Tobias Manigold (1), Nicolas Bodmer (2), Elisabeth Rosoux (4), Guillaume Fahrni (3), Deborah Markham (4), Raphael Micheroli (5), Lucas M. Bachmann (2), Jonas Brändli (6), Fabio Becce (3), Thomas Hügle (3)
Affiliation(s):
1. Department of Rheumatology, Inselspital University Hospital Bern, Bern, Switzerland.
2. Medignition, Engelstrasse 6, CH-8004 Zurich
3. Department of Radiology, Lausanne University Hospital (CHUV), University of Lausanne (UNIL), Lausanne, Switzerland
4. Department of Rheumatology, University Hospital Lausanne (CHUV), Switzerland
5. Department of Rheumatology, Zurich University Hospital, University of Zurich, Zurich, Switzerland.
6. Data Science Team, Swiss Clinical Quality Management Foundation, Zurich, Switzerland
Background: Calcium pyrophosphate deposition disease (CPPD) disease can mimic or interfere with the course of rheumatoid arthritis (RA). Previous studies suggest higher prevalence of Chondrocalcinosis (CC ), a precursor of CPPD, in seronegative RA patients. However, no significant treatment disparities were described in RA patients with or without coexisting CC.
Methods: We recently developed and validated a deep learning algorithm to classify the presence of chondrocalcinosis (CC) on hand radiographs, detecting the presence of CC at the TFCC, MCP-2, and MCP-3 sites with an accuracy of 0.86. In this study, we report the baseline characteristics of 1,344 RA patients of the Swiss Clinical Quality Management in Rheumatic Diseases (SCQM) registry who underwent CC assessment using this algorithm and number of therapy lines and with adequate quality of the radiographs to run the algorithm.
A subgroup of patients was identified for whom the results of a clinical examination were available on the same date as the radiographs. In the event that radiographs of both hands were available, one was randomly selected. The following data were extracted: age, sex, seropositivity (RF, anti-CCP) status, number of therapy lines (four or more drugs vs. less), and DAS28 (CRP) (> 5.1 vs. below). Two multivariable logistic regression analyses were conducted to assess the association of CC presence on DAS28 scores and the number of current RA medications, with adjustments made for patients’ age, sex, and serostatus.
Results: The mean age of the participants was 56.2 years (standard deviation (SD) 13.9), 954/1344 (71.0%) were female, and 916/1344 (68.2%) were seropositive. CC was present in 310 patients (23.1%) and overall not associated with the serostatus. Overall, 46 patients (3.4%) were taking at least four medications, and 82 (6.1%) exhibited high disease activity with DAS28-CRP >5.1. Overall, no association was found between CC presence and disease activity (OR 1.09 (95% CI: 0.62-1.90); p=0.771). In multivariable analyses, CC+ patients were more likely to take at least four drugs (odds ratio (OR): 2.54 (95% confidence interval (CI): 1.21-5.32); p=0.014) and were significantly older (CC-: 54.0 SD (13.5), CC+ 63.0 SD (13.0), p<0.001).
Discussion: In this cross-sectional analysis of baseline characteristics of a large cohort of patients with RA who underwent CC assessment, we found a fairly strong association between the presence of CC and high drug use. However, overall we did not find an association between the presence of CC and disease activity or serostatus. Additional analyses will be performed in subgroups based on e.g. age, disease duration and type of treatment.