All-cause mortality at the end of the follow-up by medication. Error bars indicate 95% confidence intervals. IV indicates intravenous; rt-PA, recombinant tissue plasminogen activator; NINDS, National Institute of Neurological Disorders and Stroke rt-PA Stroke Study; ECASS, European Cooperative Acute Stroke Study; ECASS II, Second European-Australasian Acute Stroke Study; ATLANTIS A, Alteplase Thrombolysis for Acute Noninterventional Therapy in Ischemic Stroke Trial, Part A; ATLANTIS B, ATLANTIS, Part B; IA, intra-arterial; PROACT, Prolyse in Acute Cerebral Thromboembolism Study; and STAT, Stroke Treatment With Ancrod Trial.
Ergin A, Ergin N. Is Thrombolytic Therapy Associated With Increased Mortality?Meta-analysis of Randomized Controlled Trials. Arch Neurol. 2005;62(3):362-366. doi:10.1001/archneur.62.3.362
Although thrombolytic therapy has shown substantial benefits in neural outcomes, concerns remain regarding the association between thrombolytic therapy and possible increased mortality.
To determine the mortality risk of certain thrombolytic agents that are a treatment option for acute ischemic stroke.
Data Sources, Extraction, and Synthesis
Studies were identified using MEDLINE, the Cochrane Central Register of Controlled Trials, and the reference lists of the articles selected. Randomized placebo-controlled trials of thrombolytic agents for the treatment of acute ischemic stroke patients were eligible. Study quality was evaluated using a previously validated scale. Data were extracted in duplicate by two independent investigators. All-cause mortality during follow-up was the main outcome. Random effects models were used to pool the individual effects of trials. Several preplanned sensitivity and subgroup analyses were completed to explain the heterogeneity among trials. Odds ratios, absolute risk differences, and numbers needed to harm were calculated.
Eleven placebo-controlled randomized trials of thrombolytic agents involving 3709 participants were included in the analysis. Thrombolytic therapy was associated with an insignificant increase in mortality (odds ratio, 1.07; 95% confidence interval, 0.8-1.39; P = .3). The treatment was associated with an absolute increased risk of mortality of 11 per 1000 persons (95% confidence interval, −24 to 48; P = .3), and the number needed to harm was 84 (the 95% confidence interval included 0).
These findings suggest that thrombolytic therapy does not significantly increase all-cause mortality.
Since the National Institute of Neurological Disorders and Stroke Recombinant Tissue Plasminogen Activator Stroke Study demonstrated substantial benefits in neurological outcomes at 3 and 12 months,1,2 thrombolytic therapy has become a subject of great interest in stroke medicine. However, neurologists’ enthusiasm for the efficacy of thrombolytic agents is tempered by their concern about a rate of early symptomatic intracranial hemorrhage that is 3.5 times higher with this treatment and about possible increased overall mortality.3,4 Sample sizes of individual thrombolytic therapy trials have been insufficiently large to evaluate the rare mortality incidents associated with the treatment. With its increased power, meta-analysis has a greater ability than individual trials to evaluate the risk of mortality. A previous meta-analysis found significantly increased death from all causes with thrombolysis.4 However, this meta-analysis pooled data from streptokinase trials. All major streptokinase trials were terminated early because of an unacceptable rate of major adverse effects as well as the inefficacy of streptokinase in acute stroke therapy.5- 7 Furthermore, a meta-analysis of individual patient data from strepokinase trials did not indicate any beneficial effect of streptokinase in any subgroup.8 As a result, future use of streptokinase in acute ischemic stroke has become very unlikely.
The objectives of this study were to determine the all-cause mortality risk of certain thrombolytic agents for acute ischemic stroke that either are currently available or may become available in the future and to assist clinicians by providing a review focused on new information from recently published and ongoing studies.
A database search of the English-language literature was conducted using MEDLINE and the Cochrane Central Register of Controlled Trials (1966–March 2003). A thorough manual review of reference lists from original research and review articles was performed. Abstract collections from pertinent conferences were sought as well. In addition, the National Institutes of Health Web sites, the National Stroke Association, and individual experts in the field of cerebrovascular diseases served as valuable resources.
Published manuscripts, abstracts, and interim reports were eligible for inclusion in this meta-analysis. Randomized controlled trials were included if they were pertinent to the research question, conducted with human subjects, and informative about all-cause mortality during follow-up. The population of interest was patients treated with thrombolytic therapy. Studies of thrombolytic agents vs another agent and studies with historical controls were excluded.
Fourteen potentially eligible studies were identified.1,9- 21 Of these, 3 were excluded from analysis. The Emergency Management of Stroke Bridging Trial19 was not eligible because it compared intravenous and local intra-arterial recombinant tissue plasminogen activator (rt-PA) therapy with intravenous-only rt-PA therapy. The Standard Treatment With Alteplase to Reverse Stroke study20 and the Canadian Activase for Stroke Effectiveness Study21 were not eligible because they did not have controls. Ancrod trials16,17 were included in the analysis because ancrod promotes the release of endogenous tissue plasminogen activator.22,23 Although thrombolysis is only one of several presumed mechanisms of ancrod’s effect, a sensitivity analysis excluding the ancrod trials from this meta-analysis showed the robustness of the pooled estimate. Therefore, we decided to keep the ancrod trials in our pooled estimate and reported a separate pooled estimate without the ancrod trials. The commercial availability of ancrod in Europe and Canada since the 1970s was a contributing factor in the inclusion of the ancrod trials in this meta-analysis.
Once studies were determined to be eligible for the meta-analysis, two investigators independently assessed study quality using a previously validated scale.24 The scale mainly evaluates trials regarding their blinding, randomization, and statistical analysis. A standard form was used for data extraction for demographic and clinical characteristics of study participants, ancillary medication, dosing regimens, and nature and definition of outcome assessments. Published reports were the primary source of the data for this meta-analysis, and there was no attempt to obtain data directly from the primary investigators of the eligible studies.
Both odds ratio (OR) and absolute risk difference were used to measure the relationship between thrombolytic therapy and all-cause mortality. Mortality rates for the treatment and placebo groups were recorded for each study using 2 × 2 tables. The ORs and absolute risk differences and their 95% confidence intervals (CIs) were calculated individually for each study. Numbers needed to harm, the number of patients that would have to be treated to cause one additional adverse outcome, were also calculated if a significant treatment effect was observed (ie, if the 95% CI of the OR did not include 1).25
The ORs from individual studies were transformed to their natural logarithm (ln[OR]) values before pooling. Pooled ln(OR) values were then calculated by assigning each a weight—the reciprocal of the variance of each study’s ln(OR).26 Homogeneity of the ln(OR) values across studies was tested using the DerSimonian and Laird Q test statistic, which sums the squared deviation of each individual ln(OR) from the pooled ln(OR) weighted by the reciprocal of its estimated variance.27 Random effects models were used to pool the individual effects of trials. Effect measures were back-transformed and re-expressed as ORs. The pooled absolute risk difference was calculated by assigning each a weight—the reciprocal of the variance of each study’s absolute risk difference. Two-tailed tests of statistical significance were used, with an α level of 0.05. Sensitivity analyses were performed to determine whether the pooled estimates were robust for different study and participant characteristics. Limited numbers of subgroup analyses were completed to explain the heterogeneity among the results of individual trials. The potential for publication bias was examined using a funnel plot, with sample size was plotted against log(OR).
Data from 11 trials with 3709 participants were analyzed. Participant and study design characteristics for these trials are presented in Table 1. With the exception of 3 studies,9,10,15 all had relatively large sample sizes. In 7 of the trials rt-PA was the thrombolytic agent used.1,9- 14 Placebo was used as control in most of the trials, except the prourokinase trials,15,16 which used heparin as control. All but 2 of the prourokinase trials15,16 used an intravenous route for drug administration. Mean follow-up time was 5 months (range, 1-12 months).
Thrombolytic therapy was associated with an insignificant increase in mortality by the end of follow-up (OR, 1.07; 95% CI, 0.8-1.39; P = .3) (Table 2). The treatment was associated with an absolute increased risk of mortality of 11 per 1000 persons (95% CI, −24 to 48; P = .3), and the number needed to harm was 84 (the 95% CI included 0).
Thrombolytic therapy was associated with an insignificant increase in mortality during the first month of therapy (OR, 1.14; 95% CI, 0.75-1.73; P = .6) (Table 3). Additionally, treatment was associated with an insignificant increase in mortality when the ancrod trials were excluded (OR, 1.13; 95% CI, 0.77-1.68; P = .2). In contrast, for rt-PA trials that had treatment within 3 hours, treatment was associated with an insignificant decrease in mortality (OR, 0.98; 95% CI, 0.63-1.53; P = .8). When only trials that used rt-PA were included, treatment was associated with an insignificant increase in mortality (OR, 1.25; 95% CI, 0.87-1.78; P = .1) (Figure).
A plot of sample size vs OR showed no obvious pattern, suggesting no publication bias (data not shown).
This meta-analysis indicates that thrombolytic therapy does not significantly increase all-cause mortality at the end of the follow-up. These findings are consistent with the findings of several phase 4 trials in the United States and Germany. The Standard Treatment With Alteplase to Reverse Stroke Study,20 which included 389 patients treated with intravenous rt-PA, had a 30-day mortality rate of 13%. In another study, 100 consecutive patients were treated with intravenous rt-PA, and the 90-day all-cause mortality rate was 10%.28 These results are comparable with the results of the control arm of the National Institute of Neurological Disorders and Stroke rt-PA Stroke Study.
In addition to failing to demonstrate any significant excess mortality among treated stroke patients, we did not show any significant mortality risk associated with thrombolytic therapy in any subgroup (Table 3). However, the CIs of the estimates became bigger, indicating that more data may be needed to reach a definitive conclusion. For example, there was an insignificant increase in mortality when only rt-PA studies were included in the meta-analysis. However, we found an insignificant decrease in mortality for trials with thrombolytic therapy within 3 hours of acute stroke.
The Third International Stroke Trial29 started patient recruitment in May 2000. This trial aims primarily to determine whether a wider range of patients might benefit from rt-PA therapy, and it targets 6000 patients worldwide. There is a strong possibility that results from this big trial will shed light on the controversy about thrombolytic therapy and mortality in acute ischemic stroke. Additionally, the European Stroke Treatment With Ancrod Trial30 results are still pending. In this trial of 1222 patients, ancrod was compared with placebo.
One major limitation of this meta-analysis is that we could not evaluate the effect of thrombolysis on cause-specific mortality, including mortality from intracranial hemorrhage. The reasons for this shortcoming were as follows:
It was difficult to determine the true causes of deaths in some trials because some patients died without a computed tomographic scan or postmortem examination. Thus, the true rate of mortality from intracranial hemorrhage may be different than that suggested by the data.
On the other hand, awareness among researchers of the association between early intracranial hemorrhage and thrombolysis may indicate an inflated number of deaths attributed to intracranial hemorrhage. For instance, review of published computed tomographic scans suggested that, at least in some trials, symptomatic intracranial hemorrhage included patients with very large swollen infarcts with trivial amounts of hemorrhage within them.4 Therefore, overestimation of the risk of death from intracranial hemorrhage is a strong possibility.
Finally, a number of the studies did not report detailed numbers for causes of mortality but did mention the most frequent cause of mortality at the end of the follow-up. A meta-analysis of individual patient data reevaluated by a group blinded to patient allocation may be a solution for this shortcoming.
There is always a possibility of publication bias in a meta-analysis. However, we conducted a comprehensive search with a systematic strategy to avoid bias. We also attempted to find unpublished trials, corresponding with experts in the field and searching abstracts from recent conferences. We believe we have identified most of the available research dealing with this issue. Another possible limitation of this study may be the question of whether the ancrod and/or prourokinase data should be included in the meta-analysis given the differences in the mechanism of action (ancrod) and the route of delivery and heparin use in controls (prourokinase). To remove the effect of this limitation we calculated separate pooled estimates for different medication groups. This approach did not change the interpretation of our pooled estimate; however, it reduced the possibility of publication bias.
In conclusion, available data do not indicate significant excess mortality among patients treated with thrombolytic agents. This meta-analysis provides further evidence of a clear need for additional randomized trials of thrombolytic therapy. In combination with the results from a meta-analysis that used individual patient data from intravenous rt-PA studies and that showed that these patients had a higher probability than nontreated patients of recovering with little or no deficit,31 our results may imply that thrombolytic treatment should be seriously considered for acute ischemic stroke patients who have the right indications.
Correspondence: Ahmet Ergin MD, PhD, MPH, Department of Public Health, Pamukkale University School of Medicine, Bursa Cad. No. 119, Kinikli, Denzili 20020, Turkey (firstname.lastname@example.org).
Accepted for Publication: March 23, 2004.
Author Contributions:Study concept and design: A. Ergin and N. Ergin. Acquisition of data: A. Ergin and N. Ergin. Analysis and interpretation of data: A. Ergin and N. Ergin. Drafting of the manuscript: A. Ergin and N. Ergin. Critical revision of the manuscript for important intellectual content: A. Ergin and N. Ergin. Statistical analysis: A. Ergin. Study supervision: N. Ergin.