Intended for healthcare professionals

Rapid response to:


Thrombolysis in acute ischaemic stroke: time for a rethink?

BMJ 2015; 350 doi: (Published 17 March 2015) Cite this as: BMJ 2015;350:h1075

Rapid Response:

Thrombolysis in acute ischaemic stroke: Author response to many responses

We have had multiple responses to our article "Thrombolysis in acute ischaemic stroke: time for a rethink?" although the responses have been scattered across the BMJ rapid response page, news media, blogs, Twitter, and personal email.

American Heart Association guideline authors or representatives provided clarifications in a blog[1] and a news article.[2] They note the data is limited at 3-4.5 hours after stroke, emphasize they rated the evidence as Level B, and agree with the call to make all trial data available for transparent and independent analysis.

They report that alteplase neither reduced nor increased all-cause mortality and implied the focus on fatal intracranial hemorrhage in the BMJ article was missing the broader view. However the data for all-cause mortality was reported in the BMJ article, and a summary statistic with a hazard ratio 1.14 (95% CI 0.95-1.36), though not statistically significant, is consistent with a 2% absolute increase in all-cause mortality. Absence of statistical significance is not equivalent to evidence of absence of an effect. The 2% absolute difference at 90 days reflecting the same absolute difference at 7 days from fatal bleeding suggests this is the same mortality risk carried forward with little effect (increase or decrease) on other causes of mortality.

The AHA emphasizes their guideline recommendations for use of tPA 3-4.5 hours after stroke are specifically limited to the population included in the ECASS III trial (i.e., patients < 80 years old without extremely severe neurologic deficits and without extensive brain injury on imaging), and suggests use of “all the data” should not be applied to interpretations for this limited subgroup. We agree and would prefer a full data analysis (from all the patients meeting these criteria across all the trials) for the subgroup for which a guideline recommendation is being made before making a strong recommendation extrapolating from some but not all of the data.

The IST-3 lead investigator responded in the Medscape news report noting that “it isn’t the case that thrombolytics work up to 3 hours but not at 1 minute past 3 hours”. We agree and the 3 hours is the time cutoff using available data. Perhaps thrombolytics “work” when given before brain injury occurs by preventing the damage and are “more harmful than beneficial” when given after the damage occurs because damaged tissue is more likely to bleed and there is less opportunity to prevent the damage. Perhaps future research will support real-time imaging at the moment of t-PA administration. But until we have such technology demonstrated, use of a 3-hour cutoff is the easiest approximation from the available data.

Dr. Janousek commented on the rapid development of technology and how enhanced imaging technology may support safer patient selection for t-PA, especially after 3 hours. Evaluation of multiple guidelines find that guidelines consistently list contraindications which are associated with increased risk for bleeding, but what specific factors are considered contraindications vary considerably across guidelines. Determination of imaging findings, stroke severity, and other factors that should be considered contraindications is another area where clarity could be improved.

Cochrane leadership supported our conclusions that researchers should have access to the complete data sets to replicate analyses to inform decision-making, and invited us to submit feedback directly on the Cochrane Library website. We have submitted feedback to inform improvements to the Cochrane review using the currently available data and support plans for using the process of updating the Cochrane review to obtain and analyse the complete data set.

The Cochrane review authors have provided a rapid response noting the use of data from all thrombolytic therapy trials (over 10,000 patients) and alteplase trials (about 7,000 patients). Distinguishing between conclusions that apply to thrombolytics in general and alteplase in particular could be a source of confusion. Either way, we continue to question the validity of “up to six hours” or “up to 4.5 hours” conclusions because:
1) The plausible explanation that time from symptom onset to time of administration has such an influence on efficacy and safety that patients at the extremes of these time windows are so dissimilar that an “average” is not a valid reflection of the whole.
2) The apparent difference in the data analyses when grouped by 0-3 hours and 3-6 hours in the Cochrane review, making this concept (previously apparent through individual trial analysis) even more striking. The Cochrane review here is very helpful to show this distinction.

The Cochrane review authors’ response also noted provision of “a detailed critique of individual trials’ risks of bias” but we are not finding this detailed critique in the portion of the review titled “Risk of bias in included studies”. For example the paragraphs on Incomplete outcome data and Selective reporting report more on the overall approach applied by the review authors and not the risk of bias in the individual trials. With a systematic approach to reporting the risk of bias for each trial, more insights may occur. Of greatest interest, the one and only trial directly supporting efficacy of alteplase at 3-4.5 hours (ECASS III) has a known risk of bias in baseline differences, and this is not included in the Cochrane review. The general statement in the Cochrane review “Trials published more recently utilised computerised randomisation, so there are less likely to be baseline imbalances than in previous versions of the review.” does not address whether the baseline imbalances are sufficient to change the results of the analyses. How much does this specific bias influence the overall meta-analysis results? In the Characteristics of included studies section, the IST3 trial is the only trial for which a full listing of Risk of bias assessment is provided.

The consistency or inconsistency of the trial results was not discussed in the Discussion/quality of evidence section of the Cochrane review. Along with consideration of the methodological limitations which may influence the results, the impact of individual trials and consistency determine the validity of the conclusions.

Couple the concern of the only trial directly reporting benefit at 3-4.5 hours being limited by a risk of bias that may be the reason for finding statistically significant benefit with the concern for overcombining data from many time periods and you can see how results for the use of alteplase (or thrombolytics in general) at 3-4.5 hours can be substantially different than implied through broad statements of efficacy up to 4.5 hours.

This is a complicated area with continued input from ongoing research, analysis of available research, and feedback from many stakeholders. We have created a synthesized view to all of these issues and will continually revise it as new information is published and additional feedback is received.[3] You can find this synthesized view made available for open access at

1. American Heart Association News. New opinion paper questions use of clot-busting drug in late time window. March 19, 2015.
2. Hughes S. Thrombolysis After 3 Hours in Stroke Needs Review? March 20, 2015. Medscape Medical News
3. Thrombolytics for acute stroke. In DynaMed [database online]. EBSCO Information Services. Updated March 26, 2015. Accessed March 29, 2015.

Competing interests: Authors of the original BMJ article; BSA, MMM, and JSM work for EBSCO Health which publishes DynaMed; no financial competing interests with any stroke-related products

29 March 2015
Brian S Alper
VP of EBM R&D, Quality and Standards
Meghan Malone-Moses, James S McLellan, Kameshwar Prasad, Eric Manheimer
EBSCO Health
10 Estes St, Ipswich, MA 01938 USA