4Discussion

Studies of Treatment of ICH

The meta-analysis performed for the surgical trials, for both death, and death and disability, showed heterogeneity within the trials. This may be the consequence of different surgical techniques used in the trials. The confidence interval associated with a pooled effect estimate for both endpoints is large and does not preclude a clinically significant treatment effect from surgical intervention particularly in certain subgroups of patients. Current clinical opinion holds that surgery be considered for patients with cerebellar hemorrhage, particularly if the diameter exceeds 3 cm, and for young patients with significant hemispheric hemorrhage experiencing clinical deterioration.²³⁴

In spite of the potential wide applicability of anti-hypertensive therapy following ICH very little data is available regarding this intervention. Both studies included had very small numbers and were significantly hampered by lack of measurement of clinical effect. The study examining intravenous ketanserin on arterial and intracranial pressures in a small group of patients raises the possibility that cerebral perfusion pressure might be affected by anti-hypertensive therapy.⁹⁰

Role of Thrombolysis in Stroke

Stroke due to vascular occlusion (ischemic stroke) comprises 80% of all stroke.²³⁵ The mechanisms of occlusion are varied but have in common the deposition of platelets and fibrin. Fibrin is a protein that is deposited in strands enmeshing red blood cells and platelets alike. Plasminogen is released from the vascular endothelium and converted to plasmin by plasminogen activator and plays a role in regulating the duration and propagation of clots within the body. The human gene can be spliced into bacteria to allow for large-scale production of tPA and harvested in commercially viable quantities. Genentech received FDA approval for its use in the coronary circulation in 1987 with subsequent use in acute occlusions of the coronary arteries.

Prior to the publication of the NINDS Trial of IV rtPA delivered within 3 hours of stroke onset,²³⁶ no acute therapy had been shown to modify the outcome of ischemic stroke in human trials. Two potential interventions for acute ischemic stroke therapy were: 1) re-establishing blood flow prior to cell death, and 2) limiting irreversible damage in ischemic brain. Efforts had been mainly directed at reducing neuronal metabolism^{49, 237} neuroprotection,²³⁸ and antioxidation.²³⁹ While the sequence of molecular events which led from energy failure through membrane depolarization, cytotoxin release, calcium influx with subsequent activation of destructive enzymes had been well worked out;²³⁹ attempts to influence the process in humans had been ineffective.^{240, 241} Preclinical work on rodent models of these interventions had been highly successful, suggesting a profound disconnection between animal models of stroke and the human syndrome studied in clinical trials. Several reasons have been suggested for this failure to translate laboratory results into clinical outcomes.

Animal models have focused on short time periods of drug administration and outcome evaluation with a focus on anatomic measurements (volume of infarct) rather than functional goals.²⁴¹ While consideration of such trials has not stopped,^{241, 242} the results suggest a need for a renewal of efforts on reversing vascular occlusion prior to the development of irreversible cellular damage.

Thrombolysis enjoys a prominence in acute stroke due to this critical role of reestablishing perfusion in the face of repeated failure of other strategies. It is the only medical treatment approved for acute ischemic stroke in the U.S. and Canada. The stroke community in North America has embraced this treatment for acute stroke.^{66, 243–245} However, the evidence supporting this treatment rests on substantially fewer trials and numbers of patients than that accumulated for cardiac ischemia.

Subsequent guidelines suggested that tPA should be used within 3 hours of stroke onset.⁵³ Similar recommendations followed in Canada from the Canadian Stroke Consortium²⁴⁶ with the recommendation that use be limited to sites with personnel with experience in acute stroke care. Post-marketing surveillance mandated by HPB and maintained by the CASES database²⁴⁷ demonstrated a treatment rate in Canada of 6% of all stroke.⁵⁹Variability was noted in treatment rates between sites. There are currently nine designated stroke centers in the Province of Ontario with rates of thrombolysis ranging from 2.9% to 20.6% for the period July to December, 2002.²⁴⁸

Our search for references examining the effectiveness of thrombolytic therapy for acute stroke identified 343 reviews. Of these, the majority did not consider intravenous tPA, acute stroke or were narrative reviews. One¹⁹⁵ analyzed the relationship between onset to treatment time and outcome and is further discussed under Intervention F. This left two reviews for inclusion.^{249, 250}

Wardlaw and colleagues²⁴⁹ reviewed 18 trials including 5727 patients. Sixteen were double blind and while a variety of thrombolytic agents were used, 50% of the patients in trials included utilized intravenous tPA. Overall, thrombolytic therapy administered in the 0 to 6 hour window decreased the proportion of patients who are dependent at three to six months (odds ratio 0.84, 95% CI, 0.75–0.95). Patients treated within this time frame also had greater odds of death at the end of 3 to 6 month follow-up (odds ratio 1.33, 95% CI, 1.15–1.53). For patients treated within the 0 to 3-hour window, death or dependency was reduced further (OR 0.66, 95% CI 0.53–0.83) with no increase in the risk of death (OR 1.13, 95% CI, 0.86–1.48). Some heterogeneity of effect was noted. Data on ethnicity was not available. The reviewers did comment that there was limited data on patients over the age of 80. Graham and colleagues²⁵⁰ conducted a meta-analysis regarding the safety of tPA for acute ischemic stroke. Fifteen published open-label studies including 2,639 treated patients were included. In spite of an overall rate of protocol violation of 19.8%, the symptomatic intracranial hemorrhage rate was 5.2% (95% CI, 4.3–6.0). This compared favorably with the NINDS Trial rate of 6.4%. Of significance, however, was the cross study correlation of mortality rate with protocol violation rate (r=0.67, p=;0.018).

Intervention A: Does Surgery Impact the Outcome in Patients with Acute Intracerebral Hematoma (ICH)?

ICH represents approximately 9% of the 700,000 strokes which occur in the U.S. every year.⁷ Significant morbidity and mortality is associated with this condition.

Older studies on surgical treatment of ICH are hampered by the low sample sizes. Only the study by Morgenstern⁸⁷ and Batjer¹²⁰ provided power calculations. Sample size was limited by poor accrual in the SICHPA trial⁸⁸ and by a perception of futility in the Batjer trial.¹²⁰ In addition to poor power, the low sample size increases the possibility of baseline imbalances. With the notable exception of Batjer,¹²⁰ the management, particularly in the control or medical arm, was poorly specified. Potentially important covariates such as intracranial pressure, hypertension, glucose control, and the institution of mechanical ventilation, were either not specified or unreported. None of the included studies specified blinded outcome assessment. In addition, all studies included patients with supratentorial hemorrhage and no studies included patients with infratentorial hemorrhage.

The STICH trial,¹²² which accrued 1,033 patients, overcame a number of the issues which limited the early trials. The time window for surgical treatment was rather large at 24 hours. Further, the method of clot evacuation was not pre-specified. The mean time between ictus and randomization was long, at 22 hours. Thus, the possibility that early treatment with minimally-invasive surgical modalities improves outcomes. Finally the primary endpoint was measured in a blinded fashion.

Our meta-analysis does not preclude a benefit for treatment. The point estimate favors surgery, with the confidence interval crossing the null (Figure 2). Given the issues identified above, a well-designed, appropriately powered prospective evaluation is required.

Intervention B: Does Antihypertensive Treatment Reduce Stroke-Related Mortality and Disability in Patients with Acute ICH?

Brain injury in ICH is due to direct mechanical injury related to expanding clot, increased intracerebral pressure (ICP), herniation secondary to mass effect and the toxic effects of extravascular blood. Patients who have an ICH often have an acute rise in systemic arterial pressure (BP) and this may be beneficial to maintain cerebral perfusion pressure (CPP). CPP is the difference between mean arterial pressure and ICP. This elevation in BP may also put the patient at increased risk of extension of the hemorrhage.

Nishiyama et al.⁹¹ was a small study, which did demonstrate a decrease in CPP. In this study this reduction did not lead to any adverse clinical outcomes suggesting some concern regarding safety; albeit with a very limited population. Current American Heart Association writing group guidelines suggest that the management of elevated blood pressure in patients with ICH be individualized based on the patient's age, history of hypertension, presumed cause of hemorrhage and interval since onset.²⁵¹ For patients with a history of hypertension it is recommended that the mean arterial pressure be maintained at less than 130 mmHg by this writing group.

In spite of wide applicability and potential simplicity of antihypertensive treatment after intracerebral hemorrhage, the data available to informed clinical decision making is extremely poor. Choice of agent, modality of treatment and timing all remain open questions. A new large trial which addresses some aspects of the clinical questions is required.

Intervention C: Does IA Thrombolysis Reduce Stroke-related Mortality and Disability in Adults with Acute Ischemic Stroke?

The study by Keris⁹³ used sequential randomization, suggesting that allocation concealment was incomplete. Further assessment was not blinded and the analysis was not performed on an intention-to-treat basis. These methodology issues raise questions about the conclusions of this trial. PROACT I and II demonstrate the possibility of using IA therapy to establish reperfusion. Further, PROACT II suggests that recanalization of vessels achieved within the six-hour window can result in a significant clinical improvement. Enlarging the time window for acute thrombolysis would potentially increase the number of patients treated. Of concern are the very large number of patients who need to be screened and subjected to angiography to achieve this result. Interventional therapy is thus a very resource-intensive course to follow in acute stroke. More accessible forms of acute imaging to establish vessel occlusion such as CT or MR angiography may be helpful in improving on the odds of receiving treatment after screening. Such noninvasive methods might decrease procedure related complications and enhance safety by eliminating individuals without vessel occlusion from potentially hazardous treatment albeit at a cost of adding time to treatment.

The EMS Bridging Trial⁹⁴ is a Phase I trial demonstrating the feasibility of combined treatment, combining IV and IA treatment as the possible advantage of reducing the time to canalization. This study demonstrates feasibility and results of a larger trial currently underway is awaited.

A new trial of intra-arterial treatment would be helpful in patient selection. In particular the timing since symptom onset, thrombus location and criteria for combining with IV therapy remain to be elucidated.

Intervention D: Does Treatment to Normalize Blood Glucose Levels Reduce Stroke-Related Mortality and Disability in Adults with Acute Stroke?

Weir and colleagues analyzed the outcomes of 750 non-diabetic patients in an attempt to correlate plasma glucose with clinical outcomes.²⁵² Of some 750 patients included in the model, 86% had ischemic stroke while the remainder were cerebral hemorrhages. Hyperglycemia with serum glucose greater than 8 mmol/L predicted a poor chance of survival and independence. This effect was persistent after adjusting for age, stroke type and severity. This observation suggests that elevated serum glucose is more than a marker for severe vascular event. As hypoglycemia can be a significant detrimental influence on neurologic outcome, randomized trial evidence should be obtained prior to advocating specific treatment protocols particularly in the setting of modest elevations of plasma glucose.

Neither of the included studies^{95, 96} addressed our question on whether treatment to normalize blood glucose levels would reduce stroke related mortality and disability in adults with acute stroke. The study of Gray et al.⁹⁵ did demonstrate that without specific intervention, plasma glucose values do decrease 24 hours after stroke but we still do not know whether treatment will lead to improved outcomes.

Current consensus guidelines in the absence of evidence recommend lowering glucose levels above 16.63 mmol/L.²⁵³ Glucose values decreased 24 hours after stroke; however, it is unknown whether treatment will lead to improved outcomes. Further trials examining this issue are required.

Intervention E: Does Mechanical Thrombus Disruption Reduce Stroke-Related Mortality and Disability in Adults with Acute Ischemic Stroke?

Both included studies compared continuous ultrasound monitoring during acute thrombolysis with no such monitoring.^{97, 111} The observation of early recanalization is consistent across both of these studies. Eggers et al. reports a benefit in the small group studied (n=25) for treatment only in the Barthel Index, with no significant benefit in mortality or mRS score.⁹⁷ Alexandrov¹¹¹ demonstrated superior efficacy for early recanalization and the prespecified endpoint of early dramatic recovery (decrease in the NIH Stroke Scale Score of 10 points on a NIH Stroke Scale Score of less than or equal to 3), but no significant differences between the two groups at 3 months. While these results are encouraging, a larger trial to ensure both the safety and efficacy of this treatment modality is required. Particularly critical is establishing an enduring benefit by reducing mortality, disability or both at 3 months. Alexandrov¹¹¹ points out that the results may not be generalizable as use of the modality requires skill and experience. Other mechanical modalities including balloons, suction devices, and mechanical clot extraction modalities have yet to report randomized results.

A phase I study of the MERCI mechanical embolus retrieval device was published recently.²⁵⁴ Twenty eight patients were studied with a mean time from onset to treatment of just over 6 hours. Successful recanalization was achieved in 43% patients using embolectomy alone. This increase of 64% when intra-arterial tPA was added. This early publication is of interest and further study of these devices is warranted. In the small number of papers published, patient selection and interaction with other modalities such as intra-arterial tPA remains unclear.

Future well-designed studies investigating mechanical thrombus disruption to reduce stroke-related mortality and disability are needed.

Intervention F: Is the Effectiveness and Safety of Thrombolytic Therapy for Adults with Acute Ischemic Stroke Affected by Time from Onset to Treatment?

Hacke and colleagues, writing for the ATLANTIS, ECASS, and NINDS tPA study group investigators, provided a patient-level meta-analysis of the relation between onset to treatment time and outcome for thrombolysis.¹⁹⁵ Data from six trials that had treatment windows between 0 and 6 hours were obtained—NINDS Parts 1 and 2, ECASS 1 and 2, and ATLANTIS Parts A and B.¹⁹⁵ These included trials encompass 99% of patient data available from RCTs of tPA. The trials varied in the time inclusion criteria. Both NINDS trials were restricted to patients treated between 0 and 3 hours. The ECASS trials included patients treated between 0 and 6 hours. ATLANTIS A enrolled from 0 to 360 minutes while ATLANTIS B initially recruited from 0 to 300 minutes, and subsequent to the publication and acceptance of the NINDS data was narrowed to 180 from 300 minutes. Inclusion and exclusion criteria in the trials were similar. A diagnosis of acute ischemic stroke with CT evidence refuting hemorrhage was required. Patients at high risk of bleeding with thrombolytic therapy, including those who had had trauma or recent surgery, were excluded. Both the ECASS trials and ATLANTIS B excluded patients with early infarct signs on the baseline CT scan. This was not a requirement in the NINDS trials. Favorable outcome for the purposes of this analysis was defined as scores on the Rankin Scale of 0 or 1, Barthel Index of 95 to 100, and a NIHSS score of 0 or 1. These ranges represent minimal or no disability. A logistic regression model was constructed to assess the relationship between onset to treatment time and favorable outcome at 90 days. A multivariable logistic regression model was constructed including possible confounding variables.

The final results were presented as an intention-to-treat analysis including patients on whom data was incomplete (n=2,775). The median age of this group was 68 years; 84.6% were White (non-Hispanic), 9.1% Black, 2% Hispanic, 0.9% Asian, with the remainder being from other ethnic backgrounds or unreported. The median baseline NIH Stroke Scale Score was 11 with a median onset to treatment time of 243 minutes. Of note, 67% of patients were treated beyond the 3-hour window. The final model included treatment, onset to treatment time, age, blood glucose, admission NIHSS, baseline diastolic pressure, interaction between age and NIHSS, and interaction between onset-to-treatment time, and treatment. An association was found between onset-to-treatment time and outcome. The odds ratio for favorable outcome with tPA treatment in the 0 to 90 minute interval was 2.81 (95% CI 1.75–4.50). This decreased to 1.15 (0.90–1.47) in the 271 to 360 minute interval. The 95% confidence interval of the adjusted odds ratio for favorable treatment remained above 1, indicating benefit for treatment until 270 minutes (4.5 hours) after onset of symptoms. Likewise, the adjusted odds ratio for death exceeded 1 only in the interval between 271 and 360 minutes. Interestingly, there was not a strong association between 3-month favorable outcome and baseline NIHSS in any time stratum. This also confirms a strong association between stroke outcome and time to treatment. Of significance, the relationship holds throughout the currently used therapeutic window of 0 to 3 hours.

Parenchial hematoma were seen in 5.3% of patients compared to 1.1% of placebo patients in multivariate modeling for hemorrhage, which included onset to treatment time (OTT), age, and NIHSS scores. The OTT and treatment interaction was not significant. Use of tPA and age increased the probability of hemorrhage but OTT and baseline NIHSS scores did not.

The prospective studies identified in this review did not prospectively validate treatment beyond the 3-hour window.^{102, 116} The study characteristics did not allow for a useful combination of results. Thus, this patient-level meta-analysis provides the best current summary of the evidence for a relationship between time to intervention and clinical outcomes.

Thus, it is imperative that systems be designed to provide treatment as soon as possible after onset of symptoms rather than aiming for treatment within 3-hours of onset. In addition, there is suggestion of benefit beyond the 3-hour window. This benefit is smaller than that in earlier time frames consistent with the overall results of the analysis. It is entirely possible that significant benefit accrues to a subgroup of patients treated beyond 3 hours. The suggestion of benefit in later time windows will require validation from prospective studies that may also provide the defining characteristics of the group most likely to benefit or conversely be harmed. Since a treatment effect in these later time windows is smaller than that observed earlier, larger groups of patients will need to be randomized.¹⁹⁵

In Canada, tPA for stroke was commissioned in 1999. The approving body required a prospective registry for safety in the context of routine care. Over 2.5 years, 1,135 patients were enrolled in 61 centers in Canada. For all patients 90 day outcome and hemorrhage rates were determine, along with baseline and 24 hour CT scans which were viewed centrally.

The median NIHSS scores were identical to that seen in the NINDS trial. Thirty six percent of the patients made a complete functional recovery and returned to the baseline state at 90 days. This result was not significantly different than the NINDS trial (p=0.15). The rate of symptomatic intracranial hemorraghe was 4.6% compared to the NINDS rate of 6.4%.²⁵⁵

Intervention G: Do Pretreatment CT Scoring Systems Affect the Safety and Efficacy of Thrombolytic Therapy for Acute Ischemic Stroke?

Prospective evaluation of CT scoring systems was not available, and both included studies are evaluations of CT's conducted during the course of prospective trials of thrombolysis in stroke. Barber and colleagues²⁵⁶ evaluated 203 consecutive patients treated with thrombolytic therapy within 3-hours of prospective scoring of their CT scans according to a 10-point scale based on an unenhanced axial CT scan. The value was calculated from two cuts within the MCA territory. A normal scan obtained a score of 10 points with one point subtracted for each area of early ischemic change. The ASPECTS Score correlated with baseline NIH Stroke Scale Score in an inverse manner (Spearman's rho = -0.56, p < 0.001). An ASPECTS Score of less than 7 had strong predictive value for death or dependence. With dependence defined as a Rankin Score of 3 to 5, the odds ratio for good functional outcome was 82 (95% CI 23–290). Similarly, an ASPECTS Score of less than or equal to 7 was predictive of the presence of symptomatic hemorrhage (odds ratio 14, 95% CI 1.8–117). Good correlation was noted between the ASPECTS Scores of pairs of stroke neurologist/radiology trainees and the neuroradiologists (kappa = 0.85, 0.71, 0.89 respectively).

Retrospective application of the ASPECTS score to CT scans used in the PROACT II Trial suggests that those with an ASPECTS Score greater than seven were three times (OR 3.2, 95% CI 1.2–9.1) more likely to have an independent functional outcome when compared with controls, whereas, those with a score less than or equal to seven were much less likely to do so.

The wide availability of CT scanners along with the ease of use of the ASPECTS Score suggests this may become an important tool in treatment decisions.

Intervention H: Do Pretreatment MRI Scoring Systems Affect the Safety and Efficacy of Thrombolytic Therapy for Acute Ischemic Stroke?

The time frames used in patient selection for thrombolysis represent surrogate measures of viable tissue. Perfusion and diffusion weighted MRI scanning raise the possibility of patient selection on the basis of the demonstration of viable tissue with impaired perfusion. Perfusion weighted imaging (PWI) allows visualization of altered areas of blood flow.²⁵⁷ Energy failure with the resulting inability to maintain water and ionic gradients results in decreased diffusion of water in the brain.²⁵⁸

Butcher and colleagues²⁵⁹ have suggested that quantitative PWI mapping may be useful in predicting the fate of tissue. Thirty-five patients with acute stroke, 17 of whom were treated with tPA, were imaged within 6 hours of onset. The mean transit time was found to be prolonged in infarcted areas relative to salvaged areas (p=0.001) with an approximately 10% reduction in regional cerebral blood flow (p=0.01). From this study, the mean transit time appeared to be the perfusion measure most predictive of outcome though delineation of absolute perfusion thresholds resulting in infarction could not be accomplished. This may, in part, be due to the dependence not only on absolute perfusion but also the duration of hypoperfusion. The interaction between these two measures is difficult to arrive at through the use of a single image during the course of an infarction.

Nevertheless, MRI definition of salvageable tissue versus tissue irreversibly committed to infarction when applied prospectively has the potential to enhance both the safety and efficacy of thrombolysis.

Intervention I: Do CT Perfusion/Angiography Affect the Safety and Efficacy of Thrombolytic Therapy for Acute Ischemic Stroke?

Prospective use of CT perfusion and angiography techniques in patient selection for thrombolysis was not identified. The two studies included demonstrate correlation of outcome with findings on CT angiography or CT perfusion techniques including Xenon CT. CT angiography permits the identification of large vessel occlusion without the risk of invasive techniques and thus has potential applicability in identifying patients who may benefit from invasive forms of revascularization (IA thrombolysis or thrombus retrieval). As noted above for Intervention H, the availability of physiologically-based criteria to select individuals with salvageable tissue presents the opportunity for significant benefit. In this regard, a CT is cheaper and more widely available than MRI scanning. Further MRI scanning has the potential to add increased time prior to treatment and decrease numbers available for treatment due to MR exclusion criteria, which include ferromagnetic foreign substances, claustrophobia and the requirement to closely monitor unstable patients.

The impact of the study by Kilpatrick¹⁰² is limited by the small number of patients and biases inherent in selecting this particular group of patients for study. Nevertheless, the hypothesis that low cerebral blood flow as demonstrated on Xenon CT or major vascular occlusion is associated with infarct, and presumably worse clinical outcome, deserves more detailed prospective study as this may aid in singling out patients who may benefit from more aggressive therapy and furthermore those in which the risks of more aggressive therapy are warranted.

Intervention J: Are Community Education Programs Effective in Reducing Stroke-Related Disability and Mortality?

Both included studies demonstrate that multi-faceted educational and system changes for the provision of acute stroke care increased utilization of tPA. Given the complexity of this intervention we are unable to determine what components of this intervention led to the increased utilization. It remains unclear if community education programs are effective in improving outcome in acute stroke. A number of similarities exist with the care of acute cardiac ischemia and it is instructive to examine the parallel and data in this field for relevance to stroke.

Ischemic heart disease is the leading cause of death in Canada and other industrialized countries.²⁶⁰ Myocardial death under ischemic conditions is a time dependent phenomenon requiring a series of biochemical events and is not simultaneous with the onset of ischemia.²⁶¹ As is the case in stroke, this time dependence provides an opportunity for therapeutic intervention and early treatment of myocardial infarction has been shown to decrease infarct size and reduce mortality.^{262, 263} Rapid access to emergency treatment requires prompt identification of symptoms by patients or bystanders and presentation to an appropriate facility. Reperfusion of ischemic tissue can then be achieved through thrombolysis or angioplasty with the effectiveness of therapy being contingent on speed of delivery.^262–264 Patient delay in seeking care in North America is several hours and is felt to contribute to lower than expected outcomes from the application of available procedures.²⁶⁴ Prompt and appropriate application of such resource intensive procedures might be expected to improve outcomes. Increased awareness of symptoms and need for urgency amongst patients might result in reduced delay and increase rates of thrombolytic treatment.

As acute myocardial infarction is the presenting symptom of ischemic heart disease in a substantial percentage of patients, a campaign aimed at unselected members of the community would be expected to reach the greatest members of people expected to benefit. Previous experience with community interventions in this area spans two decades, far longer than the experience in acute stroke. The REACT trial, reported by Luepker, was a RCT of the efficacy of community intervention on delay time employing cluster randomization of twenty U.S. cities into matched control and intervention pairs.⁶⁹ The design of the intervention was complex and involved expects in health behavior and epidemiology. The process of intervention design and its theoretical framework is extensively described in a companion publication.²⁶⁵ ED staff in study hospitals were trained in standardized questioning of patients regarding the nature and onset of acute symptoms. Matched pairs of cities were comparable in age distribution, education level, ethnic distribution, median income and baseline median delay time from symptom onset to presentation. Delay times were log transformed to obtain a more normal distribution. The trend of delay time was calculated by linear regression of log delay against calendar time adjusted for the patient level covariates: age, sex and past history of coronary artery disease. The trends were then compared pair wise with the matched communities. Delay times were available for 73% of the population of interest. A baseline period of four months was compared to the intervention period of eighteen months. The study had an 80% power to detect an end of trial difference of 30 minutes between intervention and control groups. During the study period mean delay time dropped by about 10 minutes in both the reference and intervention groups. The primary outcome response, which was the slope difference (%/yr) between intervention and reference groups, was 2.3% (95% CI -5.5 to 10.8). The intervention was consequently felt to be ineffective.

This result in a methodologically superior design to those carried out in the field of acute stroke is of concern. Application of ineffective strategies, at a minimum, diverts resources away other applications. Further, there is the possibility that a campaign which increases the number of patients presenting with potential stroke increases the probability of false positive diagnoses and, thereby, worsen outcomes.

Intervention K: Are Designated Centers Effective in Reducing Stroke-Related Disability and Mortality?

Studies evaluating the brain attack coalition definition of a stroke center were not identified. Thus, each study reviewed in this section had a different approach in development of a stroke center. Each study did demonstrate decreased time intervals in the treatment continuum from presentation through the hospital stay. The percentage of patients treated is a significant surrogate measure of outcomes and, assuming the treatment meets the NINDS protocol, one might infer that improving the percent of patients treated would result in improved outcomes. None of these studies demonstrated adequately that their strategy led to decreased stroke morbidity and mortality. At present, the potential effect of the establishment of a stroke center according to the pre-specified definition is not known and waits empiric validation. The magnitude of the effect may depend on the pre-existing infrastructure, geography and demographics of the area served. While the need for such a validation may not be readily apparent it impacts significantly on resource allocation in a competitive economic environment and has broader implications for public health.

One possible alternative with Stroke Centres strategy is dispersing the expertise through the use of Telemedicine. Technology currently exists which permits the visualization and examination of both the patient and the scan at a location remote from the examining physician. While this would transfer physician expertise in the treating center (i.e. nursing, radiology, technology, and/or laboratory support). Comparisons of each approach would be helpful in further stroke system planning.

Intervention L: Are ED Protocols for the Management of Acute Stroke Effective in Reducing Disability and Mortality?

At this time, we were unable to identify any studies that addressed our research question of whether ED protocols for the management of acute stroke are effective in reducing disability and mortality. The studies by Smith et al.¹⁰⁶ and Jahnke et al.¹¹⁸ evaluated ED protocols and document the time to treatment with thrombolytics. Smith et al.¹⁰⁶ compared the times to treatment in patients in their study with historical data and conclude that their ED protocol yields similar times to treatment than previously published in the original trials²³¹ Jahnke et al.¹¹⁸ compared the time to thrombolytic treatment between those who received treatment in the first year of the protocol and the second year. They document an improvement in time to treatment in the second year of the protocol though outcomes are not reported. The last study by Akins et al.¹¹⁷ compared patients who had their treatment initiated by the ED physician versus the neurologist. The patients had similar presentations and baseline risks and there were no differences in outcomes between these two small groups. Initially they identified more protocol violations by the ED physicians as well as a shorter time interval to treatment by the neurologist. This study did not demonstrate any difference in outcome between the ED physician treatment and neurologist treatment but the study size was very small and with the increased protocol violations and increased length to treatment there is the possibility that in a larger study worse outcomes would be found with this model.

The studies available were limited in methodology and outcomes assessed. In particular, the clinically significant outcomes of mortality and morbidity were not examined.

Limitations

Our literature searches were restricted to English language publications. Although, this limited our review it is unlikely that it biased the results in any meaningful way.^{74, 75}

Our levels of evidence “guiding principle” was to limit our review to include reports of RCTs, whenever possible. There is debate within the literature as to the merits of excluding reports of high quality observational studies from addressing questions within a systematic review.^266–269 There is little empirical data as to whether the inclusion of such studies introduced bias into the results. We elected to take the more conservative scientific position of providing ‘bias free’ estimates of effectiveness whenever possible. Quality assessing observational studies, and designs is problematic. There is no published validated scheme available. We used the NOS for assessing quality, and while we believe it has some appropriate psychometric properties, the index itself is unpublished. There is some data which suggests that assessments using unpublished systems as a grading system might introduce bias.²⁷⁰

In spite of the clear indication that early time to treatment improves outcomes and the complexity of the multilevel interventions required to accomplish these targets, a limited number of studies were available for review in all areas examined in our report. This may reflect a bias, which does not equate system interventions with medical therapies. With the significant exception of thrombolytic therapy for acute ischemic stroke, methodologic inadequacies were apparent in all topic areas. Further detailed information on the influence of patient characteristics (e.g., race) on outcome was either unavailable or unreported.

We identified a meta-analysis examining surgical intervention for ICH that was published in 2002,²³³ which presented data from two trials that were not identified by our search strategy. Our search did not capture these references, since the study by McKissock et al. was published in 1961, prior to our search date of 1964. The study by Chen (1992) is a non-English language citation and hence, was not captured since our search was limited to English language citations. Chen et al. was a large RCT investigating surgical intervention for ICH. We were able to extract enough data from the meta-analysis pertaining to mortality outcomes to include this data in our meta-analysis as a sensitivity check (Figure 4). We were less inclined to extract data from the McKissock et al. results since it was determined that the surgical methodology for ICH had substantially advanced since the time that this report was published and hence, would have been heterogeneous with other studies included in our review.

Five studies examined IA thrombolytic therapy for ischemic stroke. Due to heterogeneity of interventions and comparisons, we obtained pooled estimates based on only two studies.^{92, 110} We excluded Keris et al.⁹³ because they did not provide 90-day follow-up assessment data and there were concerns regarding methodology of the trial. Lewandowski et al.⁹⁴ randomized patients to IV tPA and IA tPA or to IV placebo and IA tPA. So the contrast between the two groups is IV tPA versus IV placebo. Conceptually, this study was not combinable with the other studies.

The effect measure selected for pooling the outcomes of death and death and disability was the odds ratio. Other effect measures suitable for dichotomous outcomes including the relative risk and absolute risk reduction could also be considered, and might show different results. However, the odds ratio has several statistical advantages, one being it has been shown to frequently exhibit less heterogeneity than other effect measures.²⁷¹ In addition, the odds ratio was used in a previous meta-analysis, which could facilitate comparison.

For surgical interventions for ICH, there was evidence of some statistical heterogeneity, although not statistically significant. It should be noted, however, that tests of heterogeneity in meta-analysis are recognized to have limited power, particularly when the number of studies is small.²⁷² We used random effects methods to adjust for statistical heterogeneity, which results in wider confidence intervals for the pooled estimate. Furthermore, random effects pooled estimates should be interpreted with some caution.

Research and Clinical Implications