The musculoskeletal pain in Ullensaker study (MUST)
Residents of Ullensaker municipality (Norway), aged 42–80 years, with self-reported OA were invited to a comprehensive data-collection including medical examination (details previously published [13]). Individuals aged between 42 and 75 were selected for this study. The MUST study was approved by the Norwegian Regional Committee for Medical and Health Research Ethics South East (reference numbers 2009/812 and 2009/1703) and the participant gave their written informed consent according to the Declaration of Helsinki prior to inclusion.
Oslo community controls (OCP)
A random sample of 329 adult inhabitants of Oslo (Norway), aged 20–70, were selected by Statistics Norway, and individuals without an inflammatory joint disease were invited to a cardiovascular risk assessment at Diakonhjemmet Hospital. ABI, radiography and US joint examinations were not performed in this cohort [14]. In total 134 (41%) consented. The OCP was approved by the Norwegian Regional Committee for Medical and Health Research Ethics South East (reference numbers 2009/1703) and the participants gave their written informed consent according to the Declaration
UK population controls (UKPC)
For external validation of our findings we used anonymised data from The Anglo-Cardiff Collaborative Trial. This was a community population-based study of ≈ 12,000 individuals drawn from General Practice lists or open access cardiovascular risk assessment units in East-Anglia and Wales, United Kingdom. The response rate was 85%. From this cohort a random sample of 963 cases, aged 40–75 years, without self-reported inflammatory rheumatic disease or OA was included.
Joint examinations (MUST only)
In MUST, hand, knee and hip joints were examined and radiographs of hands, knees and hips taken as previously described [13]. Radiographic OA severity was scored according to the Kellgren-Lawrence (KL) scale (grade 0–4) for hands (distal and proximal interphalangeal, metacarphalangeal and first carpometacarpal joints), hip and knee joints. OA was defined according to the American College of Rheumatology (ACR) classification criteria [1,2,3]. For hand OA, only clinical variables are included in the criterion, whereas both clinical and radiographic variables are included in the criteria for hip and knee OA. When radiographs were missing, the clinical ACR criterion for knee OA was used if sufficient data was available. The cut-offs for joint space narrowing (JSN) and osteophytes in the ACR criterion for hip joints were grade ≥ 1 according to the Osteoarthritis Research Society International atlas [15]. Knee osteophytes were defined as KL grade ≥ 2. Persons with hip or knee prosthesis were classified as having OA in the respective joint. Participants were categorised according to the following OA phenotypes based on the fulfilment of the ACR criteria: 1) Isolated hand OA, 2) Isolated OA in lower limbs (i.e. knee and/or hip OA), 3) Generalized OA, defined as OA in both hands and lower limbs.
The protocol for ultrasonography and conventional radiographs of bilateral hands, knees and hips has been previously described [13]. Osteophytes GS and PD were semi-quantitatively scored on 0–3 scales (0 = none, 1 = minor, 2 = moderate and 3 = major presence of US pathology). Sum scores for ultrasound defined osteophytes were calculated for hands (range 0–90) and hips/knees (range 0–30), GS and PD in the hands (range 0–90) and GS hips/knees (range 0–4). KL sum scores were calculated for the hands (range 0–120) and hips/knees (range 0–16).
Biomarkers of CVD risk
Pulse wave analyses (PWV and AIx) were performed using the Sphygmocor apparatus (Atcor®) by trained personnel using the same methodology for all cohorts (MUST, OCP and UKPC) [13, 16]. ABI was measured in a standardized fashion. To obtain the ABI, the systolic pressure was estimated by a Sonotrax - Pocket Doppler Vascular - 8 MHz probe in the posterior tibial and dorsalis pedis arteries in both legs and in the brachial artery in one arm. The highest pressure recorded in either artery was recorded as the measurement for that side. The lowest of the distal pressures (ankle) measured either at the right or the left side was then divided by the brachial systolic pressure [12].
In MUST recordings of AIx were available in 258 (70.5%), PWV in 257 (70.2%) and ABI in 174 (47.5%) persons. Missing PWA recordings were related to logistical limitations of the study, unavailability of the PWA investigator and/or poor data quality according to predefined criteria. Due to time limitations ABI was recorded in every other patient as previously described [13]. In OPC, AIx was available in 104 (96.2%) and PWV in 97 (89.8%) of the participants. All UKPC participants had available AIx and PWV. ABI was not assessed in OCP and UKPC.
Covariates
All cohorts had information concerning age, gender body mass index (BMI) measured in kg/m2, C-reactive protein (CRP) measured as mg/L and smoking status, dichotomized as daily smoker vs. non-daily smoker or non-smoker. MUST and OCP also had information on level of education, dichotomized as college or university education vs. high school or lower education, use of non-steroidal anti-inflammatory drugs (daily vs. non-daily or never), physical functioning by the Modified Health Assessment Questionnaire (MHAQ) and self-reported exercise habits (> = once a week vs. < once a week) [13].
Statistics
This study was designed as a cross-sectional case-control study, with two separate control populations. Variables were visually examined for normality, and kurtosis and skewness were calculated. Baseline demographics were compared using chi-square tests, independent samples Student’s t-test or Mann-Whitney U-test as appropriate.
In linear regression models we examined the associations between ABI, AIx, PWV (dependent variables) and the severity of OA using ultrasonography and radiographic and sum scores. The estimated marginal means of AIx and PWV were estimated and compared between OA phenotypes, OCP and UKPC. ABI was compared between OA phenotypes. The associations between AIx, PWV, ABI and all covariates were tested in age and gender adjusted models, andco-variates associated with the dependent outcome with a p < = 0.01 were included in the final models. AIx models were also adjusted for heart rate and height. PWV models were adjusted for heart rate.
Missing data was not imputed, but a comparison of demographics between participants with available vs. missing PWA and ABI was performed. Cooks distances were calculated and outliers examined.