Aims: To develop models for prognostic prediction of oral health-related quality of life (OHRQoL) for patients with temporomandibular joint osteoarthritis (TMJ OA) at 1- and 6-month follow-ups after arthrocentesis treatment with hyaluronic acid (HA) injections once a week for 4 weeks.
Methods: From a cohort of 522 adult patients with TMJ OA treated with arthrocentesis with HA injections, 510 and 463 adult patients were included in the 1- and 6-month follow-ups, respectively. Patient characteristics and history, clinical examinations, and questionnaires were recorded as potential predictors at start of treatment, and all patients underwent an identical treatment protocol. Patients' OHRQoL values at 1 and 6 months after completing the treatment were used as outcome measures. Logistic regression methods were used to develop prediction models, and the performance and validity of these models were evaluated according to state-of-the-art methods, including receiver-operating characteristics curve for the discrimination of the models and calibration plots for the calibration of the models.
Results: History of mental disease, maximal protrusion of the jaw, muscular pain with palpation, joint pain with palpation, awake bruxism, chewing-side preference, and low OHRQoL at baseline were significantly associated with OHRQoL at the 1-month follow-up, while age, pain in other joints, history of mental disease, joint pain with palpation, sleep bruxism, awake bruxism, chewing-side preference, and low OHRQoL at baseline were significantly associated with OHRQoL at the 6-month follow-up. While the performance of both models was found to be good in terms of calibration, discrimination, and internal validity, the added predictive values of the 1-month and 6-month models for ruling in the risk of low OHRQoL were 19% and 31%, respectively, while those for ruling it out were 28% and 15%, respectively.
Conclusion: Several predictors were found to be significantly associated with patients' OHRQoL after treatment. Both prediction models may be reliable and valid for clinicians to predict a patient's risk of low OHRQoL at follow-up, so the models may be useful for clinicians in decision-making for patient management and for informing the patient.