Psoriatic arthritis (PsA) can be considered as a “disease within a disease”,1 ie, the occurrence of PsA can be considered as the appearance of inflammatory arthritis (IA) on a background of pre-existing or future development of psoriasis. Usually arthritis post-dates, often by several years, the onset of psoriasis.2 It is therefore relevant to consider what factors, in an individual with psoriasis, increase their risk of developing arthritis.

The familial recurrence risk for PsA is substantially greater than for either psoriasis or rheumatoid arthritis (RA) alone1 and there are genes within the major histocompatibility complex, eg, B27 and B38 that seem to increase susceptibility to PsA.3 From an aetiological standpoint the predisposition to PsA has been noted to be associated with somewhat different factors than those related to the aetiology of RA. Specifically, two reports4 5 suggest that trauma might be associated with development of PsA. Such reports, however, do not address specifically whether physical trauma would explain the occurrence of arthritis given a background of psoriasis. This is particularly relevant as psoriasis classically develops at sites of physical trauma, the so-called Koebner reaction.6^–8

Infection may also be of relevance and HIV infection appears to be responsible for most new cases of PsA in sub-Saharan Africa.9 Similarly, hepatitis C infection has also been associated with PsA.10 Therefore, certain environmental risk factors may increase the risk of developing PsA and such risk factors may also explain susceptibility to PsA in patients with pre-existing psoriasis. We therefore report the results from a retrospective case–control study of risk factors for the development of arthritis in patients with recent onset PsA.

SUBJECTS AND METHODS

Cases

Subjects with PsA were notified by their local consultant or identified through a nationwide media campaign. We confirmed the diagnosis of PsA with their local consultant rheumatologist. All cases had documented peripheral IA and psoriasis in their medical records. We only studied patients with peripheral arthritis; we did not recruit patients with axial disease only. In order to reduce the time between arthritis onset and survey (to minimise recall error) we recruited subjects in whom the self-reported date of arthritis, defined as first observed joint swelling, was within 5 years prior to registration of cases to the study.

Controls

Psoriasis patients were recruited from the Psoriasis Clinic at the Dermatology Centre, Hope Hospital, Salford. This centre provides both secondary level of care to its local community and tertiary care to the North West of England region. We limited our recruitment to subjects attending the centre in a secondary referral capacity by examination of their postcodes. Patients who had been referred by another dermatologist were also excluded. Controls were also excluded if they had a history or clinical evidence of arthritis. This was ascertained by a negative history of arthritis on chart review and direct questioning of patients.

Method of enquiry

A questionnaire was designed for self-completion and posted to cases and controls. The questionnaire was developed from a detailed literature review that examined potential factors that have been suggested in other studies to be potential aetiological factors for the development of IA. The review included literature not only pertaining to PsA but also to RA and other seronegative arthropathies. As a result we enquired about exposures in the following groups:

• Demographic data

• Details of psoriasis history, ie, extent, severity and therapy used

• Smoking and alcohol consumption

• Comorbidities, including cardiovascular disease, recurrent mouth ulcers, inflammatory bowel disease, etc.

• Exposure to physical trauma, ie, road traffic accidents, fractures and other injuries requiring treatment by either a general practitioner or at an Accident and Emergency Department

• Exposure to infections requiring antibiotic treatment or hospital admissions

• Vaccination history, including BCG, hepatitis A, hepatitis B, influenza, pneumococcus, polio, rubella, tetanus and typhoid

• Psychologically traumatic (life) events, including family death, divorce or separation, moving house, changing job, redundancy and treatment for anxiety or depression

• Hormonal exposure data, eg, child birth, adverse pregnancy events, including abortion, still birth and miscarriage, and having used the oral contraceptive pill or hormone replacement therapy.

A pilot version of the questionnaire was then administered to 25 PsA patients not eligible for the main study, and modified for comprehensibility according to their feedback. The final version of the questionnaire asked about these specific exposures in the 10 years before the study, rather than anchoring responses to the reported arthritis onset data in the cases. The definition of relevant pre-disease exposures is discussed below. As part of the questionnaire, patients also completed a self-administered Psoriasis Area and Severity Index (SAPASI), which has been validated for use in epidemiological studies.11 In order to maximise the response rate to the questionnaire, following the first mailing subsequent attempts were made to contact the participants and for the purposes of analysis of potential non-response bias, the respondents were subsequently designated as being initial responders (responding to the first mailing) or delayed respondents (responding to subsequent mailings).

Analysis

For the cases the date of first reported joint swelling was taken as the date of onset of PsA. The self-reported exposures in the 10 years before onset were then stratified into those self-reported to have occurred prior to, and those subsequent to, that onset date. Only the former were considered in all analyses. A dummy date of onset was applied to the controls. For an individual control this was applied randomly to yield the same proportionate frequency distribution of onset dates as observed in the cases. The exposure frequencies between cases and controls were compared using logistic regression, adjusted for age and gender and the results expressed as odds ratios with a 95% confidence interval. In view of the fact that we had established our questionnaire to test a series of a priori hypotheses based on a thorough literature review, an adjustment for multiple testing was not applied and significance was set at the 5% level. The study received ethical approval from the North-West of England Multi-Centre Research Ethics Committee.

RESULTS

Recruitment

We screened 425 patients who were referred with a consultant confirmed diagnosis of IA and documented psoriasis. From this group, a total of 163 PsA cases were identified in whom the onset of first joint swelling was within 5 years of the study date. As shown in table 1, a small proportion of these were listed at the wrong address but of the 151 cases with the correct address 125 (82.7%) responded to the postal questionnaire. The database for the psoriasis controls identified 505 subjects; however, this has been compiled over several years and there were a large number of address errors (table 1). Of the 326 individuals who received a questionnaire 163 (50.0%) responded by returning a completed questionnaire. This response rate was significantly lower than that in the cases.

The duration of arthritis onset, as notified by a rheumatologist, was validated with the subject’s recall from the questionnaire data. Twenty-seven cases were unable to recall the approximate date of onset of joint swelling. In order to minimise the possibility of recall error, those individuals were excluded and thus the results presented apply to 98 (64.9%) cases with fully verified temporal data (table 1).

Disease characteristics

The characteristics of the final groups of cases and controls are shown in table 2. The cases included a higher proportion of females and were older. All subsequent analyses were therefore adjusted for age and gender. The controls had more severe psoriasis; thus a higher proportion of the controls had been hospitalised; received systemic therapy; and, by definition, had ever seen a dermatologist for their psoriasis. Their mean “worst” SAPASI was also higher than in the cases.

Environmental exposures

The results of the environmental exposure data are shown in tables 3^–6. There was a trend towards cases being less likely to have smoked or consumed excess amounts of alcohol (⩾35 units per week) although these differences were not significant. Cases were more likely to have reported a traumatic event particularly one that led to their seeking medical care (15% vs 8%; OR (95% CI) = 2.5 (1.1 to 6.0)) (table 3). The numbers with trauma were small but there was no particular type of trauma experience that was over-represented in cases. In a subgroup analysis of the small group of patients reporting fractures (six patients with PsA vs 11 controls) there was no difference in the distribution of fractures (peripheral versus axial); however, there was a higher proportion of cases who had been admitted following fracture (3 (50%) vs 1 (9%) p = 0.04)

There were no important differences in self-reported infections between the two groups apart from the superficially surprising finding of a higher frequency of infectious diarrhoea prior to the onset of arthritis in the controls than in the cases (table 4). If there had been such an illness prior to the onset of arthritis then patients may have been selectively labelled with reactive arthritis and not been given the diagnostic label of PsA. Conversely, there was also an increased reporting of oral ulceration in the cases, not seen in the controls. This is despite the fact that 39 (40%) of cases and 85 (52%) of controls had received either methotrexate or retinoid derivatives in their treatment history. A previous history of immunisation, particularly rubella and tetanus, was reported at a higher frequency in cases. In the case of the latter, the higher numbers exposed gave more robust estimates (table 4).

There were no major differences in psychological events although there was evidence for greater reporting of individuals who had either moved house or been made redundant among the cases than the controls (table 5). For the latter exposure the confidence interval spanned unity. In females there were no differences in a number of hormonal factors (table 6); however, females with PsA had a lower odds ratio of ever having been treated for fertility problems (3.8% vs 18.6%: OR (95% CI) = 0.17 (0.04 to 0.79)).

DISCUSSION

This is only the second retrospective case–control study that has compared a range of environmental exposures between individuals with PsA compared with a referent group with psoriasis alone.12 Our cohort is, however, significantly larger than that studied by Thumboo et al12 and suggests that some forms of physical or psychological trauma as well as forms of infection/immunisation may contribute to susceptibility to PsA.

Studies such as this raise a number of methodological issues. First, this was a retrospective case–control study thus subject to recall errors and recall bias. Given the low incidence of PsA13 a prospective study would not have been practical. We attempted to minimise recall errors by recruiting cases of “recent onset” PsA. As the majority of PsA patients in UK rheumatology clinics tend to have well established PsA, as a compromise between maximising the number of cases and minimising recall time errors, we restricted recruitment to individuals who were able to date their onset of arthritis to ⩽5 years. However, recall error may still have occurred, which might have led to both: (1) random misclassification making it hard to find real differences, and (2) recall bias in favour of increased reporting of exposures by the cases. To minimise the latter we framed the questionnaire over the previous 10-year period we then censored the information gathered by the date of onset of arthritis only at the time of analysis.

There was also a differential response rate in the two groups. While 83% of eligible PsA subjects responded to the questionnaire, only 50% of psoriasis controls responded. A low response raises the possibility of differential exposure rates in the non-responders. We therefore compared reporting of exposures between “initial” and “delayed” respondents both in cases and controls, but found no systematic differences (data on file).

Ideally we would have validated the self-reported exposures against contemporaneous medical and other records. However, this would have required examination of multiple hospital sources and primary care records in all subjects and was not practical. As part of the pilot of the questionnaire, we did compare the responses of a number of local participants with their medical records. While there was insufficient information to perform a statistical analysis of these; there was a good level of agreement between the patient responses and the available relevant medical records.

Our definition of PsA was a practical one based on clinical diagnosis of psoriasis and arthritis. We did not apply the recently published CASPAR criteria14 as these were not available at the time of study design. However, all patients fall within the Moll and Wright definition of PsA, ie, “an IA associated with psoriasis that is usually negative for rheumatoid factor”.15 The controls selected for this study were those individuals attending a single specialist psoriasis clinic on a secondary care basis. Despite this, controls still reported more severe skin disease. As such we are not able to analyse the role of psoriasis severity/distribution on susceptibility to PsA. An alternative approach would have been to select controls from the community. As psoriasis can wax and wane and in primary care psoriasis is more likely to be misdiagnosed or underdiagnosed, this approach is problematic. We were also anxious to ensure that cases and controls were matched for the same “level” of medical care as potential biases associated with using secondary care cases and primary care controls were unacceptable. As a result of our control selection, we cannot comment on whether patients with more severe psoriasis are more prone to developing PsA as suggested by Gelfand et al.16 Conversely, our patients are comparable with other cross-sectional studies of PsA, which shows mild–moderate psoriasis in the majority of PsA patients.17 It is also possible that some of the controls, had unrecognised arthritis, thereby attenuating the strength of any underlying associations seen. The Dermatology Centre has a high awareness of psoriatic arthritis and refers such patients to rheumatology. It is, however, still possible that subjects with less obvious forms of arthritis, eg, enthesopathy, would still have been misclassified. Overall, however, our approach actually biases against finding real differences so the positive associations seen remain valid.

Our study and that of Thumboo et al12 demonstrate how difficult it is to study aetiological factors in an uncommon arthropathy such as PsA. We originally aimed to recruit 223 cases and 446 controls, which would have given 80% power to detect a doubling in risk for exposure frequency and controls of at least 0.1. Our actual recruitment does not affect our positive results but may explain the failure to find other “significant” differences. Conversely, by analysing a large number of different exposures, some may be significant by chance alone. Our review of the literature informed the exposures studied and therefore no correction was applied for multiple significance testing. It will, however, be important for these findings to be verified in other primary and/or secondary care cohorts.

With regard to biological relevance, psoriasis is known to develop at sites of trauma6^–8 and investigations comparing PsA patients with RA controls have also suggested a role for trauma in the genesis of PsA.4 5 The association with physical trauma of sufficient severity to necessitate attendance at an Accident & Emergency department or with their general practitioner is therefore likely to be real and reflective of a “deep Koebner phenomenon”. In comparison with previous studies4 5 the current study covered all such physical traumas in the 5 years before onset of arthritis. This gave us sufficient numbers so that we were able to separate particular traumatic events into physical and psychological/life events. This is in contrast to previous work that included many types of events under a single “trauma” heading.4 Also given the enthesopathic nature of PsA onset in many cases, we were concerned not to limit the window for trauma to a short period prior to actual joint swelling developing. Our findings suggest that trauma itself may therefore have a longer-standing or latent effect that may take several years to fully manifest itself as synovitis and joint swelling. For questions on psychological trauma, only moving house was significantly associated with PsA. Psychological trauma is a recognised risk factor for psoriasis exacerbations and perhaps patients with significant psoriasis are less likely to move home. However, moving home may also be associated with a significant physical element, which may support physical over psychological trauma in the aetiology of PsA.

With regard to infection, the relationship between HIV and the development of PsA and worsening of psoriasis is well-established.9 18 In this study, exposures of significance included recurrent oral ulceration and several immunisations. Oral ulceration and conjunctivitis are part of the clinical spectrum of PsA2 but may also reflect recent exposure to infection. We speculate that immunisations may reflect a better marker of an event likely to stimulate the immune system that is less susceptible to recall error than recalled infection. There was no clear common factor between the immunisations noted but the negative association with influenza suggests that modulation or biasing of the immune system by only a certain range of triggers may favour the development of PsA. Alternatively, patients with psoriasis may avoid immunisations in order to avoid Koebner reactions. We are not aware of previous studies on immunisation and PsA; however, we have previously shown that tetanus immunisation may be a risk factor for RA as well as non-specific inflammatory polyarthritis.19 Tetanus immunisation may also of course relate to recent trauma.20

Finally, we found a negative association with treatment for previous fertility problems. The numbers in this group were small and it was the only one of the potential hormonal and pregnancy related factors that showed any association. Interestingly Thumboo et al12 found a negative association between pregnancy and onset of PsA in their case–control study. We could not confirm this finding; however, our study does lend support to the notion that significant changes to the hormonal milieu might decrease the risk of developing PsA.

In summary, while some of our findings may represent differential behaviours in psoriasis patients compared with those who eventually develop PsA, this study does suggest that there are a number of environmental risk factors that may be associated with the occurrence of arthritis in patients with psoriasis. Replication of our findings, ideally in new onset cases, is required especially for those exposures not previously reported. Further understanding of how such triggers contribute to the development of PsA may also help elucidate key pathomechanisms of this disease.