Statistics from Altmetric.com
Parker B, Urowitz MB, Gladman DD, et al. Clinical associations of the metabolic syndrome in systemic lupus erythematosus: data from an international inception cohort. Ann Rheum Dis 2013;72:1308–14.
A statistical error occurred in the above manuscript. On routine review of the analysis, the authors have uncovered a statistical error in their original baseline analysis that has resulted in an error in the attribution of metabolic syndrome (MetS) status to individuals in the cohort, such that a number of individuals have been misclassified by the algorithm reported. The error relates to how the authors used the coding commands in Stata for handling missing data. The authors’ intention was to count the total number of MetS variables present in each individual and to allow patients with missing data to be included if they had sufficient data points to make a determination of their status. Instead, the command the authors originally used (‘total’) resulted in patients with any missing data being erroneously assigned a zero total for MetS components present (instead of the true number (if ≥3) or a missing status (if ≤2). This resulted in both an underestimation of the true prevalence and individuals being erroneously included in the experimental group when they should have been excluded, and thus a larger denominator group. The authors’ initial manual checks of the data failed to identify this error.
Correction of this error (using the command ‘rowtotal’) has resulted in a reduced cohort size and an overall increase in the number of patients classified as having MetS in the cohort (now 38.2% as opposed to 16% as originally reported). The prevalence estimates in all racial/ethnic subsets are therefore higher and the estimate for patients of African ancestry is particularly affected, such that the authors now report a high prevalence in this population (55.7%). Although the point estimates in the multivariate analysis of factors associated with MetS have changed, the actual factors remain very similar to the published multivariate model (see table A) and the authors’ discussion of the impact of disease activity and exposure to steroids remains unchanged.
In summary, the estimated prevalence of MetS in this cohort is now higher than that originally reported. The multivariate model has as a result of this changed, but remains similar to the authors’ original report. Overall the key message of this paper remains the same in that “the observed ethnic variation in MetS susceptibility should help inform risk stratification in management of early disease. MetS is associated with a more severe disease phenotype and higher doses of corticosteroids, therefore balancing disease control while minimising corticosteroid exposure should be at the forefront of personalised treatment decisions in these patients”.
The authors profoundly apologise for this error and feel it is important to communicate this to the research community in the interests of accuracy and scientific validity.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement