Exploring the potential of cluster analysis in identifying disease phenotypes in CPPD: Moving from intuition to insight.

BACKGROUND

Calcium pyrophosphate deposition (CPPD) disease is a heterogeneous condition and is frequently misdiagnosed due to the different clinical presentations and variable disease progression. Aim of this study is to identify clinical phenotypes of CPPD disease by integrating real-world data from two datasets.

METHODS

Data from the COLCHICORT trial and CHRONIC-CPPD European observational cohort were analysed. Multiple Correspondence Analysis (MCA) was employed to evaluate and visualize the association between modalities of quantitative variables. Hierarchical Clustering on Principal Components (HCPC) obtained from MCA identified clusters based on Ward agglomeration method with Euclidean distance. A bivariate analysis between clusters was then performed to characterize the clusters numerically.

RESULTS

134 patients were included in the analysis, resulting in four distinct clusters. Cluster-1 had predominantly monoarticular involvement, frequent recurrent flares, older age of onset, female prevalence, and high CRP levels. Cluster-2 also involved monoarticular cases but with persistent arthritis and onset after 60 years, with about half exhibiting elevated CRP. Cluster-3 included polyarticular cases, featuring persistent arthritis and lower CRP levels, while Cluster-4 highlighted polyarticular involvement (spine and shoulder included) and primarily recurrent flares.

CONCLUSION

In conclusion, by using for the first time a robust, data-driven methodology based on real-world data, two clusters similar to the EULAR phenotypes "acute CPP crystal arthritis", and "chronic CPP crystal inflammatory arthritis" were confirmed, but in addition two more clusters that do not correspond to any existing EULAR phenotypes were also identified. Due to the characteristics of the cohorts, the Osteoarthritis plus CPPD cluster has not been explored.