Tools to evaluate the methodological quality of physical therapy trials. What items should we use from these tools? A factor analysis

Article type
Authors
Armijo-Olivo S1, Cummings G1, Fuentes J1, Saltaji H1, Ha C1, Chisholm A1, Rowe B1, Hartling L1, Rogers T2
1University of Alberta, Canada
2´University of Alberta, Canada
Abstract
Background: Numerous tools and items have been proposed to judge methodological quality of randomized controlled trials (RCT) in various health research areas. The frequency of use of these items varies according to health area, which suggests a lack of agreement regarding their relevance to trial quality. Recently, researchers have called for more in-depth analysis of the items through psychometric testing [i.e. factor analysis]; especially in the physical therapy (PT) field.

Objectives: To identify the underlying component structure of items and determine relevant items from tools used to evaluate quality/Risk of Bias (RofB) of PT trials.

Methods: RCTs used for this study were randomly selected from the Cochrane Database of Systematic Reviews. All studies were evaluated by 2 reviewers using 7 tools and their items for a combined set of 45 items. An exploratory factor analysis using Principal Axis Factoring (PAF) method followed by Varimax rotation was conducted.

Results: 214 trials were selected. PAF identified 34 items that loaded on 9 common factors as follows: 1) Selection bias; 2) Performance and detection bias; 3) Eligibility, interventions details, and outcomes measures description; 4) Psychometric properties; 5) Contamination and compliance bias; 6) Attrition bias; 7) Data analysis; 8) Sample size; and 9) Control and placebo adequacy. One item (i.e. intention to treat-ITT- performed) did not load in any of the factors mentioned; however, it was added to this set of items based on its relevance to RofB.

Conclusions: To our knowledge, this is the first study in PT to explore the underlying component items used to evaluate methodological quality/risk of bias of trials using factor analysis. The items and/or factors represent a starting point for evaluating the methodological quality/risk of bias in PT trials. Empirical evidence of the association between these items with treatment effects (meta-epidemiological approach) is required to validate these items before widespread use.