Four CPPD Phenotypes Identified via Cluster Analysis

Researchers using data-driven statistics have categorized calcium pyrophosphate deposition disease (CPPD) into four distinct clinical phenotypes, offering new insights that align closely with real-world rheumatology practice according to a study published online on September 18, 2026, in Arthritis & Rheumatology.

Medical researchers pooling baseline data from a French trial and a European observational study have identified four distinct patient clusters for calcium pyrophosphate deposition disease, or CPPD. Led by Greta Pellegrino, MD, of the Rheumatology Department at IRCCS Ospedale Galeazzi Sant’Ambrogio in Milan, Italy, the analysis utilized multiple correspondence analysis and hierarchical clustering to group 134 patients.

Methodology and Variable Grouping in the CPPD Study

The underlying investigation combined cohorts from two distinct sources: a French trial that included hospitalized patients older than 65 years with acute calcium pyrophosphate crystal arthritis, and a European observational study tracking adults with chronic or recurrent disease from centers in France, Italy, and Spain. Out of the combined pool, the final analysis evaluated 134 patients, of whom 65.7% were women.

Investigators categorized continuous measures using thresholds anchored in reference ranges from local laboratories or routine clinical practice. Age of symptom onset was split into categories of under 60, 60 to 75, and over 75 years. C-reactive protein (CRP) levels were divided at 10 mg/L or below versus above 10 mg/L. Researchers also tracked the number of affected joints—either a single joint or multiple joints—along with visual analog scale pain scores broken down into thresholds under 50, between 50 and 70, and above 70 millimeters.

By applying multiple correspondence analysis, the team condensed 29 clinical, imaging, and laboratory variables into three primary dimensions. Hierarchical clustering then grouped patients with similar features into four separate patterns.

Patient Clusters and Clinical Phenotypes Identified

The analysis sorted patients into four clusters: cluster 1 comprised 43.3% of the cohort, cluster 2 accounted for 17.9%, cluster 3 made up 19.4%, and cluster 4 represented another 19.4%. Among these, statistical analysis favored three of the clusters.

Cluster 1 closely mirrored acute calcium pyrophosphate crystal arthritis, showing strong representation among women. In this group, 81.0% of patients had a single affected joint, 93.1% experienced symptom onset after 75 years of age, and 94.8% presented with CRP levels above 10 mg/L. Cluster 2 also featured predominantly single-joint involvement in 75.0% of cases, with half of the patients showing CRP levels above 10 mg/L, while spanning all ages at onset.

Clusters 3 and 4 demonstrated a different pattern where nearly all patients had multiple affected joints. Cluster 3 matched chronic calcium pyrophosphate crystal inflammatory arthritis, frequently involving the wrists, persistent arthritis, and x-ray osteoarthritis, alongside less frequently elevated CRP levels. Cluster 4 featured more widespread joint involvement, primarily recurrent flares, often elevated CRP, and younger patients whom researchers noted could be mistaken for having polymyalgia rheumatica.

[The] findings provide a disease description that closely aligns with real-world clinical practice, highlighting for the first time the potential to subclassify CPPD disease according to the number of involved joints, similar to many other arthropathies.
Greta Pellegrino, MD, and Laurène Norberciak, BSc, Arthritis & Rheumatology study authors

Study Limitations and Disclosures

Furthermore, about 41% of the original sample was excluded due to missing data.

Broader Neurological and Pulmonary Context

Outside of arthropathies, similar data-driven clustering methods have been applied to other complex medical conditions. In a large retrospective analysis, researchers evaluated records from German patient registries to examine chronic inflammatory demyelinating polyneuropathy, commonly referred to as CIDP. By analyzing records across ten university hospitals between 2010 and 2024, investigators sorted 325 patients into a discovery cohort using an unbiased machine-learning approach, which subsequently revealed four distinct clinical subgroups.

Parallel data-driven classification efforts have likewise explored respiratory conditions, examining 319 stable chronic obstructive pulmonary disease patients to establish distinct inflammatory phenotypes characterized by varying cytokine levels and clinical severity.