Introduction

Cardiovascular disease (CVD) is composed of diseases related to the heart or blood vessels, such as ischaemic heart disease (IHD), congestive heart failure (CHF) and stroke. It accounts for a significant portion of morbidity and mortality rates worldwide, including Japan.1 Annually, approximately 17 million people globally die from CVD, particularly because of heart attacks and strokes.2 On the other hand, osteoarthritis (OA), a degenerative disease, typically occurs in middle-aged and older individuals, gradually progressing with the degeneration of joint cartilage.3 4 It can potentially affect joints throughout the body, including the knees, hips, lumbar spine and hands, and is one of the fastest growing diseases worldwide. In Japan, the number of patients with OA is increasing rapidly because of the ageing population, with approximately 25 million people suffering from knee OA (KOA) alone.5 As the disease progresses, severe pain and other subjective symptoms occur, which reduce the quality of life and limit treatment options. In Japan, among reasons for requiring caregiving services, the second most common cause is cerebrovascular diseases (16.6%), the fourth is fractures and falls (12.1%), the fifth is joint diseases (10.2%) and the sixth is heart diseases (4.6%), with OA-related and CVD-related issues comprising nearly half of the cases.6 Therefore, early treatment initiation as a preventive measure is considered crucial for extending healthy life expectancy and reducing healthcare costs.

CVD and OA can be prevented through appropriate timing of drug therapy and non-pharmacological interventions. Identifying individuals with a high risk of developing CVD, which contributes significantly to mortality rates, is of utmost importance for public health. Many people who experience CVD have multiple traditional risk factors such as obesity, hypertension, diabetes, chronic kidney disease, dyslipidaemia and smoking. Recent studies have revealed that some of these risk factors are also directly or indirectly involved in the pathogenesis of OA.7 A recent meta-analysis indicated that individuals with OA have a high CVD risk.8 However, the inter-relationship between OA and CVD and its potential mechanisms are complex, and whether OA increases the risk of CVD independent of the abovementioned risk factors is unclear. Although some studies have investigated the relationship between OA and CVD internationally, to our knowledge, no clinical studies have used large-scale data from Japan. As lifestyle, genetic factors, body type and population characteristics vary between Japan and other countries, research on this topic, specifically in Japan, is necessary.

Therefore, this study, using data with a high number of variables, such as Medical Data Vision (MDV) data, in Japan, which is experiencing super-ageing, aimed to determine whether the presence of OA is a risk factor for CVD development, whether there are differences among IHD, CHF and stroke and whether there is any difference between CVD risk and OA sites (hips, knees and hand).

MethodsStudy design and study population

This study employed a population-based case–control matching study design, and a retrospective evaluation was conducted using real-world data from Japanese MDV health insurance claims data. This dataset covers approximately 22% of diagnosis procedure combination (DPC) hospitals, which include many acute care hospitals across Japan. It is composed of electronic health insurance claims data, DPC claims and laboratory test results.9

Approximately 1 million individuals were randomly selected from 35.23 million older patients who were attending 438 DPC hospitals under contract with MDV among the entire older population in Japan. Disease-related information included the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) diagnosis codes, Japanese disease codes and diagnosis dates. Medication-related data included health insurance claims codes, prescription dates, administration routes and prescription quantities. All patient data were coded before entry into the database.

Selection of cases and controls

Cases were defined by mapping the disease codes in the dataset to the master codes for CVD, stroke, IHD and CHF and limiting them to confirmed disease names. Each master code was defined based on the following ICD-10 diagnosis codes: IHD (I20–I25), CHF (I50), stroke (I60–I64) and CVD (I20–I25, I50, I60–I64).10 Cases were defined as patients aged ≥65 years that were first diagnosed with the target disease between January 2015 and December 2020, with the diagnosis date serving as the index date. Controls were randomly assigned index dates.

The age of the patients was categorised into 5-year intervals, and 1:1 matching was performed based on age and sex. In cases where there were multiple potential controls for a single case, control individuals were randomly selected from among them.

According to Yamana et al, the validity of diagnoses within the DPC database demonstrated a sensitivity and specificity of 78.9% and 93.2%, respectively. Although variations were observed among different medical conditions, the overall results indicated favourable accuracy without significant diagnostic inaccuracies.11

Definition of OA and covariates

Hip OA (HipOA) (M160–M169) and KOA (M170–M179) and hand OA (HandOA) (M180–M189) were defined as exposure factors. Covariates included common risk factors for CVD, such as type 2 diabetes mellitus, essential (primary) hypertension, atrial fibrillation and flutter, chronic kidney disease, disorders of lipoprotein metabolism and other lipidaemias, and medications related to CVD treatment: oral antidiabetic drugs, platelet aggregation inhibitors, diuretics, beta-blocking agents, calcium antagonists, agents acting on the renin–angiotensin system and lipid-regulating/antiatheroma preparations (online supplemental appendix 1).

The types of medications and comorbidities were obtained from the MDV database. For medications, the unique nine-digit health insurance claims codes, which are assigned to each drug based on its pharmacological properties, were mapped to five-digit codes according to the Anatomical Therapeutic Chemical (ATC) Classification System managed by the European Pharmaceutical Marketing Research Association.12 The start date of medication use was set as before the index date, and medications that were being continued at the index date or had an end date within 7 days of the index date were considered. In other words, a grace period of 7 days was set after the prescription end date, and all medications that were taken at least once within 7 days from the index date were considered to affect the outcome.

Complications were included only when a confirmed diagnosis of the disease was established, and they were mapped using ICD-10 codes. Comorbidities with diagnosis dates before the index date and no end dates were registered until the index date was extracted. Based on this information, concomitant medications and comorbidities were extracted from the MDV database. The explanatory variables included sex, age, comorbidities based on ICD-10 codes and concomitant medications based on the Anatomical Classification of Pharmaceutical Products (ATC) codes.

Statistical analysis

Sex, concomitant medications and comorbidities were treated as binary variables. The characteristics of the case and control groups regarding age, sex, comorbidities and concomitant medications were summarised using mean, SD and proportions as descriptive statistics. Furthermore, a conditional logistic regression model was applied to estimate the ORs and their 95% CIs for the risk of developing CVD, IHD, CHF and stroke. Adjustments were made for the covariates mentioned above. All analyses were conducted using R V.4.1.0 (http://www.r-project.org/), and the ‘clogit’ package was used for the conditional logistic regression.

Patient and public involvement

This study will not have any patient or public involvement because it is health insurance claims data.

ResultsCharacteristics of the study population

The study population was based on approximately 1 million MDV data; initially, cases and controls meeting the selection criteria were extracted, followed by those meeting the criteria for ‘Definition of OA and covariates’. Finally, 1:1 matching based on age and sex was conducted. The total number of participants, including the case and control groups for each condition, is as follows: CVD (IHD, CHF and stroke), 39 648 in each group, totalling 79 296; stroke, 14 944 in each group, totalling 29 888; IHD, 22 996 in each group, totalling 45 992; CHF, 31 639 in each group, totalling 63 278 (figure 1). The mean age of CVD occurrence was 78.6 (SD 7.3) years, and 54.4% were males and 45.6% were females. In addition, table 1 shows the percentage of patients with comorbidities and concomitant medications. IHD, CHF and stroke are shown in online supplemental appendices 2–4, respectively.

Table 1

Characteristics of patients with CVD and controls

Figure 1

Flow chart of cases and controls. ATC, Anatomical Therapeutic Chemical Classification System; CHF, congestive heart failure; CVD, cardiovascular disease; ICD-10, International Statistical Classification of Diseases and Related Health Problems, 10th Revision; IHD, ischaemic heart disease; MDV, Medical Data Vision.

CVD and OA

In the unadjusted analysis for CVD, the OR for KOA exposure in the case group compared with the control group was 1.204 (95% CI 1.133 to 1.28). The ORs for HipOA and HandOA exposures were 1.14 (95% CI 1.001 to 1.298) and 1.06 (95% CI 0.72 to 1.56), respectively, indicating no significant association. After adjusting for covariates, the OR for KOA exposure was 1.192 (95% CI 1.115 to 1.274), indicating a significant association with CVD occurrence. The ORs for HipOA and HandOA exposures were 1.057 (95% CI 0.919 to 1.215) and 1.035 (95% CI 0.684 to 1.566), respectively, indicating no significant association (table 2). Similarly, in the unadjusted analysis for IHD, the OR for KOA exposure was 1.156 (95% CI 1.067 to 1.253), while the ORs for HipOA and HandOA exposures were 1.016 (95% CI 0.854 to 1.209) and 0.974 (95% CI 0.623 to 1.523), respectively, with no significant association. After adjusting for covariates, the OR for KOA exposure was 1.187 (95% CI 1.086 to 1.297), showing a significant association with IHD occurrence; however, the ORs for HipOA and HandOA exposures were 1.078 (95% CI 0.891 to 1.306) and 1.099 (95% CI 0.677 to 1.784), respectively, indicating no significant association (online supplemental appendix 5). For CHF, the unadjusted analysis showed ORs of 1.031 (95% CI 0.962 to 1.104), 0.966 (95% CI 0.834 to 1.12) and 0.911 (95% CI 0.597 to 1.391) for KOA, HipOA and HandOA exposures, respectively. After adjusting for covariates, the ORs for KOA, HipOA and HandOA exposures were 1.027 (95% CI 0.948 to 1.112), 0.968 (95% CI 0.816 to 1.149) and 1.139 (95% CI 0.705 to 1.841), respectively, indicating no significant association with CHF occurrence (online supplemental appendix 6). Finally, for stroke, the unadjusted analysis showed ORs of 1.26 (95% CI 1.142 to 1.39), 1.013 (95% CI 0.809 to 1.27) and 1.095 (95% CI 0.606 to 1.979) for KOA, HipOA and HandOA exposures, respectively. After adjusting for covariates, the OR for KOA exposure was 1.221 (95% CI 1.099 to 1.356), indicating a significant association with stroke occurrence, while the ORs for HipOA and HandOA exposures were 0.918 (95% CI 0.723 to 1.165) and 1.169 (95% CI 0.635 to 2.151), respectively, showing no significant association (online supplemental appendix 7).

Table 2

Crude and adjusted ORs and 95% CIs of CVD in relation to KOA, HipOA and HandOA

Discussion

To our knowledge, this study is the first large-scale population-based case–control study conducted on Japanese individuals, evaluating the relationship between OA and CVD risk using administrative claims database. In this study, after adjusting for various potential confounding factors, KOA was found to be associated with an increased risk of CVD. Furthermore, the subgroup analysis revealed that patients with KOA had an increased risk of IHD and stroke.

According to the population-based study by Nüesch et al, patients with KOA and HipOA have higher overall mortality rates than the general population, particularly with a significant association with CVD and dementia-associated mortality.13 Goel et al found a strong relationship between KOA and CVD, and the cardiovascular risk score positively correlated with the OA severity.14 On the contrary, in their prospective population-based cohort study, Hoeven et al reported no significant association between CVD risk and clinical or radiographic KOA, HipOA and HandOA.15 While the relationship between OA and CVD remains debatable, many studies have suggested that a potential increase in CVD risk is associated with OA.16

The underlying fundamental mechanisms linking the risks of OA and CVD remain unclear; however, several mechanisms have been proposed to explain the association between OA and CVD. First and foremost, a shared risk factor aspect involves metabolic syndrome components such as diabetes, dyslipidaemia and hypertension, which are common to both OA and CVD.17–21 However, these factors may act as confounding variables. Therefore, if the increased risk of CVD persists even after conducting multivariate analysis accounting for these factors, it can imply an independent association between OA and CVD. Second, the induction of inflammatory responses associated with cytokines is another point of consideration. Nüesch et al reported that walking disability, along with OA, was associated with high CVD-related mortality, and Hoeven et al reported that some forms of disability in daily activities, rather than OA itself, were associated with CVD onset.13 15 OA development can lead to severe joint pain, ultimately resulting in walking disability and decreased physical activity over time. Reduced physical activity can lead to the accumulation of visceral fat, that is, hypertrophy of fat cells, which, in turn, increases the secretion of inflammatory cytokines such as tumour necrosis factor-alpha and interleukin-6. Many of these inflammatory processes and cytokines also contribute to the development of vascular inflammation underlying conditions such as hypertension, myocardial infarction, heart failure and cerebrovascular disorders, which are components of many CVDs.22–24 Moreover, Yoshimura et al reported that the risk of KOA development increases with the presence of hypertension and impaired glucose tolerance; conversely, the coexistence of KOA increases the risk of developing hypertension and dyslipidaemia.25 These findings indicate a mutual relationship between the development and coexistence of KOA and metabolic syndrome components.

The study found associations of KOA with IHD and stroke, but no association of KOA with CHF. In a meta-analysis, Hall et al found that patients with OA had a significantly increased CHF and IHD risk compared with patients without OA; however, no significant difference was observed in stroke risk.8 A Mendelian randomisation study by Wang et al showed a potential causal relationship between HipOA and CHF along with stroke, but no association of HipOA with IHD. Although no significant associations were observed in all aspects with KOA, IHD strongly correlated with KOA onset.26 Rahman et al reported a significant association of OA with IHD and CHF in individuals aged ≥65 years, but no association of OA with myocardial infarction or stroke.27 Overall, results regarding the association with stroke, IHD and CHF vary among studies, and why only CHF among CVDs showed no association with OA in the present study remains unclear. However, one hypothesis could be that patients with OA generally have reduced physical activity28 29 and are less likely to experience overexertion in daily life. Consequently, there may be less strain on the heart, which does not increase the CHF risk.

In this study, no associations were observed in HipOA and HandOA. Macêdo et al found that both HipOA and KOA increase the risk of subclinical atherosclerosis and CVD, whereas HandOA showed no association with CVD.30 Tsuboi et al reported that when KOA coexists, the odds of death after 10 years significantly increased; however, no significant difference was observed when coexisting with lower back abnormalities.31 These authors explain this phenomenon from the perspective of weight-bearing joints. In the present study, KOA showed a significant association; however, HipOA and HandOA did not. One possible reason is that the knee and hand are weight-bearing joints more than the hip, which may lead to excessive stress on the bones and cartilage because of biomechanical load caused by the body weight, increasing the risk for KOA development and leading to more significant inflammatory reactions and a more pronounced association with CVD. Moreover, a strong correlation was found between KOA and physical activity; however, the association with HipOA is weaker than that with KOA, as suggested by previous studies.32 33 This indicates that the knee is more susceptible to physical stress than the hip.

Strengths and limitations

This study has some limitations. First, the case–control study design allows us to establish an association between KOA and CVD; however, a causal relationship could not be established. Second, the diagnosis of OA was based solely on cases diagnosed by physicians, which may have underestimated the results by not considering patients with asymptomatic OA who do not exhibit symptoms such as pain. Third, the data used in this study only included acute care hospital data from the MDV database, which may have a higher likelihood of including patients with more severe conditions and multiple comorbidities than the general population. Fourth, lifestyle factors such as alcohol consumption, smoking and exercise habits were not considered. Finally, this study was conducted using Japanese insurance claims data, and there may be limitations in generalising our research findings to other ethnic groups.

Despite these limitations, this study has several strengths that outweigh them. First, the study has a large sample size, and a case–control study design was employed while adjusting for multiple important confounding factors. In addition, health administrative data were used to verify the outcomes, providing more robust evidence in a real clinical setting.