V. Murray, Department of Clinical Sciences, Karolinska Institutet Danderyd Hospital, SE-182 88 Stockholm, Sweden. (fax: +46 8 622 6810; e-mail: email@example.com).
Murray V, Norrving B, Sandercock PAG, Terént A, Wardlaw JM, Wester P (Karolinska Institutet Danderyd Hospital, Stockholm; Department of Neurology, Lund, Sweden; Division of Clinical Neurosciences, Edinburgh, UK; Acute and Internal Medicine, Uppsala; Umeå Stroke Centre, Umeå; Sweden). The molecular basis of thrombolysis and its clinical application in stroke (Review). J Intern Med 2010; 267: 191–208.
Abstract. The rationale for thrombolysis, the most promising pharmacological approach in acute ischaemic stroke, is centred on the principal cause of most ischaemic strokes: the thrombus that occludes the cerebral artery, and renders part of the brain ischaemic. The occluding thrombus is bound together within fibrin. Fibrinolysis acts by activation of plasminogen to plasmin; plasmin splits fibrinogen and fibrin and lyses the clot, which then allows reperfusion of the ischaemic brain. Thrombolytic agents include streptokinase (SK) and recombinant tissue-type plasminogen activator (rt-PA) amongst others under test or development. SK is nonfibrin-specific, has a longer half-life than tissue-type plasminogen activator (t-PA), prevents re-occlusion and is degraded enzymatically in the circulation. rt-PA is more fibrin-specific and clot-dissolving, and is metabolized during the first passage in the liver. In animal models of ischaemic stroke, the effects of rt-PA are remarkably consistent with the effects seen in human clinical trials. For clinical application, some outcome data from the Cochrane Database of Systematic Reviews which includes all randomized evidence available on thrombolysis in man were used. Trials included tested urokinase, SK, rt-PA, pro-urokinase, or desmoteplase. The chief immediate hazard of thrombolytic therapy is fatal intracranial bleeding. However, despite the risk, the human trial data suggest the immediate hazards and the apparent substantial scope for net benefit of thrombolytic therapy given up to 6 h of acute ischaemic stroke. So far the fibrin-specific rt-PA is the only agent to be approved for use in stroke. This may be due to its short half-life and its absence of any specific amount of circulating fibrinogen degradation products, thereby leaving platelet function intact. The short half-life does not leave rt-PA without danger for haemorrhage after the infusion. Due to its fibrin-specificity, it can persist within a fibrin-rich clot for one or more days. The molecular mechanisms with regards to fibrin-specificity in thrombolytic agents should, if further studied, be addressed in within-trial comparisons. rt-PA has antigenic properties and although their long-term clinical relevance is unclear there should be surveillance for allergic reactions in relation to treatment. Although rt-PA is approved for use in selected patients, there is scope for benefit in a much wider variety of patients. A number of trials are underway to assess which additional patients – beyond the age and time limits of the current approval – might benefit, and how best to identify them.
Against a background with a summary of the consequences of stroke for the patient and society this review will describe: the formation of the thrombus; the mechanism of action of fibrinolysis; the licensing process and uptake of fibrinolysis in clinical practice; the results of a comparison between fibrinolysis data in animals and in humans (review); and some major clinical outcomes from meta-analyses in the Cochrane Database of Systematic Reviews  in relation to some molecular mechanisms.
Stroke, is one of the most common causes of death and serious disability . Stroke is the third most common cause of death after all cardiac and cancer disorders combined [3, 4]. Death from cerebrovascular disorders is declining , but in some parts of the world it is still increasing, notably in the Eastern European Countries . Stroke is the most common cause of long-term physical disablement. At 6 months, 15–20% is dead and amongst the survivors up to 60% is more or less dependent on others in their daily activities if verbal and so called invisible disabilities like intellectual impairment are included [7–9]. The risk of stroke increases exponentially with age, and – with longevity increasing in most developed economies – the proportion of patients with acute stroke who are very elderly (aged over 80 years) is set to increase substantially over the next 20 years [3, 7]. More trials in the elderly are urgently needed . Stroke is the most common physical cause of need of long-term hospital care and thus places an increasing burden on already over-stretched health care systems. Costs incurred by stroke depend to a large part on patient’s age, the younger the more expensive ; stroke severity; and grade of disablement . Effective treatment of stroke thus carries as well humanitarian as economic benefits. Fibrinolysis is the most promising pharmacological treatment to restore or improve functional capacity in ischaemic stroke [1, 7].
As ischaemic stroke is caused by occlusion of a cerebral artery by thrombus or emboli, the most logical therapeutic approach is to unblock the occluded artery as soon as possible either by intravenous (i.v.) or local intra-arterial (i.a.) pharmacological treatment or by mechanical clot removal with/or without intra-arterial drug treatment.
For many years treatment in a stroke unit (SU) and aspirin given within 48 h of stroke onset have constituted the only evidence-based acute therapeutic options. Aspirin within the first 24–48 h yields a modest but significant reduction of the combined outcome measurement, death or dependency, with 1% maybe via its secondary preventive effect . SU-care regards all patients with suspected stroke and of all ages . The positive effects of treatment in an SU are long-lasting and can be seen at follow up after 10 years . SU-care reduces death or dependency, with approximately 5–6%, and significantly increases quality of life (QoL) . An adequately organized SU seems to constitute the best environment and basis for all treatment of suspected stroke including for i.v. fibrinolytic treatment, and also for supervision after most invasive approaches, but further trial data are needed regarding the environment in which thrombolysis may best be given in routine practice.
A thrombus is the final product of the blood-coagulation step in haemostasis. Thrombus generation is a highly complex process with many components involved, both structural and regulatory [17, 18]. As a response to endothelial damage, collagen and tissue factor become exposed to the blood, thereby initiating thrombus formation via separate pathways. In the tissue factor pathway, tissue factor in presence with protein disulphide isomerase generate fibrin. Tissue factor generates thrombin by means of the blood-coagulation pathways. Platelets are captured on the vessel wall, and platelet–platelet interaction (via synapses) and platelet activation by thrombin cleavage of protease-activated receptor (Par4) follow. In the collagen pathway, on disruption of the endothelium, collagen is exposed, rapidly leading to platelet deposition resulting in merging of the collagen and the platelets. Platelets are captured on the vessel wall, and platelet–platelet interaction and platelet activation follow. Thrombin is not required for platelet activation in this pathway. In the common pathway, platelet activation is controlled by calcium mobilization. Unactivated platelets become associated with the developing thrombus. Those that are activated are detected by increases in calcium mobilization.
Exposed collagen triggers the accumulation and activation of platelets whilst exposed tissue factor initiates thrombin generation. Thrombin both converts fibrinogen to fibrin and also activates platelets (Fig. 1).
Platelet activation and platelet thrombus formation are intertwined with thrombin generation and fibrin clot propagation. Platelet accumulation is dependent on multiple proteins, including von Willenbrand Factor (vWF) and fibrinogen. Platelet-platelet form synapses via an interaction between glycoprotein IIb/IIIa and fibrinogen. Blood-borne tissue factor is delivered to the developing thrombus in a process dependent on P-selectin and PSGL-1 and the endothelial injury activates the tissue factor pathway to thrombin generation. Intracellular calcium mobilization is necessary for stable platelet interaction with the thrombus.
Mechanism of action of fibrinolysis
The mechanism of action of fibrinolysis has been extensively reviewed elsewhere [19–22]. Briefly, fibrinolysis means dissolving the fibrin meshwork in a thrombus by activation of plasminogen, a circulating single-chain glycoprotein, to plasmin. Plasmin splits both fibrinogen and fibrin into degradation products, which results in lysis of the clot (Fig. 2). Inhibition of the fibrinolytic system occurs both at the level of the plasminogen activators by plasminogen activator inhibitors (mainly plasminogen activator inhibitor-l, PAI-1) and at the level of plasmin (mainly by alpha-2-antiplasmin).
Thrombolytic agents (Fig. 3) are either approved, or under past or present clinical investigation in patients with ischaemic stroke, include streptokinase (SK), recombinant tissue-type plasminogen activator (rt-PA or alteplase), urokinase, recombinant pro-urokinase and desmoteplase [19–28]. Some biochemical and clinical specifics of the thrombolytic agents used in trials fulfilling the criteria and included in the Cochrane Database of Systematic Reviews are given in Table 1 .
Table 1. Some biochemical – fibrinolytic and molecular – and clinical properties of plasminogen activators in trials in the Cochrane Database of Systematic Reviews on thrombolysis in acute ischaemic stroke [1, 20, 22, 24–28]
Substance, generic name (abbreviation)
Molecular weight (kDa)
‘Dominant half-life’, minutes
Accumulation of fdp
kDA, kilo dalton; PLG, plasminogen, ‘dominant half-life’, time elapsed until the concentration in the circulating blood is half of the initial (an effect of the combination between the distribution and the elimination times), fdp, fibrinogen degradation products; HMW, high molecular weight; LMW, low molecular weight.
Urokinase (UK, u-PA)
HMW: 54, LMW: 32
Isolated from urine or embryonic kidney cells
Excreted in urine
No (LMW UK has a relative fibrin-specificity)
Induces an anticoagulant effect due to resulting high levels of fdp
8–25 (please see Column 6)
Isolated and purified from streptococcus bacteria
Degradation of ‘activator complex’ 18–23 min in circulation
Fdp affects platelet function increasing the bleeding risk
Liver (first passage)
No (please see column 11)
As any medication, allergic reactions observed
55 (47 when synthesized via E. coli)
Recombinant technology: mammalian cells or E. coli
Yes. Not as pronounced as t-PA
Precursor of urokinase. Thrombolytic effect ↑ by heparin
Desmoteplase (rDSPAα1) Info in animal study testing neurotoxicity, a.o.
Vampire bat saliva
Yes? (please see column 11)
Plasminogen activator, almost inactive in absence of a fibrin co-factor
Streptokinase is inactive until it forms a complex with circulating plasminogen, the streptokinase-plasminogen activator complex substitutes for tissue plasminogen activator in the fibrinolytic cascade causing plasmin activity. Recombinant t-PA is a genetically copy of the naturally occurring substance and a number of structurally related compounds have also been produced.
The effectiveness of any thrombolytic agent depends on the age and type of the thrombus and the surface area of the thrombus exposed to it. Intravenous SK treatment induces extensive systemic plasmin generations; alpha-2-antiplasmin inhibits circulating plasmin but may become exhausted during thrombolytic therapy because its plasma concentration is only approximately half that of plasminogen. As a result, plasmin, which has broad substrate specificity, will degrade several plasma proteins, such as fibrinogen, coagulation factors V, VIII and XII and vWF. Streptokinase is therefore considered nonfibrin-specific. The streptokinase-plasminogen activator complex is degraded enzymatically in the circulation. Combining with circulating neutralizing antibody clears some SK. After the use of SK, neutralizing antibodies can persist in high titre for several years and substantially reduce the effectiveness of subsequent therapy with SK. Streptokinase has a slower onset of action than t-PA owing to slow combination with plasminogen and the reperfusion of occluded vessels is slower but re-occlusion is less common because of their longer duration of action. The half-life of SK is longer than for t-PA.
On the other hand, tissue-type plasminogen activator (t-PA) is more fibrin-specific because it activates plasminogen preferentially at the fibrin surface and less in the circulation. Plasmin, associated with the fibrin surface, is protected from rapid inhibition by alpha-2-antiplasmin because its lysine-binding sites are not available and may thus efficiently degrade the fibrin of a thrombus. t-PA is metabolized in the liver.
Although all these thrombolytic agents act by converting plasminogen to plasmin, which dissolves the fibrin of blood clots, they are not all equal. Indeed, physiologic fibrinolysis is regulated by specific molecular interactions between its main component’s t-PA, plasminogen, and fibrin. As a result, plasminogen is preferentially activated at the fibrin surface, where generated plasmin is protected from rapid inhibition by α2-antiplasmin which may efficiently degrade the clot. Thus, fibrin-selective agents (rt-PA and derivatives, staphylokinase and derivatives, and to a lesser extent, scu-PA) that digest the clot in the absence of systemic plasminogen activation are distinguished from nonfibrin-selective agents (SK, urokinase), which activate systemic and fibrin-bound plasminogen relatively indiscriminately. Cardiac emboli often affecting the first or second branch of the median cerebral artery (MCA) (M1 and M2) may be examples of targets for fibrinolysis with fibrin-selective agents. Nonfibrin-selective agents are less efficient for clot dissolution and cause a systemic generation of plasmin, depletion of α2-antiplasmin, and degradation of coagulation factors, which protects against re-occlusion of the infarct-related artery.
All available thrombolytic agents suffer similar significant shortcomings, including need of large therapeutic doses, limited efficacy, re-occlusion and bleeding complications.
The licensing process and reasons for the poor uptake of fibrinolysis in clinical practice
In 1933, Tillett and Garner reported that SK, an enzyme derived from the streptococcus, could induce fibrinolysis via activation of plasminogen . Streptokinase was first approved for treatment of deep venous thrombosis and pulmonary embolism . Streptokinase also early came to use on the arterial side, in acute myocardial infarction (MI) . Tissue plasminogen activator was identified in 1947 . Despite a number of large SK trials with important implication on treatment uptake , and several rt-PA trials  it was specifically the GUSTO trial, with its angiographic substudies and comparing of rt-PA and SK, published in 1993 which established the potentials for rt-PA in MI [35, 36]. For MI there has since, along with the development of standard invasive procedures, been a move to fibrinolytic derivatives, some suitable to be administered as one single fast bolus-injection.
In MI by the late 1980s large-scale randomized controlled trials of thrombolysis had been completed. In aggregate, these trials had included over 40 000 patients and had provided proof beyond reasonable doubt that intravenous thrombolytic therapy, given within the first few hours, saved the lives of patients with acute MI. By contrast, the evidence for thrombolysis in stroke comes from a number of much smaller trials with a total of just over 7000 patients, of which 11 are trials of rt-PA (with approximately 3600 patients). Testing fibrinolysis in stroke has had a somewhat chequered history. The first ‘positive trial’ of thrombolysis, the National Institute of Neurological Disorders and Stroke (NINDS) trial was published in 1995 . The NINDS results were clearly positive with a significant 14% reduction in the absolute risk of death or dependency at 3 months in comparison to controls. These results led to the Food and Drug Administration (FDA) approving the drug for use in the USA in 1996, for patients meeting specific criteria who could be treated within 3 h of onset. However, there was an important imbalance in baseline stroke severity with fewer patients with severe stroke in those allocated rt-PA in the NINDS trial . After adjustment, the absolute benefit was still a substantial 11% [1, 39, 40] guidelines were issued . However, the next trial, ECASS-1 , did not achieve statistical significance and a long debate about the risks and benefits of the treatment ensued. A pooled meta-analysis of trials of SK in stroke was published in 1999 and included data from the MAST-E, MAST-I and ASK trials . As heparin and aspirin were allowed according to the discretion of the treating physician in MAST-E  and were also frequently used in the trial as well during the first 24 h as at higher rates later during the first 2 weeks, the true effect of SK was hard to extract. In the ASK trial aspirin was to be given within the first 4 h , and in one part of the factorial designed MAST-I within 6 h . Accordingly, the MAST-trial provided the only data on the pure effect of SK as well as the effect of SK plus aspirin in combination. It should be noted that the odds ratio for death and dependency also favours better outcome with SK, but the confidence intervals (CI) were wide and cross the line of no effect. MAST-E and ASK were prematurely terminated due to the high frequency of intracranial haemorrhage (ICH), as was MAST-I due to the general uncertainty of thrombolysis in stroke at that time. In summary, pure SK in stroke has not been investigated to its full potential. From a biochemical and clinical side, irrespective of the concomitant therapy given in such large parts of the undertaken trials, what may talk against SK is its high antigen and blood pressure lowering properties . Most probably for stroke in relation to MI, we may have an irreparable lack in our knowledge of findings in a trial similar to the GUSTO comparing the effects of a nonfibrin-specific drug with longer half-life like SK with a fibrin-specific agent with a short half-life like rt-PA .
Recombinant tissue-type plasminogen activator did not receive a conditional approval for use in Europe until 2004. There were two conditions, the SITS-MOST quality register and the randomized controlled trial of the 3–4.5 h time window, the ECASS 3 trial. Both have now been fulfilled [47, 48]. The first, the SITS-MOST showed that for symptomatic intracranial haemorrhage (SICH); death; and the combined outcome death or dependency at 3 months, the results lay well within the CI of the older rt-PA trials; for death the results were even better . A learning curve was observed interestingly confirming experiences from MI .
During this time, an individual pooled meta-analysis of rt-PA data  based on the larger rt-PA trials from the 1990s for the time interval 0–3–5/6 h: the ATLANTIS A and B [51, 52], the ECASS I and II [42, 53] and the two parts of the NINDS trial  was published. The CIs are very wide. A peak of inclusions just before 90 min in the NINDS trial makes it difficult to interpret the slope of the benefit versus time to treatment-curve, Nevertheless this pooled individual meta-analysis gave rise to two major conclusions: (i) there is certainly scope for substantial benefit from early treatment, and (ii) the upper CI suggests the worthwhile benefit out to 6 h (and further, for those treated between 181 and 270 min from stroke onset the odds of a favourable outcome was 1.40).
The ECASS 3, published autumn 2008  was the first thrombolytic trial after the NINDS trial fulfilling its chosen primary end-point. In ECASS 3, patients were randomized to rt-PA or placebo between three and 4.5 h after stroke, and it showed a significant reduction of death and dependency on its primary outcome. As demanded by the EMEA, the trial regarded the same highly selected group as defined in the EU-licence. The licence had introduced an upper and lower stroke severity grade according to clinical assessment with the National Institute of Health Stroke Scale (NIHSS), a contraindication for patients with diabetes mellitus and earlier stroke, and an upper age limit of 80 years , i.e. all patients above 80 years of age were excluded. However, ECASS 3 being positive is definitely encouraging. It reinforces the principle of the effect-curve of the pooled meta-analysis . A continued SITS register, the SITS-ISTR, also supports the ECASS 3 results .
The Cochrane Systematic Review has successively synthesized the data from each successive clinical trial of thrombolysis since the early 1990s. Despite the comprehensive evidence from these data it is still difficult to make a full risk-benefit analysis for the individual stroke patient primarily because the information was not collected in the individual trials, or because the trials only included a very narrow range of patients. Hence, data from further trials will be needed to better inform clinical decision-making.
Thrombolysis is grossly underused in clinical care. Several investigations have been undertaken to identify the reasons [56–59]. Of the theoretical target-group of 80% ischaemic stroke (of which an approximate percentage of 30–50 will have absolute or relative contraindications to treatment leaving a potential of 50–30% eligible for treatment) only as a mean 1–8% receive fibrinolytic treatment. This goes for the USA as well as for Europe and other countries. There are some centres where treatment rates reach 10–20% , but such centres are more the exception than the rule. Amongst a number of explanations, the patients’ socio-economic situation, gender, and the availability of a neurologist/stroke-specialist have been identified. The very strict criteria, especially in Europe, are limiting. Obviously vast parts of the ischaemic stroke population, the elderly, patients who cannot reach hospital within the licensed time, those with so far uninvestigated stroke types and severity-grades, as well as those with findings on imaging not in line with the licence criteria are left outside despite that several of the contraindications have not been tested in a randomized controlled trial (RCT) with sufficient power.
In a Scandinavian survey, the clinicians felt safe with thrombolysis as such . They rather assigned the low use of thrombolysis to factors outside their control such as information to the general population (to know how to act, i.e. to get an ambulance and go to hospital), the ambulance services and insufficient in-hospital logistics. According to a British Health Technology Assessment Panel report, the two main barriers to widespread use of thrombolysis were the remaining uncertainty over the effect of treatment in some categories of patient and the major investment in stroke service provision required for successful and safe implementation of treatment . An interesting observation has been a boosting effect on the randomization in the large Third International Stroke Trial, IST-3 after the publication of ECASS 3 (Sandercock PAG, personal communication). This most probably results from several reasons. The first is that physicians’ want evidence and they now envisage a larger group potentially able to be treated on evidence base after IST-3 . Also, the second barrier pointed to in the HTA report, that major investment in stroke service provision are required for successful and safe implementation of treatment, went hand in glove with findings of the Scandinavian survey where adequate stroke service provisions were obviously in shortage. In the same survey, there were some differences between the three participating countries. In Denmark fibrinolysis is given in a few specialized centres. Treatment rates in those large centres were higher than in the many smaller centres in Norway and Sweden where thrombolysis is more widely administered across the country in many more hospitals. The difference in organization is partly an adjustment to different geographical conditions, but in part also an indication of a different view on the administration of thrombolysis. In Sweden, around 3.5% of all ischaemic stroke patients get thrombolysis and of those fulfilling the licence criteria the frequency has reached 7.8% . The licensed t-PA-use for stroke increases gradually but the overall use is still low and can be much improved.
Comparison between fibrinolysis data in animals and humans (review)
The efficacy and safety of t-PA treatment in animals subjected to stroke were analysed in a systematic review and meta-analysis based on 104 reports with 389 experiments and 3379 animals (Sena et al., personal communication). The majority of studied animals were anesthetized rats or rabbits treated with intravenous t-PA [10 mg/kg] at various time after thrombotic stroke. In this large dataset there was large heterogeneity between studies, which could partly be explained by various study quality and to a certain degree to publication bias. Nevertheless, t-PA reduced the infarct volume by an average of 25.8% (95% CI 22.4–29.2, n = 3379 animals), improved neurobehavioral score by 19.1% (95% CI 13.3–25.0, n = 1165) and increased the risk of haemorrhage (OR 1.78, 95% CI 1.5–2.2, n = 2905). For infarct volume, there was an absolute reduction in efficacy of 1.1% for every 10 min delay of treatment. Overall, animal data are remarkably consistent with what is known of the clinical efficacy of t-PA which supports the notion that these data obtained from rodents are relevant to man. On the other hand, improvements in experimental design have the potential to make the process of translation from bench to bedside more efficient.
Data on clinical outcome on fibrinolysis for acute ischaemic stroke
The least biased and most precise assessment of the effects of a medical treatment is a systematic review of all the relevant randomized controlled trials. Some data from the Cochrane Database of Systematic Reviews on Thrombolysis for Acute Ischaemic Stroke with its latest searches to October 2008 can provide an illustration of the clinical application and outcome of thrombolysis. The review includes all randomized controlled trials in patients with definite ischaemic stroke and of any thrombolytic drug versus control. Since in the latest update, when new information from previously published trials was also added, the review gives all current evidence on thrombolysis .
The Cochrane review includes trials with urokinase (u-PA) [64–69], SK [44–46, 70], rt-PA [37, 42, 48, 51–53, 71–75], recombinant pro-urokinase (r-pro-UK/scu-PA) [76, 77], or desmoteplase [78–80]. Of all data more than 55% come from rt-PA trials. The majority of the trials report i.v. administration, but there are two studies on i.a. administration of pro-urokinase [76, 77] and, in two more recent reports, of i.a. urokinase [68, 69]. The trials had varying stroke onset to latest time for randomization, from 0 to 3 h to up to 24 h in one small study. The majority gave treatment up to 6 h. The early phase data includes the period of 7–10 days. The time for scheduled follow up is usually 3 months.
The main results of the Cochrane review are the apparent significant net benefit with fewer patients dead or dependent at the end of follow up, i.e. significantly more patients alive and independent, across all drugs and time windows despite a fourfold increase in SICH and a nonsignificant increase in death. This review will give the results from the Cochrane review on the major outcomes: death, symptomatic including fatal ICH, death or dependency; a brief description of the how the drugs relate to these major outcomes depending on their different molecular groups and if some conclusions on mechanisms could be made [81–83]. If not otherwise noted the results are first given for all substances together followed by rt-PA data in somewhat more detail.
Death from all causes within seven to 10 days
Thrombolysis was associated with a significant excess of early deaths (Table 2). Of those allocated to thrombolysis 11.91% died compared to 7.67% of those allocated to control. As can be seen, trials using the nonfibrin-selective drug SK showed a significant excess of early deaths, without heterogeneity, but note that this result is confounded by excessive use of antithrombotic and anticoagulant drugs as discussed above. Although all drugs contributed to the increase in death, the borderline significant heterogeneity (I2 = 44%, P = 0.04) reveals that all trials did not contribute. The analysis of available rt-PA data showed a nonsignificant excess of early deaths with no significant heterogeneity. In absolute numbers this equals 10 more (95% CI 10 fewer to 30 more) deaths per 1000 patients treated. Due to the heterogeneity, the fixed-effects analysis was complemented with the random-effects model. This confirmed the general overall excess in early death including the nonsignificant increase of the rt-PA trials.
Table 2. Any thrombolytic drug versus control: death from all causes within 7–10 days. Data from trials reporting this outcome in the Cochrane review : 4423 patients in 12 trials
Drug/no of trials
No of patients: in each study-arm (n)
Peto odds ratio Peto fixed, 95%, CI
Test for overall effect
MAST-I  appears twice in each of Tables 2–4 due to its 2 × 2 factorial design with SK or aspirin, or both, or neither. NA, not applicable; CI, confidence interval.
P < 0.00001 Test for subgroup differences: I2 = 73.0% P = 0.005
Deaths from all causes during follow up (including early deaths)
Long-term death from all causes within scheduled follow up could be analysed for all trials in 7152 patients (Table 3). Death from all causes was 16.5% amongst actively treated patients versus 13.9% amongst controls. Due to the significant heterogeneity the findings were checked with the random-effects model which confirmed the significant excess of deaths. For the rt-PA trials there was a nonsignificant increase in deaths equivalent to 10 more deaths (10 fewer to 40 more) per 1000 patients treated. As can be seen there was a nonsignificant heterogeneity (I2 = 40%, P = 0.08) and the random-effects model confirmed the findings.
Table 3. Any thrombolytic drug versus control: death from all causes within time to follow up. Data available from all trials in the Cochrane review : 7152 patients: i.v. urokinase [64–66]; streptokinase [44–46, 70]; rt-PA [37, 42, 48, 51–53, 71–75]; i.a. pro-urokinase plus i.v. heparin [76, 77]; i.a. urokinase [68, 69]; desmoteplase [78–80]
Drug/no of trials
No of patients: in each study-arm (n)
Peto odds ratio Peto fixed, 95%, CI
Test for overall effect
NA, not applicable; CI, confidence intervals.
i.v. Urokinase/trials = 4
1.15 [0.68, 1.97]
I2 = 0% P = 0.53
P = 0.60
Streptokinase/trials = 4
1.43 [1.10, 1.88]
I2 = 47% P = 0.13
P = 0.008
t-PA/trials = 11
1.14 [0.95, 1.38]
I2 = 40% P = 0.08
P = 0.16
Streptokinase + aspirin versus aspirin/trial = 1
3.02 [1.87, 4.87]
P < 0.00001
i.a. pro-urokinase plus i.v. heparin versus i.v. heparin/trials = 2
0.75 [0.40, 1.42]
I2 = 0% P = 0.49
P = 0.38
i.a. urokinase/trials = 2
1.25 [0.34, 4.57]
I2 = 0% P = 0.76
P = 0.74
Desmoteplase/trials = 3
2.17 [0.97, 4.84]
I2 = 0% P = 0.43
P = 0.06
Summary/trials = 26
1.31 [1.14, 1.50]
I2 = 43% P = 0.01
P < 0.0001 Test for subgroup differences: I2 = 68.0% P = 0.005
The pronounced effect of SK, a nonfibrin-selective agent, seems to be the summary of a sequence of events with a decrease of fibrinogen levels together with an increase in fibrinogen degradation products, confounded by excessive use of antithrombotic and anticoagulant drugs. Fibrinogen degradation products have a negative effect on platelet function. With high-dose SK haemostasis is mainly dependant on the platelets. When their function is disturbed there is not much defence left against bleeding. Adding aspirin theoretically takes away the last defence. This is a plausible mechanism for SK and aspirin to have such pronounced effects on death. The fact that urokinase, which partly also belongs to the nonfibrin-selective agents, did not display the same pattern may be due to the u-PA trials having been undertaken in a different way with strict avoidance of aspirin, and maybe importantly that the half-life of urokinase lies in the vicinity of the short half-life of rt-PA.
The reported analyses on death clearly show that new prospective data are needed, not least importantly to clarify the nonsignificant excess of deaths in rt-PA, the licensed drug.
Symptomatic (including fatal) intracranial haemorrhage within 7–10 days
For fatal ICH there was a significant approximate fivefold increase with thrombolysis (4.45% in actively treated patient versus 0.74% in controls, without any heterogeneity (I2 = 0%). In rt-PA trials this meant 27 (95% CI 20–40) extra fatal ICH per 1000 treated with no heterogeneity (I2 = 0%). In rt-PA there was a highly significant fourfold increased frequency of 7.7% symptomatic (SICH) including fatal ICH amongst thrombolysis treated patients compared to 2.1% in controls, with no significant between-trial heterogeneity (P = 0.39). In trials using rt-PA in absolute numbers there were 60 (95% CI 50–80) extra SICH/1000 patients treated with no heterogeneity between trials (I2 = 5%).
The lack of between-trial heterogeneity for ICH may, amongst other things, indicate the lack of importance of whether or not treatment is given with a fibrin-selective- or a nonfibrin-selective agent. Also, the varying half-lives of the agents seem to have produced no differences between the different thrombolytic drugs. In analysing the data, it is obvious that the patients included in the different trials are very different underpinning the importance of obtaining new trial data categorized by consistent terminology both on patient and imaging characteristics to identify determinants of ICH.
Definition of SICH. All trials provided information on ICH and most of them also in a form which made it clear how many of the patients had suffered a neurological deterioration associated with the appearance of the new ICH as seen on computed tomography (CT), magnetic resonance imaging (MR), or on postmortem examination. However, the definition of SICH varies subtly between trials and is successively being adjusted, perhaps reflecting initial increased concern about any haemorrhage in early trials, followed by more pragmatic recognition that haemorrhage might be offset by other factors such as worse swelling in later trials. From the beginning, the combination of a clinical deterioration and the discovery of any new ICH usually lead to the ICH being assigned as the cause of the deterioration/death. Now successively a more nuanced definition of SICH, including fatal ICH, is used. If for example as well an extensive oedema and a small haemorrhage or haemorrhagic transformation are found on the scan and the patient deteriorates or dies, the cause of deterioration and death will more likely be assigned the infarction and the oedema than the ICH . The distinction may at times be difficult and more knowledge is needed. At present the definition used in ECASS 3  is on the way to being adopted: SICH = any apparent extravascular blood in the brain/within the cranium associated with a clinical deterioration >4 on NIHSS, or that led to death and that was identified as the predominant cause of the neurologic deterioration. To allow comparison with published data analyses of rates of SICH should also be performed according to the definitions used in those trials. However, the new way of reasoning reflects a doctrinal shift in defining an SICH even if the exact definition may not yet have been set. For ongoing or future trials, central reading of scans may be a useful step towards conformity. All reasonable features: clinical, imaging, and possibly drug-specific molecular mechanisms and half-lives, may then be correlated to the new accepted SICH definition. Risk-identifiers should be possible to derive which as already mentioned is one of the major issues for new large trials.
Death or dependency
Outcome at the end of scheduled follow up is mainly measured as the combined outcome measure ‘death or dependency’. There was a significant reduction in death or dependency with thrombolysis: 51.8% of those allocated to thrombolytic therapy compared to 55.8% of those allocated to control (Table 4). There was significant heterogeneity of treatment effect between the trials (I2 = 38%, P = 0.04). Due to the heterogeneity analysis according to the random-effects model was performed. For all trials, as well as for the trials with rt-PA, the random-effects model confirmed the effects to be of approximately the same magnitude as with the fixed-effects analyses. Table 5 summarizes the data for rt-PA on the major outcomes in absolute numbers.
Table 4. Any thrombolytic drug versus control: death or dependency within time to follow up. Data from trials reporting this outcome in the Cochrane review  on patients available for this analysis: 6283 i.v. urokinase ; streptokinase [44–46, 70]; rt-PA [37, 42, 48, 51–53, 71, 74, 75]; i.a. pro-urokinase [76, 77]; i.a. urokinase [68, 69]; desmoteplase [78–80]
Drug/no of trials
No of patients: in each study-arm (n)
Peto odds ratio Peto fixed, 95%, CI
Test for overall effect
NA, not applicable; CI, confidence interval.
i.v. Urokinase/trial = 1
0.95 [0.64, 1.42]
P = 0.81
Streptokinase/trials = 4
0.94 [0.72, 1.24]
I2 = 0% P = 0.56
P = 0.68
t-PA/trials = 9
0.78 [0.68, 0.88]
I2 = 62% P = 0.007
P = 0.0001
Streptokinase + aspirin versus aspirin/trial = 1
1.09 [0.69, 1.73]
P = 0.71
i.a. pro-urokinase +i.v. heparin versus i.v. heparin/trials = 2
0.55 [0.31, 1.00]
I2 = 0% P = 0.84
P = 0.05
i.a. urokinase versus control/trials = 2
0.57 [0.28, 1.14]
I2 = 16% P = 0.27
P = 0.11
Desmoteplase/trials = 3
0.86 [0.53, 1.40]
I2 = 36% P = 0.21
P = 0.54
Summary/trials = 21
0.81 [0.73, 0.90]
I2 = 38% P = 0.04
P < 0.0001 Test for subgroup differences: I2 = 8.1% P = 0.37
Table 5. The three major outcomes: ‘symptomatic intracranial haemorrhage (SICH)’ (within 7–10 days); ‘death’, and ‘death or dependency’ (both within/at the end of follow up, 3–6 months). Results from meta-analyses in the Cochrane Database of Systematic Reviews on thrombolysis for acute ischaemic stroke; the odds ratio and confidence intervals are given in absolute numbers
0–6 h (all)
The numbers are given as Peto odds ratio (OR), Peto fixed, and 95% confidence intervals (CI). Bold: mean increase (favours control). Italic mean decrease (favours treatment).
60 (95% CI 50–80)
70 (95% CI 40–100)
60 (95% CI 50–80)
10 (95% CI 10–40)
0 (95% CI 50–50)
20 (95% CI 0–50)
Death or dependency
60* (95% CI 100–30)
110 (95% CI 170–50)
40 (95% CI 40–10)
Some safety issues
Antigenic properties may occur in any of the fibrinolytic agents, mainly in SK. Allergic reaction is however an issue of importance in probably any fibrinolytic agent. In treatment with rt-PA the reaction can range from any mild manifestation to full anaphylaxis. As the treatment goes along the common rules of therapy, which is also the case in the few other even more rare but general complications, these will not be discussed here. It is however one more reason for close surveillance of the patient since these reactions may occur well after the infusion.
Concomitant cardiac disorders. It is of relevance to consider the co-morbidity of stroke and a cardiac or cardio-vascular condition. Up to 20% of the stroke patients may simultaneously have a cardiac condition including a silent MI . Close cardiac surveillance to detect an MI and/or arrhythmias as well as blood pressure monitoring from a variety of clinical reasons is therefore mandatory since obviously a nontreated serious side effect or co-morbid condition can be fatal.
Extracranial haemorrhage. Extracranial haemorrhages do occur, frequently as minor manifestations of small trauma like inserting a urinary tract catheter, or at suction of the mouth, etc. Profuse extracranial haemorrhages may however be fatal. The organ-systems most often involved, the uro-genital and gastro-intestinal tracts, if quickly approached usually allow the source of the bleeding to be surgically addressed. Some SICH can be addressed with neurosurgery but generally no action is possible and the overall mortality due to SICH is approximately 3% . In any bleeding, the fibrinolytic treatment if still ongoing should immediately be discontinued and a reversing may be considered. However, this is rarely used for fear of worsening the ischaemic situation in the brain. The extracranial bleedings can generally be handled according to ordinary surgically procedures and will not be further commented upon here.
Risk of re-occlusion and effects in stroke of combining streptokinase with aspirin. In acute MI therapy, fibrinolysis is combined with platelet antiaggregators and anticoagulant treatment. In stroke this is not recommended, so far due to the findings in the randomized evidence on the combination of therapies in the SK trials. Further than that, we cannot comment at present, but the occurrence of re-occlusion may at least be elucidated when MR or CT angiography become more common. Thrombolytic agents may at a certain point be counteracting which results in the paradoxical pro-thrombotic effects . Hence, it could be this paradoxical pro-coagulation that may be responsible for clinically observed early re-occlusions after successful therapy. This kind of primary improvement followed by a deterioration need to be further elucidated in trials. The development and improvement of antithrombotic therapies have the potential to improve also the thrombolytic efficacy of the plasminogen activators. From a molecular stand point, to understand the mechanisms and be able to address them, the development of an antithrombotic as well as a pro-fibrinolytic system is as vital for the risk-benefit balance between interactions such as between SK and ASA.
In the context of re-occlusion some issues are of interest. As for thrombolytic drug, in comparing all drugs regardless of route of administration or time window there was no significant heterogeneity between trials regarding SICH (I2 = 5%, P = 0.39); no conventional heterogeneity for ‘death or dependency’ (I2 = 38%, P = 0.04) and only borderline heterogeneity for early death (I2 = 44%, P = 0.04). Random-effects model was performed. For all trials as well as for the trials with rt-PA the random-effects model confirmed the effects to be of approximately the same magnitude as with the fixed-effects analyses.
The concomitant antithrombotic drug was not of benefit in the trial with SK but has not been studied in other trials. However, in new trials like the on-going Third International Stroke Trial, IST-3, on-going treatment with antithrombotic agents prior to randomization is being studied and may give some answers to that question .
Any indirect comparison is bound to be confounded by a number of factors which are hard to interpret. As can be seen in Tables 2–5 the CI overlap. Only direct randomized comparisons would be able to answer this question of which drug has least hazard and most benefit. Accordingly, the positive differences should not either be too emphatically stressed. Further stroke severity, stroke-type, age, maybe gender, and time, although not as straightforward as mostly described, all define the risks and effects of thrombolysis.
As for death or dependency there are examples of how an imbalance between the trial arms with an enrichment of more severe stroke cases in one of the arms will imbalance the whole results by negatively affecting the outcome of that study-arm [37, 52]. Stroke severity was not included in randomization in any trials to date to balance for key prognostic factors. Stroke severity has been linked to a higher case fatality  as compared to a trial with less severe stroke . Whether severity has an effect on the frequency of ICH is not clarified. However, the reason for the so called 1/3-MCA-rule, that patients with early ischaemic findings in more than 1/3 of the MCA supply area should not get fibrinolysis comes from findings in one trial, the ECASS, and has been accepted without further randomized investigation, but was based on an un-blinded scan reading. This is definitely one objective for a new trial. From a molecular point of view, if a large ischaemic area in its border-area would have a number of small fibrin-clots a fibrin-selective agent would theoretically possibly induce bleeding also in those and not only address the target occlusion. However, it is to note that the post hoc analysis of both the NINDS trial (rt-PA) and the MAST-I (SK) suggested that the risk reduction for death or dependency varied with stroke severity, being largest for patients with moderate stroke (NIHSS score around 10–15) and least in severe stroke (NINDS score around 18–30) [1, 86]. However, these findings were based on small numbers, in post hoc analyses and they need to be verified prospectively in larger randomized trials. It will also be highly interesting to investigate the effect on seemingly mild stroke where two series of events may occur. One is that the stroke after all has not been that mild as postulated but includes symptoms of importance for QoL, like minor disabilities, e.g. mobility disturbance of the hand; sensory deficits; cognitive disturbances in the form of mild aphasia/intellectual impairment; or different varieties of pain. The other event is that the patient starts to improve rapidly and is classified as having a mild reversible stroke or transitory ischaemic attack (TIA). Then the patient suddenly deteriorates. The occurrence is not fully known but may happen in up to 15–30% of the cases .
The possibility to include also relevant mild cases in IST-3 is interesting. New trials will teach us more about stroke severity, stroke types, age [10, 63], occurrence, determinants and time frames.
The clinical effect is not dramatically different as a consequence of the thrombolytic agents used. Therefore the molecular mechanisms with regards to fibrin-specific or nonfibrin-specific agents may not be as important in practise, but at the present time we lack information to make a full interpretation. Taking into account that the trials have been undertaken at different kinds of centres with different experience and with different substances and unknown concomitant medication, one should not assign too much importance to differences. However, the most benign after the circumstances seem to be the fibrin-specific rt-PA and the nonfibrin-specific urokinase. Since so far thrombolysis is the most promising treatment in acute ischaemic stroke it is essential to get to new evidence permitting a probably much higher utilization of this treatment. The only way is via new trials of sufficient power covering clinical and imaging factors. Table 6 illustrates the impact of an increase of thrombolysis to 30%. Working under the assumption that evidence will improve our ability to detect risk-identifiers and give us the markers of good effect, this is a good illustration. Clinical trials should include, or be complemented with preclinical trials and studies on, molecular aspects. This may identify drugs that could be tailored for different types of occlusion probably correlated to the site and type of the ischaemic stroke as judged clinically and on imaging. The molecular advances should continue to tailor fibrinolytic drugs to avoid ICH and allow concomitant antithrombotic treatment and anticoagulation to avoid the risk of re-occlusion.
Table 6. The hypothetical effect of an increase in number of patients treated with thrombolysis. The calculation is based on estimated effects with aspirin , treatment in a Stroke Unit , and he effect of thrombolysis in absolute numbers for the 0–6 hour time window in the Cochrane Database of Systematic Reviews on Thrombolysis for Acute Ischaemic Stroke . Data on treatment utilisation are from the UK. How many stroke patients in the UK (130,000 stroke/year) might avoid being ‘dead or dependent’ with each treatment?
% Treated with this intervention
Number treated per year
Benefit per 1000 treated
Number who avoid death or dependency
Aspirin <48 h
Stroke unit at present (UK)
Fibrinolysis: rt-PA in acute ischaemic stroke, with treatment 0–6 h at present rate (UK)
Fibrinolysis: rt-PA in acute ischaemic stroke with treatment 0–6 h at hypothetical rate
We wish to thank the Co-chief Investigator of IST-3, Professor Richard Lindley, Sydney, for stimulating discussions on fibrinolysis and thrombolysis, Professor Björn Wiman, Department of Clin Chemistry, Karolinska University Hospital, Stockholm for discussions on molecular aspects; Mrs Brenda Thomas and Mrs Hazel Fraser of the Cochrane Stroke Group in Edinburgh for assistance with literature searching and Ms K Shuler in Edinburgh for entry of trial details into the Cochrane Thrombolysis Review. Professor Joanna Wardlaw is part-funded by the Scottish Funding Council through the SINAPSE Collaboration (Scottish Imaging Network, A Platform for Scientific Excellence, http://www.sinapse.ac.uk). The Swedish Heart-Lung Fund, the Foundation of Marianne and Marcus Wallenberg, AFA Insurances, and ALF-project grants from Karolinska Institutet and Stockholm County Council support IST-3, thrombolysis, Sweden.
Conflicts of interest statement
Dr Veronica Murray is the Swedish National Co-ordinator and member of the Swedish Management Group for IST-3.
Professor Bo Norrving is a member of the Swedish Management Group for IST-3.
Professor Peter Sandercock is the Co-chief investigator of IST-3, an independent, investigator-led trial supported by grants from among others the UK Medical Research Council.
Professor Andreas Terént has received a research grant for epidemiological research from AstraZeneca. He is a member of the Swedish Management Group for IST-3.
Professor Joanna Wardlaw is Imaging Lead for IST-3, Steering Committee Member and grant co-holder. She was on the Steering Committees of MAST-1 and contributed to the design of ECASS 3 (first Steering Committee meeting and design of scan reading). The Division of Clinical Neurosciences received payment from Boehringer Ingelheim for JMW’s work in reading scans for ECASS 3.
Professor Per Wester is a member of the Swedish Management group for IST-3.