Article type
Year
Abstract
Background: The effect of publication bias on the efficacy of pharmacological interventions has been extensively studied, but less is known on its effect in systematic reviews about non-pharmacological interventions.
Objectives: To assess the possibility of publication bias on a systematic review of the efficacy of psycho-educative interventions on the burden experienced by informal caregivers of people with dementia.
Methods: We searched randomized clinical trials in CENTRAL, PubMed, Embase, and PsycInfo reporting the efficacy of psycho-educative interventions versus usual care on dementia caregiver’s burden. We also searched lists of references of relevant qualitative and quantitative reviews on caregiver’s burden. We obtained pooled Hedges’ g effect sizes using a random effect model. We assessed the possibility of publication bias by the asymmetry of the funnel plot and quantified subsequently the likely of bias on the observed effect size by means of the trim & fill method and the Copas selection model.
Results: We obtained data for effect sizes of psycho-educative interventions from 11 trials. Although random effect model estimates showed small to moderate effects favouring the intervention (g = −0.25; 95% CI = −0.43 to −0.07), the funnel plot (Fig. 1) was clearly asymmetric (rank correlation p = 0.004; linear regression p = 0.0063). The trim and fill approach suggested 4 more trials and a smaller effect (g = −0.11; 95% CI = −0.30 to 0.08), whereas the Copas selection model suggested 11 trials might be unpublished and decreased even further the effect size (g= −0.0015; 95% CI = −0.1889 to 0.1860).
Conclusions: Our results show a huge impact of publication bias in a systematic review on the efficacy of non-pharmacological interventions, as others have suggested in this field. Since the registry of such trials is not yet mandatory and practically inexistent, the quantification of the problem is harder than for pharmacological interventions.
Objectives: To assess the possibility of publication bias on a systematic review of the efficacy of psycho-educative interventions on the burden experienced by informal caregivers of people with dementia.
Methods: We searched randomized clinical trials in CENTRAL, PubMed, Embase, and PsycInfo reporting the efficacy of psycho-educative interventions versus usual care on dementia caregiver’s burden. We also searched lists of references of relevant qualitative and quantitative reviews on caregiver’s burden. We obtained pooled Hedges’ g effect sizes using a random effect model. We assessed the possibility of publication bias by the asymmetry of the funnel plot and quantified subsequently the likely of bias on the observed effect size by means of the trim & fill method and the Copas selection model.
Results: We obtained data for effect sizes of psycho-educative interventions from 11 trials. Although random effect model estimates showed small to moderate effects favouring the intervention (g = −0.25; 95% CI = −0.43 to −0.07), the funnel plot (Fig. 1) was clearly asymmetric (rank correlation p = 0.004; linear regression p = 0.0063). The trim and fill approach suggested 4 more trials and a smaller effect (g = −0.11; 95% CI = −0.30 to 0.08), whereas the Copas selection model suggested 11 trials might be unpublished and decreased even further the effect size (g= −0.0015; 95% CI = −0.1889 to 0.1860).
Conclusions: Our results show a huge impact of publication bias in a systematic review on the efficacy of non-pharmacological interventions, as others have suggested in this field. Since the registry of such trials is not yet mandatory and practically inexistent, the quantification of the problem is harder than for pharmacological interventions.
Images
