS trial, which involved combined aspirin, clopidogrel and dipyridamole, are statistically different (p = 0.02; I² 80.5%) from those earlier trials that involved two agents when considering the outcome of recurrent stroke (Figure 16). As a result, the summary statistic at the bottom of the figure is irrelevant. In contrast, there was no difference between the TARDIS trial and earlier trials for major/fatal bleeding (p = 0.23; I² 30.8%) (Figure 17).

FIGURE 16.

Recurrent stroke in trials of intensive vs. less intensive antiplatelet therapy in patients with acute (within ≤ 3 days) ischaemic stroke or TIA. Data are updated with permission from Wong et al. 46 (© 2013 American Heart Association, Inc.) and Table 1. M–H, Mantel–Haenszel.

Data are updated with permission from Wong et al. 46

FIGURE 17.

Major bleeding in trials of intensive vs. less intensive antiplatelet therapy in patients with acute (within ≤ 3 days) ischaemic stroke or TIA. Data are updated with permission from Wong et al. 46 (© 2013 American Heart Association, Inc.) and Table 1. M–H, Mantel–Haenszel.

Data are updated with permission from Wong et al. 46

Interpretation

The TARDIS trial met its main objective of comparing the safety and efficacy of intensive triple antiplatelet therapy with guideline-based treatment. In this group of patients with acute non-cardioembolic ischaemic stroke or TIA, intensive antiplatelet therapy did not reduce stroke recurrence or its severity when compared with guideline antiplatelet therapy. Guideline antiplatelet therapy comprised either clopidogrel alone or combined aspirin and dipyridamole. Furthermore, intensive antiplatelet therapy did not improve measures of disability, cognition, quality of life or mood. However, intensive antiplatelet therapy was associated with both an increase in, and more severe, bleeding. There was no difference in mortality or the composite end point of stroke or major bleeding.

Previous meta-analyses of trials of antiplatelets in acute stroke/TIA have suggested that it is the number of drugs (i.e. two vs. one), rather than which drugs are used, that is important when determining efficacy, at least when considering aspirin, clopidogrel and dipyridamole. 45,46 If two agents are better than one, then three might be better still, providing that bleeding is not overly increased. However, the TARDIS trial demonstrated that the addition of a third agent does not reduce recurrent stroke but does increase bleeding. Because the primary outcome included haemorrhagic stroke, the failure to reduce stroke recurrence and its severity overall appears to reflect the combination of increased secondary intracranial haemorrhage and a tendency to reduced cerebral ischaemic stroke. Several factors appear to explain the results. First, participants with a severe stroke (presenting with cortical strokes) tended to do better on guideline therapy whereas intensive antiplatelets favoured those with mild stroke. The explanation for this observation is not obvious, as stroke severity did not appear to influence the effect of treatment on bleeding. Second, the type of guideline comparator appeared to be important, as there was a tendency, albeit just non-significant, for intensive therapy to have beneficial effects on the primary outcome in comparison with combined aspirin and dipyridamole, but not when compared with clopidogrel alone. In parallel, intensive antiplatelet therapy was more likely to cause bleeding when compared with combined aspirin and dipyridamole than when compared with clopidogrel. Nevertheless, these comparisons are indirect as most sites did not elect to randomise participants between the guideline groups.

A confounding factor was the use of thrombolysis that may have increased the difference in bleeding between intensive versus guideline antiplatelet therapy. Large trials such as CHANCE and Acute Stroke or Transient Ischaemic Attack Treated with Aspirin or Ticagrelor and Patient Outcomes (SOCRATES) excluded patients who received thrombolysis because of shorter recruitment time windows and inclusion of patients with only mild stroke. 47,104 In the TARDIS trial, randomised antiplatelet therapy commenced 24 hours after the completion of any alteplase maintenance dose, which might have promoted bleeding. A treatment–thrombolysis interaction was present for bleeding, which is surprising as the circulating half-life of alteplase is a few minutes (although the tissue and biological half-lives may be longer). Furthermore, antiplatelet agents were given with a loading dose following thrombolysis, as recommended in guidelines, and it is possible that this acceleration of antiplatelet activity contributed to the risk of bleeding in the presence of recent thrombolysis. Whatever the reason, this finding suggests that bleeding risk is higher following intravenous thrombolysis in spite of an interval of ≥ 24 hours between completing alteplase and randomising into the TARDIS trial.

The TARDIS trial was the first trial designed to use ordered categorical primary and safety outcomes of the type fatal event/severe non-fatal event/mild non-fatal event/no event. Empirical analyses of this approach using published data from existing trials of antiplatelets and other prophylactic interventions suggested that studies would have more statistical power, or the same power for a smaller sample size. 76 A key secondary aim of the TARDIS trial was to test this methodological approach and to compare it with analyses based on binary outcomes. As the primary efficacy analysis was non-significant, the relative merits of ordinal versus binary analysis could not be adequately assessed for stroke. However, ordinal bleeding gave comparable, if not more pronounced, results in comparison with the outcome of ‘fatal or major bleeding’.

Strengths

The present trial has several strengths, especially generalisability owing to wide inclusion criteria. In addition to motor presentations, patients with acute ischaemic stroke included those with severe stroke, dysphasia or neuroimaging-positive hemianopia. Similarly, patients with TIA included those with crescendo TIA or who were already on dual antiplatelet agents. Hence, groups of patients who are typically excluded in stroke prevention trials could be enrolled. Inclusion of patients with severe stroke meant that those with cortical syndromes, often a minority in such trials, could participate. The wide time window of 48 hours meant that patients could be enrolled after intravenous thrombolysis. Furthermore, the trial had a large sample size of > 3000 patients, concealment of treatment assignment, prospective assessment of multiple outcomes including safety measures such as haemorrhage, very high rates of follow-up (99.2% of participants had their primary outcome determined), care in specialist stroke services, and use of locally sourced aspirin, clopidogrel and dipyridamole from a variety of manufacturers (thus increasing the external validity of the trial). An additional strength is that the results define clearly that three agents do not add further efficacy over and above standard care based on one or two agents.

Limitations

Nevertheless, several limitations apply. First, the broad population might have included groups that were more likely to either respond (e.g. those with minor stroke or TIA, or atherosclerotic disease105) or have a major bleed (e.g. those receiving thrombolysis or having small vessel disease or microbleeds106), and this might explain the neutral results. Future trials of antiplatelets may need to be more specific and focus on individuals with atherosclerotic disease, who do not have significant numbers of microbleeds on MRI scanning, and who are known to respond to the antiplatelet agent(s) under test. Second, the antiplatelet agents were administered in an open-label design and participants knew which drug(s) they were on. This may have driven the reporting of known adverse events such as headache with dipyridamole and bleeding with intensive antiplatelet therapy. In mitigation, outcomes at day 90 were assessed centrally and blinded to treatment assignment to reduce the potential for bias. Third, the comparator group involved different antiplatelet agents, a situation reflecting changes in national and international guidelines that added monotherapy with clopidogrel to the existing recommendation of combined aspirin and dipyridamole. The PRoFESS mega trial compared these two strategies in 20,332 patients with chronic (not acute) ischaemic stroke or TIA and found no differential effect on stroke recurrence, although major haemorrhage occurred more frequently with combined aspirin and dipyridamole. 107 It should be noted that the TARDIS trial did not allow aspirin monotherapy (as shown to be effective for ischaemic stroke in two mega trials36,37) as a comparator because this was not recommended in UK guidelines for secondary prevention in either 2005 or 2010, largely because both aspirin and dipyridamole, and clopidogrel alone, had previously been shown to be superior to aspirin alone in three large trials. 18,20,21 Fourth, the time window for recruitment was longer than in other trials such as POINT,48 CHANCE47 and SOCRATES,104 and this led to a lower overall event rate because the risk of recurrence falls with time, even over the first 48 hours. Fifth, randomised treatments were given for 30 days and this might have been too long in view of the identified haemorrhage risk. Importantly, the risk of haemorrhage for intensive and guideline antiplatelets diverged from the start of randomised treatment and was significantly different by 14 days; hence, it is not apparent that a shorter period of intensive antiplatelets would have avoided the risk of bleeding. Last, the trial was stopped early following recommendation by the independent DMC and the results may represent a false-neutral finding related to the lower than planned statistical power. The trial recruited > 70% of its planned target of 4100 participants and the post hoc statistical power remained high at 85%. Furthermore, the prespecified effect size of 0.68 was almost ruled out (with 95% confidence). As such, it is likely that the trial’s main findings are correct (i.e. that intensive antiplatelet therapy does not appear to reduce recurrent cerebral ischaemic events but does increase the risk of haemorrhage).

Patient and public involvement

The trial was designed and delivered with active patient and public involvement (PPI) involvement by three patients, including in design, helping writing patient information sheets, membership of the TSC and interpretation of the results.

Attendance at TSCs was challenging for PPI members both in attending physically (in the presence of continuing physical impairments) and in discussion (in the presence of continuing dysphasia). Although joining by telephone resolved physical issues it was complicated by partial hearing loss. Nevertheless, these concerns were outweighed by the active involvement of PPI members, and they enjoyed the experience.

Conclusions

Contributions of authors

Philip M Bath (Stroke Association Professor of Stroke Medicine, Stroke) was chief investigator, is guarantor for the study and wrote the first draft of the manuscript. He is also a NIHR senior investigator.

Lisa J Woodhouse (Statistical Research Fellow, Stroke) ran data checks during the trial and prepared the results for publication. She revised the manuscript.

Jason P Appleton (Clinical Research Fellow, Stroke) supported medical questions and problems during the trial. He revised the manuscript.

Maia Beridze (Professor of Neurology, Neurology) was national co-ordinator for Georgia. She revised the manuscript.

Hanne Christensen (Professor of Neurology, Neurology) was national co-ordinator for Denmark. She revised the manuscript.

Robert A Dineen (Associate Professor of Neuroradiology, Neuroradiology) led neuroimaging. He revised the manuscript.

Katie Flaherty (Medical Statistician, Stroke) supported preparation of the results for publication. She revised the manuscript.

Lelia Duley (Professor of Clinical Trials Research, Clinical Trials Unit) advised on trial design and delivery. She revised the manuscript.

Timothy J England (Associate Professor of Stroke, Vascular Medicine) cowrote the protocol and supported medical questions and problems during the trial. He revised the manuscript.

Diane Havard (Senior Trials Manager, Stroke) managed the International Coordinating Centre. She revised the manuscript.

Stan Heptinstall (Professor of Haemostasis and Thrombosis, Stroke) led the P-selectin substudy. He revised the manuscript.

Marilyn James (Professor of Health Economics, Rehabilitation and Ageing) led the proto health economics substudy. She revised the manuscript.

Chibeka Kasonde (Patient Representative, Nottingham) advised on patient–carer issues, interpretation of the results and dissemination. She revised the manuscript.

Kailash Krishnan (Consultant Stroke Physician, Stroke) supported medical questions and problems during the trial, and adjudicated outcomes and SAEs. He revised the manuscript.

Hugh S Markus (Professor of Neurology, Neurology) led the TCD substudy. He revised the manuscript. He is also a NIHR senior investigator.

Alan A Montgomery (Professor of Medical Statistics and Clinical Trials, Clinical Trials Unit) supported the analyses. He revised the manuscript.

Stuart Pocock (Professor of Medical Statistics, Medical Statistics) led on statistical design and supported the analyses. He revised the manuscript.

Marc Randall (Consultant Neurologist, Neurology) adjudicated outcomes and SAEs. He revised the manuscript.

Annamarei Ranta (Associate Professor of Neurology, Neurology) was national co-ordinator for New Zealand. She revised the manuscript.

Thompson G Robinson (Professor of Stroke Medicine, Stroke) advised on trial design and delivery. He revised the manuscript. He is also a NIHR senior investigator.

Polly Scutt (Statistical Research Fellow, Stroke) supported preparation of the results for publication. She revised the manuscript.

Graham S Venables (Consultant Neurologist, Neurology) advised on trial design and delivery. He revised the manuscript.

Nikola Sprigg (Professor of Stroke Medicine, Stroke) was deputy chief investigator, and led adjudication of outcomes and SAEs. She revised the manuscript.

Publications

Bath PM, Robson K, Woodhouse LJ, Sprigg N, Dineen R, Pocock S, et al. Statistical analysis plan for the ‘triple antiplatelets for reducing dependency after ischaemic stroke’ (TARDIS) trial. Int J Stroke 2015;10:449–51.

TARDIS Trial Investigators. Safety and efficacy of intensive vs. guideline antiplatelet therapy in high-risk patients with recent ischemic stroke or transient ischemic attack: rationale and design of the Triple Antiplatelets for reducing Dependency after Ischaemic Stroke (TARDIS) trial (ISRCTN47823388). Int J Stroke 2015;10:1159–65.

Bath PM, Appleton JP, Beridze M, Christensen H, Dineen RA, Duley L, et al. Baseline characteristics of the 3096 patients recruited into the ‘Triple Antiplatelets for Reducing Dependency after Ischemic Stroke’ trial. Int J Stroke 2017;12:524–38.

Bath PM, May J, Flaherty K, Woodhouse LJ, Dovlatova N, Fox SC, et al. Remote assessment of platelet function in patients with acute stroke or transient ischaemic attack. Stroke Res Treat 2017;7365684.

TARDIS. Wiki Journal Club 2017 November. URL: www.wikijournalclub.org/wiki/TARDIS (accessed 8 August 2018).

Bath PM, Woodhouse LJ, Appleton JP, Beridze M, Christensen H, Dineen RA, et al. Antiplatelet therapy with aspirin, clopidogrel, and dipyridamole versus clopidogrel alone or aspirin and dipyridamole in patients with acute cerebral ischaemia (TARDIS): a randomised, open-label, phase 3 superiority trial. Lancet 2018;391:850–9.

Data-sharing statement

Anonymised data will be deposited with the Virtual International Stroke Trials Archive. 108 All queries and data requests should be submitted to the corresponding author for consideration in the first instance.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care.

Study investigators