|Home | About | Journals | Submit | Contact Us | Français|
Study concept and design: Gelfand, Troxel, Kurd, Azfar Acquisition of data: Gelfand, Wang, Shin, Kurd Analysis and interpretation of data: Gelfand, Shin, Kurd, Troxel, Azfar. Drafting of the manuscript: Gelfand Critical revision of the manuscript for important intellectual content: Gelfand, Shin, Troxel, Kurd, Wang, Dommasch, Azfar Statistical analysis: Gelfand, Shin, Troxel, Wang, Obtained funding: Gelfand, Administrative, technical, or material support: Kurd, Wang, Dommasch, Shin Study supervision: Gelfand, Troxel
Psoriasis is a chronic Th-1 and Th-17 inflammatory disease. Chronic inflammation has also been associated with atherosclerosis and thrombosis. The purpose of this study was to determine the risk of stroke in patients with psoriasis. We conducted a population-based cohort study of patients seen by general practitioners participating in the General Practice Research Database in the United Kingdom 1987-2002. Mild psoriasis was defined as any patient with a diagnostic code of psoriasis but no history of systemic therapy. Severe psoriasis was defined as any patient with a diagnostic code of psoriasis and a history of systemic therapy consistent with severe psoriasis. The unexposed (control) population was composed of patients with no history of a psoriasis diagnostic code. When adjusting for major risk factors for stroke, both mild (HR 1.06, 95% CI 1.0-1.1) and severe (1.43, 95% CI 1.1, 1.9) psoriasis were independent risk factors for stroke. The excess risk of stroke attributable to psoriasis in patients with mild and severe disease was 1 in 4115 per year and 1 in 530 per year, respectively. Patients with psoriasis, particularly if severe, have an increased risk of stroke that is not explained by major stroke risk factors identified in routine medical care.
Psoriasis is a chronic inflammatory disease of the skin and joints that affects about 1-3% of the general population(Gelfand, Weinstein et al. 2005; Kurd and Gelfand 2009). It is associated with impairments in health related quality of life even in mild cases, and excess mortality in severe cases(Gelfand, Feldman et al. 2004; Gelfand, Troxel et al. 2007). The cause of psoriasis remains unknown; however, its etiology involves a complex interaction between genetics and environmental exposures. Psoriasis is incurable, and once symptoms occur, the disease is characterized by a chronic course, with spontaneous, long-term remissions occurring in a minority of patients. Despite many treatment advances, the majority of patients with psoriasis, particularly if severe, have ongoing cutaneous pathology that persists for decades (Gelfand 2007; Horn, Fox et al. 2007; Nijsten, Looman et al. 2007).
Psoriasis was previously believed to affect only the skin, and in some patients, the joints; however, recent evidence suggests that chronic psoriasis may lead to other diseases in which inflammation is important(Azfar and Gelfand 2008). For example, Th-1 chronic inflammation characteristic of the psoriasis phenotype is also central to the pathophysiology of other conditions such as insulin resistance, atherosclerosis, and plaque rupture leading to thrombotic events(Hirosumi, Tuncman et al. 2002; Hansson 2005). Recent epidemiological studies support the link between psoriasis and cardiovascular disease (Gelfand, Neimann et al. 2006; Ludwig, Herzog et al. 2007; Balci, Balci et al. 2008). There is a paucity of data examining the risk of stroke in patients with psoriasis. For example, a recent study found a small increase risk of stroke in patients with psoriasis (HR 1.12, 95% CI 1.00-1.25); however, this study did not make adjustments for confounding factors or psoriasis severity (Kaye, Li et al. 2008).
Stroke is a major cause of morbidity and mortality(Donnan, Fisher et al. 2008). The major common risk factors for stroke include diabetes, hypertension, and smoking(Donnan, Fisher et al. 2008), which also predispose to cardiovascular risk in general. Less common but more specific risk factors for stroke include transient ischemic attacks and atrial fibrillation(Donnan, Fisher et al. 2008). Most strokes (80%) are ischemic in nature, with the minority of strokes being hemorrhagic (Donnan, Fisher et al. 2008). Ischemic and hemorrhagic strokes share similar risk factors and can have similar clinical presentations. They can be differentiated based on sophisticated imaging in the clinical setting, but are often difficult to differentiate in large, population-based epidemiological studies. As with myocardial infarction, chronic inflammation is thought to play an important role in the pathogenesis of stroke(Ding, Zhang et al. 2008; McColl, Allan et al. 2008). The purpose of the present study was to examine the risk of stroke in a broadly representative, population-based cohort of patients with psoriasis.
We identified 129,143 patients with mild psoriasis, 3,603 patients with severe psoriasis, and 496,666 and 14,330 matched controls, respectively (Table 1). Mild psoriasis patients were slightly younger than controls and severe psoriasis patients were slightly older than controls. Psoriasis patients, particularly if severe, were more likely to have established cardiovascular risk factors than patients without psoriasis. The majority of patients identified as having severe psoriasis were treated with methotrexate (Table 2).
The incidence of stroke in control subjects and psoriasis subjects is shown in Table 3. The unadjusted overall risk of stroke per 1000 person-years was slightly lower in mild psoriasis patients (3.7; 95% CI 3.5-3.8) compared to controls (4.05; 95% CI 4.0-4.1). The unadjusted overall risk of stroke per 1000 person-years was higher in severe psoriasis patients (6.1; 95% CI 4.8-7.6) than in controls (4.4; 95% CI 3.8-5.0).
When adjusting for age and sex, both mild and severe psoriasis were associated with an increased risk of stroke (Table 4, HR 1.07, 95% CI 1.0-1.1, and 1.44, 95% CI 1.1-1.9, respectively). When adjusting for major risk factors for stroke (age, sex, diabetes, history of stroke or TIA, hyperlipidemia, hypertension, smoking) both mild (HR 1.06, 95% CI 1.0-1.1) and severe (1.43, 95% CI 1.1, 1.9) psoriasis were independent risk factors for stroke. Based on these analyses, we estimate that each year there is approximately one excess stroke per 4115 or 530 mild or severe psoriasis patients respectively (i.e., attributable risk). Obesity and atrial fibrillation were not included in the model as their inclusion did not alter the association between psoriasis and stroke (see Table 5). Additionally, based on an external adjustment analysis, an unknown or unmeasured covariate with a prevalence of 20% in the control group would need to have a prevalence odds ratio of association with psoriasis of 2.67 and be associated with stroke with an odds ratio of 4.3 in order to render the association between severe psoriasis and stroke null. The results were robust to numerous sensitivity analyses as demonstrated in Table 5.
The results of this study demonstrate that patients with severe psoriasis have a 44% increased risk of stroke, a potentially devastating co-morbidity. The risk of stroke in patients with psoriasis was not explained by both common and rare major risk factors for stroke as identified in routine medical practice, suggesting that psoriasis may be an independent risk factor for stroke. Patients we classified as having mild psoriasis had a statistically significant increased risk of stroke, however, this association was very modest and of limited clinical significance for the individual patient. For example, based on our data, a patient with mild psoriasis has an excess risk of stroke attributable to psoriasis of 1 in 4115 per year, whereas a patient with severe psoriasis has an excess risk of stroke attributable to psoriasis of 1 in 530 per year.
The increased risk of stroke persisted in a variety of sensitivity analyses designed to insure that we were capturing incident, not prevalent strokes, and that insured that all patients were seen regularly, minimizing the risk that information bias could explain the findings. Furthermore, the results persisted when excluding patients with psoriatic arthritis, restricting the severe group to patients treated with therapy specific to severe psoriasis (e.g. oral retinoids), and when examining the risk based on different treatments that theoretically could increase (e.g. cyclosporine, oral retinoids) or decrease (e.g. methotrexate) the risk of stroke(Lebwohl and Ali 2001; Prodanowich, Ma et al. 2005). Our findings are in agreement with recent studies that have found that psoriasis is an independent risk factor for cardiovascular disease in general (e.g. MI and coronary artery disease) and cerebrovascular disease in particular(Gelfand, Neimann et al. 2006; Ludwig, Herzog et al. 2007; Balci, Balci et al. 2008). Our study also extends the findings of previous studies that observed an association between psoriasis and cerebrovascular disease or stroke but did not account for psoriasis severity or confounding factors(Kaye, Li et al. 2008).
As with all studies, there are important limitations to consider. In database studies, there remains the possibility for misclassification of stroke due to coding errors or misdiagnosis. The overall incidence of stroke in our study was similar in the two control groups and is similar to rates of stroke reported using national statistics from England and Wales (Carroll K.). Although several studies have used GPRD to evaluate the epidemiology of stroke, a small validation study demonstrated that for 25 patients in the GPRD referred for a diagnostic code of stroke, there was a confirmation of stroke by the specialist in 64% of cases and a diagnosis of TIA in 16% of cases, giving an overall agreement in the opinion of an acute cerebrovascular diagnosis of 80%(Gibbs, Newson et al. 2001). If misdiagnosis of stroke is present, such errors would be expected to be non-differential misclassification and therefore would bias our results toward the null. Furthermore, it is difficult to differentiate between ischemic and hemorrhagic strokes in large population-base studies of medical record data, and therefore cannot determine directly how psoriasis influences the risk of various subtypes of strokes. Since the majority of strokes are ischemic in nature, the results are likely driven by the association of psoriasis and ischemic stroke. The association between severe psoriasis and stroke was similar in a sensitivity analysis restricting the outcome of stroke to those whom received therapy consistent with an ischemic etiology. Another potential limitation of our study is that we did not examine patients with exclusively incident (new onset) psoriasis. Therefore, we could not determine how duration of psoriasis affects stroke risk, and it is possible that some bias towards null findings could be introduced through the depletion of susceptible patients (e.g., some patients with psoriasis will have died from stroke prior to being captured in our study population). Ideally, an inception cohort study could be performed, however, in diseases such as psoriasis, which may not come to medical attention for many years; it is difficult to validly identify truly incident (new onset) cases in a medical records database setting. Furthermore, we did not directly determine severity of psoriasis based on extent of skin disease, which may introduce misclassification of mild and severe psoriasis when using therapy as a marker of psoriasis phenotype. It is likely that our mild psoriasis group contains a small subset of patients with severe skin disease, as systemic therapies are used infrequently for psoriasis in the general population, and that our severe group contains a small subset of patients with mild skin disease despite the use of systemic therapy. This misclassification would result in an over estimation and under estimation of the risk of stroke in the mild and severe psoriasis patients, respectively. Moreover, the generalizability of our results to patients with severe psoriasis who are not treated with the agents we used to identify severe disease needs to be determined in future studies. Finally, although our study suggests that psoriasis, particularly if severe, is an independent risk factor for stroke, it is possible that incomplete measurement of confounders or unknown confounding factors could explain some of the observed association. We did not adjust for alcohol in this analysis as validly measuring alcohol intake in any setting is challenging, the relationship between alcohol consumption and stroke is a complex “J shaped” curve, and the relationship between stroke and alcohol intake is believed to be mediated by hypertension (which was adjusted for in our models)(Reynolds, Lewis et al. 2003; CDC 2007). Analyzing our data using an external adjustment approach suggests that such an unknown or unmeasured confounder would have to be common in the general population, and have a strong association with psoriasis and a stronger association with stroke than the known stroke risk factors in order to render our findings null(Greenland 1996).
This study adds to the growing literature suggesting that patients with psoriasis, particularly if disease is severe, are at increased risk of cardiovascular events that is not explained by traditional risk factors. Clinicians are advised to alert psoriasis patients, particularly if disease is severe, to have assessment and treatment of their cardiovascular risk factors as recommended by current guidelines(Friedewald, Cather et al. 2008; Kimball, Gladman et al. 2008). More research is necessary to directly determine the mechanisms by which psoriasis may lead to adverse cardiovascular outcomes, as well has how psoriasis severity and treatment influence cardiovascular risk.
We conducted a cohort study using the General Practice Research Database (GPRD), a medical records database in the United Kingdom (UK) that was established for epidemiologic research in 1987 (Gelfand, Dattani et al. 2005). The GPRD is representative of the UK population in terms of age, sex, and geographic distributions. Approximately 5% of the UK population is represented in the GPRD and cumulative data are available for over 9 million patients with over 40 million person-years of follow up time. In the UK, more than 99% of patients are registered with a GP through the National Health Service, which coordinates all of the patient’s medical care. Data on diagnoses and prescriptions are recorded by the GPs as part of the patient’s electronic medical record. Patients with complex medical conditions are seen be seen by specialists (at the request of the GP) who may initiate a new treatment; however, patients are then referred back to the GP for chronic monitoring as necessary. Certain treatments, such as PUVA and oral retinoids, are restricted to the dermatologist in the UK; however, GPs capture these treatments through their electronic medical record. The GPRD has been shown to capture information on diagnoses and treatments from specialists through the GP’s electronic medical record (Jick, Jick et al. 1991; Jick, Kaye et al. 2003). GPs received specific training and were subject to inducements and penalties in order to ensure high quality data. The data are also audited for completeness and a practice receives an up-to-standard (UTS) designation when at least 95% of relevant prescriptions and diagnoses are captured electronically. Over 250 peer-reviewed scientific papers have been published using GPRD data(Gelfand, Dattani et al. 2005). The GPRD has also been studied in numerous validation studies, including those of psoriasis and cardiovascular disease, to demonstrate that it captures these outcomes accurately (Gelfand, Berlin et al. 2003; Gelfand, Dattani et al. 2005; Gelfand, Weinstein et al. 2005; Neimann, Shin et al. 2006; Huerta, Rivero et al. 2007).
Data were utilized from 1987-2002 for patients who were age 18 or older at the date their person-time began.
We defined mild psoriasis patients as those with a diagnostic code of psoriasis, but no history of systemic therapy at any time point. Severe psoriasis patients were defined as those with a diagnostic code of psoriasis, and history of systemic therapy consistent with severe psoriasis. Systemic therapy included phototherapy, PUVA, methotrexate, azathioprine, cyclosporine, oral retinoids (etretinate, acitretin), hydroxyurea, and mycophenolate mofatile. Of note, during the time period that this study was conducted, biologic therapies were not approved for use for psoriasis in the UK. The unexposed (control) population was composed of patients with no history of a psoriasis diagnostic code.
All patients defined as having mild or severe psoriasis (as defined above) who were age 18 or older at their index date, and had at least 1 day of observation time were included. For psoriasis patients, the index date was the first date on or after registration with the practice in which a psoriasis diagnosis was recorded. For patients with severe psoriasis, their index date was the first date on or after the first diagnosis of psoriasis in which the patient received a code for treatment consistent with severe disease. If psoriasis occurred prior to registration, then the registration date was the index date. For patients without psoriasis, their index date was the date of a medical record entry which was within 60 days of the psoriasis index date. Up to 4 control subjects who were age 18 or older at their index date were randomly selected for each psoriasis patient, matched on practice, date of registration in the practice, and psoriasis index date (corresponding to a medical record date of the control patient). The matching on dates occurred as follows: Registration: +/− 90 days (if registration date ≥ 1980, otherwise +/− 5 years), Index date: +/− 60 days. The purpose of matching on practice, registration and index dates was to ensure that patients with and without psoriasis were evaluated by similar physicians during the same time period to account for potential variations in medical practice and to allow for similar degrees of opportunity for GPs to identify medical conditions in psoriasis and non-psoriasis patients.
For mild psoriasis patients, follow-up started at the latest date of when the patient first received a psoriasis code, registered in the practice, or the practice was deemed UTS. For severe psoriasis patients, follow-up started at the latest of the date of when the patient could first be defined as having severe psoriasis (e.g., received a treatment code consistent with severe disease), was registered with the practice, or the practice was deemed UTS. For unexposed subjects (controls), follow-up started at the latest date of when the patient was registered in the practice, and the practice was deemed UTS. For all groups, follow-up ended at the earliest date of developing the outcome of interest (e.g. stroke), death, transfer out of the practice, or end of UTS.
The outcome of interest was stroke occurring after the start date. Stroke was identified using diagnostic codes (READ or OXMIS) entered by the GP into the medical record. Use of diagnostic codes to identify stroke has been a well accepted methodology in previous GPRD studies(Gibbs, Newson et al. 2001; Cleary, Shorvon et al. 2004; Nightingale and Farmer 2004; Souverein, Berard et al. 2004; Mulnier, Seaman et al. 2006; Weiner, Barnhart et al. 2007) (Tannen, Weiner et al. 2006). We defined any stroke code that occurred after the start date as incident, whereas any stroke code which occurred on or before the start date was defined as a prevalent (e.g. history of) stroke. Co-variables of interest: Major epidemiological risk factors for stroke including age, sex, hypertension, diabetes, hyperlipidemia, atrial fibrillation, body mass index, and smoking (current, former, non) were identified by the presence of diagnostic codes for the conditions or direct calculation from data in the medical record (e.g. BMI).
The sample size was determined by including the maximum eligible number of patients with severe psoriasis based on age criteria. We randomly selected up to 4 control subjects per patient with psoriasis as additional matching yields minimal increases in statistical power. Data were summarized descriptively. Dichotomous variables were tested with Fisher’s exact test and continuous variables were tested with a T test. We fit age- and sex-adjusted Cox models to determine the overall hazard ratio of stroke in psoriasis patients(Cox 1972). When indicated by an association of psoriasis with stroke based on the Cox models, we fit models with covariates for stroke included (described above) and models with age and sex interaction terms to determine if the relative risk of stroke in psoriasis patients was different based on sex or age characteristics. The primary model includes major cardiovascular risk factors. We did not include BMI in the primary model as it is recorded in only about 65% of patients. Each dichotomous variable in the model was checked for proportionality while adjusting for the other co-variates in the model by examining diagnostic log-log plots. A method of external adjustment was used to determine the degree to which our findings could be explained by an unknown or unmeasured confounder(Greenland 1996). Multiple sensitivity analyses were performed to test the underlying assumptions of our primary analysis. All analyses were performed using STATA 10.0.
This study was approved by the University of Pennsylvania Institutional Review Board and by the Independent Scientific Advisory Committee of the Medicines and Healthcare Products Regulatory Agency of the United Kingdom Department of Health. The study was conducted in accordance with the Declaration of Helsinki.
We are indebted to Jean Liu for her assistance in creating the analytical dataset.
Funding Source/Role of Sponsors: Supported by an unrestricted grant to the Trustees of the University of Pennsylvania from Centocor (JMG) and grant K23AR051125 from the National Institute of Arthritis, Musculoskeletal, and Skin Diseases (JMG). The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
Conflict of Interest Statement
Financial Disclosures (Potentially Relevant): Dr. Gelfand receives grant support or is an investigator for AMGEN, Centocor, and Pfizer. He is a consultant for Pfizer, Genentech, Celgene, AMGEN, Centocor, and Luitpold. Ms. Dommasch, Kurd, and Wang, and Mr. Shin report no financial relationships.