Article Text
PDF
Abstract
Aims To assess and compare the incidence of cardiovascular (CV) events, by CV phenotype, between patients with ankylosing spondylitis (AS), rheumatoid arthritis (RA) and the general population.
Methods Using linkages of national and population-based registers, we identified one cohort of prevalent patients with AS (n=5358), one with RA (n=37 245) and one with matched general population subjects (n=25 006). These cohorts were identified in 2006 through 2011 and were followed in 31 December 2012, for first ever occurrence of acute coronary syndromes (ACS), deep venous thromboembolism, pulmonary embolism and stroke, respectively. For each outcome, we calculated incidence rates standardised to the age and sex distribution of the AS cohort, as well as relative risks using Cox proportional hazards models.
Results Based on 69 ACS events during 20 251 person-years of follow-up of the patients with AS, and 966 events during 127 014 person-years in the RA cohort, the age/sex-adjusted relative risks for ACS compared with the general population was 1.3 (95% CI 1.0 to 1.7) for AS and 1.7 (1.4 to 2.0) for RA. For thromboembolic events, the corresponding risks were 1.4 (1.1 to 1.9) in AS and 1.8 (1.5 to 2.1) in RA. Finally, for stroke, the relative risks were 1.5 (1.1 to 2.0) in AS and 1.5 (1.2 to 1.8) in RA, compared with the general population.
Conclusions Prevalent patients with AS are at a 30%–50% increased risk of incident CV events. When compared with patients with RA, this level of increase was similar for stroke, but only half as high for ACS and thrombotic events.
- Ankylosing Spondylitis
- Cardiovascular Disease
- Psoriatic Arthritis
- Epidemiology
Statistics from Altmetric.com
Introduction
Patients with rheumatoid arthritis (RA) are at a well-documented increased risk of several different types of cardiovascular (CV) diseases, including acute coronary syndromes (ACS), deep venous thromboembolism (DVT), pulmonary embolism (PE) and cerebrovascular diseases. Although the relative risks reported have varied across studies, partly due to differences in patient with RA characteristics and disease duration, risk increases in the order of 50%–100% for ACS, 0%–50% for stroke and 100%–200% for DVT/PE have typically been reported.1–3 Evidence to date suggests that traditional CV risk factors alone do not account for the entire risk increase observed. Instead, RA disease activity has been proposed to be of critical importance.4
Ankylosing spondylitis (AS) is a chronic inflammatory disease with a different sex and age distribution, pathology and phenotype than RA, including an often less pronounced systemic inflammation (as measured, eg, by acute phase reactants). In contrast to the literature on CV incidence in RA, data on CV incidence in AS are scarce and somewhat heterogeneous (summarised in table 1). Most, but not all, studies of ischaemic heart disease (defined somewhat differently across available studies) in AS have reported some level of risk increase, as have studies on stroke and less-defined composite CV outcomes, when compared with the general population.5 In a recent study of CV incidence in AS, no increase at all for myocardial infarction was noted.6
Previously published estimates of relative risk (incidence or mortality) of CV disease in patients with AS, compared with subjects without AS
The above-described heterogeneity in relative risks associated with AS, in combination with the differences in age and sex distribution between patients with AS and RA, and typical effect modification of relative risks by age (higher relative risks in younger age groups, where the baseline risk is the lowest) leaves some uncertainty with regard to the ‘true’ level of risk increase for CV disease in AS, and how these risks compare with the corresponding and better established risks in RA.
The aim of this study was therefore to assess and compare age-adjusted and sex-adjusted CV incidences, by CV phenotype, in contemporary patients with AS and RA, and to put these risks in relation to the corresponding risks in the general population, all identified from the same source population and subjected to similar assessment of CV outcomes.
Methods
Setting
The Swedish healthcare system is public and tax funded. Patients with inflammatory diseases such as AS and RA are typically diagnosed by rheumatologists7 and, in the case of RA and for treatments with disease-modifying anti-rheumatic drugs (DMARDs) and biologics, treated by rheumatologists rather than by general practitioners. The vast majority of rheumatologists in Sweden are working at hospital-based clinics rather than as private practitioners.
Register sources
The register source used for identification of the prevalent AS and RA cohorts was the National Patient Register (data capture 1969–2011). Data from the National Patient Register were also used to identify events of ACS, DVT/PE and stroke, both before and after cohort entry, as well as baseline comorbidities of other CV diseases, chronic obstructive pulmonary disease, diabetes and malignancy (data capture 1969–2012). International classification of diseases (ICD) codes used for the definition of AS and RA, baseline comorbidities and the diagnoses used as exclusion criteria are listed in online supplementary table 1.
Supplemental material
The Prescribed Drug Register (data capture July 2005 through 2012) was used to identify individuals with dispensed prescriptions for non-steroidal anti-inflammatory drugs (NSAIDs), diuretics, β-blockers, ACE inhibitors, calcium blockers, aspirin, anti-diabetics and insulin and statins (Anatomic Therapeutic Chemical classification system (ATC) codes used are listed in online supplementary table 2) prior to start of follow-up.
Mortality data were retrieved from the Causes of Death Register, while general population comparator subjects were sampled from the Register of the Total Population, from which we also collected data on emigration. Linkage of these registers was performed using each individual's personal identity number, a unique identifier carried by each Swedish resident (online supplementary figure 1).
Study population
Three cohorts were assembled; prevalent patients with AS, prevalent patients with RA and general population comparators.
Prevalent patients with AS
Individuals above 18 years of age with ≥1 visit in non-primary outpatient care listing a diagnosis code for AS at an internal medicine or rheumatology department in 2006–2011 were identified. Of the identified patients, those who had ≥1 diagnosis code for RA, psoriatic arthritis or systemic lupus erythematosus were excluded. This definition has been found to have a high validity, with a positive predictive value of close to 90% for axial spondyloarthritis according to the Assessment of SpondyloArthritis (ASAS) criteria.8
Prevalent patients with RA
Individuals with ≥2 visits in non-primary outpatient care listing RA in 2006–2011, whereof ≥1 visit at an internal medicine or rheumatology department and ≥1 visit listing RA as primary diagnosis were identified. Patients were excluded if they had ≥1 diagnosis code listing AS, psoriatic arthritis, systemic lupus erythematosus, spondyloarthritis or inflammatory bowel disease. This definition of RA has previously been linked to a positive predictive value close to 90%.9
General population comparators
Five general population comparator subjects were individually matched on birth year, sex and county of residence to every patient with prevalent AS by the year of first identification of the index patient with AS. Matching the population comparator to the RA cohort yielded similar relative risks, though with lower statistical precision for the AS versus general population comparison cohort (and vice versa).
Follow-up
Cohort entry was defined as the first date after 1 January 2006 when the inclusion criteria to the study cohorts were met, while start of follow-up was set to 1 year after cohort entry to avoid inclusion of patients for reasons related to an imminent CV occurrence. For the AS cohort start of follow-up was therefore 1 year after the first visit listing a diagnosis of AS at an internal medicine or rheumatology department, and for the RA cohort the start of follow-up was 1 year after the last of the second visit listing RA, the first visit listing RA at an internal medicine or rheumatology department or first visit listing RA as primary diagnosis. The individually matched general population comparator cohort received the same start date as their matched index person with AS. Study subjects were followed until first of an outcome event, death, emigration or end of follow-up (31 December 2012).
Outcomes
Three outcomes were defined using previously devised algorithms for register-based definitions of ACS,10 stroke11 and DVT/PE12 (definitions of diagnosis groups are listed in online supplementary table 1):
First ever ACS, defined as first ever hospitalisation with a primary diagnosis of ICD-10 I20-I21 in 2007–2012, or having a cause of death listing ICD-10=I21. Study subjects with a history of ischaemic or congestive heart disease (primary or contributory diagnosis) were excluded.
First ever stroke, defined as first hospitalisation with a primary or contributory diagnosis of ICD-10 I60-I61, I63 in 2007–2012. Study subjects with a history of ischaemic stroke, haemorrhagic stroke or unspecified stroke (primary or contributory diagnosis) were excluded.
First ever DVT/PE, defined as first ever non-primary outpatient care visit or hospitalisation with a primary or contributory diagnosis of DVT/PE in 2007–2012. Subjects with a history of DVT/PE were excluded.
In the analysis of ACS, individuals with a history of stroke or DVT/PE were not excluded, and vice versa for the other outcomes.
Statistical analyses
Incidence rates for the three different outcomes were calculated in each of the three cohorts, both overall as well as per age and sex categories. We also computed standardised incidence rates for the RA cohort, using the age and sex distribution of the AS cohort as standard. All incidence rates were estimated with 95% CIs assuming a Poisson distribution among the observed cases. Relative risks comparing the study cohorts were estimated by Cox proportional hazard regression13 adjusted for age and sex.
Sensitivity analyses
To increase statistical precision, at the expense of our restriction to patients seen in contemporary daily clinical practice, we modified the inclusion criteria and date of earliest start of follow-up to 1 January 2001, retaining end of follow-up at 31 December 2012.
Results
Characteristics of the study population
A total of 5358 patients with AS, 37 245 patients with RA and 25 006 general population comparator subjects were identified (table 2). As expected, there was a male predominance among the patients with AS, a female predominance in RA and a lower mean age at cohort entry in AS. Compared with the general population, the crude baseline prevalence of selected comorbidities, and of baseline prescriptions of the corresponding medications (with the exception of NSAIDs), was somewhat higher in AS, and more markedly elevated in RA (table 3), but taking age and sex into account, and with the exception of NSAIDs, there was no major differences comparing patients with AS and patients with RA. With respect to NSAIDs, 70% of the patients with AS had filled at least one prescription during the year prior to start of follow-up, compared with 58% of those with RA, and 15% in the general population.
Demographic characteristics at start of follow-up of the cohorts (N (%) if not otherwise stated) of Swedish patients with ankylosing spondylitis (AS), rheumatoid arthritis (RA) and general population comparator subjects (GenPop) identified 2006 through 2011
Medical characteristics at start of follow-up of the cohorts of Swedish patients with AS, RA and GenPop identified 2006 through 2011
Incidence of ACSs
Among 4898 patients with AS at risk of a first ever ACS, 69 events occurred during 20 251 person-years of follow-up (crude incidence 3.4 per 1000; table 4). In the RA group, 966 ACS events occurred (crude incidence 7.6 per 1000, which corresponded to an incidence of 4.9 per 1000 once standardised to the AS population). In the general population, the incidence of ACS was 2.4 per 1000 (table 4).
Overall number of study subjects, events and sum of follow-up years, and overall, sex-specific and age-specific incidence rates per 1000 person-years (95% CI) of ACS, stroke and DVT/PE in the three cohorts of Swedish patients with AS, RA and GenPop identified 2006 through 2011 and followed for CV incidence through 31 December 2012
Compared with the general population, the relative risk (HR) of ACS in AS was 1.3 (95% CI 1.0 to 1.7) and 1.7 (1.4 to 2.0) in RA. The point estimates were the highest in the youngest age groups and tended to be lower in AS versus RA. We noted no sex difference in relative risks (figure 1 and online supplementary table 3).
Sex-adjusted and age-adjusted HRs, overall, and by sex and age for ACS, stroke and DVT/PE in patients with AS and in patients with RA, using general population rates as reference. ACS, acute coronary syndromes; AS, ankylosing spondylitis; DVT, deep venous thromboembolism; GenPop, general population comparators; PE, pulmonary embolism; RA, rheumatoid arthritis.
Incidence of stroke
Based on 65 stroke events in the AS cohort and 1036 stroke events in the RA cohort, the incidences were 3.0 per 1000 in AS and 3.0 in RA (standardised to AS), which, compared with 1.8 per 1000 in the general population, resulted in a relative risk of 1.5 (1.1 to 2.0) in AS and 1.5 (1.2 to 1.8) in RA versus the general population. For stroke, we noted higher relative risk estimates for women than for men (figure 1 and online supplementary table 3). Excluding individuals with atrial fibrillation did not alter the relative risks.
Incidence of thromboembolic events
Based on 68, 1016 and 213 events in the AS, RA and general population cohorts, respectively, the incidences of DVT/PE were 3.2 (AS), 4.4 (RA, standardised to AS) and 2.1 (general population). The corresponding relative risks were 1.4 (1.1 to 1.9) for AS and 1.8 (1.5 to 2.1) for RA. The relative risk estimates were typically lower in AS compared with RA, and the highest point estimates of relative risks were noted among men (figure 1 and online supplementary table 3).
Sensitivity analyses
When the study period was extended to encompass patients and follow-up from 2001 (instead of 2006) and later, the person-time and the number of events increased substantially (eg, 160 ACS events, 135 DVT/PE events and 128 stroke events in the AS cohort, followed for approximately 45 000 person-years). All associations remained, however, largely similar to those of the main analysis; the relative risk of ACS in AS was 1.3 and 1.6 in RA, the relative risk of thromboembolic events was 1.4 in AS and 1.9 in RA and the relative risk of stroke was 1.4 in AS and 1.4 in RA (online supplementary tables 4 and 5). Similarly, when the two study periods 2001–2006 and 2007–2011 were compared, the relative risks (RA vs general population, AS vs general population and AS vs RA) in each were similar (data not shown). Defining AS as at least two (instead on at least one) registered visit with a diagnosis code for AS did not materially alter the relative risks.
Discussion
In this study based on linkage of contemporary cohorts of patients with AS and RA to nationwide registers on incident CV events, we made a series of observations. (a) Taking age and sex into account, the pattern of baseline CV comorbidities and CV medications in the AS and the RA populations were reasonably similar. (b) Patients with AS were at moderately increased (30%–50%) risk of all three types of incident CV events under study compared with the general population. (c) Taking age and sex into account, the corresponding risks in patient with RA were all higher (or, for stroke: similar). (d) The higher incidence in RA than in AS was observed in almost all strata as defined by age and sex.
Although the existing literature on CV disease risks in AS is considerably smaller than that in RA, most14–21 but not all studies6 ,22 to date have reported some level of risk increase for different types of incident CV diseases in patients with AS compared with non-AS subjects, though estimates for, for example, ischaemic heart disease, often broadly defined, have varied from a 20% to a 120% increase in incidence. As such, our 30% increase in risk for ACS is within the range of previously reported estimates. For other incident CV endpoints, such as stroke, reported risks have ranged from non-elevated22 to 140% increased.18 Again our 50% increase in risk is thus within the broad range of previously reported risks. For other CV endpoints, such as thromboembolic endpoints, the existing literature is scarce; hence, our finding of a 40% increase in risk is difficult to benchmark. Some studies have assessed CV mortality rather than incidence in patients with AS.23 Although seemingly robust, the observed mortality (relative) risks must be viewed in light of the fact that CV mortality is the resultant of incidence and prognosis, and may be affected by any increase in underlying mortality attributable to the index disease (here AS).
For several reasons related to study methodology, background risks, AS cohort characteristics and CV outcome definitions, the comparison of relative risk estimates across studies such as those described above is not straightforward. By including a cohort of patients with RA subjected to the same follow-up and CV ascertainment as our AS cohort, we could, however, compare the risks observed in AS with those in RA, for which the level of risk increase is better known. In doing so, we could both ‘verify’ the increased risks for ACS, stroke and thromboembolic event previously reported in RA at a level on par with previous studies, and offer context to the risks observed in AS. The discrepancy in level of risk increase between AS and RA in our study is in the same order of magnitude as in the only other study that have made similar comparisons in cardiovascular disease (CVD) risks in patients with RA and AS.16
While we set out to assess, quantify and compare CV incidences in AS and RA, relative to the general population, we abstained from attempts to partition and attribute the observed risks in either disease to different components or to investigate the extent to which any such were shared across AS and RA. For instance, previous studies have illustrated that the increased CV risk in RA most likely represents interplay between traditional CV risk factors and RA disease/treatment characteristics.24 In AS, the pattern of CV risk factors, both traditional and disease related, may be different than in RA and in the general population. For instance, NSAID use, which was higher in the AS cohort than in the RA cohort, and certainly than in the general population cohort, may affect CV risk.25 We deliberately abstained from modelling any specific effect of NSAIDs on our risks, mainly since the differential channelling to NSAIDs in the populations under study put straightforward adjustment into question. Whereas NSAID in AS is common and a mainstay treatment, NSAID in the general population is less common and used for different indications, which in turn may be differently linked to CV risks. Adjustment for, for example, NSAIDs in our study, whether at baseline or time dependently, is thus likely to change the relative risk for ACS but without guaranteeing that the bias accommodated would be larger than the bias introduced. In light of our results, an important next step will, however, be to disentangle different types of private and/or shared risk factors or pathways for CV occurrence in AS and RA.
Our study has several limitations. With our study design, we were neither able nor aimed to assess risks related to accumulated disease severity or treatment. The latency from first symptoms to diagnosis is 10–15 years longer in AS than in RA.8 We studied prevalent patients with AS (and RA) and therefore did not compare trajectories in relative risks as a function of disease. We took caution to exclude the first 12 months from the inclusion-defining visits with AS or RA to avoid inflation of risks due to reversed causality caused by visits for reasons more related to an impending CV event than to the AS (or RA) disease. However, since the overall mortality in RA is higher than in AS, and since CV morbidity is linked to disease severity in RA,26 we cannot exclude that competing risks (‘depletion of susceptibles’) might have underestimated the true difference in relative risks for AS and RA, at least in the older age groups.
Our study also has some strengths; instead of assessing composite CV endpoints, or CV mortality, we could study the incidence of distinct CV phenotypes, using validated algorithms applied to prospective, data sources that were independent of the AS and RA exposures (both of which were also based on validated algorithms). The statistical precision permitted comparisons between AS (or RA) and the general population, and between AS and RA, per each CV event type. The register-based setting provided a nationwide and population-based sample, ensuring generalisability to (average) patients in clinical practice, and minimised losses to follow-up.
In conclusion, we demonstrate that prevalent patients with AS are at a 30%–50% increased risk of acute coronary, cerebrovascular and thromboembolic events compared with the general population. For acute coronary and thromboembolic events, this level of increase is around half of that seen in the corresponding age and sex groups with RA, while the increased risk of cerebrovascular events in patients with AS was similar to that in RA.
References
Footnotes
Handling editor Tore K Kvien
Contributors JKE had full access to all of the data in the study and take responsibility for the accuracy of the data analysis. JA, JKE and LJ: trial design. JA, KB, JKE and LJ: quality control of data, analysis and interpretation of data and critical revision of the manuscript for important intellectual content. JA and JKE: drafting of the manuscript.
Funding This study was funded through grants from the Swedish Research Council, the Swedish Foundation for Strategic Research and from the Stockholm County Council (ALF).
Competing interests JA has received research grants from Pfizer, AstraZeneca, Merck and UCB in relation to work based on the Swedish Biologics Register ARTIS. JA and JE reported participating in research projects fully or partly funded by Novo Nordisk. JE had served as an external consultant to AbbVie.
Ethics approval Stockholm Ethics Review Board.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The authors have access to raw data for this study and may be contacted for inquiries. According to national data protection rules, these linked raw data cannot be distributed further.