Systematic review of evidence on thrombolytic therapy for acute ischaemic stroke Authors Wardlaw J, Warlow C, Counsell C. Source Lancet. 350:607-14. August 30, 1997. Institution Department of clinical neurosciences, University of Edinburgh, UK. Support UK Stroke Association; UK Medical Research Council; Wellcome Trust. Background Prior to 1992, six randomized controlled trials evaluating the use of thrombolytic therapy for ischemic stroke had been carried out; since 1992 another 8 RCT's were published. Based on the results of one trial (the NINDS study, reviewed here), the US FDA has recommended the licensing of tPA for the treatment of ischemic stroke, if given within 3 hours of symptom onset. The purpose of this systematic review was to present all of the data available from randomized controlled trials of thrombolytic therapy for ischemic stroke. Methods Selection of trials Trials were identified by multiple methods (this article is a summary of a review originally prepared for the Cochrane collaboration). They were included if: They were randomized. They were unconfounded (treatment and control groups differed only by thrombolytic therapy). They were or could be analyzed by intention to treat methodology. Strokes were confirmed by CT or MR scanning. Data extracted Data extracted from the trials included: All-cause mortality (early and at the end of follow-up). Dependence on others for activities of daily living, at the end of follow-up. Death or dependence at final follow-up. Symptomatic or fatal intracranial hemorrhage during the acute stage. Analysis Odds ratios were calculated for unfavorable outcomes in the the treatment groups vs. control groups. Heterogeneity between trials was calculated, and a univariate analysis of various factors (thrombolytic agent, time to treatment, length of follow-up) that might contribute to heterogeneity was performed. Results Trials analyzed There were 12 trials that met the inclusion criteria and that were analyzed. A further 9 trials were excluded for various reasons (no CT scans, not enough data, not published yet). Five trials have not been completed yet. The trials included several recent, well-publicized studies: NINDS (United States trial, tPA vs. placebo, 624 patients) MAST-I and MAST-E (European studies, 623 patients, streptokinase vs. placebo; also 309 patients streptokinase + aspirin vs. aspirin) ASK (Australian trial of streptokinase vs. placebo, 340 patients) ECASS (European trial of tPA vs. placebo, 620 patients) A total of 3345 patients were randomized, 1777 to thrombolytic therapy, 1658 to placebo. There were significant differences between the 12 trials. These included: Lytic agent: Urokinase in 3 trials (from Japan, prior to 1992), streptokinase in 4, tPA in 5. Dosing was variable (half the trials used doses roughly comparable to those used in the treatment of MI). Time to treatment: Within 6 hours in 6 trials, within 4 or 3 hours in 3 trials. The 3 Japanese urokinase trials allowed treatment within 5 days or 2 weeks. Antithrombotic drugs: Forbidden during the first 10 days in 6 trials and during the first 24 hours in 2 trials. In the remaining 4 trials they were given during the first 24 hours either at the discretion of the attending physician or by protocol. Mortality Total mortality at end of follow-up was increased in the thrombolytic group, 22% vs. 18.3%, for an absolute increase in risk of 3.7%. The OR was 1.36 (95% CI: 1.14-1.62). Looking at the odds ratio for mortality by thrombolytic agent used (ignoring the urokinase trials): Streptokinase vs. control (MAST-I, MAST-E, ASK and another very small trial): 1.43 (1.10-1.88) Streptokinase + aspirin vs. aspirin (MAST-I): 3.02 (1.87-4.87) tPA vs. control (ECASS, NINDS and one other small trial): 1.06 (0.80-1.39) NS Early mortality was reported in all the major trials except NINDS (where the early mortality was not included in the report and apparently could not be obtained by the authors of this analysis). There was a highly significant absolute increase of 9.1% in early mortality in the treatment group (20.9% vs. 11.8%). Functional outcome Functional outcome assessed by Rankin or Barthel scales was available for analysis in 7 trials (2567 patients), with final follow-up at 1, 3 or 6 months depending on the trial. Poor functional outcome was assessed as "death or dependency". Overall, the number of patients dead or dependent at final follow-up was 61.5% in the thrombolytic groups, vs. 68% in the control groups, for an absolute risk reduction of 6.5%. The odds ratio was 0.75 (95% CI: 0.63-0.88). Looking at these results by thrombolytic agent used: Streptokinase vs. control: 0.94 (0.72-1.24) i.e. NS Streptokinase + aspirin vs. aspirin: 1.09 (0.69-1.73) NS tPA vs. control: 0.57 (0.45-0.72) Intracranial hemorrhage The risk for symptomatic and fatal intracranial hemorrhage was increased in the treatment group, in all recent trials. That risk was 9.6% symptomatic and 6.2% fatal in the thrombolysis group, vs. 2.6% and 1.1% in the control groups. The excess risk was thus 7% for symptomatic hemorrhage and 5.1% for fatal hemorrhage. There were no major differences between tPA and streptokinase. Heterogeneity between trials Except for the outcome of intracranial hemorrhage, there was significant heterogeneity in the results among the trials. The best way to assess the causes of this heterogeneity would be by analyzing individual patient data from the original trials, which the authors were not able to do. They did undertake exploratory (univariate) analysis of a number of factors which might have been responsible for the differences, concentrating on the outcome of total mortality (which exhibited the greatest heterogeneity). Time to randomization: Differences in time to randomization accounted for some, but not all of the heterogeneity between trials. Limiting the analysis to patients randomized within 3 hours eliminated the overall excess in mortality in the treatment group, but significant heterogeneity persisted. On the other hand, limiting analysis to these patients eliminated the heterogeneity in the outcome "death or dependency". Thrombolytic agent and dose: Contrary to what a superficial reading of the results would lead one to believe, the choice of thrombolytic agent alone was not enough to explain the differences between trials. The difference between streptokinase (OR 1.43) and tPA (1.06) was not statistically significant in the analysis performed here. Use of antiplatelet agents: Some of the heterogeneity may have been due to differing use of antiplatelet agents in the trials. These agents appeared to confer a worse prognosis when added to lytic therapy. Baseline risk: There was little evidence that differing baseline risks contributed significantly to heterogeneity. Author's discussion The authors draw a number of conclusions from their results. They feel that thrombolytic therapy is potentially beneficial, but that the exact patient groups and treatment strategies that will provide the most benefit remain unclear. Early treatment appears to convey benefit. They feel that the superiority of tPA over streptokinase has not been conclusively demonstrated, since multiple other differences between trials may account for the better outcomes that were seen with tPA. Future randomized trials should clarify the host of questions that remain unanswered: drug to use, dose of drugs, how to deal with patients already on aspirin, time limit to thrombolysis, other criteria for risk stratification. They conclude: "We need to define and quantify the risks and benefits of thrombolysis much more clearly before it finds its place in the treatment of acute ischemic stroke".

Comment The concept of trial heterogeneity is interesting and its application is well illustrated by this review. When a number of trials studying the same question are performed, they are unlikely to have the exact same outcomes, simply on the basis of sampling error. Just as a small difference between two treatment groups may be due to chance alone, a difference between two trials may also be due to chance. What we need to know is whether discrepancies between trials are greater than we would expect by chance. A formal analysis of heterogeneity, as was performed in this review, uses statistical tests (such as Chi-square) to determine whether the results of various trials are more different than one would expect by chance alone. If so, further exploratory analyses can help elucidate the specific factors that are at the root of these differences. Such formal analysis of heterogeneity is one difference between a systematic review such as this one and a more traditional literature review. In the more traditional review, the reviewer will look at a number of trials and try to explain perceived differences between them. This is by nature a very subjective exercise. Simply inspecting the data from the trials presented here, one would be tempted to concentrate on the differences between trials that showed benefit and those that did not. The difference between these groups seems to be related mainly to the use of streptokinase or tPA. Formal analysis, on the other hand, forces one to consider not only differences between trials that show an effect and those that don't, but also to examine trials that show a markedly different degree of effect. It also reveals that the choice of thrombolytic agent may not be the primary determinant of outcome. The review presented here concludes that there is insufficient evidence to state that streptokinase is worse than tPA. This may well be the case, but three trials of streptokinase in acute stroke were halted before completion because of safety concerns. If the authors had obtained individual patient data and had been able to show that the entire benefit of tPA over streptokinase was accounted for by other factors (time to reperfusion, use of antiplatelet agents, baseline risk), it might be justifiable to include streptokinase as a treatment arm in future trials. Until such data are obtained, however, it seems unlikely that streptokinase will be tested on a large scale for acute ischemic stroke. September 30, 1997 References References related to this article from the NLM's PubMed database.

Reader Comments Date: Mon, 13 Oct 97 From: Masatoshi Matsumoto <matmo10@jb3.so-net.or.jp> Hello! I am a Japanese resident of family practice. I think the heterogeneity of RCTs in this review implies no conclusion about the use of thrombolytic agents for ischemic stroke. A good RCT which is not biased and has a large number of samples should be done in the near future. Submit a comment

Journal Club homepage