Prevalence of malignancy in psoriatic arthritis

Authors


Abstract

Objective

To determine the prevalence and types of malignancy in a large cohort of patients with psoriatic arthritis (PsA), and to compare this rate with that in the general population.

Methods

A cohort analysis of patients who were followed up prospectively from 1978 to 2004 at the University of Toronto Psoriatic Arthritis Clinic was performed. Patients were followed up at 6–12-month intervals according to a standard protocol, which included recording of malignancy, and tracked on a computer database. The cohort was linked with a provincial database to find malignancies that may have been missed by the protocol or developed after patients were lost to followup. Data were presented and analyzed using descriptive statistics and the Cox regression model with robust estimate of variance. Rates of first malignancy in the cohort were compared with rates in the population to derive standardized incidence ratios (SIRs).

Results

Of the 665 patients included, 68 (10.2%) developed a malignancy at an average age of 62.4 years. The most frequently seen malignancies were breast (20.6%), lung (13.2%), and prostate (8.8%) cancer. The SIR for all cancers was 0.98 (95% confidence interval 0.77–1.24). Overall cancer type–specific SIRs were 0.69 (95% CI 0.26–1.83) for hematologic and 0.88 (95% CI 0.46–1.69) for lung cancer. In females, the SIR for breast cancer was 1.55 (95% CI 0.92–2.62), and in males, the SIR for prostate cancer was 0.65 (95% CI 0.29–1.44).

Conclusion

Overall, 10.2% of patients in the Toronto PsA cohort developed cancer. The most frequent cancers were breast, lung, and prostate cancer. The incidence of malignancy in the large PsA cohort did not differ from that in the general population.

Psoriatic arthritis (PsA) is a chronic inflammatory joint disease that affects 5–42% of those with psoriasis (1). Its rheumatic manifestations range from mild enthesopathy to a mutilating polyarthritis (2–5). Historically, PsA has been thought to be a benign disease compared with other forms of inflammatory arthritis, such as rheumatoid arthritis (RA). However, recent studies have revealed the destructive nature of PsA. Studies of radiographic evaluation, both at the Psoriatic Arthritis Clinic at the University of Toronto and in a clinic in Spain, have shown erosive changes in 67% of patients (2, 4). In the Toronto cohort, 20% of PsA patients had complete joint destruction or ankylosis. Clinically, 11% of this cohort had pronounced physical disability (2). The prevalence of severe disease in this study approached that of RA. Studies in Britain and Spain also support the notion that there is progressive joint damage in PsA over time, and that severe disease at presentation is predictive of disease progression (4–7).

RA and other inflammatory joint diseases have been linked to an increased prevalence of malignancy (8–11). Psoriasis is also known to predispose patients to malignancy, particularly nonmelanoma cancers of the skin (12–15). Studies have also found increased rates of lung cancer (12, 15) and hematologic malignancy such as lymphoma (13, 16, 17) in the psoriatic population. However, it is difficult to attribute the increased risk of malignancy in either inflammatory arthritis or psoriasis to disease activity alone. Traditionally, both conditions have been treated with highly cytotoxic medications. The same cytotoxic and immunosuppressive medications taken for RA, such as azathioprine, methotrexate (MTX), and tumor necrosis factor α inhibitors, are used in the treatment of PsA. Evidence suggests that these treatments are themselves associated with the development of malignancy (13, 18–24).

The associations between rheumatic disease, psoriasis, immunosuppressive medication, and malignancy remain enigmatic. There are few current data on the baseline prevalence of malignancy in patients with PsA and their risk of developing cancer. These data would be most useful, and would allow important prognostic information to be determined. Furthermore, they would allow for investigation into potential links between the immunosuppressive therapy used for PsA and malignancy risk.

The aim of this study was to determine the prevalence of malignancy in patients with PsA, using a cohort of 680 Toronto PsA patients followed up prospectively since 1978. There are extremely detailed data on this cohort of patients. Thus, the objectives of this study were to determine the prevalence of malignancy and the type of cancers that developed in that cohort of patients, to determine if there was an increased risk of malignancy in the Toronto PsA cohort when compared with the general Ontario population, to determine which types and sites of malignancy occurred at different rates in the Toronto PsA cohort compared with the general Ontario population, and to determine whether there was any evidence of prognostic factors that have an impact on the time to development of malignancy in PsA.

PATIENTS AND METHODS

Patient population and protocol.

A cohort of >680 patients with PsA was followed up prospectively from 1978 to 2004, using standard research protocols, at the University of Toronto Psoriatic Arthritis Clinic. The cohort consisted of patients referred to a specialty clinic for the management of PsA. The clinic serves as primary, secondary, and tertiary referral centers. The spectrum of disease activity and duration in these patients was very broad. Patients entered the cohort at their initial visit to the clinic, at which time other forms of arthritis were ruled out, and were then reevaluated at 6–12-month intervals (2). At each visit, a standardized protocol was completed and used to record historical, physical, and laboratory findings. Information on the presence of malignancy was collected. Radiologic evaluation was performed every 2 years. The data were then entered and stored in a computerized database.

Data abstraction.

The database and paper charts were searched for any reports of malignancy in any of the patients in the cohort. Pathology, laboratory, or radiography reports were obtained to confirm the diagnosis. Cancer type and site were coded according to the International Classification of Diseases, Ninth Revision. The date of diagnosis and the location of the malignancy were also recorded. In instances of death, coroner's and autopsy reports were searched, if available.

The cohort data were confirmed by a simultaneous search of the Ontario Cancer Registry for all reported malignancies in any of the patients who had been included in the cohort. The Ontario Cancer Registry is a computerized database of all Ontario residents, and tracks the incidence of and mortality from cancer. All new cases are registered, but nonmelanoma skin cancers are excluded.

Statistical analysis.

The standardized incidence ratios (SIRs) of observed to expected first malignancies are reported with 95% confidence intervals (95% CIs). The Ontario general population provided the reference population for calculating the SIRs. Cancer incidence data were obtained from Cancer Care Ontario, and were analyzed in 5-year age groups for each calendar year of occurrence from 1978 to 2004 and by sex. Cancer incidence rates for 2003 and 2004 were not available. Thus, the 2002 incidence rates were used for these 2 years. Persons were considered at risk only when they came under active followup in the clinic. Time-dependent Cox regression models were fitted to investigate the effects of disease-related variables on the age at diagnosis of first malignancy. Delayed entry was incorporated into the Cox regression models.

RESULTS

Through computer and manual review of the clinic's accumulated protocols and linkage with the provincial cancer database, we identified 90 separate malignancies in 83 patients (Figure 1). Fifteen patients were subsequently excluded from analysis because they had a malignancy prior to or upon enrollment into the cohort. Of these, 10 had a malignancy that occurred at or after arthritis diagnosis. Thus, 665 patients were included in the analysis, of whom 68 patients (10.2%) developed a malignancy after entering the cohort. Of these 68 patients, 2 were found to have 2 histologically distinct malignancies, and 66 patients had a single malignancy. Thus, a total of 70 cancers occurred in these 68 patients. However, we considered only first malignancies in the following analyses.

Figure 1.

Identification of cancers in the Toronto psoriatic arthritis (PsA) cohort.

In the cohort, the mean ± SD age at diagnosis of first malignancy was 62.4 ± 11.7 years. Of the 68 patients in whom malignancies were detected, 38 were female (55.9%) and 30 were male (44.1%). The highest proportions of first cancers found in the cohort were breast cancer (20.6% of overall cases, 36.8% of female cases), lung cancer (13.2% of overall cases), and prostate cancer (8.8% of overall cases, 20% of male cases). Figure 2 outlines the breakdown of the incidence of malignancy by site. There were a number of other types of first malignancies that occurred only once in the cohort, thus comprising only 1.5% of the first malignancies. These infrequent tumors included uterine, multiple myeloma, ovarian, vaginal, brain, pancreatic, cervical, esophageal, spinal cord, gastric, connective tissue, vulvar, and endocrine malignancies.

Figure 2.

Prevalence of cancers detected among patients in the Toronto psoriatic arthritis cohort.

SIRs were calculated to compare the rates of cancer in the Toronto PsA cohort with those of the general Ontario population (Table 1). There was no evidence of any difference in the cancer incidence rates when compared with the general Ontario population both overall and by sex. There were also no discernible differences in incidence rates for site-specific cancers between the 2 groups.

Table 1. Standardized incidence ratios (SIRs) for malignancies in the Toronto psoriatic arthritis cohort compared with the general Ontario population
Malignancy*SIR (95% CI)
  • *

    Cancer incidence data were obtained from Cancer Care Ontario and analyzed in 5-year age groups for each calendar year of occurrence from 1978 to 2002. Data on cancer incidence rates for 2003 and 2004 were not available; thus, 2002 data were used.

  • 95% CI = 95% confidence interval.

All cancers 
 Both sexes0.98 (0.77–1.24)
 Males0.81 (0.56–1.15)
 Females1.18 (0.86–1.62)
Hematologic0.69 (0.26–1.83)
Lung0.88 (0.46–1.69)
Female breast1.55 (0.92–2.62)
Prostate0.65 (0.29–1.44)

To assess the impact on the SIRs of excluding the 10 patients who had a malignancy that occurred before entry into the cohort but at or after PsA diagnosis, we performed a sensitivity analysis by including these patients, first assuming that for these 10 patients their date of entry into the clinic was the date of PsA diagnosis, and then assuming that for the 675 patients (including the 10 patients above) the period at risk began from the date of PsA diagnosis. The first assumption generally resulted in a biased inflation of the SIR estimates obtained in Table 1, while the second assumption generally led to a biased reduction in the SIR estimates obtained in Table 1.

Overall, the results of the sensitivity analysis produced SIR estimates and 95% CIs that did not change the conclusions obtained earlier. However, for “female breast cancer” under the first assumption, there was an inflation of the SIR estimate from 1.55 to 1.77 (95% CI 1.08–2.89). Consistently for “all cancers (female)” under this assumption, the SIR estimate was inflated somewhat to 1.33 (95% CI 0.99–1.79).

Table 2 shows the disease characteristics at presentation (and before) and medication “ever use” in the 665 patients in the Toronto PsA cohort, by malignancy group. Information regarding medication “ever use” in patients with malignancies was obtained from the medication history prior to the diagnosis of the first malignancy, whereas for the patients without malignancies, the medication history was up to the last visit. Medication history was recorded upon entry into the clinic. For most types of medication, “ever use” was recorded for >50% of the patients in each group. In particular, MTX was taken by 58% of the patients without malignancies and by 44% of the patients who developed malignancies. However, biologic agents had been ever used by only 9.7% of patients in the nonmalignancy group and by 2.9% of patients in the malignancy group, prior to malignancy occurrence.

Table 2. Characteristics of the patients at presentation and medication use in patients with and those without malignancy*
 No malignancy groupMalignancy group
  • *

    PsA = psoriatic arthritis; ESR = erythrocyte sedimentation rate; NSAIDs = nonsteroidal antiinflammatory drugs; DMARDs = disease-modifying antirheumatic drugs.

Age at presentation, mean ± SD years42.4 ± 13.151.5 ± 12.6
Age at onset of PsA, mean ± SD years35.1 ± 12.842.6 ± 13.6
Age at onset of psoriasis, mean ± SD years28.0 ± 14.034.0 ± 15.6
ESR, mean ± SD mm/hour26.7 ± 21.029.8 ± 22.2
No. of actively inflamed joints, mean ± SD10.4 ± 9.79.6 ± 9.1
No. of clinically deformed joints, mean ± SD3.3 ± 7.82.0 ± 4.0
No. of effused (swollen) joints, mean ± SD3.1 ± 4.23.2 ± 4.3
Ever used NSAIDs, %94.588.2
Ever used DMARDs, %57.150.0
Ever used immunosuppressive drugs, %64.851.5
Ever used methotrexate, %58.144.1
Ever used biologic agents, %9.72.9
Ever used intraarticular steroids, %51.955.8

Table 2 is descriptive, and a more formal survival analysis with time-dependent covariates was undertaken to investigate the possible prognostic effects of these variables on the age at first malignancy.

Table 3 shows the Cox regression models for each of the potential prognostic variables, controlled for sex. None of the variables was found to be statistically significant at the 5% level, except the erythrocyte sedimentation rate (ESR) (P = 0.02), a measure of inflammation. Besides ESR, the only variable that came close to approaching statistical significance was the number of effused joints (P = 0.07), another measure of disease activity. A multivariate Cox regression analysis, adjusted for sex and disease-related variables (age at onset of psoriasis and PsA, ESR, effused joints, and clinically deformed joints), generated similar findings.

Table 3. Results of Cox regression analysis of variables assessed as potential predictors of age at diagnosis of first malignancy, controlled for sex*
VariableHR (95% CI)
  • *

    95% CI = 95% confidence interval (see Table 2 for other definitions).

  • Hazard ratio (HR) reported per 10-cm/hour increase, rather than as 1 mm/hour.

  • P = 0.02.

Age at onset of PsA1.01 (0.99–1.03)
Age at onset of psoriasis1.00 (0.99–1.02)
ESR, cm/hour1.13 (1.02–1.25)
No. of active joints at assessment1.00 (0.97–1.03)
No. of effused joints at assessment1.04 (1.00–1.09)
No. of clinically deformed joints at assessment0.99 (0.97–1.01)
NSAIDs, past or current use at assessment0.99 (0.48–2.04)
DMARDs, past or current use at assessment1.18 (0.73–1.91)
Immunosuppressive drugs, past or current use at assessment1.07 (0.66–1.72)
Methotrexate, past or current use at assessment1.25 (0.77–2.03)
Biologic agents, past or current use at assessment2.39 (0.58–9.86)
Intraarticular steroids, past or current use at assessment0.91 (0.56–1.49)

DISCUSSION

We found no evidence that patients in this large, closely studied PsA cohort were at a greater risk of developing malignancy than those in the general local Ontario population. Joint activity, as measured by the number of joints with active disease and the number of joints with effusions, was not found to be associated with an increased hazard of malignancy. In addition, both age at onset of psoriasis and age at onset of PsA were not found to be associated with increased cancer rates in the Toronto PsA cohort. This finding is important, both for clinicians and for patients with PsA, in terms of prognosis and treatment. Moreover, there was no evidence that the treatment used in this patient cohort increased the risk of malignancy.

Our findings are in contrast to those of similar studies of RA. Several retrospective and prospective studies of RA patients have found an increased risk of lymphoproliferative malignancy in those with RA. A recent study of associations between RA and malignancy reviewed 10 publications and found that patients with RA had an increased SIR of both colorectal tumors and lymphoproliferative disorders (25). This review included data from a Canadian RA cohort of 862 patients similar to the Toronto PsA cohort (26). In this Canadian RA study, the overall incidence of malignancy was reduced, but there was an increased incidence of hematologic malignancy, particularly leukemia (26). Of the excessive malignant lymphomas that occur in RA patients (11), there is an increased association with diffuse large B cell lymphoma in particular (27). Interestingly, these lymphomas tend to occur more frequently in patients with medium or high levels of disease activity (27, 28). Several other RA studies also showed an association between increased disease activity and the development of cancer (25). This differs from our findings in PsA, which showed no association between the number of actively inflamed or effused joints at presentation to the clinic and the risk of developing malignancy.

In contrast to the extensive literature regarding malignancy risk in RA, few published data exist about cancer risk in the seronegative inflammatory arthritides. A large Swedish study examined the incidence of cancer among patients with ankylosing spondylitis (AS) (29). That study found no increased overall cancer risk in AS patients, although the risk of rectal cancer was decreased and that of kidney cancer increased. The authors theorized that these site-specific differences might be due to the use of nonsteroidal antiinflammatory drugs and therapeutic pelvic irradiation, respectively (29). Contrary to studies in RA populations, this large cohort study found no increased incidence of hematologic malignancy (29). A second study of the same Swedish AS cohort specifically examined the issue of lymphoma incidence in hospitalized patients (30). This investigation also found no elevation in the average lymphoma risk in hospitalized AS patients (30). Furthermore, there were no increased risks of the major lymphoma types (30). On the other hand, a retrospective cohort study (published in abstract only) based on the General Practice Research Database in the UK between 1994 and 2001 identified increased incidence risk ratios for lymphoma among patients with AS that were similar to those calculated for RA (2.8 and 3.0, respectively) (31).

From the above data, as well as the data from our study, there appears to be a notable difference in malignancy risk in seropositive inflammatory RA when compared with seronegative inflammatory arthritis (AS and PsA). This may be related to differences in systemic inflammation, since generally only half of the patients with PsA and AS demonstrated elevated levels of acute-phase reactants. However, the current study shows that an elevated ESR is predictive of the development of malignancy among patients with PsA.

In patients with psoriasis, the reported risk of malignancy has been a subject of controversy. Stern et al (33) reported a nonsignificant increase in nonskin cancers (SIR 1.2) among 1,380 patients with psoriasis treated with psoralen ultraviolet A (PUVA), but a higher prevalence of colon cancer and primary neoplasms of the central nervous system. A Danish study found a 2.5-fold increased risk in nonmelanoma skin cancer, excesses of lung cancer, cancer of the larynx and pharynx in men, and of colon and kidney cancer in women (15). The same authors confirmed their findings in a later study (12).

Margolis et al (34) reported that patients with psoriasis are at an increased risk of developing a malignancy compared with patients with hypertension. The increased risk is greatest for those with severe disease (i.e., patients with psoriasis treated with systemic agents) and minimal (if an increased risk at all) for those with less severe disease compared with those in the hypertension group. The increased risk was mainly for lymphoproliferative cancers and nonmelanoma skin cancers. A population-based study in the UK (17) also demonstrated a low prevalence but increased risk of lymphomas among patients with psoriasis. The risk for lymphoma was found to be associated with the use of MTX or cyclosporine in patients with psoriasis treated with PUVA (35). In contrast, a population study in Sweden found no difference in the prevalence of malignancy among patients with psoriasis as compared with that of the general population (36). The incidence of nonskin cancer was not increased compared with the general population in a cohort of 1,252 patients with psoriasis treated with cyclosporine (37).

The immunosuppressive/cytotoxic medications used in PsA may themselves impact malignancy rates (13, 18–24, 34, 35). However, we found that the “ever use” of these medications was not associated with age at onset of first malignancy. Thus, in our study we cannot implicate medications in the development of malignancy. Note, however, that very few patients had ever taken biologic agents.

Our study is the first to examine malignancy rates in a large, prospectively followed cohort of patients with clearly established PsA. We have found no evidence to suggest that there is a higher risk of malignancy in our study population than in the general Ontario population. Importantly, neither joint activity nor medication use was associated with an increased cancer risk in our PsA patients. These findings differ from those in RA, but are consistent with the known data regarding other seronegative disorders (AS). These data may serve as background information for the calculation of risk associated with biologic agents.

AUTHOR CONTRIBUTIONS

Dr. Gladman 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 design. Hassa, Schentag, Gladman.

Acquisition of data. Hassa, Rohekar, Schentag, Gladman.

Analysis and interpretation of data. Rohekar, Tom, Schentag, Farewell, Gladman.

Manuscript preparation. Rohekar, Tom, Schentag, Farewell, Gladman.

Statistical analysis. Tom, Schentag, Farewell.

Ancillary