Clinical risk factors for the development of psoriatic arthritis among patients with psoriasis: A review of available evidence

Abstract

Psoriatic arthritis (PsA), a chronic inflammatory arthritis, affects about 10% of patients with psoriasis with higher prevalence seen in patients with more extensive skin disease. Early identification of PsA may result in improved outcomes. While it remains unclear which patients with psoriasis will develop PsA, several studies have identified potential risk factors for PsA among patients with psoriasis. This review examines the basic epidemiologic principles of identifying risk factors and reviews the evidence to date about risk factors for PsA among patients with psoriasis.

Introduction

Psoriasis is a chronic inflammatory skin disease affecting 1–3% of the general population.[1] Psoriasis has been associated with systemic inflammation and medical comorbidities such as cardiovascular disease and diabetes. Psoriatic arthritis (PsA), a chronic inflammatory arthritis, affects about-10% of patients with psoriasis overall,[2, 3] with a higher prevalence in patients with more extensive skin disease and a prevalence as high as 30% in dermatology clinics (where patients tend to have more extensive/severe psoriasis).[4–6] PsA is often a progressive disease resulting in joint damage, even early in the disease course. Earlier identification of PsA may result in improved outcomes. While it remains unclear which individual patients with psoriasis will develop PsA, several studies have identified potential risk factors for PsA among patients with psoriasis. Identification of such risk factors inform our understanding of the pathophysiology of PsA and may help clinicians identify groups of patients with psoriasis that should be followed more closely for symptoms of PsA. In this paper, epidemiologic principles for identifying risk factors will be discussed and studies of clinical risk factors for PsA will be reviewed. Genetic determinants of psoriatic arthritis are not included in this review but have been recently discussed in another review.[7]

Basic epidemiologic principles

What is a risk factor?

Risk factors are variables that are associated with development of a disease state or outcome.[8] Risk factors can occur at a single point in time (e.g. traumatic motor vehicle accident) or can occur over a prolonged period of time (e.g. smoking) with risk of the outcome tied to exposure time. Identifying risk factors can be difficult for a variety of reasons: 1) there may be a long latency between exposure to the risk factor and development of the disease, requiring a long follow up period; 2) if either the risk factor or the disease are relatively uncommon in the population or the risk factor has a relatively small effect on development of the disease, it may take very large numbers of patients to see an association between the risk factor and the outcome; 3) often, risk factors do not act alone – multiple risk factors may be related and it may be that the combination of risk factors together is associated with the outcome rather than any one risk factor individually; and 4) risk of developing the outcome can change over time, a phenomenon frequently observed in examining drugs as risk factors for adverse outcomes. Most adverse outcomes occur early after initiation of the medication, often the result of a phenomenon known as “depletion of the susceptible.” [8, 9]

Causation versus association

Interpreting studies of risk factors requires an understanding of the difference between causation (the risk factor directly causes the disease) and association (a relationship between the proposed risk factor and the disease exists). Determining whether causation in fact exists requires consideration of several factors outlined by the Bradford Hill criteria: strength of the association, consistency (i.e., identified by more than one study in different settings), specificity (i.e., there is no other explanation for the findings), temporality (i.e., the risk factor occurs before the disease), biological gradient (i.e., otherwise known as dose-effect), plausibility, coherence (i.e., between different types of studies), interventional or experimental evidence when possible (e.g., randomized controlled trials), and analogy (i.e., similar factors have been considered).[10]

Study Design

Risk factors occur prior to development of the outcome.[8] The ideal observational study design is a cohort study in which patients without the outcome of interest are followed longitudinally until they develop the outcome of interest. In cohort studies, risk factors are measured at baseline and may be updated over time (known as a time-varying exposure or covariate). One of the disadvantages of cohort studies is the requirement for large numbers of patients followed over long periods of time, limiting the ability to conduct prospective studies (particularly given the cost of such studies). Alternatively, case-control studies are often used to identify risk factors. Cases and controls are identified at a single point in time (after the outcome has developed) and researchers then look back in time to ascertain potential risk factors or exposures. Selection of cases and controls is of particular importance in the design of a case-control study (e.g. selection of patients with severe disease as the cases leads to selection bias). Cases and controls should be from the same source population. For example, selecting cases with PsA from one university clinic and comparing to psoriasis controls from a different university would represent two different source populations. One important consideration in the interpretation of case-control studies is that they rely on patient recall when assessing risk factors, potentially inducing recall bias (particularly if patients with the disease are more likely to recall a particular risk factor or vice versa). Finally, cross-sectional studies, in which risk factors and disease are measured at the same point in time, frequently claim to identify risk factors. For example, when a potential risk factor is measured at the time of enrollment in a patient with psoriatic arthritis of several years duration, the risk factor may have developed after disease onset. These studies lack a temporal association (in which the risk factor precedes the disease) and thus are mainly useful for generating hypotheses about which risk factors should be tested in future studies.[8]

Challenges in identifying risk factors for PsA

Many challenges exist in studying the epidemiology of PsA.[11] Studies of risk factors often suffer from insufficient power to detect a difference given the relatively low prevalence and incidence of the disease in the general population. Given the chronicity and often insidious onset of PsA, determining when the disease truly began is often not possible, particularly from retrospectively collected data. Additionally, patients with psoriasis and undiagnosed PsA may be misclassified as having psoriasis without PsA.[6] All of these factors may lead to imprecision in estimating the effect of a risk factor on the development of PsA.

Studies to date: Risk Factors for the Development of Psoriatic Arthritis

Studies examining the risk for PsA among patients with psoriasis are shown in Table 1. Three additional studies have been recently published assessing risk factors for the development of PsA and psoriasis (jointly) among the population studied.[12–14] As the goal of this paper is to discuss risk factors among patients with psoriasis, these studies are mentioned in the text but not included in the table.

Table 1.

Studies Examining Risk Factors for PsA Among Patients with Psoriasis

Study	Study Design	Source Population	Definition of PsA	Patients	Assessment of Risk Factors	Risk Factors Identified
Thumboo 2002[15]	Nested Case-Control	Rochester Epidemiology Project (Population-based medical record database)	Psoriasis with Inflammatory Arthritis	Cases: 60 Controls: 120	Database Self-administered questionnaire	Corticosteroid use in two years prior to psoriasis onset Pregnancy in 2 years prior to psoriasis onset (neg association)
Pattison 2008[16]	Case-Control	Psoriasis patients from dermatology clinic PsA patients from rheumatology clinic	Rheumatologist-diagnosis of PsA	Cases: 98 Controls: 163	Self-administered questionnaire	Trauma leading to medical care Vaccination for rubella Moving to a new home Treatment for fertility (neg assoc)
Wilson 2009[17]	Nested Case-Control	Rochester Epidemiology Project (Population-based medical record database)	CASPAR	Cases: 97 Controls: 1593	Database	Scalp psoriasis Intergluteal and/or perianal psoriasis >3 affect sites Nail dystrophy
Soltani-Arabshahi 2010[18]	Case-Control	Utah Psoriasis Initiative (dermatology psoriasis clinic)	Patient report of PsA	Cases: 250 Controls: 693	Database Self-administered questionnaire	BMI at age 18 Female sex Younger age at psoriasis onset Worst-ever BSA Koebner phenomenon Nail involvement
Tey 2010a[19]	Case-Control*	Psoriasis and photo-medicine clinic in Singapore	Diagnosis by rheumatologist (psoriasis + seronegative inflammatory arthritis)	Cases: 134 Controls: 266	Questionnaire	Family history of PsA Maximum BSA 76–100% Indian ethnicity
Eder 2011[20]	Case-Control	Psoriasis patients from Toronto psoriasis cohort PsA patients from Toronto PsA cohort	CASPAR	Cases: 159 Controls: 159	Database Self-administered questionnaire	Any injury Infectious diarrhea Infections requiring hospitalization Heavy lifting Smoking (ever vs never)
Eder 2012[21]	Case-Control	Psoriasis patients from dermatology clinic PsA patients from rheumatology clinic	Psoriasis with Inflammatory Arthritis (98% met CASPAR)	Cases: 728 Controls: 404	Database	Smoking (inverse) Interaction between smoking and HLA-Cw06
Love 2012[22]	Cohort study	THIN/UK primary care patients with psoriasis (defined by medical code)	Medical code	Patients with psoriasis: 75,395 Incident PsA: 976	Database	BMI with dose response
Li 2012[23]	Cohort study	Nurses’ health study II	Patient self-report verified by PASE tool	Psoriasis: 1,213,693 person years Incident PsA cases: 146	Database of Questionnaire Responses	BMI with dose response
Li 2012[24]	Cohort study	Nurses’ health study II	Patient self-report verified by PASE tool	Psoriasis: 1,303,970 person years Incident PsA cases: 157	Database of Questionnaire Responses	Smoking status with dose response
Wu 2015[25]	Cohort study	Nurses’ health study II	Patient self-report verified by PASE tool	Psoriasis: 1,137, 763 person years Incident PsA cases: 141	Database of Questionnaire Responses	Heavy drinking (compared to moderate)

^a. Some variables from Tey et al were collected cross-sectionally (e.g. alcohol and smoking) and are not included in this table.

Abbreviations: CASPAR = ClASsification of Psoriatic ARthritis; BMI = body mass index; BSA = Body Surface Area; PASE= psoriatic arthritis screening and evaluation.

Skin predictors of psoriatic arthritis

Severe psoriasis is more common among psoriasis patients with PsA than patients without PsA.[26, 27] It is plausible that a higher burden of skin inflammation may lead to increased systemic inflammation and trigger development of PsA. Three studies have reported psoriasis severity as a risk factor for PsA, although measured in different ways: worst ever body surface area (HR 1.01 for each one unit increase in BSA; 95%CI 1.00–1.01), three or more sites of psoriasis compared to one site (HR 2.24; 1.23–4.08 in a univariate model but was not significant in the multivariate model) and 76–100% body surface of psoriasis at its worst compared with 0–25% maximum BSA (the OR for 26–50% and 51–75% were 1.53 and 1.64 respectively, the relatively wide confidence intervals included the null) were associated with development of PsA.[17–19] All three studies were performed using retrospective collection of worst-ever BSA using a case-control design rather than a baseline or time-updated variable for BSA. While these studies suggest that psoriasis severity is associated with development of PsA, additional well powered studies with information on the time between psoriasis severity and development of arthritis are needed to establish this relationship.

Beyond psoriasis severity, location of psoriasis may be associated with development of PsA. Wilson et al found that psoriasis involving the scalp and intergluteal and/or perianal region were associated with development of PsA.[17] However, this has not been replicated in additional studies. This finding could also be a reflection of more severe psoriasis or more thorough examination of the skin (and likewise a more thorough examination of the joints). Finally, earlier age of onset of psoriasis has also been associated with the development of PsA in one study, suggesting that disease duration (and possibly inflammatory burden over time), may be important.[18] It may also be that patients with earlier onset of psoriasis have different genetic determinants of disease which may also contribute to development of PsA.[28] Furthermore, previous studies have suggested a relationship between age at onset of psoriasis with severity of skin disease which may be a confounder.[29, 30] However, in this study, age at onset was independently associated with PsA after adjusting for worst ever BSA.

Nail disease

Psoriatic nail disease is associated with enthesitis, or inflammation at the site where the extensor tendon attaches to the nail unit.[31] Inflammation of the enthesis likely precedes joint involvement and it may be similarly true that nail disease precedes development of psoriatic arthritis (particularly DIP arthritis). Nail disease may be a marker of increased “immunoreactivity” which leads to the development of PsA or nail disease may not be a “risk factor” for disease but rather an early phase of PsA, and thus a predictor of the onset of clinically overt PsA. Cross-sectional studies have found an increased prevalence of nail disease among patients with PsA compared to patients with psoriasis alone.[32] Additionally, Scarpa et al. found ultrasound evidence of nail bed inflammation in nearly all patients with PsA, even if nail disease was not clinically evident.[33] Two studies have reported the presence of nail disease as a predictor for the future development of PsA.[17, 18] While the association was strong in these studies (HR 2.24; 95% CI: 1.26–3.98 and HR 1.76; 95% CI 1.24–2.47), Thumboo et al did not find a statistically significant association (OR 1.16; 95%CI: 0.46–2.92) although this may have been due to lack of power to detect this association.[15] The confidence interval includes the point estimates from the other two studies. In these studies, nail involvement was ascertained either from a medical record, patient report, or at the time of enrollment in the cohort (when some already had active PsA). Patients are often unaware of nail disease, particularly if mild, so patient recall may be problematic. Additional studies addressing the timing of nail disease onset and PsA onset will be critical in determining whether nail disease is a risk factor or a manifestation of early PsA.

Obesity

Obesity, among the most common health problems worldwide, has been identified as a risk factor for both psoriasis and osteoarthritis.[34, 35] Three studies now support an increased risk of PsA among obese patients with psoriasis and two, Love et al and Li et al, found a dose-effect of BMI on the development of PsA.[18, 22, 23] While the pathophysiologic link is not entirely clear, hypothesized explanations include increased systemic inflammation induced by adipose tissue, increased mechanical loading on the joints, and a possible link with obesity-related dyslipidemia as suggested in OA.[34] In fact, a recent study by Wu et al found that a diagnosis of hyperlipidemia, independent of obesity, was associated with development of PsA.[13] Studies demonstrating improvement in disease activity and medication response with weight loss further support the concept of obesity as a driver of inflammation in psoriasis and PsA. However, the association between obesity and PsA could also potentially be related to misclassification of OA as PsA.[36]

Smoking

The role of smoking in the development of PsA is unclear. Smoking is associated with oxidative stress that may stimulate chronic inflammation. Given the association between smoking and the development of psoriasis and rheumatoid arthritis, this is a reasonable assumption.[37, 38] Conversely, nicotinic receptor activation inhibits intracellular proinflammatory pathways hypothesized to be at play in the development of inflammatory arthritis (analogous to the “protective” role of smoking in the development of ulcerative colitis).[21]

Studies examining the relationship between smoking and PsA have reported opposing results. Eder et al suggested an inverse association between smoking and development of PsA.[20, 21] Pattison found a similar association although not statistically significant in a univariable model (OR 0.68; 95%CI: 0.39–1.17).[16] However, when Eder et al stratified by HLA-Cw6 (a psoriasis susceptibility gene), the inverse association only held among patients negative for the gene.[21] No association was found for those with HLA-Cw6. In contrast, Li et al reported a positive association, particularly among women with psoriasis that smoked >=15 cigarettes/day.[24] This association was stronger and dose-dependent when examining the risk for PsA in the entire population, as opposed to only among patients with psoriasis. Why the difference in results? Different study designs were used (clinic-based case-control and a population-based cohort studies respectively), smoking was measured at different time points (at first visit in Eder et al which could have been up to 6 months after developing PsA and as a time varying exposure in Li et al), and Li et al was limited to women (Nurses’ Health Study II). Additional studies are needed to better determine the role of smoking in the development of PsA.

Alcohol

Alcohol has a complicated relationship with many diseases, often demonstrating a “J-shaped dose-effect curve” where moderate alcohol intake is beneficial but excessive intake is harmful (in cardiovascular disease, for example).[25] Excessive alcohol intake may be a risk factor for the development of psoriasis but results from studies examining this association are mixed. Furthermore, in some studies, alcohol has an inverse association with the development of RA.[39] Similar to smoking, alcohol has been assessed in studies of PsA but has been recorded at various time points including at the development of PsA, limiting assessment of the temporal relationship.[19] Information on the amount of alcohol consumed (none, moderate, excessive) has likewise been limited. Wu et al recently found that women with the highest level of alcohol intake have a substantially elevated risk of PsA among all individuals in the study compared to non-drinkers at baseline (HR 4.45; 2.07–9.59) and when using a time-updated alcohol exposure status (HR 3.96; 2.03–7.70).[25] However, this association was not significant among patients with psoriasis, only among the entire study cohort, although there was a significant association among excessive drinkers compared to moderate drinkers (HR 2.79; 1.24–6.26).

Environmental Triggers: Trauma, stress and infection

Theories of the pathogenesis of PsA, and autoimmune diseases in general, hypothesize that inflammation may be preceded by an environmental trigger in the setting of a genetically susceptible host. Enthesitis is hypothesized to begin from biomechanical sheer stress and microtrauma triggering an innate immune response.[40] Similarly, infection is also theorized to spark inflammation.[41] While these environmental factors are hypothesized to trigger PsA among patients with psoriasis, and many case series exist, few longitudinal studies have addressed these risk factors. Environmental factors associated with the development of PsA among patients with psoriasis in case-control studies have included trauma leading to medical care,[16] any injury, heavy lifting,[20] Koebner phenomenon,[18] moving to a new home (possibly related to the stress or physical labor of moving)[16], infectious diarrhea, and infections requiring hospitalization.[20] However, longitudinal cohort studies are needed to better address these potential risk factors.

Conclusion

In summary, identifying risk factors for the development of PsA among patients with psoriasis is a complex epidemiologic problem and limited evidence exists to date. If we return to the Bradford-Hill criteria for causation, all of the factors identified above have biologic plausibility (although ‘moving homes’ may not be directly related but rather indirectly associated through trauma and stress potentially). However, while some of the risk factors identified have strong associations, most have not yet been replicated (with the exception of obesity and psoriasis severity) and in some cases there are conflicting results (e.g. smoking). A dose-effect has been demonstrated for obesity in two studies and smoking in one study. Most studies have been case-control designs, limiting the ability to assess temporality (particularly as many risk factors were obtained at the time of diagnosis of PsA rather than before). A few more recent studies have used a cohort study design. Recording potential risk factors at baseline, before the onset of psoriatic arthritis, is critical to identifying such risk factors. An ongoing prospective cohort study at University of Toronto may be able to address some of these issues.[42] Ultimately, identification of risk factors could help identify patients likely to develop PsA which will enable us to treat their disease early in its course, improving prognosis. Moreover, identification of modifiable risk factors holds the promise of preventing the development for psoriatic arthritis. For example, if severity of psoriasis is a causal risk factor for PsA then it can be hypothesized that successful treatment of psoriasis would lower the risk of developing PsA; a hypothesis that requires careful testing in prospective studies.