No abstract

ObjectiveTo investigate sex differences in spinal radiographic progression in axial spondyloarthritis (axSpA).MethodsAxSpA patients in the Swiss Clinical Quality Management cohort with available spinal radiographs every 2 years were included. Paired radiographs were scored by two readers according to the modified Stoke Ankylosing Spondylitis Spinal Score (mSASSS). Progression was defined as an increase of ≥2 mSASSS units in 2 years. The relationship between sex and progression was investigated with binomial generalised estimating equation models, considering baseline spinal damage as an intermediate covariate. Additional analyses included adjustments for explanatory variables and multiple imputations for missingness.ResultsIn a total of 505 axSpA patients (317 men and 188 women), mean±SD radiographic progression over 2 years was 1.0±2.8 years in men and 0.3±1.1 years in women (p<0.001). Male sex was associated with enhanced progression in a small model not including baseline damage (OR 3.41, 95% CI 1.87 to 6.21). Both a direct effect of male sex on spinal progression, and an indirect effect, via enhancement of baseline spinal damage were significant (OR 2.06, 95% CI 1.15 to 3.67 and OR 1.04, 95% CI 1.01 to 1.07, respectively). A significant impact of male sex on spinal radiographic progression was still demonstrated after multiple adjustments for covariates known to potentially affect spinal radiographic progression (OR 1.97, 95% CI 1.04 to 3.71).ConclusionsSpinal radiographic progression in axSpA is more severe in men than in women, with three times higher odds of progression in male patients and an effect that is mediated in part through an increase in baseline radiographic damage.

No abstract

Ensembles improve prediction performance and allow uncertainty quantification by aggregating predictions from multiple models. In deep ensembling, the individual models are usually black box neural networks, or recently, partially interpretable semi-structured deep transformation models. However, interpretability of the ensemble members is generally lost upon aggregation. This is a crucial drawback of deep ensembles in high-stake decision fields, in which interpretable models are desired. We propose a novel transformation ensemble which aggregates probabilistic predictions with the guarantee to preserve interpretability and yield uniformly better predictions than the ensemble members on average. Transformation ensembles are tailored towards interpretable deep transformation models but are applicable to a wider range of probabilistic neural networks. In experiments on several publicly available data sets, we demonstrate that transformation ensembles perform on par with classical deep ensembles in terms of prediction performance, discrimination, and calibration. In addition, we demonstrate how transformation ensembles quantify both aleatoric and epistemic uncertainty, and produce minimax optimal predictions under certain conditions.

Institute for Data modalities. Overall, we highlight that DTMs provide a powerful, interpretable approach to analyzing semistructured data and that they have the potential to support clinical decision-making.