Background Osteoarthritis (OA) is associated with an increased risk of CVD1, but confounder bias has not been sufficiently explored in observational studies. Pulse wave velocity (PWV) and augmentation index (AIx) are predictors of future CVD events.
Objectives To investigate the relationship between OA and CVD, PWV and AIx, accounting for confounders.
Methods All residents of Ullensaker municipality, aged between 40 and 79 years were mailed a questionnaire concerning general health and musculoskeletal pain (Musculoskeletal pain in Ullensaker Study (MUST)). Respondents who reported having OA were invited to an extended clinical examination at Diakonhjemmet Hospital, Oslo, Norway. OA was diagnosed according to the ACR clinical classification criteria. Demographics and co-morbidities were recorded on questionnaires. AIx and PWV were measured (Sphygmocor apparatus, Atcor®). Participants in the Oslo control cohort were recruited by Statistics Norway, and underwent a similar data-collection (response rate 41%). Data from the Anglo-Cardiff Collaborative Trial (ACCT), a community based population study (response rate 85%) from the United Kingdom, formed a secondary control population. In separate logistical regression models, adjusted for age, sex, we compared the odds ratio (OR) for CVD and elevated AIx and PWV (dichotomized at median) between participants in MUST, ACCT controls and Oslo controls. Patients with hand OA were compared against patients with lower limb OA (hip and/or knee OA). AIx was adjusted for heart rate and height, whereas PWV was adjusted for heart rate. A confounder was defined as a variable which changed the estimate of the exposure effect by ≥20% when entered into the model.
Results Out of 583 patients in MUST with self-reported OA, 153 (26.2%) had isolated hand OA according to the ACR criteria and 153 (26.2%) had lower limb OA. 134 persons consented to participate in the Oslo control cohort. 1111 patients without inflammatory joint disorders from the ACCT were randomly selected as controls. The age and sex adjusted logistical regression models are presented in the table. Analyses with additional adjustment for smoking, BMI, CRP, anti-hypertensive and statin treatment, respectively, gave similar results.
Conclusions Patients with OA have an increased frequency of CVD, and higher levels of AIx and PWV, compared to ACCT controls. Lower limb OA seems to be associated with CVD to a greater extent than hand OA. No confounder variables were identified in these analyses.
Nuesch BMJ. 2011;342:d1165.
Disclosure of Interest None declared