Statistical testing against baseline in orthodontic research: a meta-epidemiologic study

Background/objectives To assess the prevalence of within-group comparisons from baseline to follow-up in published orthodontic articles and to identify potential associations between this statistical problem and a number of study characteristics. Materials/method The most recent 24 issues of four leading orthodontic journals with highest impact factor (American Journal of Orthodontics and Dentofacial Orthopedics; AJODO, European Journal of Orthodontics; EJO, Angle Orthodontist; ANGLE, Orthodontics and Craniofacial Research; OCR) were electronically searched until December 31st 2017. The proportion of articles using comparisons against baseline and interpretation of findings according to within-group comparisons were recorded. The association of this practice with journal, year of publication, study design, continent of authorship, number of centres and researchers, statistical significance of results, and statistical analysis was tested. Univariable and multivariable modified Poisson regression were used to identify significant predictors. Results Overall, 339 articles were eligible for inclusion with the majority published in ANGLE (n = 157, 46%), followed by AJODO (n = 75, 22%), and EJO (n = 75, 22%). A total of 60 studies (18%) presented interpretation of their findings based on within-group comparisons against baseline in isolation. Statistical significance of the primary outcome was a very strong predictor of the prevalence of this flawed approach (RR: 2.33, 95% CIs: 1.22, 4.43; P = 0.01). Limitations The effect of time since publication was not addressed. Conclusions/implications Statistical testing and interpretation within groups is prevalent in orthodontic research. Endorsement of accurate conduct and reporting of statistical analyses and interpretation of research findings is important in order to promote optimal inferences to support clinical decision-making