The utility of novel accelerometer metrics for characterising clinical features in patients living with rheumatoid arthritis.

Abstract

Research-grade accelerometers are commonly used to measure physical activity (PA) in rheumatology research, demonstrating superior reliability and validity relative to self-report methods. Several accelerometers offer manufacturer software and embedded proprietary algorithms to reduce the complexities of data processing. However, algorithms vary between device brands, and hinder standardisation of data processing and analysis. Best practice in PA research is now therefore considered to be the collection and analysis of raw accelerometer data, to which transparent and replicable data transformation methods can be carried out post-processing. Novel metrics include average acceleration, intensity gradient, and MX metrics, which represent PA volume, intensity and patterns, and have not been examined in rheumatic diseases. To explore the utility of novel accelerometer metrics for characterising clinical features in Rheumatoid Arthritis (RA), i.e. disease activity and severity, cardiovascular disease (CVD) risk. People living with RA (n = 104) provided demographic data, medical history, a fasted blood sample, and completed the health assessment questionnaire (HAQ, disease severity). Disease activity was measured using the Disease Activity Score 28-CRP (DAS28-CRP), and CVD risk determined using the QRISK3. Participants wore a GT3X Actigraph accelerometer on their right hip for 7-days during waking hours. Accelerometer data were analysed using GGIR (v.2.1-1) to determine average acceleration (AA, mg, proxy for daily volume of PA), intensity gradient (IG = distribution of PA across the day) and MX metrics (acceleration above which a person's most active �X� mins are accumulated e.g., M5 = most active 5 mins). A higher AA and more positive IG indicate a favourable activity profile.�Statistical analyses:�Participants were grouped according to DAS28-CRP (remission = <2.6, low = 2.6 - 3.1, moderate = 3.2 � 5.1, high = >5.1) disease severity (HAQ; low = <1, moderate = 1 � 1.9, high = ?2) and QRISK3 (low = <10%, moderate = 10 - 19%, high = ?20%. Between group differences in AA and IG were analysed using analysis of variance, adjusted for accelerometer wear time. Radar plots were produced in R, to illustrate differences in MX metrics according to clinical features. Valid accelerometer data (?10 hr on ?4 days), were available for n = 102 participants (M �SD, AA = 13.7 �5.2 mg, IG = -2.91�.36). Participants in remission and with low CVD risk, demonstrated a better activity profile (i.e. [M �SE, all�p<.05] significantly higher AA [DAS28-CRP = 17.9 �1.1; QRISK3 = AA = 15.4 �0.6] and more positive IG [DAS28-CRP = -2.62 �0/07, QRISK3 = 2.80 �0.04], compared to patients with moderate or high disease activity and CVD risk (DAS28-CRP [moderate, AA = 12.8 �0.6, IG = -2.99�.04] and [high, AA = 11.2 �1.1, IG = -3.07�.08], QRISK3 [moderate, AA = 12.0 �1.0; IG = -3.0 �0.07] and [high AA = 10.3 �1.1; IG = -3.16 �0.08]). For disease severity IG was significantly more positive in patients with low (-2.78 �0.05) vs. high (-3.10 �0.08) HAQ scores. Radar plots (Figure 1) showed the intensity of the most active accumulated 2-45 mins (M2-M45) was greater (with M10 exceeding 75mg) among participants with better disease profiles (i.e. remission/low DAS28-CRP, HAQ and QRISK3 scores vs. moderate/high). This is the first study to demonstrate the clinical utility of novel accelerometer metrics in RA. Results suggest higher AA, more positive IGs, and accumulating ?10 mins at an intensity indicative of a slow walk (M10 >75mg), is characteristic of more favourable disease profiles. Future studies utilising these raw accelerometer metrics could provide valuable, standardised accelerometer data that can be used to deliver more personalised care (i.e. precision medicine).