Article type
Year
Abstract
Does the inverse of Freeman‐Tukey double arcsine transformation provide misleading results in meta‐analysis of single proportions?
Bardach A, Ciapponi A, Glujovsky D, Luz Gibbons
Background: Standard generic inverse variance methods for the combination of single proportions are based on transformed proportions using the logit, arcsine, and Freeman‐Tukey double arcsine transformation and more recently generalized linear mixed models. Irrespective of the approach, meta‐analysis results are typically back‐transformed to the original scale in order to facilitate interpretation. A recent study suggested that the Freeman‐Tukey double arcsine transformation, one of the most frequently used approaches, could provide misleading results after the back‐transformation.1
Objectives: To evaluate the consistency of the back‐transformed results obtained with the inverse of Freeman‐Tukey double arcsine transformation with alternative methods of meta‐analyses of single proportions.
Methods: We will analyze the results of five meta‐analyses of single proportions obtained by the inverse of Freeman‐Tukey double arcsine transformation2 and we will compare with the results obtained by the logit transformations.
Additionally, we will analyze if the sample size could be influential in the potential discrepancies
Results: The comparison of the results by method will be presented at the Colloquium.
References
1. Schwarzer G, et al. Seriously misleading results using inverse of Freeman-Tukey double arcsine transformation in meta-analysis of single proportions. Res Synth Methods. 2019;10(3):476-483.
2. Ciapponi A, Glujovsky D, Virgilio SA, Bardach AE. Conducting and Disseminating Epidemiological Systematic Reviews in Latin America and the Caribbean: Pitfalls and Lessons Learned. Value in Health Regional Issues. 2017;14:64-72.
Patient or healthcare consumer involvement:
Bardach A, Ciapponi A, Glujovsky D, Luz Gibbons
Background: Standard generic inverse variance methods for the combination of single proportions are based on transformed proportions using the logit, arcsine, and Freeman‐Tukey double arcsine transformation and more recently generalized linear mixed models. Irrespective of the approach, meta‐analysis results are typically back‐transformed to the original scale in order to facilitate interpretation. A recent study suggested that the Freeman‐Tukey double arcsine transformation, one of the most frequently used approaches, could provide misleading results after the back‐transformation.1
Objectives: To evaluate the consistency of the back‐transformed results obtained with the inverse of Freeman‐Tukey double arcsine transformation with alternative methods of meta‐analyses of single proportions.
Methods: We will analyze the results of five meta‐analyses of single proportions obtained by the inverse of Freeman‐Tukey double arcsine transformation2 and we will compare with the results obtained by the logit transformations.
Additionally, we will analyze if the sample size could be influential in the potential discrepancies
Results: The comparison of the results by method will be presented at the Colloquium.
References
1. Schwarzer G, et al. Seriously misleading results using inverse of Freeman-Tukey double arcsine transformation in meta-analysis of single proportions. Res Synth Methods. 2019;10(3):476-483.
2. Ciapponi A, Glujovsky D, Virgilio SA, Bardach AE. Conducting and Disseminating Epidemiological Systematic Reviews in Latin America and the Caribbean: Pitfalls and Lessons Learned. Value in Health Regional Issues. 2017;14:64-72.
Patient or healthcare consumer involvement: