Can alternative means of phenotyping rheumatoid arthritis reduce its apparent heterogeneity? A comparison of three disease activity classifications with DAS28.

OBJECTIVES

Rheumatoid arthritis (RA) is a disease with a heterogeneous phenotype. Partly, this heterogeneity may be concealed by metrics such as 28-joint disease activity score (DAS28) that may assign similar scores to dissimilar phenotypes. To explore how individual patient profiles may be separated, we developed three classification schemes and compared these with DAS28 in newly diagnosed RA.

METHODS

We selected patients aged 18-100 years, newly diagnosed with RA between 2012 and 2022 and registered in the Swedish Rheumatology Quality Register. We devised three alternative classifications based on (i) a subjective-objective decomposition of the DAS28 (seven levels), (ii) clinical experience-based cut-offs for each of the DAS28 components (five levels) and (iii) a data-driven Gaussian Mixture Model (five clusters). For each classification, we calculated descriptive statistics for clinical characteristics, demographic variables and comorbidity histories, and contrasted these with those based on DAS28 categories in our study population.

RESULTS

We identified 6624 patients with complete data on all included variables. In each of the alternative classifications, subjectively dominated and objectively dominated subsets captured differences regarding comorbidity histories not detectable in the DAS28 categories. Across all alternative classifications, subjectively dominated subsets had lower work ability than objectively dominated subsets and included 50%-300% more patients with pain and psychiatric diagnoses prior to RA diagnosis.

CONCLUSIONS

Separating subjective and objective dimensions reveals different RA patient profiles that are grouped together by DAS28, demonstrating the extent to which DAS28 is confounded by factors beyond RA disease activity. Disentangling and treating heterogeneous RA patient profiles may therefore require more granular disease activity classification systems.