Article type
Year
Abstract
Background: Exclusion of non-English and grey literature (i.e. literature not published in peer-reviewed journals or having limited distribution) observational studies from systematic reviews may introduce bias and impact on overall treatment effect estimates.
Objectives: We have investigated the potential effect of including or excluding grey and non-English literature on effect estimates using non-randomized add-on topiramate, levetiracetam and gabapentin studies in refractory epilepsy.
Methods: Electronic searches of MEDLINE (1966-2005), EMBASE (1974-2004), CINAHL (1982-2005), ZETOC (1993-2005), and handsearching congress proceedings were performed to identify relevant studies. Data were extracted for study attributes, 50% responder and seizure freedom rates. Effect estimates were compared for non-English versus English studies, and for studies published in full versus those published in abstract form only. Funnel plots were used to assess publication bias.
Results: Ninety-six non-randomized studies (42 published in full, 54 abstracts) were identified for add-on topiramate therapy, 80 nonrandomized studies (36 published in full, 44 abstracts) for add-on levetiracetam therapy and 47 (26 published in full, 21 abstracts) for add-on gabapentin therapy. Median reported responder and seizure freedom rates were higher in topiramate studies published in full versus those from abstracts (62% vs. 55%) and (16% vs. 8%) respectively. Combining studies resulted in smaller median responder and seizure freedom rates compared to summary estimates from studies published in full. Including non-English topiramate studies published in full did not impact on the overall treatment effect estimates. Similar results were noted when analysing levetiracetam and gabapentin add-on studies.
Conclusions: The exclusion of grey literature from systematic reviews could lead to over inflated summary effect estimates for an intervention. An attempt to retrieve all relevant data is paramount to systematic reviews.
Objectives: We have investigated the potential effect of including or excluding grey and non-English literature on effect estimates using non-randomized add-on topiramate, levetiracetam and gabapentin studies in refractory epilepsy.
Methods: Electronic searches of MEDLINE (1966-2005), EMBASE (1974-2004), CINAHL (1982-2005), ZETOC (1993-2005), and handsearching congress proceedings were performed to identify relevant studies. Data were extracted for study attributes, 50% responder and seizure freedom rates. Effect estimates were compared for non-English versus English studies, and for studies published in full versus those published in abstract form only. Funnel plots were used to assess publication bias.
Results: Ninety-six non-randomized studies (42 published in full, 54 abstracts) were identified for add-on topiramate therapy, 80 nonrandomized studies (36 published in full, 44 abstracts) for add-on levetiracetam therapy and 47 (26 published in full, 21 abstracts) for add-on gabapentin therapy. Median reported responder and seizure freedom rates were higher in topiramate studies published in full versus those from abstracts (62% vs. 55%) and (16% vs. 8%) respectively. Combining studies resulted in smaller median responder and seizure freedom rates compared to summary estimates from studies published in full. Including non-English topiramate studies published in full did not impact on the overall treatment effect estimates. Similar results were noted when analysing levetiracetam and gabapentin add-on studies.
Conclusions: The exclusion of grey literature from systematic reviews could lead to over inflated summary effect estimates for an intervention. An attempt to retrieve all relevant data is paramount to systematic reviews.