Background We previously identified 5 clinically relevant phenotypes from a heterogeneous knee OA population (Osteoarthritis Initiative [OAI] cohort), which needs to be replicated for validation.
Objectives The present study aims to replicate this phenotype identification for validation in a clinical setting using data from the Amsterdam Osteoarthritis (AMS-OA) cohort.
Methods K-means clustering analysis was performed in 374 knee OA patients from the AMS-OA cohort, using 4 clinically relevant and easily obtainable patient characteristics or clustering variables: radiographic severity (Kellgren/Lawrence [K/L] grade), body mass index (BMI), upper leg muscle strength, and depression (from Hospital Anxiety and Depression Scale [HADS] questionnaire). The cluster solution with the highest Pseudo F value was considered the most adequate number of clusters or phenotypes in the dataset. This solution was compared with the original study.
Results The most adequate number of phenotypes was 5, similar as in the original study. Moreover, the 5 identified phenotypes represented in general similar phenotypes compared to the original study, namely a “minimal joint disease phenotype”, a “strong muscle phenotype”, a “non-obese weak muscle phenotype” (although obesity was more prevalent compared to the original study), an “obese weak muscle phenotype”, and a “depressive phenotype” (although depression was less prevalent compared to the original study).
Conclusions The identification of phenotypes from the OAI cohort could be replicated, except for two minor differences that are presumably attributable to differences in study population. The phenotype identification seems therefore a valid finding, with phenotypes possibly representing different etiological subtypes of knee OA in which phenotype-specific interventions may be needed.
Disclosure of Interest None declared