Letter to the Editor on “The Prevalence of Positive Findings on Metal Artifact Reduction Sequence Magnetic Resonance Imaging in Metal-on-Metal Total Hip Arthroplasty”: Part 2.

We read with great interest the article by Lindgren et al [1]. Unfortunately, we have concerns regarding the statistical methods used in their study.

In the study, the authors use ordinal logistic regression (OLR) to correlate predictor variables against primary outcome variables. OLR serves as an extension and alternative to binary logistic regression analysis, should the primary outcome be an ordinal variable (ie, Likert scaled poor/fair/good/excellent) rather than a dichotomous type (yes/no). This is an important distinction, and with it, arises many assumptions that must be met by the data.