Invited Speaker Australian and New Zealand Stroke Organisation Conference 2025

Should mean- or percentile-based methods be used for the meta-analysis of process-of-care time measures in hyperacute stroke?       (123197)

Dominic Italiano 1 2 , Hannah Johns 1 2 , Bruce Campbell 3 , Guillaume Turc 4 5 6 7 , Leonid Churilov 1 2
  1. Melbourne Medical School, University of Melbourne, Melbourne, VIC, Australia
  2. Australian Stroke Alliance, Melbourne Brain Centre, Royal Melbourne Hospital, Melbourne, VIC, Australia
  3. Department of Medicine and Neurology, Melbourne Brain Centre at the Royal Melbourne Hospital, University of Melbourne, Melbourne, VIC, Australia
  4. Department of Neurology, GHU Paris Psychiatrie et Neurosciences, Paris, France
  5. Université Paris Cité, Paris, France
  6. Inserm, Paris, France
  7. FHU NeuroVasc, Paris, France

Background: Process-of-care time measures provide vital information about hyperacute stroke interventions. These measures are often highly skewed, and relevant mean-based statistics may be misleading. Percentile-based statistics (e.g. median) conveniently and unbiasedly summarise the proportion of patients treated within a given timeframe. Despite this, mean-based synthesis methods are primarily recommended by the Cochrane Handbook for Systematic Reviews of Interventions for conducting meta-analyses. It’s also recommended that percentile-based outcomes are either excluded from meta-analyses or included by imputing mean-based measures using transformation methods.

 

Aims: To provide methodological foundation for the appropriate meta-analysis of process-of-care time measures in hyperacute stroke systematic reviews and meta-analyses.

 

Methods: We evaluated the performance of transformation-based methods using process-of-care time data from hyperacute stroke clinical trials of thrombolysis, thrombectomy and MSU interventions. We also compared the performance of various mean- and percentile-based meta-analysis techniques for process-of-care time measures.

 

Results: Transformation-based methods showed strong agreement in estimating the true mean from different sets of percentile-based summary statistics (LCCC = 0.9987). These methods became less reliable as the skew of the underlying distribution increased. Mean-based and percentile-based meta-analysis methods provided comparable results when analysing summary measures of between-arm treatment effects and varied results when analysing individual-arm summary measures.  

 

Conclusion: As the observed variability affects the interpretability of the results, we recommend that stroke researchers utilise percentile-based approaches for the reporting, analysis, and meta-analysis of process-of-care time measures in hyperacute stroke studies.