The longitudinal relationship between hand, hip and knee osteoarthritis and cardiovascular events: a population-based cohort study

  • T. Kendzerska
    Affiliations
    The Ottawa Hospital Research Institute, University of Ottawa, Ottawa, ON, Canada

    Institute for Clinical Evaluative Sciences, Ottawa, ON, Canada

    Institute for Clinical Evaluative Sciences/Sunnybrook Research Institute, Toronto, ON, Canada
    Search for articles by this author
  • P. Jüni
    Affiliations
    Applied Health Research Centre, The Li Ka Shing Knowledge Institute of St. Michael's Hospital, Toronto, ON, Canada

    Department of Medicine, University of Toronto, Toronto, ON, Canada
    Search for articles by this author
  • L.K. King
    Affiliations
    Department of Medicine, University of Toronto, Toronto, ON, Canada
    Search for articles by this author
  • R. Croxford
    Affiliations
    Institute for Clinical Evaluative Sciences/Sunnybrook Research Institute, Toronto, ON, Canada
    Search for articles by this author
  • I. Stanaitis
    Affiliations
    Women's College Research Institute, Women's College Hospital, Toronto, ON, Canada
    Search for articles by this author
  • G.A. Hawker
    Correspondence
    Address correspondence and reprint requests to: G.A. Hawker, Sir John and Lady Eaton Professor and Chair, Department of Medicine, University of Toronto, 190 Elizabeth Street, Suite RFE 3-805, Toronto, Ontario, M5G 2C4, Canada. Tel.: 1-416-946-8071; Fax: 1-416-978-7230.
    Affiliations
    Institute for Clinical Evaluative Sciences/Sunnybrook Research Institute, Toronto, ON, Canada

    Department of Medicine, University of Toronto, Toronto, ON, Canada

    Women's College Research Institute, Women's College Hospital, Toronto, ON, Canada
    Search for articles by this author
Open AccessPublished:August 08, 2017DOI:https://doi.org/10.1016/j.joca.2017.07.024

      Summary

      Objective

      In this population-based cohort study, we examined the association between the presence of symptomatic osteoarthritis (OA) and risk for cardiovascular (CV) events.

      Method

      A cohort aged ≥55 years recruited from 1996 to 98 was followed through provincial health administrative data to 2014. Demographics, joint complaints and functional limitations were collected. Hip, knee and hand OA were defined using a validated definition. Using Cox-regressions, the relationship between OA and a composite CV outcome (myocardial infarction (MI), stroke, angina, heart failure, revascularization) was assessed controlling for age, body mass index (BMI), sex, pre-existing metabolic factors, comorbidities, income status, primary care exposure and functional limitations.

      Results

      18,490 participants were included: median age was 68 years, 60.3% were female; 24.4% met criteria for OA (10.0% hip, 15.3% knee, 16.0% hand), 16.3% self-reported limitation in grip and 25.4% in walking. Over a median 13.4 years, 31.9% experienced a CV event. Controlling for all but walking limitation, a dose–response relationship was observed between number of joints affected by knee/hip OA and CV risk (HR 2 hips/knees vs none: 1.13, 95% CI 1.03–1.23; 3+ hips/knees: 1.22, 95% CI 1.09–1.36). This relationship became non-significant additionally controlling for difficulty walking. Self-reported difficulty walking was associated with a 30% increased hazard for CV events. The effect of hand OA was not significant.

      Conclusion

      In a large population cohort, a greater burden of hip/knee OA was associated with higher CV risk; the relationship was explained by OA-related difficulty walking. Increased attention to management of OA with a view to improving mobility has potential to reduce CV events.

      Keywords

      Introduction

      North Americans are witnessing an unprecedented rise in the number of people living with obesity and surviving well into old age
      • Perruccio A.V.
      • Power J.D.
      • Badley E.M.
      Revisiting arthritis prevalence projections--it's more than just the aging of the population.
      • Ackerman I.N.
      • Osborne R.H.
      Obesity and increased burden of hip and knee joint disease in Australia: results from a national survey.
      , and thus with multiple chronic conditions
      • Hawker G.A.
      • Gignac M.A.
      • Badley E.
      • Davis A.M.
      • French M.R.
      • Li Y.
      • et al.
      A longitudinal study to explain the pain-depression link in older adults with osteoarthritis.
      . At a primary care and population level, the most common clusters of conditions include hypertension, diabetes, cardiovascular disease (CVD), and osteoarthritis (OA)
      • Puenpatom R.A.
      • Victor T.W.
      Increased prevalence of metabolic syndrome in individuals with osteoarthritis: an analysis of NHANES III data.
      • Yoshimura N.
      • Muraki S.
      • Oka H.
      • Tanaka S.
      • Kawaguchi H.
      • Nakamura K.
      • et al.
      Accumulation of metabolic risk factors such as overweight, hypertension, dyslipidaemia, and impaired glucose tolerance raises the risk of occurrence and progression of knee osteoarthritis: a 3-year follow-up of the ROAD study.
      • Sowers M.
      • Karvonen-Gutierrez C.A.
      • Palmieri-Smith R.
      • Jacobson J.A.
      • Jiang Y.
      • Ashton-Miller J.A.
      Knee osteoarthritis in obese women with cardiometabolic clustering.
      • Boeckxstaens P.
      • Peersman W.
      • Goubin G.
      • Ghali S.
      • De Maeseneer J.
      • Brusselle G.
      • et al.
      A practice-based analysis of combinations of diseases in patients aged 65 or older in primary care.
      .
      The clustering of OA with hypertension, diabetes and CVD
      • Puenpatom R.A.
      • Victor T.W.
      Increased prevalence of metabolic syndrome in individuals with osteoarthritis: an analysis of NHANES III data.
      has led to speculation about a ‘metabolic OA’ phenotype. Obesity, dyslipidemia, impaired fasting glucose, and hypertension have been hypothesized to contribute to OA incidence and progression
      • Zhuo Q.
      • Yang W.
      • Chen J.
      • Wang Y.
      Metabolic syndrome meets osteoarthritis.
      • Berenbaum F.
      Osteoarthritis as an inflammatory disease (osteoarthritis is not osteoarthrosis!).
      and, in turn, OA has been hypothesized as an independent ‘metabolic’ risk factor for CVD. A number of cross-sectional studies have demonstrated a positive association between the presence of OA, variably defined, and both metabolic syndrome and CVD
      • Puenpatom R.A.
      • Victor T.W.
      Increased prevalence of metabolic syndrome in individuals with osteoarthritis: an analysis of NHANES III data.
      • Yoshimura N.
      • Muraki S.
      • Oka H.
      • Tanaka S.
      • Kawaguchi H.
      • Nakamura K.
      • et al.
      Accumulation of metabolic risk factors such as overweight, hypertension, dyslipidaemia, and impaired glucose tolerance raises the risk of occurrence and progression of knee osteoarthritis: a 3-year follow-up of the ROAD study.
      • Liu S.H.
      • Waring M.E.
      • Eaton C.B.
      • Lapane K.L.
      Association of objectively measured physical activity and metabolic syndrome among US adults with osteoarthritis.
      ; these relationships appear stronger for hand and knee OA than for hip OA. More recently, it has also been shown in longitudinal studies that people with symptomatic hip, knee, and possibly hand OA are at increased risk for CVD
      • Nuesch E.
      • Dieppe P.
      • Reichenbach S.
      • Williams S.
      • Iff S.
      • Juni P.
      All cause and disease specific mortality in patients with knee or hip osteoarthritis: population based cohort study.
      • Haugen I.K.
      • Ramachandran V.S.
      • Misra D.
      • Neogi T.
      • Niu J.
      • Yang T.
      • et al.
      Hand osteoarthritis in relation to mortality and incidence of cardiovascular disease: data from the Framingham heart study.
      • Hawker G.A.
      • Croxford R.
      • Bierman A.S.
      • Harvey P.J.
      • Ravi B.
      • Stanaitis I.
      • et al.
      All-cause mortality and serious cardiovascular events in people with hip and knee osteoarthritis: a population based cohort study.
      relative to the age- and sex-matched general population or to non-OA controls. However, interpretation of studies to date has been limited by inadequate control for elevated body mass index (studies have often controlled for obesity as a dichotomous variable), and components of the metabolic syndrome (hypertension, dyslipidemia and hyperglycemia). Also, two of the three studies were restricted to persons with OA and thus could not compare CV risk in persons with vs without OA
      • Nuesch E.
      • Dieppe P.
      • Reichenbach S.
      • Williams S.
      • Iff S.
      • Juni P.
      All cause and disease specific mortality in patients with knee or hip osteoarthritis: population based cohort study.
      • Hawker G.A.
      • Croxford R.
      • Bierman A.S.
      • Harvey P.J.
      • Ravi B.
      • Stanaitis I.
      • et al.
      All-cause mortality and serious cardiovascular events in people with hip and knee osteoarthritis: a population based cohort study.
      . Finally, studies to date have not simultaneously controlled for metabolic factors and OA severity, in particular the effects of painful, disabling OA on mobility and other comorbid conditions that may affect physical function and which may also increase CVD risk by contributing to weight gain and metabolic alterations
      • Caporali R.
      • Cimmino M.A.
      • Sarzi-Puttini P.
      • Scarpa R.
      • Parazzini F.
      • Zaninelli A.
      • et al.
      Comorbid conditions in the AMICA study patients: effects on the quality of life and drug prescriptions by general practitioners and specialists.
      • Sale J.E.
      • Gignac M.
      • Hawker G.
      How “bad” does the pain have to be? A qualitative study examining adherence to pain medication in older adults with osteoarthritis.
      • van Dijk G.M.
      • Veenhof C.
      • Schellevis F.
      • Hulsmans H.
      • Bakker J.P.
      • Arwert H.
      • et al.
      Comorbidity, limitations in activities and pain in patients with osteoarthritis of the hip or knee.
      . Thus, it remains unclear if OA is an independent risk factor for CVD events and, if so, why. Given the high population prevalence of OA and CVD, elucidation of the relationship between these conditions is important.
      In this large population-based longitudinal cohort study, we examined the association between presence of symptomatic hand, hip and knee OA at baseline and subsequent risk for CV events and the effects of common risk factors, sex, metabolic factors, comorbidity, health care utilization, income status and functional limitations on these relationships. We hypothesized an independent association between OA at baseline and CV events, which would be explained by reduced mobility in individuals with OA.

      Methods

       Study participants

      The Ontario Hip and Knee Cohort was recruited from 1996 to 1998 through a survey of all 39,333 individuals aged 55+ years living in two regions – one rural and one urban – of Ontario, Canada
      • Hawker G.A.
      • Wright J.G.
      • Coyte P.C.
      • Williams J.I.
      • Harvey B.
      • Glazier R.
      • et al.
      Determining the need for hip and knee arthroplasty: the role of clinical severity and patients' preferences.
      . We excluded individuals who did not self-complete the screening questionnaire or who self-reported lower-limb amputation or wheelchair use (Fig. 1). Respondents who did not self-complete their questionnaire were excluded as responses on symptoms completed by a proxy may not be valid. In addition, these individuals may be unable to complete the questionnaire due to cognitive impairment or illness. We also excluded respondents with inflammatory arthritis (IA) or other rheumatologic diseases based on self-report or health administrative data as high inflammatory burden associated with IA is a known risk factor for CVD
      • Choy E.
      • Ganeshalingam K.
      • Semb A.G.
      • Szekanecz Z.
      • Nurmohamed M.
      Cardiovascular risk in rheumatoid arthritis: recent advances in the understanding of the pivotal role of inflammation, risk predictors and the impact of treatment.
      . The study was approved by the institutional ethics review boards at Women's College Hospital and Sunnybrook Health Sciences Centre, Toronto, Canada.

       Patient involvement

      Prior qualitative research (focus groups and one-on-one interviews) by our group in people with OA directly informed the current study
      • Sale J.E.
      • Gignac M.
      • Hawker G.
      How “bad” does the pain have to be? A qualitative study examining adherence to pain medication in older adults with osteoarthritis.
      • Hawker G.A.
      • Stewart L.
      • French M.R.
      • Cibere J.
      • Jordan J.M.
      • March L.
      • et al.
      Understanding the pain experience in hip and knee osteoarthritis–an OARSI/OMERACT initiative.
      • Gignac M.A.
      • Davis A.M.
      • Hawker G.
      • Wright J.G.
      • Mahomed N.
      • Fortin P.R.
      • et al.
      “What do you expect? You're just getting older”: a comparison of perceived osteoarthritis-related and aging-related health experiences in middle- and older-age adults.
      . This work identified major barriers to OA care, including the high prevalence of comorbidities, which present competing demands and are deemed ‘more important’. However, we know from our work and others that people with multiple conditions, including those with OA, prioritize pain relief and mobility to maintain independence over other health outcomes. This is concerning since our qualitative research has also shown that people with OA give up moving (walking) to manage their OA symptoms rather than use ‘risky pain killers’ or because they have been offered nothing else. Despite these prior findings, which speak to the need to address chronic painful OA to improve people's quality of life, we have experienced a remarkable lack of interest among health professionals in OA. Given that lack of physical activity is a known risk factor for diabetes and heart disease, we decided to use existing data on a population cohort to examine the relationship between OA and CV events and whether or not the relationship, if found, was explained by reduced mobility as measured by walking difficulty. Through this research our hope is to increase understanding of the importance of managing OA to maintain mobility, and in this way, help to advance management of this common chronic condition.
      Recruitment of the original cohort was in the 1990's and involved active engagement of the two communities (one urban, one rural) – we sought support from local government, arthritis support groups, religious institutions, local papers, and other resources to assist in recruitment. We spoke at community forums for the lay public about the study – and how best to make participation rewarding for their communities. We used regular cohort newsletters and an annual luncheon in each community to provide feedback to participants regarding study progress and findings. At the annual events, we incorporated education on OA provided by the local Arthritis Society of Canada therapists and others. The cohort was 55+ years of age at recruitment in 1996 and thus the youngest participants are now in their mid to late 70's. For this reason, and because we are no longer following the cohort, the current results will not be disseminated to these participants directly. However, our research group is actively engaged in the Arthritis Alliance of Canada (arthritisalliance.ca). Formed in 2002 to improve the lives of Canadians with arthritis, the Alliance involves more than 36 organizations representing arthritis health care professionals, researchers, funding agencies, governments, voluntary sector agencies, industry and, most importantly, representatives from arthritis patient organizations from across Canada. Our findings have been shared with this group and through this group, may be disseminated broadly.

       Data sources

      A standardized mail/telephone survey assessed demographics, body mass index (BMI) from self-reported height and weight, joint complaints, functional limitations, and self-reported comorbidity.
      Baseline data were linked to data available from the Institute for Clinical Evaluative Sciences (ICES) using unique encoded identifiers and analyzed at ICES. The ICES data repository consists of individual-level, longitudinal, coded and linkable health databases on most publicly funded health services for the Ontario population (http://www.ices.on.ca/Data-and-Privacy/ICES-data). These include information on discharge summaries of emergency department (ED) visits and hospital stays, outpatient visits, and for those 65 years and older, medical drug claims to the Ontario Drug Benefit Program. To document deaths that occurred out of hospital, the Registered Persons Database was used. Details on these databases and variables included are available at https://datadictionary.ices.on.ca/Applications/DataDictionary/Default.aspx.

       Exposures

      Symptomatic OA was defined as (1) self-reported swelling, pain, or stiffness in any joint lasting ≥6 weeks in the past 3 months; and (2) indication on a joint homunculus that a hip, knee or hand (any of fingers, wrist or base of thumb) was “troublesome”. Previously, on a random subsample, we showed that 98% of study subjects who met these criteria had hip and/or knee OA confirmed on physical examination (sensitivity of 74.4% and specificity of 76.1%) and 66% met American College of Rheumatology (ACR) criteria for OA in a hip or a knee joint based on musculoskeletal examination and radiographs (sensitivity of 78.3% and specificity of 24.6%)
      • Hawker G.A.
      • Wright J.G.
      • Coyte P.C.
      • Williams J.I.
      • Harvey B.
      • Glazier R.
      • et al.
      Determining the need for hip and knee arthroplasty: the role of clinical severity and patients' preferences.
      . OA burden was assessed by the number of hips, knees and hands affected (0–2 for each of hip, knee and hand OA) as well as number of hips and knees with OA (0–4).

       Outcome

      Our outcome was time from baseline to any hospitalization or ED visit for CVD (myocardial infarction (MI), stroke, angina, congestive heart failure (CHF), and revascularization procedures). Hospitalizations and ED visits were chosen as a robust outcome measure of interest for patients, physicians and policy makers.

       Risk factors and confounders

      Potential baseline risk factors and confounders were grouped as: (1) common risk factors: age, BMI and sex; (2) metabolic factors: hypertension, diabetes and CVD (MI, angina, CHF, revascularization procedures and stroke) defined from health administrative data; (3) other factors: neighborhood income status, additional comorbidities considered as separate conditions (chronic obstructive pulmonary disease (COPD) and dementia defined from health administrative data, and self-reported malignancy) and as an aggregated score, prior receipt of hip or knee arthroplasty, and primary care exposure; and (4) OA-related functional limitations (Fig. E1).
      Validated algorithms were used to define comorbidities at baseline from health administrative data
      • Tu K.
      • Campbell N.R.
      • Chen Z.L.
      • Cauch-Dudek K.J.
      • McAlister F.A.
      Accuracy of administrative databases in identifying patients with hypertension.
      • Alter D.A.
      • Chong A.
      • Austin P.C.
      • Mustard C.
      • Iron K.
      • Williams J.I.
      • et al.
      Socioeconomic status and mortality after acute myocardial infarction.
      • Hux J.E.
      • Ivis F.
      • Flintoft V.
      • Bica A.
      Diabetes in Ontario: determination of prevalence and incidence using a validated administrative data algorithm.
      . Severity of comorbidities at baseline was approximated using an aggregated score, the Johns Hopkins' Aggregated Diagnosis Groups categories (The Johns Hopkins ACG® System, Version 10), which has been shown to be associated with all-cause mortality
      • Austin P.C.
      • Walraven C.
      The mortality risk score and the ADG score: two points-based scoring systems for the Johns Hopkins aggregated diagnosis groups to predict mortality in a general adult population cohort in Ontario.
      , and, for those eligible for drug benefits during the pre-baseline year, by treatment prescribed for CVD and diabetes. Number of outpatient visits to primary care physicians in the 2 years prior to the baseline assessment was considered as a measure of health care utilization, and prior receipt of hip or knee arthroplasty as a proxy for OA severity. Limitation in grip was defined as self-reported difficulty with gripping and holding in the last 3 months. Walking limitation was defined as self-reported difficulty standing or walking in the last 3 months. Details on definitions are provided in the Data Supplement.

       Analyses

      Baseline cohort characteristics and the proportions that experienced the outcome were described using means (SD), medians (IQR) and proportions as appropriate. Event-free survival was estimated using the Kaplan–Meier method and compared between subjects with and without OA at baseline using the log-rank test.
      The primary analyses used Cox-regressions to assess the relationship between baseline self-reported OA (number of hips, knees and hands with OA, separately and combined, and then the total number of upper (hands) and lower extremity joints (hips and knees) affected by OA) and the primary outcome, controlling for covariates. Covariates were entered into the statistical models sequentially, as outlined above to assess the change in the estimated effect for exposure-related variables after each step. The Cox proportional hazards regression assumptions for each variable were tested using a graphical approach (visually examining the distribution of Schoenfeld residuals) and an analytic method (a global test of the proportional hazards)
      • Harrell F.
      Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis.
      • Grambsch P.
      • Therneau T.
      Proportional hazards tests and diagnostics based on weighted residuals.
      . Interactions between OA and both sex and baseline CVD, specified a priori, were also evaluated. We investigated whether functional limitations explained the association between OA and outcomes by inclusion of walking disability in our final models, and by investigating separately the effect of functional limitations on the risk of CV events controlling for confounders.
      To elaborate more, we used the approach of Lange et al.
      • Lange T.
      • Vansteelandt S.
      • Bekaert M.
      A simple unified approach for estimating natural direct and indirect effects.
      to test if the observed effect of hip or knee OA on CV events is at least partially mediated through walking limitation. The corresponding code written in R statistical programming language published by Rochon et al. was utilized
      • Rochon J.
      • du Bois A.
      • Lange T.
      Mediation analysis of the relationship between institutional research activity and patient survival.
      .
      To estimate the incremental value of OA over traditional CV risk factors (common and metabolic risk factors outlined above) on absolute risk of CV events, a clinical nomogram was constructed. A nomogram is a graphical representation of a mathematical model which generates individualized predictions by converting a predictive model (e.g., Cox proportional hazards model for survival data) into a single numerical estimate of the probability of an event
      • Iasonos A.
      • Schrag D.
      • Raj G.V.
      • Panageas K.S.
      How to build and interpret a nomogram for cancer prognosis.
      . The five- and 10-year predicted CV event-free survival probabilities and median survival time were calculated using the multivariable Cox regression model, which includes the number of hip/knee joints affected by OA and traditional CV risk factors, underlying the nomogram. Based on the estimated regression coefficients, points were assigned to each predictor system (from 0 to 100 points). The predictor with the biggest impact was identified as the reference (assigned 100 points); other predictors were then assigned points based on their proportional impact on the outcome of interest relative to the reference.
      To account for the competing risk of death, we also investigated the effect of OA and functional limitations on CVD in a secondary analysis using Fine and Gray competing-risk regressions
      • Fine J.P.
      • Gray R.J.
      A proportional hazards model for the subdistribution of a competing risk.
      . Death was considered as a competing event, because of expected high all-cause mortality in our population, which may preclude hospitalization or ED visits for CV events or alter the chances to observe it, resulting in a biased estimate of CV risk as a stand-alone outcome
      • Austin P.C.
      • Lee D.S.
      • Fine J.P.
      Introduction to the analysis of survival data in the presence of competing risks.
      .
      Missing data were observed on income status (0.1%), height (18.0%) and weight (7.2%). We used multivariate imputation by chained equations to generate twenty complete data sets
      • Azur M.J.
      • Stuart E.A.
      • Frangakis C.
      • Leaf P.J.
      Multiple imputation by chained equations: what is it and how does it work?.
      . Thirty two variables including both the event status and the survival time were included in the imputation model (details are presented in the Data Supplement). We used Rubin's rules to pool the separate estimates and standard errors from imputed datasets
      • Rubin D.
      Multiple Imputation for Nonresponse in Surveys.
      .
      Analyses were conducted using R Version 2.15.2 (www.r-project.org).

      Results

      Of 28,451 respondents (72.3% response rate), 25,388 individuals were linked to provincial health administrative data to enable assessment of health care use from 1991 to 2014, and 18,490 participants were included after applying exclusion criteria (Fig. 1). Participants median baseline age was 68 years (IQR: 62–75), 60.3% were female and median BMI was 25.5 kg/m2 (IQR: 23.0–28.4). 1,856 participants (10%) met criteria for hip OA, 2,830 (15%) for knee OA, and 2,961 (16.0%) for hand OA (Table I). Self-reported limitation in grip and walking were present in 3,020 subjects (16.3%) and 4,690 subjects (25.4%), respectively. Individuals with hip, knee or hand OA were more likely than those without OA to be female, have comorbidities, and to report functional limitations (Table I).
      Table IBaseline cohort characteristics for the total sample and by presence of osteoarthritis
      CovariatesOverall (n = 18,490)Any OA (hand/knee/hip)
      Yes (n = 4,516)No (n = 13,974)
      Socio-demographics
      Age (years), median (IQR)68 (62–75)68 (62–75)68 (62–75)
      Sex (female), n (%)11,122 (60)3,053 (68)8,069 (58)
      Income, n (%)
       Quintile 12,993 (16)783 (17)2,210 (16)
       Quintile 22,625 (15)637 (14)2,048 (15)
       Quintile 34,402 (24)1,024 (23)3,378 (24)
       Quintile 44,217 (23)1,036 (23)3,181 (23)
       Quintile 54,178 (23)1,034 (23)3,144 (23)
      Health conditions
      BMI (kg/m2), median (IQR)25.5 (23.0–28.4)26.2 (23.5–29.4)25.2 (22.8–28.0)
      Categories, n (%)
       Underweight (<18.5 kg/m2)331 (2)57 (1)274 (2)
       Normal (18.5–24.9 kg/m2)6,150 (33)1,276 (28)4,874 (35)
       Pre-obese (25.0–29.9 kg/m2)5,751 (31)1,350 (30)4,401 (31)
       Obese (≥30.0 kg/m2)2,236 (12)768 (17)1,468 (11)
       Missing4,022 (22)1,065 (24)2,957 (21)
      Cardiovascular disease, n (%)
       AMI405 (2.2)92 (2.0)313 (2.2)
       Angina574 (3.1)166 (3.7)408 (2.9)
       CHF409 (2.2)106 (2.3)303 (2.2)
       CABG/PCI270 (1.5)73 (1.6)197 (1.4)
      Hypertension, n (%)7,787 (42)2069 (45.8)5,718 (40.9)
      Stroke, n (%)360 (1.9)87 (1.9)273 (2.0)
      Diabetes, n (%)2,067 (11)551 (12.2)1,516 (10.9)
      COPD, n (%)859 (4.6)448 (9.9)1,112 (8.0)
      Malignancy (self-report), n (%)383 (2.1)101 (2.2)282 (2.0)
      Mental Health, n (%)157 (0.8)49 (1.1)108 (0.8)
      ADG categories
       High2,925 (16)974 (22)1951 (14)
       Medium6,419 (35)1,716 (38)4,703 (34)
       Low9,146 (49)1,826 (40)7,320 (52)
      Osteoarthritis (1 + joint), n (%)
      Hand2,961 (16)2,961 (66)
      Knee2,830 (15)2,830 (63)
      Hip1,856 (10)1,856 (41)
      Prior hip or knee total joint arthroplasty506 (2.7)238 (5.3)268 (1.9)
      Functional limitations, n (%)
      In gripping/holding3,020 (16.3)1,745 (38.6)1,275 (9.1)
      In walking/standing4,690 (25.4)2,460 (54.5)2,230 (16.0)
      Outcomes
      CV events5,898 (31.9)1,529 (33.9)4,369 (31.3)
      All-cause mortality9,689 (52.4)2,388 (52.9)7,301 (52.3)
      ADGs – Johns Hopkins' Aggregated Diagnosis Groups, BMI – body mass index, AMI – acute myocardial infarction, CHF – congestive heart failure, CABG – coronary artery bypass grafting, CV – cardiovascular, IQR – interquartile range, COPD – chronic obstructive pulmonary disease, TJA – total joint arthroplasty.

       Primary analyses

      Over a median follow-up of 13.4 years (IQR: 6.2–17.8 years), 5,898 (31.9%) individuals experienced one or more CV events (2,759 hospitalizations for CHF, 1,865 for MI, 1,844 for stroke, 1,344 for angina and 1,238 for a cardiac revascularization procedure) (Table I; Table E1).
      In unadjusted analyses, individuals with knee or hip, but not hand, OA were significantly more likely to experience a CV event (Table E1). Compared with individuals with no hips/knees with OA, a dose–response relationship was observed between number of joints affected by knee/hip OA and CV risk in the univariable model (Fig. 2, Fig. 3). Adjustment for each of age, BMI, sex, metabolic factors and covariates other than walking limitation only modestly attenuated this relationship (adjusted HR vs none: two hips/knees 1.13, 95% CI 1.03–1.23; 3+ hips/knees 1.22, 95% CI 1.09–1.36) (Fig. 3 and Table E2). Further adjustment of the model for difficulty walking attenuated these relationships such that they became non-significant (adjusted HR for two hips/knees vs none: 1.00, 95% CI 0.91–1.10; 3+ hips/knees 1.05, 95% CI 0.94–1.19) (Fig. 3, Table E2). Self-reported difficulty walking was associated with a 30% increased hazard of CV events development (adjusted HR of 1.30, 95% CI 1.23–1.38) (Fig. 2, Table E3).
      Fig. 2
      Fig. 2Unadjusted Kaplan–Meier estimates of cardiovascular event-free survival by (A) the number of joints affected by knee/hip osteoarthritis (0, 1, 2, 3+) and (B) the presence of self-reported walking limitation.
      Fig. 3
      Fig. 3Caterpillar plot of the effects of the number of joints affected by knee/hip OA (0, 1, 2, 3+) (hazard ratio and 95% confidence interval) on cardiovascular even-free survival unadjusted and adjusted for confounders. Model 1: Age, sex and body mass index (BMI); Model 2–Age, sex, BMI, pre-existing cardiovascular disease (CVD), hypertension and diabetes; Full model–Age, sex, BMI, pre-existing CVD, hypertension, diabetes, pre-existing other comorbidities, income status and prior health care utilization.
      Taking into account baseline confounders included in the full model, the overall adjusted hazard ratio of CV events associated with the presence of hip or knee OA was 1.10 (95% CI: 1.01 to 1.18). The proportion of patients that reported walking limitations was 63% in individuals with hip or knee OA and 17% in those without (adjusted OR = 7.72, 95% CI: 7.00 to 8.51). The total HR of 1.10 was decomposed into a direct HR = 0.98 and an indirect HR, which corresponds to the mediator effect, of 1.12. This suggests that the entire effect of hip or knee OA was mediated by difficulty walking.
      No significant interactions were found between OA and sex or baseline CVD.

       Clinical nomogram

      Clinical nomograms were based on the statistical model that included the number of hip/knee joints affected by OA and traditional CV risk factors (age, sex, BMI, prior hypertension, diabetes, and pre-existing CVD) (Fig. 4). Variables with the widest point range in the nomogram provide the greatest discrimination. The effect of 3+ OA hips/knees (6 points) was comparable to the effects of obesity (7 points) and hypertension (7 points) on CVD risk. The points assigned to each predictor are presented in the Data Supplement (Table E4).
      Fig. 4
      Fig. 4Clinical nomogram. To obtain the nomogram predicted probability of 5- and 10-year CV event-free survival and to estimate median event-free survival, locate patient values at each axis, then draw a vertical line to the ‘‘Point’’ scale (axis) to determine how many points are attributed for each predictor. Sum the points for all predictors. Locate the sum on the ‘‘Total Points’’ scale. Draw a vertical line towards the ‘‘5-year Survival,’’ ‘‘10-year Survival,’’ and ‘‘Median Survival Time’’ axes to determine the 5-year CV outcome-free survival, the 10-year event-free survival, and to estimate median survival respectively.
      To demonstrate the effect of hip/knee OA on absolute CVD risk, we compared the estimated risk for a 65-year-old obese woman with hypertension and diabetes (a typical cohort participant) with and without pre-existing CVD and with and without hip/knee OA. As shown in the nomogram, the base case (no CVD or hip/knee OA) has a score of 57 points indicating a 10-year event-free survival of 70–80%. If this same woman has 3+ hip/knees with OA, but no CVD, her score is 63, giving a 10% lower estimated 10-year event-free survival of 60–70%. If she has pre-existing hypertension, diabetes and CHF, but no hip/knee OA or other CVD, her score is 83, giving a 10-year event-free survival of 40–50%. Finally, if she has pre-existing hypertension, diabetes, CHF, and 3+ hips/knees with OA, her score is 89, which reduces her 10-year event-free survival to 30–40%.

       Secondary analyses

      When analyzed using a competing risk approach, after further adjustment of the model for difficulty walking the relationship between number of joints affected by hip/knee OA was attenuated but remained significant (adjusted HR for two hips/knees vs none: 1.07, 95% CI 0.97–1.19; 3+ hips/knees 1.16, 95% CI 1.02–1.31) (Table E2). The effect of knee/hip OA on the competing event, all-cause mortality, was non-significant (Table E2).
      Similar results were obtained when knee OA or hip OA were considered separately as exposures (Table E5).

      Discussion

      In a large population cohort aged ≥55 years, one in five participants had symptomatic OA in at least one hip or knee. Hip and knee OA, but not hand OA, was a significant independent predictor of incident or recurrent CV events after controlling for common risk factors, metabolic factors, sex, and additional comorbid conditions. A dose–response relationship was observed; individuals with a greater number of hips and knees affected experienced a higher risk for CV events. The absolute effect of having at least three hip/knee joints with OA was comparable to the effects of obesity or hypertension on risk for subsequent CV events. This relationship became non-significant when we further controlled for difficulty walking at baseline. Self-reported difficulty walking was associated with a 30% increased hazard of CV events development controlling for confounders. These findings provide compelling evidence to suggest that hip/knee OA-related walking difficulty is a clinically relevant and potentially modifiable risk factor for CVD.
      Consistent with other studies, we found no increase in mortality risk in patients with knee or hip OA, although this was not the primary outcome for our study
      • Turkiewicz A.
      • Neogi T.
      • Bjork J.
      • Peat G.
      • Englund M.
      All-cause mortality in knee and hip osteoarthritis and rheumatoid arthritis.
      • Xing D.
      • Xu Y.
      • Liu Q.
      • Ke Y.
      • Wang B.
      • Li Z.
      • et al.
      Osteoarthritis and all-cause mortality in worldwide populations: grading the evidence from a meta-analysis.
      .
      Study strengths include reliance on a large population cohort, long and complete follow-up, validated assessment of outcomes using health administrative data, use of validated measures for self-reported hip/knee OA and disability, consideration of OA affecting weight-bearing vs non-weight-bearing joints (mobility vs systemic inflammation), careful control for myriad risk factors for CVD and sedentary behaviour, evaluation of both the relative and absolute effects of OA on CVD risk, and use of a competing risk approach to account for the competing risk of death. Given these strengths, we believe our findings to be widely generalizable to the broader population of older adults living with OA and other chronic conditions. However, some study limitations should be noted. First, as in all observational studies, there is the potential for unmeasured confounders (e.g., lack of information on lipid levels, abdominal obesity, smoking status and family history of CVD). We did not have information on NSAID use or smoking, which are known risk factors for CVD events. As NSAIDs are used to control OA-related pain, their use may be greater in those with vs without walking disability; thus, we cannot rule out a further mediating effect of NSAIDs on the relationship between OA and CV events. Although we did not have information on smoking, we controlled for COPD, which is strongly related to smoking. We did not control for changes over the follow-up period in OA symptom severity, or other covariates, as we were interested in the association between presence of symptomatic hand, hip and knee OA and functional limitations at baseline on the outcomes of interest. Our criteria for hand OA have not been validated. The presence of walking difficulty was based on self-report, which may over- or under-estimate capacity. However, concordance between performance-based and self-reported measures of mobility is generally high
      • Capistrant B.D.
      • Glymour M.M.
      • Berkman L.F.
      Assessing mobility difficulties for cross-national comparisons: results from the World health organization study on global AGEing and adult health.
      . Further, the strength of the relationship between our simple measure of walking difficulty and CV events suggest its potential usefulness in clinical practice. Finally, we attributed difficulty walking to hip and knee OA. In prior work we showed hip/knee OA to be the main contributor to walking difficulty among older adults
      • Kendzerska T.
      • King L.
      • Croxford R.
      • Stanaitis I.
      • Wall A.
      • Hawker G.
      The impact of hip and knee osteoarthritis and associated walking limitations on all-cause mortality: a population-based cohort study [abstract].
      and in the current study, we comprehensively controlled for other contributing factors. Thus, we feel this assumption is valid.
      The relationship between OA and a composite CVD outcome was recently summarized by Hall AJ et al.
      • Hall A.J.
      • Stubbs B.
      • Mamas M.A.
      • Myint P.K.
      • Smith T.O.
      Association between osteoarthritis and cardiovascular disease: systematic review and meta-analysis.
      in a meta-analysis of fifteen cross sectional and longitudinal studies. The authors reported an increased risk of incident or recurrent CV events in subjects with OA vs matched non-OA controls (adjusted relative risk 1.69, 95% CI 1.13–2.53). However, the studies included variably controlled for a number of potential confounders, e.g., pre-existing CVD and diabetes, had relatively short durations of follow-up, and did not examine for differential effects on CVD risk based on joint site or burden of OA.
      Potential explanation for a causal relationship between OA and CVD includes the impact of OA on physical activity. In people with hip and knee OA, baseline level of walking difficulty has been shown to be a significant independent predictor of subsequent risk for CV events and all-cause death
      • Nuesch E.
      • Dieppe P.
      • Reichenbach S.
      • Williams S.
      • Iff S.
      • Juni P.
      All cause and disease specific mortality in patients with knee or hip osteoarthritis: population based cohort study.
      • Hawker G.A.
      • Croxford R.
      • Bierman A.S.
      • Harvey P.J.
      • Ravi B.
      • Stanaitis I.
      • et al.
      All-cause mortality and serious cardiovascular events in people with hip and knee osteoarthritis: a population based cohort study.
      and post-baseline receipt of total hip or knee replacement has been shown to be protective
      • Ravi B.
      • Croxford R.
      • Austin P.C.
      • Lipscombe L.
      • Bierman A.S.
      • Harvey P.J.
      • et al.
      The relation between total joint arthroplasty and risk for serious cardiovascular events in patients with moderate-severe osteoarthritis: propensity score matched landmark analysis.
      • Pedersen A.B.
      • Baron J.A.
      • Overgaard S.
      • Johnsen S.P.
      Short- and long-term mortality following primary total hip replacement for osteoarthritis: a Danish nationwide epidemiological study.
      . However, these studies have not included non-OA controls and thus could not examine whether walking difficulty explains the relationship between OA and CVD. The current study explicitly addressed these limitations. Our findings of a dose-effect relationship between number of hips and knees with OA and CV risk after rigorously controlling for CV risk factors, and attenuation of the relationship after controlling for walking difficulty, lend strong support for a causal relationship between OA and CVD that is explained, at least in part, by OA-related mobility.
      Given the high population burden and substantial CVD risk associated with hip and knee OA, we believe these findings are important. They suggest that symptomatic hip and knee OA may be an important modifiable risk factor for CVD. From our nomogram, the effect of having OA in 3+ hips and knees on CVD risk was comparable to the effects of obesity and hypertension. People with painful hip and knee OA often give up activities, like walking, as a strategy to manage their pain rather than take ‘risky pain killers’ or because they have not been offered other options
      • Sale J.E.
      • Gignac M.
      • Hawker G.
      How “bad” does the pain have to be? A qualitative study examining adherence to pain medication in older adults with osteoarthritis.
      . Given the central role of physical activity in the prevention and management of chronic diseases, including OA, this approach is problematic. Clinical trials consistently shown that therapeutic exercise reduces OA symptoms and improves metabolic alterations, e.g., accumulation of advanced glycation end-products, dyslipidemia, excess adipose tissues/fatty acids, hyperglycemia, and systemic inflammation (e.g., IL-6 levels)
      • Piva S.R.
      • Susko A.M.
      • Khoja S.S.
      • Josbeno D.A.
      • Fitzgerald G.K.
      • Toledo F.G.
      Links between osteoarthritis and diabetes: implications for management from a physical activity perspective.
      • Messier S.P.
      • Mihalko S.L.
      • Legault C.
      • Miller G.D.
      • Nicklas B.J.
      • DeVita P.
      • et al.
      Effects of intensive diet and exercise on knee joint loads, inflammation, and clinical outcomes among overweight and obese adults with knee osteoarthritis: the IDEA randomized clinical trial.
      , improving CV risk profile. Based on these findings, we believe increased support for, and access to, community-based exercise, as has been implemented in Denmark
      • Skou S.T.
      • Odgaard A.
      • Rasmussen J.O.
      • Roos E.M.
      Group education and exercise is feasible in knee and hip osteoarthritis.
      , is warranted, together with additional non-pharmacologic therapies (local heat, weight management, self-management & education and strength training), biomechanical interventions, local and topical agents, e.g., intra-articular corticosteroids in active OA, and duloxetine, which may be both safe and effective in OA with multi-morbidity
      • Berenbaum F.
      Osteoarthritis as an inflammatory disease (osteoarthritis is not osteoarthrosis!).
      • Myers J.
      • Wielage R.C.
      • Han B.
      • Price K.
      • Gahn J.
      • Paget M.A.
      • et al.
      The efficacy of duloxetine, non-steroidal anti-inflammatory drugs, and opioids in osteoarthritis: a systematic literature review and meta-analysis.
      .
      Despite the high co-occurrence of OA with other common chronic conditions, there remains a paucity of evidence to support the management of hip and knee OA among people with multi-morbidity. Large, pragmatic, controlled clinical trials are needed to establish the safety, efficacy and effectiveness of existing and novel OA therapies on OA-related symptoms, and the impact of symptom improvement on performance-based measures of mobility, sedentary behavior and physical activity, and on markers of systemic inflammation and metabolic alterations, in the majority of people living with OA who have at least one other chronic condition.

      Acknowledgements

      The authors wish to thank Angela Wall for her substantial role in cohort recruitment and project management. We would like also to express our sincere thanks to all participants of the Ontario Hip and Knee Cohort without whom this research could not have taken place.
      Severity of comorbidities at baseline was approximated using an aggregated score, the Johns Hopkins' Aggregated Diagnosis Groups categories (The Johns Hopkins ACG® System, Version 10).

      Appendix A. Supplementary data

      The following is the supplementary data related to this article:

      Contributions

      All authors contributed to study design and interpretation of data. TK, PJ, RC, and GAH were responsible for the statistical analysis. TK and GAH drafted the manuscript. All authors critically revised the manuscript for important intellectual content, read and approved the final manuscript and the decision to submit for publication. TK and GAH are the guarantors and had full access to all of the data in the study and take responsibility for the integrity of the data and accuracy of the data analysis.

      Competing interests

      All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf and declare: no support from any organization for the submitted work; Gillian Hawker has received research support as the Sir John and Lady Eaton Professor and Chair of Medicine, Department of Medicine, University of Toronto; Tetyana Kendzerska was supported by the Canadian Respiratory Research Network (CRRN); no other relationships or activities that could appear to have influenced the submitted work.

      Role of the funding source

      This study was funded by an operating grant from the Canadian Institutes of Health Research , grant # MOP-15468 .
      This study was also supported by the Institute for Clinical Evaluative Sciences (ICES), which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC). No endorsement by ICES or the Ontario MOHLTC is intended or should be inferred. ICES had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or in the decision to submit the report for publication. Parts of this material are based on data and information compiled and provided by the Canadian Institute for Health Information (CIHI). The opinions, results and conclusions reported in this paper are those of the authors and are independent from the funding sources and CIHI.

      References

        • Perruccio A.V.
        • Power J.D.
        • Badley E.M.
        Revisiting arthritis prevalence projections--it's more than just the aging of the population.
        J Rheumatol. 2006; 33: 1856-1862
        • Ackerman I.N.
        • Osborne R.H.
        Obesity and increased burden of hip and knee joint disease in Australia: results from a national survey.
        BMC Musculoskelet Disord. 2012; 13: 254
        • Hawker G.A.
        • Gignac M.A.
        • Badley E.
        • Davis A.M.
        • French M.R.
        • Li Y.
        • et al.
        A longitudinal study to explain the pain-depression link in older adults with osteoarthritis.
        Arthritis Care Res Hob. 2011; 63: 1382-1390
        • Puenpatom R.A.
        • Victor T.W.
        Increased prevalence of metabolic syndrome in individuals with osteoarthritis: an analysis of NHANES III data.
        Postgrad Med. 2009; 121: 9-20
        • Yoshimura N.
        • Muraki S.
        • Oka H.
        • Tanaka S.
        • Kawaguchi H.
        • Nakamura K.
        • et al.
        Accumulation of metabolic risk factors such as overweight, hypertension, dyslipidaemia, and impaired glucose tolerance raises the risk of occurrence and progression of knee osteoarthritis: a 3-year follow-up of the ROAD study.
        Osteoarthr Cartil. 2012; 20: 1217-1226
        • Sowers M.
        • Karvonen-Gutierrez C.A.
        • Palmieri-Smith R.
        • Jacobson J.A.
        • Jiang Y.
        • Ashton-Miller J.A.
        Knee osteoarthritis in obese women with cardiometabolic clustering.
        Arthritis Rheum. 2009; 61: 1328-1336
        • Boeckxstaens P.
        • Peersman W.
        • Goubin G.
        • Ghali S.
        • De Maeseneer J.
        • Brusselle G.
        • et al.
        A practice-based analysis of combinations of diseases in patients aged 65 or older in primary care.
        BMC Fam Pract. 2014; 15: 159
        • Zhuo Q.
        • Yang W.
        • Chen J.
        • Wang Y.
        Metabolic syndrome meets osteoarthritis.
        Nat Rev Rheumatol. 2012; 8: 729-737
        • Berenbaum F.
        Osteoarthritis as an inflammatory disease (osteoarthritis is not osteoarthrosis!).
        Osteoarthr Cartil. 2013; 21: 16-21
        • Liu S.H.
        • Waring M.E.
        • Eaton C.B.
        • Lapane K.L.
        Association of objectively measured physical activity and metabolic syndrome among US adults with osteoarthritis.
        Arthritis Care Res Hob. 2015; 67: 1371-1378
        • Nuesch E.
        • Dieppe P.
        • Reichenbach S.
        • Williams S.
        • Iff S.
        • Juni P.
        All cause and disease specific mortality in patients with knee or hip osteoarthritis: population based cohort study.
        BMJ. 2011; 342: d1165
        • Haugen I.K.
        • Ramachandran V.S.
        • Misra D.
        • Neogi T.
        • Niu J.
        • Yang T.
        • et al.
        Hand osteoarthritis in relation to mortality and incidence of cardiovascular disease: data from the Framingham heart study.
        Ann Rheum Dis. 2015; 74: 74-81
        • Hawker G.A.
        • Croxford R.
        • Bierman A.S.
        • Harvey P.J.
        • Ravi B.
        • Stanaitis I.
        • et al.
        All-cause mortality and serious cardiovascular events in people with hip and knee osteoarthritis: a population based cohort study.
        PLoS One. 2014; 9e91286
        • Caporali R.
        • Cimmino M.A.
        • Sarzi-Puttini P.
        • Scarpa R.
        • Parazzini F.
        • Zaninelli A.
        • et al.
        Comorbid conditions in the AMICA study patients: effects on the quality of life and drug prescriptions by general practitioners and specialists.
        Semin Arthritis Rheum. 2005; 35: 31-37
        • Sale J.E.
        • Gignac M.
        • Hawker G.
        How “bad” does the pain have to be? A qualitative study examining adherence to pain medication in older adults with osteoarthritis.
        Arthritis Rheum. 2006; 55: 272-278
        • van Dijk G.M.
        • Veenhof C.
        • Schellevis F.
        • Hulsmans H.
        • Bakker J.P.
        • Arwert H.
        • et al.
        Comorbidity, limitations in activities and pain in patients with osteoarthritis of the hip or knee.
        BMC Musculoskelet Disord. 2008; 9: 95
        • Hawker G.A.
        • Wright J.G.
        • Coyte P.C.
        • Williams J.I.
        • Harvey B.
        • Glazier R.
        • et al.
        Determining the need for hip and knee arthroplasty: the role of clinical severity and patients' preferences.
        Med Care. 2001; 39: 206-216
        • Choy E.
        • Ganeshalingam K.
        • Semb A.G.
        • Szekanecz Z.
        • Nurmohamed M.
        Cardiovascular risk in rheumatoid arthritis: recent advances in the understanding of the pivotal role of inflammation, risk predictors and the impact of treatment.
        Rheumatol Oxf. 2014; 53: 2143-2154
        • Hawker G.A.
        • Stewart L.
        • French M.R.
        • Cibere J.
        • Jordan J.M.
        • March L.
        • et al.
        Understanding the pain experience in hip and knee osteoarthritis–an OARSI/OMERACT initiative.
        Osteoarthr Cartil. 2008; 16: 415-422
        • Gignac M.A.
        • Davis A.M.
        • Hawker G.
        • Wright J.G.
        • Mahomed N.
        • Fortin P.R.
        • et al.
        “What do you expect? You're just getting older”: a comparison of perceived osteoarthritis-related and aging-related health experiences in middle- and older-age adults.
        Arthritis Rheum. 2006; 55: 905-912
        • Tu K.
        • Campbell N.R.
        • Chen Z.L.
        • Cauch-Dudek K.J.
        • McAlister F.A.
        Accuracy of administrative databases in identifying patients with hypertension.
        Open Med. 2007; 1: e18-26
        • Alter D.A.
        • Chong A.
        • Austin P.C.
        • Mustard C.
        • Iron K.
        • Williams J.I.
        • et al.
        Socioeconomic status and mortality after acute myocardial infarction.
        Ann Intern Med. 2006; 144: 82-93
        • Hux J.E.
        • Ivis F.
        • Flintoft V.
        • Bica A.
        Diabetes in Ontario: determination of prevalence and incidence using a validated administrative data algorithm.
        Diabetes Care. 2002; 25: 512-516
        • Austin P.C.
        • Walraven C.
        The mortality risk score and the ADG score: two points-based scoring systems for the Johns Hopkins aggregated diagnosis groups to predict mortality in a general adult population cohort in Ontario.
        Can Med Care. 2011; 49: 940-947
        • Harrell F.
        Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis.
        Springer, New York2001
        • Grambsch P.
        • Therneau T.
        Proportional hazards tests and diagnostics based on weighted residuals.
        Biometrika. 1994; 81: 515-526
        • Lange T.
        • Vansteelandt S.
        • Bekaert M.
        A simple unified approach for estimating natural direct and indirect effects.
        Am J Epidemiol. 2012; 176: 190-195
        • Rochon J.
        • du Bois A.
        • Lange T.
        Mediation analysis of the relationship between institutional research activity and patient survival.
        BMC Med Res Methodol. 2014; 14: 9
        • Iasonos A.
        • Schrag D.
        • Raj G.V.
        • Panageas K.S.
        How to build and interpret a nomogram for cancer prognosis.
        J Clin Oncol. 2008; 26: 1364-1370
        • Fine J.P.
        • Gray R.J.
        A proportional hazards model for the subdistribution of a competing risk.
        J Am Stat Assoc. 1999; 94: 496-509
        • Austin P.C.
        • Lee D.S.
        • Fine J.P.
        Introduction to the analysis of survival data in the presence of competing risks.
        Circulation. 2016; 133: 601-609
        • Azur M.J.
        • Stuart E.A.
        • Frangakis C.
        • Leaf P.J.
        Multiple imputation by chained equations: what is it and how does it work?.
        Int J Methods Psychiatr Res. 2011; 20: 40-49
        • Rubin D.
        Multiple Imputation for Nonresponse in Surveys.
        Wiley, New York1987
        • Turkiewicz A.
        • Neogi T.
        • Bjork J.
        • Peat G.
        • Englund M.
        All-cause mortality in knee and hip osteoarthritis and rheumatoid arthritis.
        Epidemiology. 2016; 27: 479-485
        • Xing D.
        • Xu Y.
        • Liu Q.
        • Ke Y.
        • Wang B.
        • Li Z.
        • et al.
        Osteoarthritis and all-cause mortality in worldwide populations: grading the evidence from a meta-analysis.
        Sci Rep. 2016; 6: 24393
        • Capistrant B.D.
        • Glymour M.M.
        • Berkman L.F.
        Assessing mobility difficulties for cross-national comparisons: results from the World health organization study on global AGEing and adult health.
        J Am Geriatr Soc. 2014; 62: 329-335
        • Kendzerska T.
        • King L.
        • Croxford R.
        • Stanaitis I.
        • Wall A.
        • Hawker G.
        The impact of hip and knee osteoarthritis and associated walking limitations on all-cause mortality: a population-based cohort study [abstract].
        Osteoarthr Cartil. 2016; 24: S421
        • Hall A.J.
        • Stubbs B.
        • Mamas M.A.
        • Myint P.K.
        • Smith T.O.
        Association between osteoarthritis and cardiovascular disease: systematic review and meta-analysis.
        Eur J Prev Cardiol. 2016; 23: 938-946
        • Ravi B.
        • Croxford R.
        • Austin P.C.
        • Lipscombe L.
        • Bierman A.S.
        • Harvey P.J.
        • et al.
        The relation between total joint arthroplasty and risk for serious cardiovascular events in patients with moderate-severe osteoarthritis: propensity score matched landmark analysis.
        BMJ. 2013; 347: f6187
        • Pedersen A.B.
        • Baron J.A.
        • Overgaard S.
        • Johnsen S.P.
        Short- and long-term mortality following primary total hip replacement for osteoarthritis: a Danish nationwide epidemiological study.
        J Bone Joint Surg Br. 2011; 93: 172-177
        • Piva S.R.
        • Susko A.M.
        • Khoja S.S.
        • Josbeno D.A.
        • Fitzgerald G.K.
        • Toledo F.G.
        Links between osteoarthritis and diabetes: implications for management from a physical activity perspective.
        Clin Geriatr Med. 2015; 31: 67-87
        • Messier S.P.
        • Mihalko S.L.
        • Legault C.
        • Miller G.D.
        • Nicklas B.J.
        • DeVita P.
        • et al.
        Effects of intensive diet and exercise on knee joint loads, inflammation, and clinical outcomes among overweight and obese adults with knee osteoarthritis: the IDEA randomized clinical trial.
        JAMA. 2013; 310: 1263-1273
        • Skou S.T.
        • Odgaard A.
        • Rasmussen J.O.
        • Roos E.M.
        Group education and exercise is feasible in knee and hip osteoarthritis.
        Dan Med J. 2012; 59: A4554
        • Myers J.
        • Wielage R.C.
        • Han B.
        • Price K.
        • Gahn J.
        • Paget M.A.
        • et al.
        The efficacy of duloxetine, non-steroidal anti-inflammatory drugs, and opioids in osteoarthritis: a systematic literature review and meta-analysis.
        BMC Musculoskelet Disord. 2014; 15: 76