A P Verhagen1, H C W de Vet, S Willemsen, Th Stijnen. 1. Department General Practice, ErasmusMC, University Medical Centre Rotterdam, PO Box 2040, Room Ff318, 3000 CA Rotterdam, The Netherlands. a.verhagen@erasmusmc.nl
Abstract
OBJECTIVES: In the conduct of a systematic review or meta-analysis, many possible sources of bias exist, such as bias caused by design characteristics. We studied the influence of the methodological study characteristics of randomized clinical trials (RCTs) on the outcome in a systematic review on conservative treatments in patients with tension-type headache (TTH). STUDY DESIGN AND SETTING: Included were RCTs from a systematic review on TTH, which were a control group receiving a placebo or no treatment and presented data on recovery or headache severity, intensity, or frequency. Design characteristics were assessed using the Delphi list. Regression analysis is performed on separate design characteristics on size of treatment effect. RESULTS: Out of the original data set of 146 trials, 61 trials fulfilled our selection criteria. The number of trials presenting only dichotomous data was larger than trials presenting only continuous data. All study characteristics show a nonsignificant relation with the effect estimate. Whether outcome is presented dichotomous or continuous appears to have a significant impact on treatment effect estimates. CONCLUSION: In this study, sample design characteristics do not show to have an impact on treatment effect estimates, but the way the treatment effect is measured has a significant impact.
OBJECTIVES: In the conduct of a systematic review or meta-analysis, many possible sources of bias exist, such as bias caused by design characteristics. We studied the influence of the methodological study characteristics of randomized clinical trials (RCTs) on the outcome in a systematic review on conservative treatments in patients with tension-type headache (TTH). STUDY DESIGN AND SETTING: Included were RCTs from a systematic review on TTH, which were a control group receiving a placebo or no treatment and presented data on recovery or headache severity, intensity, or frequency. Design characteristics were assessed using the Delphi list. Regression analysis is performed on separate design characteristics on size of treatment effect. RESULTS: Out of the original data set of 146 trials, 61 trials fulfilled our selection criteria. The number of trials presenting only dichotomous data was larger than trials presenting only continuous data. All study characteristics show a nonsignificant relation with the effect estimate. Whether outcome is presented dichotomous or continuous appears to have a significant impact on treatment effect estimates. CONCLUSION: In this study, sample design characteristics do not show to have an impact on treatment effect estimates, but the way the treatment effect is measured has a significant impact.
Authors: An-Wen Chan; Jennifer M Tetzlaff; Peter C Gøtzsche; Douglas G Altman; Howard Mann; Jesse A Berlin; Kay Dickersin; Asbjørn Hróbjartsson; Kenneth F Schulz; Wendy R Parulekar; Karmela Krleza-Jeric; Andreas Laupacis; David Moher Journal: BMJ Date: 2013-01-08
Authors: Susan Armijo-Olivo; Humam Saltaji; Bruno R da Costa; Jorge Fuentes; Christine Ha; Greta G Cummings Journal: BMJ Open Date: 2015-09-03 Impact factor: 2.692