Can knee osteoarthritis pain flares be predicted?

Abstract

Purpose: Knee pain is the main cause of disability and reduced function in knee osteoarthritis (KOA). Though it is established that KOA pain fluctuates, there is currently no method to predict knee osteoarthritis pain flares. This project aims to identify risk factors which predict KOA pain flares in the following 30 days. Methods: Study participants were selected from SPARK, a 90-day web-based longitudinal study that examined risk factors for KOA pain flares. Participants were requested to complete an online questionnaire at 10 day intervals from day 0 (control period assessment points) and at time points whenever they experienced knee pain flare (case period assessment points) during the follow-up period. A KOA pain flare was defined as current pain >=2 points on a 0-10 point numeric rating scale compared to usual background KOA pain intensity reported at day 0. The SPARK dataset was partitioned into 3 sections: Day 0–30, Day 31–60 and Day 61–90. Exposures to the following risk factors/derived variables from Day 0–30 (D0–30) were used to identify associations/ predict KOA flares in the next/following 30 days-that is Day 31–60. Predictors assessed for the D0-30 period were as follows: (1) any episodes of buckling of index knee between D0–30 (2) any injury to index knee between D0-30 (3) percentage of days of medication used (4) physical activity, classified as mild physical activity or any moderate or any vigorous physical activity performed between D0-30 and (5) heel height worn >2/3 of assessed time period (heel height categories <2.5 cm, 2.5–5 cm, >5 cm). The (6) positive and negative affect score, (7) Intermittent and Constant Osteoarthritis Pain Score (ICOAP)/100 were assessed at D30. The participant's (8) age, (9) body mass index, (10) duration of having osteoarthritis, (11) usual background pain score and (12) worst pain score were assessed at recruitment. Patients who reported flares at the point of enrolment/baseline and those with missing observations were omitted. Logistic regression was used to identify significant risk factors. Step-wise elimination was done and only variables with P <0.1 were used for the prediction modelling. Data was analysed using STATA version 13. Results: 315 persons (62.3% females) with a mean age of 62.1 years (SD±8.2) with complete data on the variables of interest were selected. The mean body mass index was 29.8 kg/m2 (SD±6.5). The participants rated their baseline pain (on a visual analogue score) as 4.36 (SD±2) and worst pain as 7.91 (SD±1.74). 115 persons (36.3%) reported at least one flare from D31-60. The majority of variables were eliminated by step-wise elimination. Only the ICOAP constant pain sub scale at D30, the usual level of pain reported at baseline and history of any knee buckling 30 days prior were significant predictors for the occurrence of any flare in the ensuing 30 days (Table 1). Area under the receiver operating characteristic curve was 0.72 (95% Confidence Interval 0.66-0.78), suggesting a moderate predictive ability for the occurrence of knee pain flare. Conclusions: Higher ICOAP constant sub scale values at baseline, higher background pain levels reported by the patient at recruitment and any knee buckling 30 days prior predicted KOA pain flares in the ensuing 30 days.