To estimate the excess risk of cardiovascular and cerebrovascular diseases among individuals with ankylosing spondylitis (AS) in Quebec compared with the general population of Quebec.
To estimate the excess risk of cardiovascular and cerebrovascular diseases among individuals with ankylosing spondylitis (AS) in Quebec compared with the general population of Quebec.
A retrospective cohort study was conducted using population-based administrative data from Quebec. The cohort included all adult individuals with at least 1 AS diagnosis on physician billing or hospital discharge records between 1996 and 2006. A comparison cohort was generated using a 1% random sample of individuals without AS. Cardiovascular and cerebrovascular diseases, and associated hospitalizations, were classified into 1 of 6 subcategories: congestive heart failure, valvular (aortic or nonaortic) heart disease, ischemic heart disease, cerebrovascular disease, or “other” cardiovascular disease. The age- and sex-stratified prevalence estimates, and standardized prevalence ratios, of cardiovascular or cerebrovascular disease in patients with AS, compared to that in the general population, were calculated.
The AS cohort included 8,616 individuals diagnosed over the period 1996–2006. The prevalence of cardiovascular and cerebrovascular diseases increased with increasing age for all cardiovascular disease subgroups, and was similar for individuals of both sexes. Age- and sex-stratified prevalence ratios were highest in younger individuals with AS. The age- and sex-standardized prevalence ratios comparing the risk among those with AS to the risk in the general population were as follows: for aortic valvular heart disease 1.58 (95% confidence interval [95% CI] 1.31–1.91), for nonaortic valvular heart disease 1.58 (95% CI 1.43–1.74), for ischemic heart disease 1.37 (95% CI 1.31–1.44), for congestive heart failure 1.34 (95% CI 1.26–1.42), for “other” cardiovascular disease 1.36 (95% CI 1.29–1.44), for cerebrovascular disease 1.25 (95% CI 1.15–1.35), and for any hospitalization for a cardiovascular or cerebrovascular disease 1.31 (95% CI 1.22–1.41).
Compared with the general population, patients with AS are at increased risk for many types of cardiovascular and cerebrovascular diseases, and are more likely to be hospitalized for these diseases. The excess risk is greatest in younger patients with AS.
It is well established that individuals with ankylosing spondylitis (AS) are at increased risk of cardiac disease (1). This may occur as the chronic inflammation that is characteristic of AS manifests itself in cardiac structures (1, 2). Sequelae in cardiac valves were the first to be identified in patients with AS (1). More recent evidence suggests that the risks of other forms of cerebrovascular and cardiovascular disease (including conduction disturbances, left ventricular dysfunction, and myocardial infarction) may also be elevated (3–10). It is likely that inflammatory processes are implicated in disease development. The role of inflammation in atherogenesis and plaque formation is now well recognized (11), and inflammation has been linked to an atherogenic lipid profile among individuals with AS (12, 13). Other contributory factors may include the use of nonsteroidal antiinflammatory drugs (NSAIDs) (14), decreased physical activity (15), genetics (16), and the higher frequency of metabolic syndrome among those with AS (13).
While the link between AS and cardiovascular or cerebrovascular disease is now evident, less is known about the magnitude of that increased risk. Investigators have largely based their findings on AS patient populations derived from rheumatology clinics (1, 2, 5–7, 9, 10), a setting in which patients are likely to have more severe disease (17). It is unclear how these estimates may be generalizable to the entire AS population. Recent guidelines from the European League Against Rheumatism (EULAR) recommend that physicians conduct an annual assessment of cardiovascular risk in individuals with AS (although the EULAR guidelines also acknowledge that the evidence base to support this recommendation is lacking) (14). Therefore, the objectives of the present study were to 1) describe the prevalence of cardiovascular and cerebrovascular diseases in a population-based cohort of individuals with AS in the Canadian province of Quebec between 1996 and 2006, and 2) estimate the age- and sex-specific increased risks of these conditions relative to those in the general population.
A retrospective cohort study was conducted using longitudinal, patient-level, population-based administrative data from the Régie de l'Assurance Maladie du Québec (RAMQ) database and the Quebec provincial hospital discharge (MED-ÉCHO) database. These data were used to estimate the period-specific prevalence of cardiovascular and cerebrovascular diseases among 2 cohorts, comprising individuals in whom AS was diagnosed (AS cohort) and those in whom it was not (general population cohort).
The RAMQ database comprises insurance data on all provincial health plan registrants in Quebec (∼99% of the population, or 7,731,600 residents in 2006) for medical and hospital services. The databases of the RAMQ include data on claims for physician services (18, 19) offered by the ∼92% of Quebec physicians who work on a fee-for-service basis (19). These data were linked by unique patient identifier to acute- care hospital discharge abstracts from the MED-ÉCHO database. These databases have been widely validated for epidemiologic research (20, 21). However, they only record drug utilization for the ∼43% of residents who are insured by the provincial drug plan (18, 22).
We also obtained counts of persons in each age and sex stratum from Canadian national census data collected in 1996, 2001, and 2006. This allowed us to obtain estimates of the population at risk (general population) (23,24). Ethics approval of the protocol and data release were provided by the Commission d'acces à l'information du Quebec.
The sampling frame was all adult registrants (age >19 years) in the RAMQ. Members of the AS cohort were selected on the basis of whether the individual had at least 1 diagnosis of AS defined according to the International Classification of Diseases, Ninth Revision (ICD-9) code 720.0 between January 1, 1996 and December 31, 2006, inclusive. This single ICD-9 code–based definition has high sensitivity (91%) and specificity (99%) for identifying individuals with AS from administrative data (25). All AS cohort members were considered at risk as of January 1, 1996, irrespective of their date of first recorded diagnosis of AS. Thus, the AS cohort was prevalence-based and, at the start of the period, included individuals who were diagnosed as having AS at a later date. As a sensitivity analysis, we selected individuals for the AS cohort by requiring that the diagnosis of AS be made at least twice with ICD-9 code 720.0 within the time period.
The general population cohort included a 1% random sample of RAMQ registrants without AS. All general population cohort members had at least one interaction with the health care system during the period for inclusion, and were also enrolled as of January 1, 1996.
Censoring occurred at the time that the diagnosis of the cardiovascular or cerebrovascular disease was made or at the time that an individual's RAMQ coverage was terminated (presumably at death or on moving out of the province). In these cases, date of censoring was assigned at the end of the year prior to an individual's being deregistered.
Sample demographics were summarized by age, sex, and the year of first reported diagnosis of AS. The frequency of codiagnoses of other inflammatory conditions was calculated for ulcerative colitis (2 diagnoses with an ICD-9 code of 556.x without Crohn's disease), Crohn's disease (ICD-9 code 555.x without unspecified intestinal obstruction [ICD-9 code 560.9]), and psoriatic arthritis (ICD-9 code 696.x) (26).
Since it was not possible to individually match members of the AS cohort and general population cohort based on the data provided by the RAMQ, we calculated age- and sex-stratified and standardized, measures of risk and increased risk. We estimated the age- and sex-stratified prevalence of cardiovascular and cerebrovascular diseases using 2 definitions. The primary definition required that at least 1 diagnosis of a cardiovascular or cerebrovascular disease be made according to the relevant ICD-9 code within the time period (details on diagnoses and codes are available from the corresponding author upon request). The secondary definition required that at least 2 diagnoses be made by the relevant ICD-9 code within the time period. Because of the nature of the study design, these prevalence estimates may be an overestimate of the risk of AS-related cardiovascular or cerebrovascular disease, since some diagnoses could have preceded the AS diagnosis. Therefore, we also estimated the increased risk among those with AS compared to the general population by calculating sex- and age-stratified prevalence ratios, as well as overall age- and sex-standardized prevalence ratios, with 95% confidence intervals (95% CIs). These measures were calculated for individuals in any of the following disease categories: nonaortic or aortic valvular disease, ischemic heart disease, congestive heart failure, cerebrovascular disease, and “other” cardiovascular disease (excluding hypertension).
In addition, standardized prevalence ratios were calculated for hospitalization for a cardiovascular or cerebrovascular disease diagnosis, defined according to a primary ICD-9 code for 1 or any of the above-described subgroups. These hospitalizations were identified on the hospital discharge abstract.
The prevalence of AS during 1996–2006 was estimated using the number of AS cases over the period as the numerator, and Quebec population estimates as the denominator (23,24). We divided the number of AS cases with a first diagnosis of cardiovascular or cerebrovascular disease by the number of AS cases overall, to calculate the prevalence of cardiovascular or cerebrovascular disease (per 100 AS cases). The overall prevalence, as well as age- and sex-stratified prevalence, of cardiovascular and cerebrovascular disease was calculated using the following age categories (selected for their clinical relevance): age 20–39 years, age 40–59 years, and age ≥60 years. Similar calculations were performed for the general population cohort. No differences in the distribution of ages between the AS cohort and general population cohort were observed.
We also calculated 2 measures of increased risk of cardiovascular or cerebrovascular disease among individuals with AS. The first measure was the calculation of the age- and sex-stratified prevalence ratios of cardiovascular or cerebrovascular disease in patients with AS relative to the general population cohort, for which the 95% CIs were approximated by exponentiating the 95% confidence bounds for the logarithm of the prevalence ratio (27). The second measure was the calculation of the age- and sex-standardized prevalence ratios using generalized linear models, assuming a Poisson distribution with a log-link function (28).
Each model included terms for age group (in 5-year bands), sex, and whether an individual had AS. Age group was treated as a continuous variable, assuming the impact of age was linear. Within each stratum defined by unique combinations of age group, sex, and AS status, the number of cases was the outcome variable, assumed to follow a Poisson distribution. The log of total person–time contributed within the stratum was included as an offset, and the 95% CIs for the standardized prevalence ratios were estimated by exponentiating the end points of confidence intervals for the regression coefficient associated with AS status. All statistical analyses were performed using Stata statistical software.
The RAMQ data included a sample of 8,616 registrants with AS. The characteristics of the sample at cohort entry are presented in Table 1. The median age at first recorded diagnosis was 42.5 years, and 56% of the patients (4,836) were male. The median age at cohort entry for the general population cohort was also 42.5 years. The mean length of followup was similar between patients with AS (10.2 years) and the general population (10.5 years). The prevalence of AS over the time period 1996–2006 is presented in Table 2.
|Male (n = 4,836)||Female (n = 3,780)|
|Year of first recorded AS diagnosis|
|1996||584 (12.1)||305 (8.1)|
|1997||468 (9.7)||297 (7.9)|
|1998||356 (7.4)||255 (6.7)|
|1999||436 (9.0)||336 (8.9)|
|2000||478 (9.9)||394 (10.4)|
|2001||425 (8.8)||319 (8.4)|
|2002||421 (8.7)||353 (9.3)|
|2003||450 (9.3)||418 (11.1)|
|2004||445 (9.2)||434 (11.5)|
|2005||398 (8.2)||298 (7.9)|
|2006||375 (7.8)||371 (9.8)|
|Age group at cohort entry|
|20–39 years||2,061 (42.6)||1,590 (42.1)|
|40–59 years||1,976 (40.9)||1,489 (39.4)|
|≥60 years||799 (16.5)||701 (18.5)|
|Codiagnosis of other inflammatory conditions|
|Ulcerative colitis||137 (4.4)||167 (3.5)|
|Crohn's disease||279 (8.0)||3.4 (6.3)|
|Psoriasis||310 (11.5)||435 (9.0)|
|2006†||4,455 (0.15)||3,590 (0.12)||8,045 (0.14)|
Age- and sex-stratified prevalence estimates for all cardiovascular and cerebrovascular disease subgroups, using the primary ICD-9 code–based definition, are presented in Table 3. Among both the AS cohort and the general population cohort, ischemic heart disease was the most frequent diagnosis, followed by “other” cardiovascular disease and congestive heart failure. In the AS cohort, prevalence estimates for these 3 diagnoses were 26.4%, 19.2%, and 16.6%, respectively, among male patients, and 21.5%, 17.9%, and 13.5%, respectively, among female patients. In the general population cohort, prevalence estimates for these diagnoses were 20.6%, 14.3%, and 13.3%, respectively among male subjects, and 17.1%, 14.9%, and 11.4%, respectively, among female subjects.
|CV Definition 1||CV Definition 2|
|Patients with AS (n = 8,616)||General population (n = 50,699)||Patients with AS (n = 8,616)||General population (n = 50,699)|
|Aortic valvular heart disease||2.0||1.1||1.5||1.4||1.1||0.5||0.8||0.8|
|Age 20–39 years||0.4||0.3||0.2||0.3||0.2||0.2||0.1||0.1|
|Age 40–59 years||2.4||1.2||1.4||1.1||1.5||0.3||0.6||0.4|
|Age ≥60 years||5.0||2.7||4.8||3.8||2.6||1.4||2.4||1.3|
|Nonaortic valvular heart disease||5.7||6.8||4.0||3.9||3.0||2.7||2.0||1.7|
|Age 20–39 years||1.5||4.3||0.9||1.4||0.5||1.5||0.2||0.3|
|Age 40–59 years||6.2||6.3||3.9||3.8||3.1||2.6||2.0||1.6|
|Age ≥60 years||15.3||13.4||12.1||8.9||9.1||5.7||6.3||4.3|
|Ischemic heart disease||26.4||21.5||20.6||17.1||18.9||13.0||15.0||11.0|
|Age 20–39 years||9.1||7.1||5.2||3.6||4.1||2.0||2.2||1.3|
|Age 40–59 years||30.4||24.6||23.9||16.0||21.7||14.1||16.8||8.6|
|Age ≥60 years||61.3||47.8||52.6||44.3||49.9||35.4||43.3||33.1|
|Congestive heart failure||16.6||13.5||13.3||11.4||11.0||7.6||9.3||6.9|
|Age 20–39 years||4.9||4.1||2.7||2.3||2.6||1.4||1.2||0.8|
|Age 40–59 years||18.2||12.7||14.0||8.9||11.8||6.4||9.3||4.4|
|Age ≥60 years||42.9||36.8||38.9||32.4||30.4||24.1||29.8||22.7|
|“Other” cardiovascular disease||19.2||17.9||14.3||14.9||12.0||9.6||9.0||8.4|
|Age 20–39 years||8.4||10.5||4.9||6.3||3.9||4.5||2.0||2.4|
|Age 40–59 years||20.6||16.1||13.4||12.9||12.3||8.4||7.7||6.0|
|Age ≥60 years||43.4||38.2||40.2||34.2||31.9||23.4||30.0||23.7|
|Age 20–39 years||1.9||2.9||1.6||1.7||0.7||1.2||0.6||0.6|
|Age 40–59 years||7.8||6.2||6.6||5.4||5.4||3.3||4.1||2.9|
|Age ≥60 years||27.2||24.5||24.6||21.6||18.7||13.8||17.4||14.6|
Using the secondary definition for diagnosis of the cardiovascular and cerebrovascular diseases (at least 2 diagnoses made by the relevant ICD-9 code within the time period), a lower prevalence of cardiovascular and cerebrovascular diseases was observed in both cohorts; nevertheless, the trends noted with the primary definition remained consistent. In both male patients and female patients with AS, ischemic heart disease occurred most frequently, followed by “other” cardiovascular disease and congestive heart failure (Table 3). This was similar in the general population cohort, except among male subjects, in whom the risk of congestive heart failure (9.3%) was slightly higher than the risk of “other” cardiovascular disease (9.0%) (Table 3).
The prevalence of cardiovascular and cerebrovascular diseases increased with age among both cohorts, when either set of definitions was used. Using the primary definition for diagnosis, ischemic heart disease was the most frequent diagnostic subtype among male patients with AS in all age groups, increasing from 9.1% among male patients age 20–39 years to 61.3% among male patients age ≥60 years. Although ischemic heart disease was also the most frequent diagnosis among female patients age ≥40 years (24.6% among those age 40–59 years and 47.8% among those age ≥60 years), the prevalence of “other” cardiovascular disease was highest among the younger female patients (10.5% in those age 20–39 years) than in the younger male patients (Table 3). Among male subjects without AS, ischemic heart disease was also the most frequent diagnosis, being observed in 5.2% among those age 20–39 years, 23.9% among those age 40–59 years, and 52.6% among those age ≥60 years. Finally, among female subjects without AS, the most frequent condition was “other” cardiovascular disease among those age 20–39 years (6.3%), and ischemic heart disease among those age 40–59 years (16.0%) and those age ≥60 years (44.3%) (Table 3). The same age-specific trends were observed with the secondary definition of cardiovascular or cerebrovascular diseases, although the magnitude of the prevalence estimates was smaller (Table 3).
Individuals with AS in almost all of the age groups were at significantly elevated risk of all cardiovascular (but not cerebrovascular) diseases considered, compared to individuals without AS of the same age and sex (Table 4). The increased risk among individuals of younger age was consistent across cardiovascular or cerebrovascular disease subcategories. Of note, younger male patients with AS (age 20–39 years) were at a particularly greater increased risk of aortic valvular disease, congestive heart failure, and ischemic heart disease, compared to younger male subjects without AS (prevalence ratio 2.97 [95% CI 1.31–6.70], 1.80 [95% CI 1.44–2.25], and 1.75 [95% CI 1.50–2.05], respectively). In contrast, younger female patients with AS (age 20–39 years) were at a particularly high risk of nonaortic valvular heart disease, ischemic heart disease, and congestive heart failure (prevalence ratio 3.01 [95% CI 2.28–3.97], 1.97 [95% CI 1.61–2.41], and 1.75 [95% CI 1.34–2.29], respectively). In general, older individuals with AS were at increased risk of some, but not all, cardiovascular and cerebrovascular diseases, compared to the general population (Table 4).
|Aortic valvular heart disease|
|Age 20–39 years||2.97 (1.31–6.70)||1.13 (0.39–3.26)|
|Age 40–59 years||2.31 (1.64–3.24)||1.53 (0.92–2.55)|
|≥60 years||1.27 (0.91–1.78)||0.87 (0.54–1.38)|
|Nonaortic valvular heart disease|
|Age 20–39 years||1.54 (1.03–2.30)||3.01 (2.28–3.97)|
|Age 40–59 years||1.47 (1.20–1.78)||1.57 (1.26–1.95)|
|Age ≥60 years||1.20 (1.00–1.43)||1.42 (1.16–1.74)|
|Ischemic heart disease|
|Age 20–39 years||1.75 (1.50–2.05)||1.97 (1.61–2.41)|
|Age 40–59 years||1.27 (1.18–1.37)||1.54 (1.39–1.70)|
|Age ≥60 years||1.17 (1.10–1.24)||1.08 (0.99–1.17)|
|Congestive heart failure|
|Age 20–39 years||1.80 (1.44–2.25)||1.75 (1.34–2.29)|
|Age 40–59 years||1.30 (1.17–1.44)||1.42 (1.23–1.65)|
|Age ≥60 years||1.10 (1.01–1.21)||1.13 (1.02–1.26)|
|Age 20–39 years||1.22 (0.87–1.71)||1.69 (1.23–2.33)|
|Age 40–59 years||1.19 (1.00–1.40)||1.15 (0.93–1.43)|
|Age ≥60 years||1.10 (0.97–1.25)||1.14 (0.99–1.31)|
|“Other” cardiovascular disease|
|Age 20–39 years||1.72 (1.46–2.03)||1.67 (1.42–1.96)|
|Age 40–59 years||1.54 (1.39–1.70)||1.25 (1.10–1.42)|
|Age ≥60 years||1.08 (0.99–1.18)||1.12 (1.01–1.24)|
Standardized prevalence ratios, comparing the prevalence of cardiovascular or cerebrovascular diseases among patients with AS to that in the general population, after adjusting for age and sex, are presented in Figure 1A. Significant elevations in the risk of cardiovascular and cerebrovascular diseases were observed across all disease categories considered, and not only valvular diseases. Standardized prevalence ratios ranged from 1.25 (95% CI 1.15–1.35) for cerebrovascular disease, 1.34 (95% CI 1.26–1.42) for congestive heart failure, 1.36 (95% CI 1.29–1.44) for “other” cardiovascular disease, 1.37 (95% CI 1.31–1.44) for ischemic heart disease, 1.58 (95% CI 1.31–1.91) for aortic valvular heart disease, and 1.58 (95% CI 1.43–1.74) for nonaortic valvular heart disease. This indicates that patients with AS are at a 25%, 34%, 36%, 37%, 58%, and 58%, increased risk of developing cerebrovascular disease, congestive heart failure, “other” cardiovascular disease, ischemic heart disease, aortic valvular heart disease, and nonaortic valvular heart disease, respectively, compared to those without AS of the same age and sex. Use of the secondary set of definitions of cardiovascular and cerebrovascular diseases produced results similar to those with the primary definition (Figure 2B).
A significantly elevated risk of hospitalizations for cardiovascular and cerebrovascular diseases was also observed (Figure 2A), both overall (i.e., any hospitalization) and for all subgroups except valvular heart disease (there were few hospitalizations for valvular diagnoses in either cohort). Standardized prevalence ratios for hospitalization ranged from 1.24 (95% CI 1.05–1.46) for cerebrovascular disease, 1.26 (95% CI 0.83–1.93) for nonaortic valvular disease, 1.31 (95% CI 1.06–1.61) for congestive heart failure, 1.34 (95% CI 0.16–11.6) for aortic valvular disease, 1.37 (95% CI 1.24–1.51) for ischemic heart disease, and 1.56 (95% CI 1.37–1.79) for “other” cardiovascular disease. The standardized prevalence ratio for any hospitalization was 1.31 (95% CI 1.22–1.41), which means that patients with AS are at a 31% increased risk of hospitalization for any cardiovascular or cerebrovascular disease, compared to those without AS of the same age and sex.
Sensitivity analyses were performed by varying the definition of AS, and when the more stringent definition was imposed, requiring at least 2 relevant ICD-9 code–based diagnoses of AS within the time period, prevalence estimates of AS declined. Nevetheless, the direction and magnitude of the prevalence ratios for cardiovascular or cerebrovascular diseases did not change materially (Figure 1B).
The central goals of treating patients with AS include reduction of pain and stiffness, maintenance or improvement of function, mobility, and quality of life, prevention of disability and structural damage, and management of the many comorbid medical conditions that can result (29). Despite the recognition that the risk of cardiovascular and cerebrovascular diseases is increased in patients with AS (1,2,5–7,9,10,14), comprehensive population-based information on the risk of these diseases in AS has not been available as yet. In the present study, we analyzed population-level data from Quebec, which showed that over a 10-year period, individuals with AS are at significantly increased risk of developing a cerebrovascular or cardiovascular disease. The increased risk was highest for younger individuals with AS, and was similar between the sexes.
After adjustment for age and sex, individuals with AS had at least a 25% increased risk of developing any of the cardiovascular disease diagnoses, when compared to individuals without AS. The highest increase in risk, of ∼60%, was observed for valvular heart disease, with more moderate increases in risk of between 25% and 40% for ischemic heart disease, congestive heart failure, cerebrovascular disease, and “other” cardiovascular disease. Increases in risk were consistent, irrespective of the definition of cardiovascular or cerebrovascular disease used. These increases in risk were mirrored in the increased risk of hospitalization for these conditions. Patients with AS had at least a 30% increased risk of hospitalization for a cardiovascular or cerebrovascular disease, compared to those without AS of the same age and sex.
The need for well-designed and adequately powered cohort studies to precisely quantify the risk of cardiovascular and cerebrovascular disease in patients with AS has been cited (4, 30), since information on the magnitude of the risk is conflicting (1–3, 29–32). Many studies have lacked the appropriate comparison groups, while in other studies there are issues of external validity, because the studies have enrolled patients with AS from specialty clinics, a setting in which patients are likely to have more severe disease. Whereas several studies have shown an increased frequency of nonvalvular cardiovascular disease in patients with longstanding AS (3, 6–8), other investigators have concluded that routine cardiologic evaluation is not indicated (2, 30). The results of the present study are consistent with the findings from a recent case–control study, in which a 3-fold increase in nonfatal myocardial infarction was observed among patients with AS attending specialty clinics (4), and with findings from a registry-based study from Sweden (32). Our results are also consistent with a recent cross-sectional US claims-based analysis showing an increased risk of congestive heart failure (prevalence 1.8 [95% CI 1.2–2.6]), peripheral vascular disease (prevalence 2.6 [95% CI 1.2–2.6]), cerebrovascular disease (prevalence 1.7 [95% CI 1.3–2.3]), hyperlipidemia (prevalence 1.2 [95% CI 1.1–1.3]), and hypertension (prevalence 1.3 [95% CI 1.1–1.4]) (3). Our results represent an advance in providing comprehensive longitudinal and population-based estimates of the risk of cardiovascular and cerebrovascular diseases among a representative sample of individuals with AS.
Adequately defining AS is critical to analyses based on population data from administrative data sets. Most prior investigators defined AS with the use of diagnostic criteria or radiographic evidence, rather than diagnostic codes provided by physicians in administrative billing data, as was used herein. Population-based administrative data may be a valuable source from which to estimate AS prevalence and for predicting outcomes in patients with AS, as these sources are not susceptible to the reporting or selection biases common to clinic-based studies (27). The present study operationalized the definition of AS to include individuals in whom at least 1 diagnosis was made within the period, a definition found in a US Department of Veterans Affairs study to have the highest validity for classifying AS (sensitivity 91% and specificity 99%) when compared to a diagnosis made by 2 relevant ICD-9 codes during the same period (sensitivity 82%) (25). Additional arguments supporting the use of 1 diagnostic code to identify AS from the RAMQ database were that estimates of lifetime prevalence (∼0.4%) are consistent with estimates from sources other than administrative data (27, 33), and that the prevalence estimates of AS-related comorbid conditions (including psoriasis and inflammatory bowel disease) were in line with published estimates.
As a sensitivity analysis, we explored alternate definitions for diagnosis, including identifying patients with AS by at least 2 diagnostic codes over the time period. The prevalence of AS was, as expected, lower when the more stringent definition was applied, but the overall pattern of prevalence and ratios calculated for cardiovascular and cerebrovascular diseases remained largely unchanged. A rheumatologist-based diagnosis was also considered, but use of this method identified only 40 cases of AS over the 10-year period, indicating that most patients with AS in Quebec are seen by nonrheumatologists.
To select diagnostic codes to identify cardiovascular and cerebrovascular outcomes, we reviewed relevant articles and obtained consensus from clinical experts. The codes that we utilized for diagnosis of congestive heart failure, ischemic heart disease, atherosclerosis, and cerebrovascular disease were also used by Han et al to identify comorbidities in their study (3). The accuracy of these codes in identifying these conditions from Canadian administrative data sets has been confirmed in a number of pharmacoepidemiologic studies (sensitivity >80% and specificity >90%) (21, 34, 35). A coding study from the US Medicare database identified our codes for valvular heart disease to be excellent for ruling in a diagnosis of valvular heart disease, but poor for ruling out the diagnosis (sensitivity 41% and specificity 97%) (36). While the use of 1 diagnostic code to identify cases of cardiovascular or cerebrovascular disease has been done in other studies (3, 35, 36), it may overestimate the true prevalence of those conditions if the physician had coded it as a “rule-out” diagnosis.
The prevalences of cardiovascular or cerebrovascular disease estimated herein using only 1 diagnostic code were slightly higher than other literature-based estimates obtained in studies that also used a single ICD code to identify cardiovascular or cerebrovascular disease (3, 37). We also implemented a more specific, but less sensitive, secondary definition of cardiovascular and cerebrovascular disease, requiring 2 ICD-9 codes over the same period (27). Prevalence estimates using the secondary definition were consistent with or lower than other population-based estimates of cardiovascular or cerebrovascular disease (3, 37–39).
We incorporated a prevalence-based approach in defining the AS and general population cohorts, for 2 reasons. First, as we could not match the members of the general population sample directly, we did not have an innate date of study entry for the members of that cohort. More importantly, we were uncertain of the accuracy of the first date of diagnosis of AS within the period. Clinically, AS is difficult for many nonspecialist physicians to diagnose accurately, and long delays (from 8 to 11 years) can occur between symptom onset and diagnosis (40). An individual's true date of disease onset could therefore have easily occurred prior to the start of the 10-year study period, and applying a washout period may be insufficient to eliminate such prevalent cases. If the first date of AS diagnosis within the period was used to define cohort entry, it would lead to misclassification of both exposure and outcomes, as individuals who developed cardiovascular disease after the AS diagnosis might appear as though the onset of cardiovascular disease occurred before the start of AS, due to the inaccurate date of first AS diagnosis. The prevalence-based approach utilized herein minimizes misclassification and makes best use of the complete followup period.
Strengths of this study include the large, population-based approach that was undertaken to estimate the prevalence of all types of cardiovascular and cerebrovascular disease in AS from a large number of representative individuals over a 10-year period. To our knowledge, this is the first population-based study to comprehensively investigate the risks of cardiovascular and cerebrovascular diseases in AS. The Quebec billing databases are large, population-based, and have been widely used for health services research (20, 21).
Given that the AS cohort comprised all adult individuals in Quebec over the study period, and the general population cohort comprised a random sample of those without AS, selection bias is unlikely to have impacted our findings. We obtained data from the 10 most recent years available to generate robust and timely estimates. The prevalence-based approach to defining AS avoided an effect of immortal time bias, which might impact the results, as may be introduced when members of the exposed (AS) cohort, by definition, must survive a period of time without experiencing the outcome to be considered exposed. This bias would make measures of increased risk appear smaller than the actual risk (41). Lastly, this study measured a comprehensive set of diagnoses to bring awareness to a broad range of cardiovascular and cerebrovascular conditions for which individuals with AS are at increased risk.
The RAMQ databases are designed for reimbursement, and their use for research is associated with potential limitations related to the data availability within them, which could allow information bias to affect the study results. One potential source of misclassification is the identification of AS cases based solely on ICD codes, without clinical verification or the use of classification criteria (42). However, a similar strategy was used to identify AS in the US cross-sectional claims-based analysis (3). Although this algorithm was shown to have excellent sensitivity and specificity for identifying AS (25), we recognize that it has not been evaluated in the context of physician billing data from Canada. Furthermore, as some individuals with mild AS may go long periods without seeing a physician (17, 43, 44), they may have been misclassified into the general population cohort (if the diagnostic visit occurred prior to the start of the period). Such misclassification would attenuate the measures of risk estimated, and therefore represents a conservative bias. Nevertheless, due to the 10-year study period, we believe that this bias is unlikely to have a major impact on the results.
There is also potential bias due to misclassification of outcomes, in particular, a misclassification of individuals as having cardiovascular or cerebrovascular disease. While the ICD-9 codes for most of the cardiovascular and cerebrovascular diseases considered were validated in Canadian provincial (21, 34, 35) or US Medicare (36) data sets, we could not identify studies validating codes for “other” cardiovascular diseases. Use of the medical services database of RAMQ to assess general health status is limited by the requirement that physicians enter only 1 diagnosis code per visit (22). Thus, for individuals over the age of 65 years, in whom there are potentially multiple comorbid conditions, the single ICD-9 code does not provide a complete description of health status.
Furthermore, individuals with a chronic condition like AS may be more likely to have their visit coded as AS, leading to underreporting of cardiovascular and cerebrovascular diseases. However, this would result in lower risk estimates, and therefore the approach that we have taken is conservative.
Finally, there is the potential for ascertainment bias, as it may be more likely that those with AS, as compared with healthy individuals, will be investigated for cardiovascular or cerebrovascular diseases, particularly since the awareness among rheumatologists about the risks of cardiovascular and cerebrovascular disease in AS has been increasing (3). This effect would be particularly pronounced among younger individuals, who would not normally be investigated for cardiovascular or cerebrovascular disease, and may result in a bias toward higher prevalence ratios.
Data to measure cardiovascular and cerebrovascular disease risk factors, such as lipid levels, smoking, or use of NSAID therapy, were not available from the RAMQ. We therefore could not determine whether the increased risk of cardiovascular and cerebrovascular diseases was attributable to AS itself, to the increased prevalence of risk factors in this population, or to medications used in the treatment of AS. Although data on the importance of cardiovascular risk factors in AS are conflicting (3, 45), one study measuring these factors and mortality concluded that those with AS have an enhanced cardiovascular risk profile (13). Significantly higher risks of mortality in individuals with other inflammatory conditions, such as rheumatoid arthritis, have been attributed to an increased occurrence of cardiovascular disease, independent of cardiovascular and cerebrovascular risk factors (14, 46). It has been suggested that the same may hold true for AS (47).
It is interesting that while the absolute risk of cardiovascular or cerebrovascular disease among younger patients with AS remained low, the increased risk was greatest among this age group. The absolute risk of cardiovascular or cerebrovascular disease was much higher in older individuals, in both the AS and comparison cohorts. This suggests that AS as an important component cause of cardiovascular or cerebrovascular disease (48) is much greater among younger, as compared to older, adults with AS. Although not measured directly, our risk estimates support the view that AS should be considered a risk factor for cardiovascular and cerebrovascular disease (4, 49, 50).
Thus, this population-based analysis shows that AS is associated with an increased risk for many types of cardiovascular and cerebrovascular diseases, and increases the likelihood of hospitalization for these conditions. The excess risk of cardiovascular disease is greatest among younger individuals with AS, but is observed in both sexes and across all age strata. These data therefore address an important gap in the management of AS, and provide support for the EULAR recommendation advising regular cardiovascular and cerebrovascular risk assessments for patients with AS.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Levy had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Szabo, Levy, Rao, Lacaille, Cifaldi, Maksymowych.
Acquisition of data. Szabo, Levy.
Analysis and interpretation of data. Szabo, Levy, Rao, Kirbach, Lacaille, Cifaldi, Maksymowych.
Authors Rao, Kirbach, and Cifaldi are employees of Abbott. Drs. Rao and Cifaldi contributed to the study conception and design, while all 3 authors participated in the analysis and interpretation of the data. Abbott had no role in the writing of the manuscript or the decision to submit the manuscript for publication. Publication of this article was not contingent upon approval by Abbott.