Exploring heterogeneity in meta-analyses using summary and individual patient data methodologies from stroke trials
Until recently, meta-analyses have usually been performed based on summary data methods. Individual patient data methods are becoming more popular, but the advantages of using these methods have not been fully investigated with regards to assessing and exploring heterogeneity. This thesis has assessed whether there are any clinically important differences in the results from analysing data from three meta-analyses of randomised controlled trials in the area of stroke medicine, using summary and individual patient data methodologies. Blood pressure ill Acute Stroke Collaboration (BASC) The management of blood pressure during the acute phase of stroke remains an enigma, therefore a systematic review and meta-analysis of existing randomised controlled trials was conducted to assess the effects of vasoactive drugs on outcome. Trends towards an increase in the risk of death, and death or dependency at the end of trial were found in patients randomised to a vasoactive drug as compared to those randomised to control. When baseline systolic blood pressure was taken into account in the analyses, patients randomised to a vasoactive drug had a significantly higher risk of death at the end of trial. Analyses also indicated that patients recruited early and within 48 hours has significant increases in the risk of death at the end of trial. However, no significant effects were seen for early change in systolic blood pressure. Community occupational therapy in stroke patients An evaluation of the efficacy of occupational therapy given in the community to stroke patients was performed using a systematic review and meta-analysis of randomised controlled trials. At the end of intervention, patient randomised to occupational therapy had significantly higher scores for extended and personal activities of daily living, and non-significantly higher scores for leisure participation. These effects appeared to be maintained over time. No effects were seen between the groups for death or minor psychiatric disorders as measured in patients or their carers. Subgroup analyses revealed that the benefits of occupational therapy were greatest when targeted interventions were used. Also, being male or independent at baseline was found to be important predictor of extended activities of daily living scores. Dipyridamole in Stroke Collaboration (DISC) Results from randomised controlled trials of dipyridamole, given with and without aspirin, for secondary prevention after stroke or transient ischaemic attack have given conflicting results; therefore, we performed a systematic review and meta-analysis. The risk of subsequent fatal or non-fatal stroke was reduced using the dual treatment of aspirin and dipyridamole as compared to either aspirin alone, dipyridamole alone, or control. Additionally, the dual treatment lowered the risk of non-fatal stroke, and subsequent vascular events defined by a composite outcome (non-fatal stroke, non fatal myocardial infarction, or vascular death). Analyses indicated that these results were independent of method of formulation of dipyridamole, dose of aspirin, type of qualifying event, and gender of the patients. However, increasing age was found to be an important predictor of subsequent stroke. These systematic reviews demonstrate that collaborations within the area of stroke medicine can be successful and much data can be shared. The findings from meta-analyses can be informative about the effectiveness of particular treatment and about which patients should be targeted for treatment; and may help steer the direction of future trials. Although summary data meta-analyses are practically easier to perform, it is important that assessments and explorations of heterogeneity should always be performed. Meta-analyses based on individual patient data may be needed to allow for more in depth investigations of heterogeneity, especially of patient characteristics. However, they themselves are not the panacea to all difficulties since they are subject to particular problems, mainly related to obtaining individual patient data to enable these in depth analyses to be performed.