Article Text

Extended report
Disability and not osteoarthritis predicts cardiovascular disease: a prospective population-based cohort study
  1. T A Hoeven1,2,
  2. M J G Leening2,3,
  3. P J Bindels1,
  4. M Castaño-Betancourt2,4,
  5. J B van Meurs4,
  6. O H Franco2,
  7. M Kavousi2,
  8. A Hofman2,
  9. M A Ikram2,5,6,
  10. J C M Witteman2,
  11. S M Bierma-Zeinstra1,7
  1. 1Department of General Practice, Erasmus University Medical Center, Rotterdam, The Netherlands
  2. 2Department of Epidemiology, Erasmus University Medical Center, Rotterdam, The Netherlands
  3. 3Department of Cardiology, Erasmus University Medical Center, Rotterdam, The Netherlands
  4. 4Department of Internal Medicine, Erasmus University Medical Center, Rotterdam, The Netherlands
  5. 5Department of Neurology, Erasmus University Medical Center, Rotterdam, The Netherlands
  6. 6Department of Radiology, Erasmus University Medical Center, Rotterdam, The Netherlands
  7. 7Department of Orthopaedics, Erasmus University Medical Center, Rotterdam, The Netherlands
  1. Correspondence to Theun A Hoeven, Department of General Practice, Erasmus University Medical Center, PO Box 2040, Rotterdam 3000 CA, The Netherlands; t.hoeven{at}


Background Previous studies found an association between osteoarthritis (OA) and risk of cardiovascular disease (CVD) and therefore suggested intensive treatment of cardiovascular risk factors in OA patients. However, prospective population-based data is lacking.

Objectives To investigate the association between OA and CVD longitudinally in a general population and examine the role of disability in this association.

Methods This study was embedded in the Rotterdam Study, a prospective population-based cohort study in Rotterdam, the Netherlands that started in 1989. At baseline 4648 persons aged ≥55, free of CVD were classified into those with and those without radiographic or clinical OA. HRs adjusted for traditional cardiovascular risk factors for developing CVD (a composite of fatal and non-fatal coronary heart disease and stroke) were calculated.

Results During a median follow-up of 14.4 years, 1230 cardiovascular events occurred, of which 101 were in participants with clinical OA. Presence of radiographic OA at baseline was not related to future CVD (HR 0.99, 95% CI 0.86 to 1.15), neither was presence of clinical OA (HR 1.09, 95% CI 0.88 to 1.34). However, persons with increasing disability were more likely to suffer a cardiovascular event compared with non-disabled persons (HR 1.26, 95% CI 1.12 to 1.42); this was independent of the presence of OA.

Conclusions In this large population-based study, participants with OA were not at increased risk of CVD. The close relation between disability and osteoarthritis may explain previous findings. Further studies are required in order to clarify whether OA patients need more intensive treatment of their cardiovascular risk factors.

  • cardiovascular disease
  • disability
  • population-based

Statistics from



Osteoarthritis (OA) is the most frequent joint disorder in the elderly and causes a considerable burden of pain, disability, and ever increasing costs to society.1 Several cross-sectional2 ,3 and disease-specific mortality4–6 studies suggest an association between OA and cardiovascular disease (CVD). The strongest body of evidence comes from a recent study comparing CVD mortality of selected OA patients with mortality statistics from the general population.6 Therefore more intensive treatment of cardiovascular risk factors has been proposed to prevent premature CVD in OA patients, similar to conditions like rheumatoid arthritis or gout.7–9 Before implementing OA as another red flag into guidelines, critical assessment of such claims is warranted, especially if robust prospective population-based comparisons are lacking.

Previous studies were hampered by methodological issues,6 ,10 such as small or selected patient groups, performed in occupational settings, or included reference groups based on data from national statistics. These study designs may thus be liable to mechanisms that increase the possibility of false positive results.

In the large ongoing Rotterdam Study, an unselected population, we studied the association between various definitions of OA and long-term risk of CVD. We additionally examined the role of disability in explaining the apparent association between OA and CVD occurrence.


Study design, setting and population

The Rotterdam Study is a prospective population-based cohort study that started in 1990 to investigate the occurrence and determinants of diseases in a middle aged and elderly population.11 All 10 275 inhabitants aged 55 years and older who had been living for at least 1 year in the Ommoord district of the city of Rotterdam were invited to participate, of which 7983 agreed. Trained interviewers obtained information at home on current health status, medical history, activities of daily living, joint complaints and cardiovascular risk factors. Subsequently, all participants were invited to visit the research centre for clinical examinations, laboratory assessments and radiographs, regardless of their health status.

For this study, 1428 participants with known CVD (defined as a history of myocardial infarction, surgical or percutaneous coronary revascularisation, or stroke) at baseline were excluded. Furthermore, 1907 persons who either did not visit the research centre or had no radiographs taken due to logistic or technical problems were excluded, leaving a total population of 4648 persons for analyses in whom we assessed CVD at follow-up.

All participants provided written informed consent to participate in the Rotterdam Study and to obtain information from their treating physicians, separately. The study was approved by the medical ethics committee of the Erasmus Medical Center, Rotterdam, the Netherlands.

Assessment of OA

In total, three definitions of OA were formulated for knees, hips and hands separately. The first one, radiographic OA (K&L graded score greater or equal to 2), is described in online supplementary appendix 1. Second, we defined clinical OA as the presence of radiographic OA and reported complaints of the same joint in the last month. Finally, we analysed self-reported OA (verified by trained interviewers; ‘Did you suffer from joint complaints last month which were attributed to OA by a physician?’) as a potential phenotype of OA. Since knee OA is highly prevalent and most clearly defined, we decided to show results with this phenotype in the analyses. However, we also analysed hand and hip OA (see online supplementary appendix 2).

Assessment of disability

For the assessment of disability the Stanford Health Assessment Questionnaire (HAQ) was used.12 The HAQ measures disability in eight fields (dressing and grooming, rising, reach, hygiene, eating, walking, grip and activity). Each field comprises two to four items. Per item the status of the respondent is scored as able to do without difficulty (0), with some (1) or much (2) difficulty, or unable to do with or without assistance (3). The highest item score determines the final field score. The mean score of all fields constitutes the disability index, ranging from 0.00 to 3.00.13 In the present study we dichotomised disability as having any disability present in one of the eight fields versus none, irrespective of the presence of OA. In addition, we similarly analysed lower limb disability involving the fields rising and walking only. All values given in our analyses refer to the dichotomised disability variable.

Assessment of covariates

Data on medication use and smoking habits were obtained during the home interview. Smoking was classified as never, former or current smoking. Established cardiovascular risk factors were measured at the research centre. Body mass index was computed as weight divided by height squared. Blood pressure was measured at the right upper arm using a random-zero sphygmomanometer. We used the average of two measurements measured on one occasion and defined hypertension as a systolic blood pressure of ≥140 mm Hg higher, a diastolic blood pressure of ≥90 mm Hg or the use of blood pressure lowering medication for hypertension. Diabetes mellitus was defined as the use of antidiabetic medication or a non-fasting or post-load serum glucose level exceeding 11.0 mmol/L. Serum total cholesterol and high-density lipoprotein (HDL) cholesterol were determined by means of an automated enzymatic procedure in non-fasting blood samples.

Assessment of CVD

Subjects were followed for the occurrence of a first coronary or cerebrovascular event, including myocardial infarction, surgical or percutaneous coronary revascularisations, coronary death and stroke (both ischaemic and haemorrhagic) confirmed by one of the patients’ treating physicians. Methods of data collection and adjudication of events have been described previously in detail.14 ,15 In short, Rotterdam Study participants are followed-up continuously through direct digital linkage of the study base with medical files from the general practitioners working in the research area. Moreover, the entire medical record of each participant is checked on a regular basis for diagnoses of interest. All available information, such as discharge reports, ECGs and neuroimaging results are copied from the medical records or are obtained from the hospitals. Subsequently each potential CVD event is adjudicated according to standardised definitions by two independent research physicians and either an experienced cardiologist or neurologist.

Statistical analysis

Baseline characteristics are summarised as percentages for categorical data. Continuous data are presented as mean values and SDs.

We used 18 Cox proportional hazard models to estimate the association between OA and risk of incident CVD. All models met the proportional hazards assumption. In the first model, we estimated the association of radiographic knee OA with the occurrence of CVD adjusted for age and gender. In the second model, we additionally adjusted for the following cardiovascular risk factors: hypertension, smoking status, body mass index, diabetes and total cholesterol/HDL cholesterol ratio. Likewise, clinical knee OA and self-reported OA were analysed as different phenotypes of OA. Next, we checked whether there was evidence of effect modification by gender, since previous studies have described associations between OA and subclinical atherosclerosis in women only.2 ,16 This was done by assessing the interaction terms between gender and OA status.

Since a recent study reported an association between walking disability and risk of CVD within OA patients,6 we ran similar Cox models as described above to estimate the associations between overall disability and CVD, and lower limb disability and CVD; this was done independent of the presence of OA. Last, in order to reduce the potential diluting effects of misclassification of the outcome we performed a sensitivity analysis restricted to hard atherosclerotic CVD outcomes: myocardial infarction, ischaemic stroke and coronary death.

Persons free of radiographic knee OA, free of clinical knee OA and those without self-reported OA were reference categories, respectively. All analyses were performed for hip OA and hand OA as well in an identical manner. Persons were censored at the date of non-coronary or non-cerebrovascular death, loss to follow-up or the end of the study period defined as the last date of follow-up. Approximately 1% of our study population had missing values for one or more cardiovascular covariates. These missing values were handled by single imputation using an expectation-maximisation algorithm.17 All measures of association are presented with 95% CIs. We used the level of significance of p<0.05 for all statistical analyses and the data were analysed using PASW V.17.0.2 (SPSS).


Characteristics of the study population

Baseline characteristics of the study population free of CVD at baseline are summarised in table 1. The mean age of the 4648 persons in this study was 67.6 years, and 61% were women. During a median follow-up time of 14.4 years (IQR 7.6 years) a total of 1230 CVD events occurred, of which 101 were in the participants with clinical knee OA. These events consisted of 304 myocardial infarctions, 64 coronary artery bypass grafts, 69 percutaneous coronary interventions, 215 other fatal coronary heart disease events and 578 strokes (of which 60 were confirmed haemorrhagic). Only 41 (<1%) persons were lost to follow-up.

Table 1

Baseline characteristics of the study population

Osteoarthritis and CVD

The associations between OA of the knees and incident CVD are shown in table 2. After adjustment for age and gender, we found no association between radiographic knee OA and incident CVD (HR 0.99, 95% CI 0.86 to 1.15), neither between clinical knee OA and CVD (HR 1.08, 95% CI 0.88 to 1.33) nor between self-reported OA and CVD (HR 1.08, 95% CI 0.93 to 1.24). Further adjustment for cardiovascular risk factors did not change the results. Interaction by gender could not be demonstrated (p=0.34) and the sensitivity analyses with hard CVD only did not alter the results. We performed identical analyses for hand and hip OA; the corresponding results are presented in online supplementary appendix 2. Results for hand OA or hip OA were not different from those for knee OA.

Table 2

Knee osteoarthritis and risk of incident cardiovascular disease

Disability and CVD

After adjustment for age and gender, disability was strongly associated with incident CVD (HR 1.30, 95% CI 1.15 to 1.46). Full adjustment for body mass index, diabetes mellitus, hypertension, total cholesterol/HDL cholesterol ratio and smoking slightly attenuated the association (HR 1.26, 95% CI 1.12 to 1.42). Sensitivity analyses with lower limb disability only were not substantially different from overall disability. This is shown in table 3. Age and gender adjusted survival curves according to disability status, stratified on the presence of clinical knee OA, are shown in figure 1. There was no evidence for interaction between disability and clinical knee OA status (p=0.89).

Table 3

Disability and risk of incident cardiovascular disease

Figure 1

(A, B) Age and gender adjusted cardiovascular disease-free survival curves for non-disabled and disabled participants in participants with clinical knee osteoarthritis (OA).


In the present study, we examined the association of different phenotypes of OA with the long-term risk of CVD in a prospective population-based cohort. We show that the presence of OA is not related to incident CVD in adults aged 55 years and older. However, overall and lower limb disabilities are risk factors for CVD, independent of OA status.


Our study was prompted by the recent report by Nüesch et al,6 who found that OA patients were at higher risk of cardiovascular death compared with their age and gender matched peers. This study was performed in a group of OA patients compared to national statistics where data on covariates were not available. As a result, they could not correct for significant confounding variables such as co-morbidities or disability. Furthermore, selection of diseased patients into their study may have led to inflated estimates. This is referred to as the sick-person effect.18

We used a population-based cohort with detailed data on cardiovascular covariates and disability, also for persons without OA, and were thereby able to provide a less biased estimate of the association. Given that we did not find an association between several phenotypes of OA and cardiovascular endpoints in our study population, we also investigated other phenotypes. When we analysed summed K&L scores of the knees, hands or hips separately or included all total joint replacements due to OA as markers of severity of OA, the results did not alter. This was also the case when we compared participants with generalised OA (at least two out of three joint groups affected) to participants without any OA. Moreover, the association of OA with the various components of the composite CVD outcome (coronary heart disease and stroke) analysed separately did not change our results (data not shown). In the present study disability—defined by the HAQ questionnaire—strongly predicted CVD independent of clinical OA status. Our results therefore support the view that disability, and not OA, predicts CVD.19 Disability might represent a marker of the general health status rather than a representation of specific underlying chronic diseases such as OA. Further studies are needed to investigate whether HAQ disability contributes to improved cardiovascular risk stratification in the population at large and to identify underlying causes of disability in order to ameliorate excess cardiovascular risk.


The relation between OA and increased risk of future CVD mortality, as found by Nuesch et al,6 suggested intensive treatment of cardiovascular risk factors in OA patients. Before implementing OA as another red flag into CVD prevention guidelines, we considered it necessary to critically assess this claim in a robust population-based study. In contrast to Nuesch et al, we excluded persons with known CVD at baseline to investigate whether OA patients were at high risk of CVD. We demonstrated that OA is not an independent risk factor for the development of incident CVD in a population-based setting.

Strengths and limitations

The major strengths of the present study are its vast size, the long-term highly detailed follow-up, the community-based setting, and moreover the use of an internal reference population for comparing people with OA to those free of OA. A substantial number of people with OA—defined both by radiographs and by symptoms—were included and the median follow-up time was 14.4 years.

However, residual confounding has to be considered. As we did not measure physical fitness and had no reliable information on medication use without a prescription, additional confounders might exist. In addition, participation bias, occurring in all studies requiring active participation, may have influenced our results.

The main aim of this study was to investigate whether physicians should consider OA patients at high risk of CVD. Our study population comprised persons aged 55 years and older and we were therefore unable to look into the association between OA and CVD in younger individuals. A cardiometabolic phenotype of OA20 in younger women characterised by low-grade inflammation by visceral adipose tissue21 might exist. Although both conditions share aetiological features3 ,16 and risk factors, we did not focus on a common pathophysiology by investigating subclinical measures of atherosclerosis. This complex and promising field of research merits further elucidation and will be investigated in depth in future studies with currently gathered data from younger women.11


In this population-based study, persons with OA were not at increased risk of developing CVD compared to those without OA. The close relation between disability and OA may explain earlier findings of a relation between OA and risk of CVD. Further studies are required to clarify whether younger OA patients than in the present study benefit from more intensive treatment of their cardiovascular risk factors.


The dedication, commitment and contribution of the inhabitants, family physicians and pharmacists of the Ommoord district to the Rotterdam Study are gratefully acknowledged.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Handling editor Tore K Kvien

  • Contributors The first author had full access to all the data in the study and takes full responsibility for the integrity of the data and the accuracy of the data analysis. All authors were involved in drafting the manuscript and critical revision of the manuscript for important intellectual content. They are all guarantors. All co-authors contributed to the realisation of this paper and revised the paper several times. Contributions to different parts of the paper are categorised in order of importance. ML: data management cardiovascular outcomes, methodology, abstract, discussion; PB: methodology, abstract, discussion; MC: introduction, discussion, data management osteoarthritis; JM: data management osteoarthritis, statistical analysis, abstract, discussion; OF: study design, abstract, discussion; MK: data management cardiovascular outcomes, methodology (bias); AH: study design, abstract; MI: methodology (bias), statistical analysis, introduction, discussion; JW: study design, introduction, discussion, abstract; SBZ: study design, introduction, discussion, abstract.

  • Funding The Rotterdam Study is supported by the Erasmus Medical Center and Erasmus University Rotterdam, The Netherlands Organization for Scientific Research (NWO), The Netherlands Organization for Health Research and Development (ZonMw), the Research Institute for Diseases in the Elderly (RIDE), The Netherlands Genomics Initiative (NGI), the Ministry of Education, Culture and Science, the Ministry of Health, Welfare and Sports, the European Commission (DG XII), and the Municipality of Rotterdam. This study is supported by grants from the Reumafonds (Dutch Arthritis Foundation), the Netherlands Organization for Scientific Research (NWO) and The Netherlands Organization for Health Research and Development (ZonMw) (Vici grant number 918-76-619; ZonMw grant number 80-82500-98-10208).

  • Competing interests None.

  • Ethics approval Medical ethics committee of the Erasmus Medical Centre, Rotterdam, the Netherlands.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.