Effects of alteplase within 6 hours of ischaemic stroke on survival up to 3 years: Results from the Third International Stroke Trial (IST-3)
Short title:Alteplase for acute ischaemic stroke
Eivind Berge, MD1;Geoffrey Cohen, MSc2; Melinda B. Roaldsen, MD3, Erik Lundström, MD4;Eva Isaksson, BN4; Ann-Sofie Rudberg, MD4; Karsten Bruins Slot, MD1; John Forbes, PhD5; Joel Smith, PhD6; Jonathan Drever, BSc2, Joanna M. Wardlaw, MD2, Richard I. Lindley, MD7, Peter A. G. Sandercock, DM2, William N. Whiteley, MD2,on behalf of the IST-3 collaborative group
1Department of Internal Medicine, Oslo University Hospital, Oslo, Norway
2Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, Scotland
3Department of Neurology, University Hospital of North Norway, and Institute of Clinical Medicine, University of Tromsø, Tromsø, Norway
4Institute of Clinical Neuroscience, Karolinska Institute, Stockholm, Sweden
5Health Research Institute, University of Limerick, Limerick, Ireland
6Health Economics Research Centre, Nuffield Department of Population Health,University of Oxford, Oxford, England
7George Institute for Global Health, and Discipline of Medicine, University of Sydney, Sydney, Australia
Address for correspondence: Prof Eivind Berge, Department of Internal Medicine, Oslo University Hospital, Kirkeveien 166, NO-0407 Oslo, Norway.
Tel.: +47 22119100. E-mail: .
Abstract
Background and purpose: A secondary objective of the Third International Stroke Trial (IST-3) was to assess the effect of intravenous alteplase on long-term survival after ischaemic stroke. The effect of alteplase on survival has been the subject of debate.
Methods: IST-3 was an international multicentre, randomised-controlled, open-label trial of intravenous alteplase (0.9 mg/kg) plus standard care versus standard care alone within 6 hours of ischaemic stroke. We followed participants in the UK and Scandinavia for survival up to 36 months after randomisation using data from national registries, and compared survival in the two groups with proportional hazards survival analysis, adjusting for key prognostic variables.
Results: Of all 3,035participants in the trial,1948 came from UK or Scandinavia, of whom1946 were included in the analysis. By 36 months after randomisation, 453 of 967 alteplase participants (46.8%) and 494 of 979 controls (50.5%) had died (risk reduction 3.6%, 95% CI -0.8 to 8.1%).Participantsallocated to alteplase had a significantly higher hazard of death during the first 7 days (HR 1.46, 95% CI 1.06 - 2.01) and a significantly lower hazard in the period 8 days to 36 months (HR 0.78, 95% CI 0.68 - 0.90). The results did not differ significantly between those aged ≤80 versus >80 years, with NIHSS score ≤10 versus >10, or treated ≤3 versus >3 hours (all p≥0.3).
Conclusions: Alteplase within 6 hours was associated with a small, non-significant reduction in the risk of death at 36 months, but among those who survive the acute phase, treatment was associated with a significant increase inlong-term survival.
Introduction
Timely thrombolytic treatment with alteplase for acute ischaemic stroke improves functional outcome by 3-6 months(1;2), but has not been shown to improve survival, and questions have been raised about the long-term benefit of treatment(3-5). Alteplase leads to an increased risk of death during the first week, chiefly due to intracranial haemorrhage, but this increase is later offset by a risk reduction, so that, by 3-6 months, there is no excess of deaths among patients allocated alteplase(1;2).The National Institute of Neurological Disorders and Stroke (NINDS) trial(6) and the Third International Stroke Trial (IST-3)(7)followed participantsbeyond the first 3-6 months, but found no significant effect of alteplase on case fatality at 12 or 18 months(8-10). However, a difference mightarise after longer follow up,if improvements in functional outcome translate into survival benefits over time, as suggested by observational studies(11;12).To examine this hypothesis, and to provide further evidence on long-term effects of alteplase, we report survival up to 3 years among participants recruited into IST-3 in the United Kingdom (UK) and Scandinavia (Sweden and Norway), where long-term follow up was possible by linking to routinely collected data in national death registries.
Methods
IST-3 was a randomised-controlledtrial of alteplase (a recombinant tissue plasminogen activator)0.9 mg/kg plus standard care versus standard care alone within 6 hours of ischaemic stroke(7). Details of the protocol, the baseline characteristics and the main results have been published previously(7;13;14).The randomisation procedure was stratified by region(Sweden and Norway was one region, and UK participants constituted 91% of the North-West European region) and used a minimisation algorithm to balance treatment groups with respect to key prognostic variables(15). Continued follow-up of participants up to 3 years was planned in these countries because linkage to routinely collected death records was possible, and was approved by research ethics committees.
Follow-up in IST-3 was by clinical visit at 7 days (or at discharge, whichever occurred first) and by post or telephone interview at 6 and 18 months(7;9;10). In the UKand the Scandinavian countries participants were also followed for survivaluntil 36 months, by flagging in the central death registers in the UK and Norway, and data extraction from the national patient registry in Sweden.For the present analysis we also included participants from the UK who were recruited after June 2010 and thereforecould not be included in a previous18 months’ analysis of functional outcome and case fatality up till January 2012(9). Since registration of deaths is complete in the UK and Scandinavia we assumed that participantswere still alive if they had no death record up to 1096 days, and we did not take account of emigration, which we assumed to be small.For all deaths in the first seven days cause of death was assigned by the local investigator. For deaths after seven days a broad classification of cause of death was made based on the extent of information received from national registries or general practitioners. The hypothesis was that participants randomised to alteplase who survive the early risk of intracranial haemorrhagehave a better long-term survivalthan participants randomised to control.
We compared case fatality up to 36 months in the alteplase and control groups using Kaplan-Meier survival estimates, and examined whether the effect of alteplase was modified by age (≤80 versus > 80 years), stroke severity (NIHSSscore ≤10 versus >10), or time to randomisation (≤3 versus >3 hours) by determining (with likelihood ratio test) whether multiplicative interaction terms improved the fit of a logistic regression model with survival status at 36 months as the dependent variable. The thresholds for the definition of subgroups were based on the alteplase licence (age and NIHSS) or prior work(1). We also compared survival in the two groups with Cox proportional hazards regression analysis. We examined whether the proportional effect of alteplasewas constant over time,both graphically and with a formal test of a multiplicative interaction term between treatment and time(≤7 or >7 days) in a Cox regression model, which also included the main effect of treatment. We anticipated that hazards would be non-proportional for the whole time period, so we separately calculated hazard ratios in the early time period (≤7 days) and later (7 days). For survival beyond 7 dayswe censored follow-up at 1096 days after randomisation (36 months), and adjusted for the linear effects of age, stroke severity (NIHSS score) and time to randomisation. We also examined whetherthe effect of treatment in these time periods was modified by age(≤80 versus > 80 years), stroke severity (NIHSS score ≤10 versus >10), or time to randomisation (≤3 versus >3 hours), by determining whether interaction terms improved the fit of theCox proportional hazards model.In a post-hoc analysis, in order to understand why there were differences in long-term survival after 7 dayswhile differences in overall case fatality from day 0 were not apparent, we compared age, stroke severity and time to treatment at baselinebetween participants who died during the first 7 days and participants who survived the first 7 days, both in the alteplase and the control groups.Analyses were done with SAS (version 9.4).
Results
Of all 3035participants in the trial, 1948 participants(64%) were recruited in the UK and Scandinavia. Two participants were randomised after 6 hours and were excluded from the present analysis, and of the remaining 1946 participants, 1446 were recruited in the UK, 297 in Sweden and 203 in Norway. Participantsfrom the UK included 449 who were recruited after June 2010 and who were therefore not eligible for a previous 18 months’ analysis (Figure 1)(9). Baseline characteristics were well balanced between participants in the alteplase and control groups (Table 1).
At 7 days, 99 of 967 alteplase participants (10.2%) and 65 of 979 control participants (6.6%) had died (difference 3.6%, 95% CI 1.1 to 6.1%). During follow-up,the proportion of participantsdeadat 6, 18 and 36 months was non-significantly lower with alteplase by 1.4% (95% CI -2.6 to 5.4%), 2.6% (-1.7 to 6.9%), and 3.6% (-0.8 to 8.1%), respectively, and at 36 months 453 of 967 alteplase participants (46.8%) and 494 of 979 controls(50.5%) had died(Figure 2, Table 2).Subgroup analyses of differences in case fatality at 36 months did not showeffect modification by age(≤80 versus >80 years), stroke severity (NIHSS score ≤10 versus >10) or time to treatment(≤3 versus >3 hours, all interaction p-values ≥0.6, Table 2).
Hazards were non-proportional for the whole time period (p<0.001), so we separately calculated hazard ratios in the early time period (≤7 days) and later (>7 days). Within the later time periods (8 days to 18 months, and 18 to 36 months), there was no evidence that hazards were non-proportional (p=0.98). Compared to the hazard of death in control participants, thehazard with alteplase was significantly higher in the period 0 to 7 days (HR 1.46, 95% CI 1.06–2.01), and significantly lower in the period 8 daysto 36 months (HR 0.78, 95% CI 0.68–0.90,Figure 3). There were no effect modifications by age, stroke severity or time to treatmentusing the same subgroup classifications (all interaction p-values ≥0.3, Figure 3), or by using alternative subgroup classifications or modelling with continuous variables (data not shown).
The proportion of deaths up to 7 days due to intracranial haemorrhage or swelling of the original infarct was significantly greater in the alteplase group (p<0.0001). There was no significant difference in causes of death between the alteplase and control groups after 7 days (p=0.18, Table 3).
We checked whether differences in survival after 7 days could be explained by imbalances in important prognostic variables between treatment groups at 7 days (Table 4). As expected, participants who died during the first 7 dayshad more severe stroke at baseline than participants who survived the first 7 days, but this was similar in the alteplase and the control groups, so there was no evidence of an imbalance in age or stroke severity among survivors at 7 days.
Finally, since our analysis was performed in a subset of all participants in IST-3, we compared participants in our dataset with other participants in IST-3, and found that participants in our analysis were older, had more severe strokes and were treated earlier (Supplementary Table).We also analysed survival beyond 7 days in the datasets previously used for analysis of survival up to 18 months(9;10), and found hazard ratios that were broadly similar to those of the present dataset (Supplementary Figures 1 and 2).
Discussion
In this analysis of the two-thirds ofparticipantsin IST-3 scheduled for long-term follow-up, we found that alteplase was associated with an absolute reduction of 3.6% in the risk of death at 3 years, which was not statistically significant.We also found thatparticipantsallocated alteplase who survived to 7 days had a significantly better long-term survival up to 3 years.This means that, even if alteplase increases the risk of intracranial haemorrhage and death in the acute phase, patients can be assured that in the longer-term risk of death is not increased with alteplase, and that, if they survive to one week,alteplase increasesthe chance of survival.Many patients are willing to trade off an early increased hazard with a later improved chance of a good outcome(16). These result are reassuring both for patients considering the treatment and for those clinicians who have expressed concerns about the benefit of alteplase, based on previously published estimates of effect on survival(3;5). Although the effect on survival after the first 7 days is modest, it adds weight to the evidence for long-term benefits of alteplase on functional outcome and quality of life(9), and so reinforces the case for thrombolytic treatment.
The improved survival beyond the acute phasecan probably be explained by the effect on brain ischaemia of alteplase given within 6 hours of stroke onset. Alteplase improves functional outcome at 3-6 months(11;12), and it is plausible that these improvements in functional outcome can translate into survival benefits over time, as indicated by other studies(11;12). In line with this theory, the primary IST-3 analysis found that the early increase in deaths was offset by a later reduction, so that, by 6 months, the numbers of deaths were similar in the two groups(7). Our later analyses of survival up to 18 months did not show a significant overall survival benefit from alteplase(9;10), but analysis of survival beyond 7 days in these datasets shows hazard ratiosthat are broadly similar to those of the present analysis (Supplemental Figures 1 and 2).An alternative explanation for the increased survival after the acute phase might be that alteplase participants who survived the first week were younger or had less severe stroke than control participants (because older participants and those with severe stroke might have died from intracranial haemorrhage), but we found no evidence of any differences to support this hypothesis, and in any case our analysis of survival beyond 7 days wasadjusted for age and baseline stroke severity. We therefore find it reasonable to infer that the increased survival beyond the acute phase is an effect of alteplase that becomes evident during long-term follow-up.
We found no evidence that theproportional effect of alteplase on survival to 3 years was dependent on participantage, stroke severity or time to treatment up to 6 hours.For functional outcome at 3-6 monthsthere is strong evidence of better effects of alteplase in patients treated early(1;7). Our previous analysis of case fatality rates at 18 months in IST-3 also suggested a better effect of early treatment(10), butthis was not found in an analysis of early deaths(17), and not in the present analysis of case fatality rates at 36 months.Re-analysis ofsurvival beyond the acute phase also showed no subgroup differences in either of the 18 months’ datasets (Supplementary Figures 1 and 2), as in the present analysis of survival up to 36 months.An explanation for the lack of effect modification by time to treatment might be that our analysisof long-term survival includedfewer participants than the analyses of functional outcome at 3-6 months(1;7), which will have affected statistical power for subgroup analyses. Additionally, participantsin our analysis were older, had more severe strokes and were treated earlier than other participants in IST-3 (Supplementary Table), which may also have restricted the scope for effect modifications to be detected.An alternative explanation might be that differences between subgroups may become attenuated with time, so that differences that were present at 3-6 months can no longer be detected at 36 months.
The main strength of this analysis isthe randomisation of a large number of participants, with complete and blinded recording of deaths, and over a long period of time.The survival analysis also accounted forthe non-proportionality ofhazards and provided period-specific hazard ratiosand adjusted for key prognostic variables. In addition,the core IST-3 dataset (with 6 and 18 months’ data) was made available by 15th April 2016 through a controlled access process(18),and publication of the present dataset will in due course beconsidered by the IST-3 publication committee(19).
There are also some limitations. Firstly, we were not able to follow the entire IST-3 population up to 36 months, and therefore we have analysed a pre-specified subgroup.However, randomisation was stratified by region, and using a minimisation algorithm a good balance between treatment groups was achieved for the known importantprognostic factors. Secondly, the sample size was relatively small (by comparison to the large-scale trials of thrombolytic treatment for myocardial infarction), so the statistical power to detect moderate, but worthwhile differences in case fatality was limited, as was the power to detect subgroup interactions. Thirdly, participantsin this analysis were somewhat older, had more severe strokes and were randomised a little earlier than other participants in IST-3, but the result of the present analysis was not different from the result of the re-analysis of survival up to 18 months among all participants in the trial. Fourthly, although we did not find differences in age or stroke severity of 7-day survivors in the two groups, we cannot rule out the possibility of a selection bias leading to other, unmeasured differences in favour of the alteplase group. Finally, management of acute ischaemic stroke is changing rapidly, and these results will not be applicable for patients receiving complementary intra-arterial treatments(20). However, intravenous alteplase is likely to remain the primary treatment for patients without large artery occlusions and in other cases where intra-arterial treatment cannot be given, so the results will be applicable to many stroke patients world-wide for some time to come.
In summary, although alteplase given within 6 hours of ischaemic stroke was not associated with a clearreduction in the risk of death at three years, among participants surviving the acute phase there was a significantlong-term survival advantage. These results are reassuring for clinicians who have expressed concerns about the effect of alteplase on survival, and will make it easier for clinicians to inform patients and their relatives, and to discuss with them the early hazards and long-term benefits of treatment.
Research in context
We updated the literature search used for our Cochrane review(2). Among all trials of alteplase for ischaemic stroke, only theNINDS trial(6) and IST-3(7) followed participants longer than 3-6 months. The NINDS trial included 624participants and followed to 12 months(8), while IST-3 included 3035 participants and followed to 18 months(9;10), with an additional follow-up to 36 months of 1948 participants from the UK and Scandinavia. The analysis of survival to 36 months showed a small, non-significant reduction in the risk of death by end of follow-up, but among those who survived the acute phase, there was a significant effect on survival up to 36 months. This may be reassuring for clinicians who have been unsure about the benefits of alteplase, and for patients who have an option to trade an earlyincreased hazard with later and long-term benefits. The effect on survival adds to the long-term beneficial effects of alteplase on functional outcome and quality of life, and reinforces the case for alteplase for ischaemic stroke