XVII. European Stroke Conference
Nice, France

Oral Session:

Meta-analysis and review papers  
Date: Friday 16 May 2008  
Time: 8:00 - 8:10 -  Room: Euterpe
Chair: A. Algra, The Netherlands and P. Sandercock, United Kingdom

01
Publication bias in animal studies leads to overstatement of neuroprotective efficacy
E.S.Sena    M.R.Macleod    D.W.Howells                                                 
on behalf of the CAMARADES collaboration

University of Edinburgh

UNITED KINGDOM

Background – Over 500 drugs have efficacy in animal models of stroke, but no candidate neuroprotective drug has successfully made the transition from bench to bedside. Systematic review and meta-analysis can provide a robust assessment of the evidence for efficacy. However, if neutral and negative studies remain unpublished they are less likely to be identified in systematic review, and such publication bias will lead to the overstatement of neuroprotective efficacy in meta-analysis. The scale of publication bias is unclear, and often the number of studies in any individual meta-analysis is too small to provide a robust assessment of whether publication bias is likely. Our aim here is to seek evidence for publication bias in previously conducted systematic reviews of candidate neuroprotective drugs. Methods – 11 of 15 drug datasets held on the CAMARADES database were large enough to allow a reasonable analysis of publication bias. For each experiment, effect size and its standard error were extracted. Publication bias was identified using funnel plotting and Egger regression. We used METATRIM (a non-parametric method enabled in STATA) to estimate efficacy in the absence of publication bias. Results – Egger Regression of the complete dataset (describing 991 experiments involving 14792 animals) was consistent with the presence of a substantial publication bias. For individual drugs, imprecise studies effects consistent with publication bias were seen for NXY-059, IL1-RA, Melatonin and FK506. Using METATRIM, overall efficacy was reduced from 31.8% (95% CI 30.0 to 33.6%) to 25.6% (95% CI 23.7 to 27.5%). For hypothermia, METATRIM suggests a reduction in efficacy from 43.5% (95%CI 40.1 to 47.0) to 35.6% (95%CI 31.9 to 39.3). Discussion – These data suggest a publication bias may lead to an absolute overstatement of neuroprotective efficacy of around 5%. Measures to promote publication of all data, and systems to register unpublished neutral or negative studies, may lead to improvements in translation of neuroprotective efficacy from bench to bedside.

Meta-analysis and review papers  
Date: Friday 16 May 2008  
Time: 8:10 - 8:20 -  Room: Euterpe
Chair: A. Algra, The Netherlands and P. Sandercock, United Kingdom

02
A Systematic Review of Stem Cell Therapies in Animal Models of Focal Cerebral Ischemia
J.S.Lees    E.S.Sena    D.W.Howells    S.A.Koblar    M.R.Macleod                                   
 

University of Edinburgh

UNITED KINGDOM

Background Stem cells have been proposed for the treatment of diverse neurological conditions including stroke. For many candidate stroke treatments, efficacy in animal models has not been replicated in clinical trials. This may be due to problems in the conduct and reporting of animal studies and/or in the design of clinical trials. A robust and systematic review of the conditions under which efficacy is observed in animals is desirable in the development of new treatments. Methods Systematic review with searching of 4 online databases to identify reports of experiments using stem cells in animal models of focal ischemia, reporting outcome as infarct volume (IV) and/or neurological score (NS). Effect sizes were calculated using DerSimmonian and Laird random effects meta-analysis. Statistical significance of differences between groups was assessed by partitioning of heterogeneity. Results Electronic searching identified 4111 publications of which 85 met our pre-specified inclusion criteria. IV and NS were reported by 55 and 76 papers respectively. IV was improved by 28.8% (95% CI: 24.5–33.2); NS by 34.1% (95% CI: 29.0-39.3). Effect size increased with cell dose and reduced when outcome was assessed at later time points. The effect on IV, but not NS, diminished with increasing delays to treatment. Median study quality score was 4/10 (IQR 3-5): poor quality studies reported greater improvements in outcome than did higher quality studies. Randomisation and blinded outcome assessment were reported by 25% and 24% of publications respectively. Discussion Notwithstanding concerns regarding the quality of some of this literature, these data suggest that stem cell based therapies have a potential role in the treatment of stroke, even when delivered some time after stroke has occurred.

Meta-analysis and review papers  
Date: Friday 16 May 2008  
Time: 8:20 - 8:30 -  Room: Euterpe
Chair: A. Algra, The Netherlands and P. Sandercock, United Kingdom

03
Systematic review and meta-analysis of validations of the ABCD and ABCD2 scores in prediction of stroke risk after transient ischaemic attack
M.F.Giles    P.M.Rothwell                                                        
 

University of Oxford

UNITED KINGDOM

BACKGROUND: Stroke risk is high in the hours and days after TIA. The ABCD and ABCD2 scores are used to predict this risk and aid triage. However, although these scores have been validated in Oxford and California, and performed well, it is important that their performance in other truly independent validations is assessed. We have therefore systematically reviewed all validation studies of the ABCD and ABCD2 scores. METHODS: Articles reporting the validation of either score were identified from MEDLINE, EMBASE (to November 2007), relevant reference lists and reviews. All independent cohorts of TIA patients reporting stroke outcome stratified by score were included. Receiver operating characteristic (ROC) curves were calculated to determine predictive power. RESULTS: 11 articles reported validations of one or both scores in 13 cohorts of TIA patients, including 5938 subjects with 332 strokes at 7 days. Pooled estimates of the areas under the ROC curves for the ABCD and ABCD2 respectively were 0.70 (0.66- 0.73) and 0.70 (0.66- 0.74) for 7-day stroke risk, 0.68 (0.65- 0.71) and 0.69 (0.66- 0.72) for 90-day risk, and 0.64 (0.60- 0.69) and 0.65 (0.61- 0.70) for 8-90 day risk. Predictive power was independent of clinical setting (emergency departments, specialist neurovascular units, and population based studies), but was greater in two additional cohorts that included patients with both suspected and confirmed TIA, (ROC for 7-day risk – 0.9 (0.81-0.99) and 0.75 (0.61-0.88)). CONCLUSION: The ABCD and ABCD2 scores reliably predict stroke risk within the first 7 days after TIA regardless of clinical settings, but perform less well for strokes in the subacute phase.

Meta-analysis and review papers  
Date: Friday 16 May 2008  
Time: 8:30 - 8:40 -  Room: Euterpe
Chair: A. Algra, The Netherlands and P. Sandercock, United Kingdom

04
Elevated troponin after stroke – a systematic review
G. Kerr    G. Ray    O. Wu    D.J.Stott    P. Langhorne                                   
 

Academic Section of Geriatric Medicine, Glasgow University

UNITED KINGDOM

Background Circulating markers of myocardial damage (especially troponins) can be elevated in acute stroke patients. However it remains unclear what the underlying pathophysiology is, which patients are at risk, and whether this is a marker of poor outcome. Methods We carried out a systematic review of published observational studies that examined troponin I or T levels (checked within 1 week) in acute stroke inpatients (either ischaemic or haemorrhagic). EMBASE, MEDLINE, CENTRAL and CDSR were searched using the terms Troponin, Stroke and Cerebrovascular Accident. Two independent reviewers extracted the data. Those with and without raised Troponin levels were compared in a meta-analysis. Results We initially identified 171 citations; 145 were deemed not relevant, 6 were reviews and 5 did not fulfil the criteria above. This left 15 studies (2768 subjects) for full review. Overall 18.8% (95% CI 14.1-23.5) of acute stroke patients had an elevated troponin I or T. On univariate analysis this group were more likely to have an ischaemic ECG (OR 3.0; 1.5-6.2); 43% of those with raised troponin had an ischaemic ECG compared to 22% of those who did not. Elevated troponin was associated with increased mortality (OR 6.0; 3.1-11.6) during study follow-up. Discussion Approximately 1-in-5 of hospitalised acute stroke patients have an elevated troponin, consistent with acute myocardial damage. This finding is associated with an increased prevalence of ischaemia on the ECG and it is likely that a significant proportion of these patients have underlying ischaemic heart disease. It is not clear whether the increased mortality rate with elevated troponin is due to associated cardiac disease or some aspect of the acute stroke event.

Meta-analysis and review papers  
Date: Friday 16 May 2008  
Time: 8:40 - 8:50 -  Room: Euterpe
Chair: A. Algra, The Netherlands and P. Sandercock, United Kingdom

05
Comparisons of different systems of organised (stroke unit) care: a secondary analysis of systematic review
L. Govan    C.J.Weir    P. Langhorne    A.E.Ades                                          
on behalf of Stroke Unit Trialists Collaboration

University of Glasgow

UNITED KINGDOM

Background: Although organised inpatient (stroke unit) care is well established in acute stroke management it has not been established which system of care is most effective. We explored whether any one system of care is most effective in improving patient outcomes. Methods: We updated the Stroke Unit Trialists Collaboration systematic review that currently contains 32 clinical trials (7194 subjects). Six care systems were identified: semi-intensive; comprehensive; rehabilitation; mixed-rehabilitation; mobile stroke teams; and general medical wards. Our main outcomes were death, dependency, and need for institutional care. We also extracted subgroup data for age and severity. Using data from direct and indirect comparisons and Bayesian hierarchical models we estimated odds ratios for outcomes in treatment wards versus general medical wards. Results: Rehabilitation stroke units showed a reduction in death (OR 0.60; 95%CI 0.41-0.87) and non-significant reduction in death or dependency (0.66; 0.42-1.05); comprehensive stroke units showed non-significant reductions in death (0.84; 0.70-1.01); and semi-intensive stroke units showed reductions in death or dependency (0.57; 0.33-0.99). Older patients and more severe strokes were at greater risk of poor outcome. After adjusting for these subgroups in the subset of studies in which they were recorded the effects of systems of care were no longer significant. Discussion: Rehabilitation, comprehensive and semi-intensive stroke units improve patient outcomes compared to general medical wards. However these effects are less significant when age and severity is accounted for.

Meta-analysis and review papers  
Date: Friday 16 May 2008  
Time: 8:50 - 9:00 -  Room: Euterpe
Chair: A. Algra, The Netherlands and P. Sandercock, United Kingdom

06
Plasma Triglyceride Level and Risk of Stroke : A Systematic Review of Epidemiological Studies
J. Labreuche    P.-J. Touboul    P. Amarenco                                                 
 

INSERM U-698 and Denis Diderot University-Paris VII Department of Neurology and Stroke Center, Bichat Hospital

FRANCE

Background and Purpose —The etiological role of lipids in stroke risk remains uncertain and little attention has been paid to the effect of triglycerides. The objective of present study was to summarize the epidemiologic evidences for triglycerides exerting a detrimental effect on stroke risk. Methods— Systematic PubMed literature search of epidemiological studies that examined the association between triglycerides and stroke. We calculated a pooled estimate of relative risk (RR) of stroke per one standard deviation (SD) increase in triglycerides among the prospective cohort studies. Results—We identified and selected 17 prospective cohort studies that enrolled 140 788 subjects followed for 3 to 18 years with 4144 incident strokes, and 14 case-control studies that enrolled 3422 cases and 4125 controls. Although study design varied largely (population characteristics, blood-sample fasting status, stroke endpoint definitions, statistical models), nine of prospective cohort studies (Figure) and 6 of case-control studies supported a significant association of elevated triglyceride level with an increase risk of stroke. Of the negatives studies, two prospective studies and one case-controls studies showed an opposite trend. Based on all available data from prospective studies (n=9), the combined RR of main stroke outcome per one SD increase was 1.10 (95% Confidence Interval, 1.07-1.13) with no heterogeneity. In case-control studies, odds ratio of stroke increased inversely with mean triglyceride levels in control groups. One prospective study and one case-control study reported negative relationship with hemorrhage stroke. Conclusions— The weight of evidence favors a positive association between triglycerides and stroke in the prospective cohort studies and supports the evaluation of triglyceride-lowering therapy in stroke prevention. However, this review shows the need for additional large prospective studies, especially in stroke subtype, to firmly establish an association between triglyceride levels and stroke risk.

Meta-analysis and review papers  
Date: Friday 16 May 2008  
Time: 9:00 - 9:10 -  Room: Euterpe
Chair: A. Algra, The Netherlands and P. Sandercock, United Kingdom

07
Meta-analysis of Randomised Controlled Trials (RCTs) of Surgery Versus Conservative Treatment in Spontaneous Supratentorial Intracerebral Haemorrhage
B.A.Gregson    A.D.Mendelow                                                        
International Intracerebral Haemorrhage Group

Newcastle University

UNITED KINGDOM

Background: Treatment for spontaneous intracerebral haemorrhage (ICH) remains anecdotal and inconsistent. A number of RCTs have been undertaken in recent years in an attempt to gain a rigorous answer but the evidence remains controversial. Objective: The objective of this study was to pool data from all trials of surgery versus conservative treatment in spontaneous intracerebral haemorrhage (ICH) since 1985 in order identify subpopulations of ICH patients and test their response to surgery. Methods: The authors of published papers were contacted by mail, email and at conferences. Full data sets have been supplied by six authors (1640 patients) and grouped data by a seventh author (100 patients). Data supplied included patient’s age and sex, GCS at presentation, volume and site of haematoma and presence of intraventricular haemorrhage (IVH), method of evacuation of haematoma and outcome at three to six months. Results These data demonstrate very different patient groups and different methods of haematoma removal. Some trials included patients with lower GCS and larger haematomas, others patients with higher GCS and smaller haematomas. The age group of the patients varied from trials with 70% aged less than 60 to those with more than 70% older than 60. Meta-analysis showed that there was a trend towards a favourable outcomes with surgery compared to conservative treatment in patients with lobar haematomas (n=509) (OR 0.72 CI 0.49, 1.06) or no IVH (n=1126) (OR =0.77 CI 0.60, 1.00). However for patients with a lobar haematoma and no IVH (n=328) surgery was most effective OR = 0.57 (CI 0.36, 0.91). Conclusions This work was instrumental in identifying the subgroup of patients for whom surgery would provide an appropriate treatment option: those with lobar haematomas and no IVH. This subgroup is clinically intuitive and is the treatment group being studied in STICH II.

Meta-analysis and review papers  
Date: Friday 16 May 2008  
Time: 9:10 - 9:20 -  Room: Euterpe
Chair: A. Algra, The Netherlands and P. Sandercock, United Kingdom

08
Association between hormone replacement therapy and subsequent arterial and venous vascular events: a meta analysis
G.M.Sare    L.J.Gray    P.M.W.Bath                                                 
 

Institute of Neuroscience, University of Nottingham

UNITED KINGDOM

Background: Observational studies have suggested that hormone replacement therapy (HRT) may be beneficial in the prevention of arterial thrombotic events. However, randomised controlled trials (RCT) have shown that the risk of stroke and venous thromboembolism (VTE) is increased with HRT; the effect on coronary heart disease (CHD) remains unclear. Methods: RCTs of HRT were identified from electronic searches and reference lists. Event rates for cerebrovascular disease, CVD, (stroke and TIA), CHD (myocardial infarction, unstable angina and sudden cardiac death) and VTE (pulmonary embolism, deep vein thrombosis) were extracted and analysed using Stata statistical software. Sensitivity analyses were performed by type of HRT: oestrogen alone versus combined oestrogen and progesterone, and subject age. Results: 35 trials (62,079 subjects) were identified. HRT was associated with increases in stroke (odds ratio, OR, 1.20, 95% confidence intervals, CI, 1.07-1.34) and VTE (OR 1.84, 95% CI 1.40-2.40). In contrast, CHD events were not increased (OR 0.99, 95% CI 0.90-1.10). Ordinal analyses confirmed that stroke severity was increased with HRT (OR 1.19, 95% CI 1.06-1.34). Age did not significantly affect risks of CVD, CHD or VTE. The addition of progesterone to oestrogen doubled the risk of VTE events. Conclusions: HRT is associated with an increased risk of stroke, stroke severity, and VTE events, but not of CHD events. Although most trials studied older patients, increased risk was not related to age. Combined HRT increases the risk of VTE compared to oestrogen monotherapy.

Meta-analysis and review papers  
Date: Friday 16 May 2008  
Time: 9:20 - 9:30 -  Room: Euterpe
Chair: A. Algra, The Netherlands and P. Sandercock, United Kingdom

09
Variation in the PDE4D gene and ischaemic stroke: a systematic review and meta-analysis on 5200 cases and 6600 controls
S. Bevan    M. Dichgans    A. Gschwendtner    G. Kuhlenbaumer    E. Ringelstein    H. Markus                            
 

St Georges, University of London

UNITED KINGDOM

Background The identification of PDE4D as a susceptibility gene for ischaemic stroke led to a great deal of optimism in the stroke genetics community. Subsequent replication studies have failed to confirm the initial findings however, with many being small, underpowered and focused on only a subset of the significant variants originally identified in disparate populations. Meta-analysis allows these studies to be combined to determine, in a large sample size, whether the association between PDE4D and ischaemic stroke can be considered truly replicated. Methods A meta-analysis of all PDE4D variants investigated in relation to ischaemic stroke has been undertaken. Analysis of any variant appearing in 2 or more replication studies was performed, comprising 6 SNPs, allele 0 of AC008818 and the G0 haplotype. A total of 16 studies were identified totalling up to 5216 cases and 6615 controls for a single variant. Analyses were performed including all data, excluding data from the original report (a true replication analysis) and for stroke subtypes. Results No individual SNP was associated with ischaemic stroke cases. Allele 0 of AC008818 and haplotype G0 carriers were associated with increased risk (RR 1.12 1.01-1.25 p=0.03 and RR 1.18 1.05-1.33 p=0.007 respectively) but these associations became non-significant after exclusion of the original study from the analysis (RR 1.06 0.94-1.20 p=0.34 and RR 1.16 1.00-1.34 p=0.06). Analysing only Caucasians, the majority of cases studied, did not result in a significant association for any analysis, while few robust associations were identified with individual stroke subtypes. Discussion No genetic variant examined in PDE4D showed a robust and reproducible association with ischaemic stroke. Any association which may exist is likely to be weak and potentially restricted to specific populations.

Meta-analysis and review papers  
Date: Friday 16 May 2008  
Time: 9:30 - 9:40 -  Room: Euterpe
Chair: A. Algra, The Netherlands and P. Sandercock, United Kingdom

10
Stress hyperglycemia in aneurysmal subarachnoid hemorrhage: a systematic overview
N.D.Kruyt    G.J.Biessels    R. de Haan    M. Vermeulen    G.J.Rinkel    B. Curt    Y.B.Roos                     
 

Academic Medical Center

THE NETHERLANDS

Background: Hyperglycemia after aneurysmal subarachnoid haemorrhage (aSAH) seems frequent and may be related to poor clinical outcome. We performed a systematic review to characterize the level of glucose and the rate of hyperglycemia on admission and we related the latter to clinical outcome in this patient group. Methods: We included published cohort studies or clinical trials reporting on a consecutive series of patients with aSAH admitted within 72 hours. For analysis with clinical outcome, we only included studies that assessed outcome prospectively after at least three moths. The glucose level and the rate of hyperglycemia on admission was calculated by weighting for the number of patients included in each study. We pooled the odds ratio’s (OR) and 95% confidence interval (95%CI) of poor clinical outcome in patients with or without hyperglycemia. In a separate analysis we pooled the studies with a definition of hyperglycemia above or below the median. Findings: Mean admission glucose was 9.2 mmol/L (range 7.4–10.9 mmol/L; 14 studies) and the rate of hyperglycemia was 70% (range 29–100%; 17 studies). The pooled OR for poor outcome associated with hyperglycemia was: 3.46 (2.44–4.91; 8 studies). For studies with a definition of hyperglycemia below the median (7.0 mmol) the pooled OR was 2.96 (1.89–4.64). For studies with a definition above the median this was 4.10 (2.47–6.82). Interpretation: After aSAH, the level of glucose and the rate of admission hyperglycemia are high and the latter is associated with an increased risk on poor clinical outcome. A randomized clinical trial is warranted to study the potential beneficial effect of tight glycemic control.

Meta-analysis and review papers  
Date: Friday 16 May 2008  
Time: 9:40 - 9:50 -  Room: Euterpe
Chair: A. Algra, The Netherlands and P. Sandercock, United Kingdom

11
Cooling therapy for acute stroke
H.M.den Hertog    H.B.van der Worp    M. Tseng    D.W.Dippel                                          
 

Erasmus MC University Medical Center Rotterdam

THE NETHERLANDS

Background: Increased body temperatures are common in patients with acute stroke and are associated with poor outcome. Cooling therapy may therefore reduce brain tissue damage incurred by stroke. Aim: To assess the effects of cooling therapy in patients with acute stroke. Methods: We updated the 1999 Cochrane review "Cooling therapy for acute stroke". Relevant trials were identified in the Cochrane Stroke Group’s Trials Register (last search November 2007). Additional searches were performed in MEDLINE and EMBASE (January 1998 to December 2007). We included all completed randomised or non-randomised controlled clinical trials, published or unpublished, where medical and or physical strategies to reduce body temperature were applied in acute ischaemic or haemorrhagic stroke. Outcome measures were death or dependency (modified Rankin Scale score >=3) at the end of follow up, and adverse effects. Results: Five medical cooling trials and three physical cooling trials involving 423 patients were included. We found no statistically significant effect of medical cooling compared to placebo in reducing the risk of death (OR 1.2, 95% CI 0.5-3.0) or dependency (OR 0.9, 95% CI 0.5-1.4). We also found no significant effects of physical cooling in reducing the risk of death (OR 1.0, 95% CI 0.4-2.4) or dependency (OR 0.9, 95% CI 0.3-2.5). Both interventions were associated with a non-significant increase in the occurrence of infections. Conclusion: There is currently no evidence from randomised trials to support routine use of medical or physical cooling therapy in patients with acute stroke. Further large randomised clinical trials are needed to study the effect of such strategies. The updated Cochrane review will be published in the Cochrane Database of Systematic Reviews in 2008.

Meta-analysis and review papers  
Date: Friday 16 May 2008  
Time: 9:50 - 10:00 -  Room: Euterpe
Chair: A. Algra, The Netherlands and P. Sandercock, United Kingdom

12
Improved Prediction of Treatment Success in Stroke Based on Pooled Control Arms
P. Mandava    T. Kent                                                        
 

Michael E DeBakey VA Medical Center and Baylor College of Medicine

USA

Background: Success in phase I/II trials of treatment for stroke have not been confirmed in follow-up. Difficulties in randomization for key baseline characteristics and lack of robust effect may be factors. We hypothesized that comparison with pooled results from control arms with different baseline characteristics would provide a more accurate prediction. We performed a proof of concept analysis of published trials to see if such a function could be generated and how it performs. Methods: Randomized controlled trials (RCT) for acute stroke with > 10 subjects including baseline NIHSS, age and 3 month outcomes were identified. Matlab© functions were written to accomplish predictive function minimization/optimization. The novel feature was generating multi-dimensional +95% prediction intervals along this function's surface permitting assessment of whether the study was significantly different from placebo even if its own placebo group was not representative. Results: A 3-dimensional function based on the control arms of 17 RCTS (n=5849) was generated (figure; R2=0.89, p<0.0001; mortality figure not shown) relating outcome to age and baseline NIHSS. Two studies were above the 95% surface, NINDS IV rt-PA ('a', figure) as expected, and Phase 1 NEST-I scalp surface laser ('J') treatment arm. All other treatment arms were within the prediction surfaces. Notably, ABESTT and SAINT-I (and their follow up ABESTT-II and SAINT-II) treatment arm outcomes fell well within prediction surface bounds hence correctly predicting eventual futility. Non-randomized series were also tested for potential success and results will be presented. Conclusion: A pooled placebo group derived function and prediction intervals provide an accurate method to predict success or failure of early trials.

http://www.eurostroke.org/ni_graphics/g_aid3056.htm