Background: Occurrence and progression of age-related irreversible degradations of skeletal joints, osteoarthritis (OA), has a stochastic nature. However, it is commonly described using polynomial models, which may not necessarily be optimal.
Aim: To implement a stochastic model of gradual accumulation of the distinct changes for estimating individuals' putative age at onset and risk of the process advancing in the OA longitudinal data.
Subjects and methods: The model was formulated as a discrete Markov process. It was applied to radiographic knee osteoarthritis (RKOA) data: 243 Kellgren-Lawrence (K/L) and 207 osteophytes (OP) score histories from the 15-year follow-up Chingford study.
Results: The model performance was examined in Monte-Carlo simulations. The mean age at onset of knee osteoarthritis was: 53.04 and 53.23 years and the average annual risk of one K/L and one OP grade appearance was: 0.066 and 0.025, respectively. The analysis also suggested that there is 3-4 years difference between the inferred age at onset and the age when knee osteoarthritis becomes detectable on radiograph.
Conclusion: The stochastic model provides more accurate description of the empiric data compared with the corresponding polynomial model. The model-based individual's estimates could be used as an important tool to fit age-related patterns of the corresponding diseases and conditions.