Ankylosing Spondylitis, Psoriatic Arthritis, and Risk of Malignant Lymphoma: A Cohort Study Based on Nationwide Prospectively Recorded Data From Sweden

Authors


  • The views presented herein are those of the authors and do not represent an official position of the Swedish Medical Products Agency.

Abstract

Objective

Data on lymphoma risk in ankylosing spondylitis (AS) and psoriatic arthritis (PsA) are scarce. This study was undertaken to assess the risk of lymphoma in AS and PsA overall and in relation to therapies, including tumor necrosis factor inhibitor (TNFi), for which lymphoma risks are a concern.

Methods

Through the Swedish National Patient Register we assembled nationwide prevalence cohorts of patients with AS (n = 8,707) and patients with PsA (n = 19,283) for whom data were obtained between 2001 and 2010. Each cohort member was matched to 5 population comparator subjects. Linkage with the nationwide Cancer Register identified all lymphomas recorded from 2001 to 2010. Through the Swedish Biologics Register (Anti-Rheumatic Therapy in Sweden [ARTIS]), we identified patients exposed to TNFi in the AS cohort (n = 1,908) and the PsA cohort (n = 2,605) before lymphoma diagnosis. Hazard ratios (HRs) for lymphoma were estimated by Cox regression. Crude incidences of lymphoma in TNFi-exposed and TNFi-naive patients were compared.

Results

For AS patients, the HR of having lymphoma versus the general population was 0.9 (95% confidence interval [95% CI] 0.5–1.6) (14 lymphomas). For PsA patients, the corresponding HR was 1.2 (95% CI 0.9–1.7) (45 lymphomas). For PsA patients treated with methotrexate and/or sulfasalazine, the HR of having lymphoma was 1.7 (95% CI 1.0–3.1). The numbers and incidence of lymphoma were not materially different in TNFi-exposed versus TNFi-naive AS and PsA patients, although the numbers of lymphomas were small.

Conclusion

In contrast to rheumatoid arthritis, the average risks of lymphoma in AS or PsA are not elevated, although increased risks in a subset of PsA patients cannot be excluded. Our findings indicate that TNFi does not affect the risk of lymphoma in AS or in PsA.

Excess risks for malignant lymphomas have been reported for several chronic inflammatory rheumatic diseases. In rheumatoid arthritis (RA), there is a well-established doubled average risk, a strong correlation with inflammatory activity and disease severity, and a particularly increased risk for the diffuse large B cell lymphoma (DLBCL) subtype ([1, 2]).

Compared to RA, data on lymphoma risk in patients with spondyloarthritides (SpA) are limited, but indicate potential differences between different SpA entities. Cutaneous psoriasis has repeatedly been linked to increased lymphoma risks ([3-8]), in particular T cell lymphoma, although other studies have failed to show any increased risk ([9-11]). For inflammatory bowel disease (IBD), the lymphoma risk overall does not seem to be increased, although recently, there have been observations of an increased occurrence of hepatosplenic T cell lymphoma in patients treated with a combination of tumor necrosis factor inhibitor (TNFi) and azathioprine ([12]). In ankylosing spondylitis (AS) and psoriatic arthritis (PsA), data on lymphoma risks are limited (Table 1).

Table 1. Reports presenting relative risks of lymphoma in patients with ankylosing spondylitis (AS) and patients with psoriatic arthritis (PsA)*
Author, year (ref.)Disease studiedType of studyNo. of AS or PsA patientsNo. of lymphoma casesNo. of lymphoma cases with ASNo. of controlsNo. of controls with ASRelative risk or incidence rate (95% CI)a
  1. Search terms used were psoriatic arthritis + lymphoma, malignancy, comorbidity; ankylosing spondylitis or Bechterew + lymphoma, malignancy, comorbidity; English, and 1980–2012. 95% CI = 95% confidence interval; NA = not available.
  2. aThe relative risk is the standardized incidence ratio for all hematopoietic malignancies for the studies by Feltelius et al and Rohekar et al, the incidence rate ratio for lymphoma for the study by Shibata et al, the odds ratio (OR) for lymphoma combined for the study by Becker et al, the OR for lymphoma for the study by Askling et al, the OR for non-Hodgkin's lymphoma for the studies by Mellemkjaer et al and Anderson et al, and the incidence rate for the study by Gross et al.
Feltelius et al, 2003 ([22])ASCohort study6,621121.3 (0.9–1.9)
Shibata et al, 2004 ([25])ASCohort study3,262NA2.8 (1.4–5.6)
Becker et al, 2005 ([11])ASPopulation-based case–control study710371013.0 (0.1–29)
Askling et al, 2006 ([23])ASPopulation-based case–control study50,6152392,928411.0 (0.6–1.7)
Mellemkjaer et al, 2008 ([24])ASPopulation-based case–control study25,9411958,551411.1 (0.6–1.8)
Anderson et al, 2009 ([8])ASPopulation-based case–control study33,72140122,5311281.1 (0.7–1.5)
Rohekar et al, 2008 ([26])PsACohort study665NA0.7 (0.3–1.8)
Gross et al, 2011 ([27])PsACohort study2,97730.4 (0.1–1.2)

Apart from its etiologic importance (i.e., which aspects of chronic inflammation drive the increased lymphoma risk?), a proper understanding of overall lymphoma risks, including heterogeneities across subsets of patients, is important for the evaluation of drug safety. This is of particular relevance following treatment with a TNFi, for which lymphoma risks have been a concern ([1]). To date, increased pretreatment risks, together with channeling of patients with severe RA to TNFi therapy, have made posttreatment evaluation difficult in RA ([13-15]). Further, there are indications that the lymphoma risk in RA remains increased to the same magnitude in incident RA and in patients diagnosed as having RA in the last 5–10 years ([16]), raising the question of whether modern RA therapy might have replaced one type of lymphoma risk with another. If the underlying lymphoma risks in AS and PsA are not elevated, then patients with these conditions, for which TNFi therapy is increasingly used, would provide important populations for risk assessments of TNFi therapy.

The aims of this study were therefore as follows: 1) to assess the occurrence and the underlying risk of lymphoma in patients with AS and PsA compared with that of the general population, 2) to assess the distribution of lymphoma subtypes in these conditions, and 3) to assess the overall association between therapy, including TNFi, and lymphoma risk in AS and PsA. To do this, we linked data from Swedish population-based health care registers, mandatory cancer reporting, and clinical registers including the nationwide Swedish Biologics Register (Anti-Rheumatic Therapy in Sweden [ARTIS]).

PATIENTS AND METHODS

Study design

A population-based cohort study of patients with AS, patients with PsA, and matched population comparator subjects was conducted. The main exposures of interest were AS or PsA and TNFi. The outcome of interest was malignant lymphoma.

Study setting

Swedish health care is public and tax funded. Patients diagnosed as having AS or PsA who are treated with disease-modifying antirheumatic drugs (DMARDs) are typically cared for by rheumatologists, most of whom work in hospital-based care. However, in particular prior to the era of treatment with biologic agents, a significant proportion of these patients were jointly cared for by internists and primary care physicians. Linkage across different nationwide and population-based health and census registers is enabled through the national registration number, a unique 10-digit number issued to all Swedish residents and used in all official registers and medical files. The study was approved by the Ethics Committee at Karolinska Institutet.

Study population

Cohorts of Swedish patients with AS and patients with PsA

The Swedish National Patient Register covers hospital discharges since 1964. Since 2001, nationwide data from specialist outpatient care (non–general practice visits) have also been added to the register, including information on all outpatient visits with primary and contributory discharge diagnoses assigned by the physician and coded according to the calendar year–specific version of the International Classification of Disease (ICD; versions 7–10, with ICD-10 after 1997). For 2006, the assessed average coverage for somatic care in the outpatient register was almost 80% (with lower coverage for private surgical care and higher coverage for public hospital-based care) ([17]).

We identified all individuals who had one or more outpatient visits at a rheumatology or internal medicine department for which a diagnosis code for AS (ICD-10 code M45) or PsA (ICD-10 codes M07.0-3 and L40.5) was listed and who were alive and lymphoma free at entry. Patients with these codes or the corresponding codes for juvenile disease (codes M08 and M09) before 18 years of age (the Swedish cutoff between pediatric and adult care) were excluded. The followup period started at the first specialist outpatient visit with a diagnosis of AS or PsA (i.e., at least 1 visit was required) and ended at the first of either a lymphoma diagnosis, death, emigration, or end of the study period (December 31, 2010).

To avoid the possibility that an underlying lymphoma was misclassified as AS or PsA or was the reason for the visits that led to inclusion in the cohorts under study, we excluded all person-time (i.e., days of followup) and all events during the first 90 days of followup.

In total, we identified 8,707 patients with AS and 19,283 patients with PsA for whom data were obtained between 2001 and 2010 (Table 2). In sensitivity analyses, we excluded all patients with a diagnosis code of RA ever recorded in the outpatient register, leaving 7,629 AS patients and 15,929 PsA patients, and applied a stricter exposure definition, requiring ≥2 outpatient specialist visits for patients with AS (n = 6,415) or PsA (n = 14,088). In the latter analysis, the start of followup began at the date of the second specialist visit with a diagnosis of AS or PsA.

Table 2. Characteristics of the Swedish cohorts of patients with AS and patients with PsA with data recorded between 2001 and 2010*
 AS patients (n = 8,707)PsA patients (n = 19,283)
  1. AS = ankylosing spondylitis; PsA = psoriatic arthritis; IQR = interquartile range.
  2. aStart of followup was defined as the first outpatient visit with records listing a diagnosis of AS or PsA in 2001 or later.
  3. bTreatment with disease-modifying antirheumatic drugs (DMARDs) was defined as ≥2 prescriptions of methotrexate or/and sulfasalazine within 6 months registered in the Prescribed Drug Register between 2005 and 2010.
  4. cOf the AS patients (n = 8,536) and PsA patients (n = 19,031) who were at risk from July 1, 2005 and later. Treatment was defined as ≥2 prescriptions of the drug within 6 months as registered in the Prescribed Drug Register between 2005 and 2010.
  5. dDefined as exposure to tumor necrosis factor inhibitor (TNFi) agents ever during the entire study period as registered in the Swedish Biologics Register (Anti-Rheumatic Therapy in Sweden [ARTIS]).
Sex, no. (%) female2,990 (34)10,739 (56)
Age at start of followup, yearsa
Mean (range)47 (18–93)52 (18–96)
Median (IQR)58 (47–66)61 (52–69)
Years of followup, median (IQR)a8 (6–9)8 (5–9)
DMARDs, no. (%)b, c2,606 (30)9,380 (49)
Methotrexate, no. (%)c1,373 (16)8,072 (42)
Sulfasalazine, no. (%)c1,407 (16)2,372 (12)
Oral glucocorticoids, no. (%)c1,537 (18)4,126 (22)
TNFi, no. (%)d1,908 (22)2,605 (14)
Median (IQR) age, years51 (42–59)55 (46–61)
No. (%) female586 (31)1,362 (52)

General population comparator cohort

For each patient with AS or PsA, 5 population comparator subjects were randomly selected (matched for age, sex, and county of residence) from the Swedish Population Register (which contains information on births, civil status, emigrations, and deaths for all Swedish residents). The comparator subjects had to be alive and lymphoma free at the first entry into the cohort (the date of entry of their corresponding index patient with AS/PsA).

For a small proportion of patients (6%), fewer than 5 comparator subjects could be identified. In total, we identified 41,092 comparator subjects for the AS patients and 92,684 for the PsA patients.

By linkage of all individuals in the study population to the nationwide and virtually complete Cause of Death register and to the Swedish Population Register, we collected information on date of emigration and date of death during the study period.

Exposure of AS and PsA patients to DMARDs and TNFi therapy

From the Swedish Prescribed Drug Register (virtually complete capture of dispensed drugs; in operation since July 2005) ([18]) we collected information on prescriptions filled for methotrexate, sulfasalazine, and oral glucocorticoids in 2005 and later. DMARDs were defined as methotrexate and/or sulfasalazine. To be defined as treated, patients had to have ≥2 prescriptions dispensed within 6 months registered in the Prescribed Drug Register in 2005 and later.

Since 1999, information on adult patients with rheumatic diseases starting biologic therapy has been recorded in the Swedish Biologics Register (ARTIS) at treatment initiation and at regular followup visits. The register contains information on the type of rheumatic condition, prescribed biologic therapy, including treatment duration, and concomitant antirheumatic therapy. From our underlying national AS and PsA cohorts, we identified 1,908 patients with AS (22% of all AS patients), and 2,605 patients with PsA (14% of all PsA patients) who had at least one period of TNFi therapy registered in ARTIS between 2001 and 2010 (Table 2).

Occurrence of lymphoma

Through linkage to the Swedish Cancer Register, all registered malignant lymphomas from 1958 to 2010 were identified. The Swedish Cancer Register, to which reporting is mandatory (coverage >95%), includes information on incident malignancies since 1958 ([19]). The register includes information on cancer type using ICD versions 7–10, cancer date, and morphology. For this study, malignant lymphomas were defined as non-Hodgkin's lymphoma (NHL), including chronic lymphocytic leukemia (CLL), and Hodgkin's lymphoma, i.e., ICD-7 codes 200, 201, 202, and 204.1. By using available diagnosis codes of ICD-10 as registered in the Swedish Cancer Register, we further categorized the NHLs into subtypes according to the World Health Organization (WHO) classification ([20]).

To specifically validate the diagnoses of AS or PsA and lymphomas in the patients exposed to TNFi, we confirmed the AS and PsA diagnoses from medical records and the lymphoma diagnoses, including subtype, through pathology reports. The lymphoma subtypes were further validated by one of the authors (CS) through review of the tissue specimens and were classified according to the WHO classification ([20]).

Statistical analysis

Incidence and relative risks of lymphoma in patients with AS and patients with PsA compared with the comparator subjects

The incidence of lymphoma in AS patients and PsA patients was calculated from the start of followup (the first date the patient fulfilled the entry criteria for each cohort and the corresponding date for their comparator subjects) until the first of either a lymphoma diagnosis, death, emigration, or end of the study period (December 31, 2010). Relative risks were assessed as hazard ratios (HRs) derived from Cox regression analysis (using SAS version 9), and stratified by age (5-year age categories up to 85 years and then 85 years and older), sex, and county of residence, with followup time as the time scale. The proportional hazards assumption was tested by introducing an interaction term between the exposure and the log of followup time for AS and PsA, respectively, and was not violated (P value from the primary models 0.3 for AS and 0.6 for PsA).

Lymphoma risks were presented overall and stratified by sex, age at the start of followup (younger than 50 years, 50–74 years, or ≥75 years), calendar period at the start of followup (before or after 2005), and duration of PsA or AS (≤1 year or >1 year). To limit the risk of exposure misclassification, the analyses were also performed excluding all patients for whom a diagnosis code for RA was ever recorded.

In the analyses of lymphoma risks in relation to antirheumatic drug treatment (other than TNFi), the start of followup was defined as the date of the second recorded prescription (from the Prescribed Drug Register) of each drug for the AS patients and PsA patients (corresponding date for their comparator subjects). The patients were followed up until the first of either a lymphoma diagnosis, death, emigration, start of first TNFi (to minimize the risk that the TNFi affected the results of these analyses), or the end of followup.

Incidence of lymphoma in TNFi-exposed versus TNFi-naive patients

The crude incidences of lymphoma in AS patients and PsA patients exposed to TNFi therapy were calculated, counting the time from the start of the first TNFi until the first of either a lymphoma diagnosis, death, emigration, or end of followup (December 31, 2010). Patients were considered unexposed until the first TNFi was started and were considered exposed thereafter. Sensitivity analyses were performed, excluding all patients with a diagnosis code for RA ever recorded and applying a stricter definition of AS and PsA (i.e., requiring ≥2 outpatient visits).

RESULTS

Incidence and risk of lymphoma in AS patients and PsA patients compared with the comparator subjects

In the AS cohort, we observed 14 lymphomas during 51,240 person-years (crude incidence 29 per 100,000 [95% confidence interval (95% CI) 15–46]) versus 75 lymphomas during 242,255 person-years (crude incidence 31 per 100,000 [95% CI 25–39]) among the comparator subjects. This resulted in an HR of having lymphoma of 0.9 (95% CI 0.5–1.6). Excluding all AS patients who had at least one visit code listing a diagnosis of RA (12% of the cohort) resulted in a virtually unchanged HR of 0.8 (95% CI 0.4–1.5) (Table 3).

Table 3. HRs of malignant lymphoma in the Swedish cohorts of patients with AS (n = 8,707) and patients with PsA (n = 19,283) compared with their comparator subjects with data recorded between 2001 and 2010, overall and stratified by sex, age, calendar period, and duration of followup time*
 No. of lymphomas among AS patients/no. of lymphomas among general populationHR (95% CI)No. of lymphomas among PsA patients/no. of lymphomas among general populationHR (95% CI)
  1. Hazard ratios (HRs) and 95% confidence intervals (95% CIs) were determined in a Cox regression analysis taking sex, county of residence, and age (in 5-year age categories until 85, thereafter 85 and older) into account. All person-time and all events during the first 90 days of followup were excluded.
  2. aAll patients with a diagnosis of rheumatoid arthritis (RA) recorded before or after the ankylosing spondylitis (AS) or psoriatic arthritis (PsA) diagnosis in the Outpatient Register were excluded from the analysis.
Overall14/750.9 (0.5–1.6)45/1751.2 (0.9–1.7)
Stratified by sex    
Female2/180.5 (0.1–2.3)21/791.3 (0.8–2.0)
Male12/571.0 (0.5–1.8)24/961.2 (0.8–1.9)
Stratified by age at start of followup    
18–49 years2/210.4 (0.1–1.9)6/221.3 (0.5–3.2)
50–74 years11/511.2 (0.5–2.0)33/1361.2 (0.8–1.7)
≥75 years1/31.5 (0.2–15)6/171.6 (0.6–4.0)
Stratified by calendar period at start of followup    
2001–200412/650.9 (0.5–1.6)36/1261.3 (0.9–2.0)
2005–20102/101.0 (0.2–4.4)9/491.0 (0.6–2.0)
Stratified by duration of followup    
≤1 year3/131.1 (0.3–3.1)8/241.2 (0.6–2.6)
>1 year11/620.8 (0.5–1.6)37/1511.2 (0.8–1.7)
Ever RA diagnosis excludeda11/750.8 (0.4–1.5)28/1751.0 (0.6–1.4)

In the PsA cohort, 45 lymphomas were observed during 109,143 patient-years (crude incidence 41 per 100,000 [95% CI 30–55]) versus 175 lymphomas during 520,069 person-years (crude incidence 35 per 100,000 [95% CI 29–39]) among the comparator subjects, resulting in an HR of having lymphoma of 1.2 (95% CI 0.9–1.7). Excluding all PsA patients who had at least 1 visit code for RA (17% of the cohort) yielded an HR of 1.0 (95% CI 0.6–1.4) (Table 3). Subset analyses by sex, age, calendar period, or duration of followup time resulted in HRs largely consistent with the overall HR both in AS and in PsA (Table 3).

Applying a stricter case definition (i.e., requiring at least 2 outpatient visits with a diagnosis of AS or PsA) resulted in an HR of having lymphoma of 0.9 (95% CI 0.5–1.6) for AS and 1.3 (95% CI 0.9–1.9) for PsA (Table 4).

Table 4. HRs of lymphoma in the Swedish cohorts of patients with AS (n = 6,415) and PsA (n = 14,088) compared with their comparator subjects with respect to treatment with DMARDs and oral glucocorticoids*
 No. of lymphomas among AS patients (person-time)aHR (95% CI)No. of lymphomas among PsA patients (person-time)aHR (95% CI)
  1. Hazard ratios (HRs) and 95% confidence intervals (95% CIs) were determined in a Cox regression analysis taking sex and age (linear variable within the regression model) into account. Analyses were based on a stricter definition of case exposure, i.e., requiring ≥2 outpatient specialist visits with a diagnosis of ankylosing spondylitis (AS) or psoriatic arthritis (PsA). DMARDs = disease-modifying antirheumatic drugs.
  2. aPerson-time refers to days of followup.
  3. bTreatment was defined as ≥2 prescriptions of either methotrexate or sulfasalazine within 6 months as registered in the Prescribed Drug Register between 2005 and 2010.
  4. cTreatment was defined as ≥2 prescriptions within 6 months as registered in the Prescribed Drug Register between 2005 and 2010.
Overall11 (39,640)0.9 (0.5–1.6)37 (83,468)1.3 (0.9–1.9)
DMARDsb3 (10,261)1.2 (0.3–4.3)14 (39,131)1.7 (1.0–3.1)
Oral glucocorticoidsc0 (5,604)3 (19,668)0.6 (0.2–2.0)

Risk of lymphoma in relation to DMARD therapy

Analyses restricted to AS patients who were treated with DMARDs (methotrexate and/or sulfasalazine) or with oral glucocorticoids did not reveal any significantly different HRs compared to those of the original analyses. In contrast, in PsA patients, treatment with DMARDs was associated with a borderline significant increased risk of lymphoma compared with their corresponding comparator subjects, whereas HRs in relation to oral glucocorticoids were not (Table 4). Excluding patients who had a RA diagnosis ever recorded from this DMARD-treated cohort resulted in an HR of having lymphoma for PsA patients treated with DMARDs of 1.9 (95% CI 1.1–3.7).

Incidence of lymphoma in relation to TNFi therapy

Among the TNFi-treated patients with AS, 2 lymphomas occurred during 7,790 person-years (crude incidence 25 per 100,000 [95% CI 3–93]) versus 12 lymphomas during 43,905 person-years (crude incidence 27 per 100,000 [95% CI 14–48]) among non–TNFi-treated AS patients. Applying a stricter AS case definition did not change the crude incidence (Table 5). For both of the patients with AS who had lymphomas, the AS and the lymphoma diagnoses could be confirmed from the medical records and from the pathology reports.

Table 5. Crude incidences of lymphoma in the Swedish cohorts of patients with AS (n = 6,415) and PsA (n = 14,088) compared with their comparator subjects with respect to treatment with a TNFi*
 No. of lymphomas among patients with AS (person-time)aCrude incidence (95% CI) per 100,000 person-yearsNo. of lymphomas among patients with PsA (person-time)aCrude incidence (95% CI) per 100,000 person-years
  1. Analyses were based on a stricter definition of case exposure, i.e., requiring ≥2 outpatient specialist visits with a diagnosis of ankylosing spondylitis (AS) or psoriatic arthritis (PsA). 95% CI = 95% confidence interval.
  2. aPerson-time refers to days of followup.
  3. bTreatment with a tumor necrosis factor inhibitor (TNFi) was defined as ever having been exposed to a TNFi during followup (between 2001 and 2010), as registered in the Swedish Biologics Register (Anti-Rheumatic Therapy in Sweden).
Overall11 (39,640)28 (14–50)37 (83,468)44 (31–61)
TNFi-treated patientsb2 (7,028)28 (3–102)5 (9,600)52 (17–122)
Non–TNFi-treated patients9 (32,882)27 (12–52)32 (74,230)43 (29–61)

Among the TNFi-treated patients with PsA, we identified 8 lymphomas during 10,912 person-years, corresponding to a crude incidence of 73 per 100,000 (95% CI 34–144), versus 37 lymphomas during 98,710 person-years (crude incidence 37 per 100,000 [95% CI 26–52]) among non–TNFi-treated PsA patients. When applying a stricter definition of PsA (i.e., requiring at least 2 outpatient visits with PsA), the crude incidence in the subset treated with TNFi was 52 per 100,000 (95% CI 17–122) (Table 5). Indeed, validation through the medical records revealed that 3 of the original 8 patients with PsA and lymphoma actually fulfilled the American College of Rheumatology 1987 classification criteria for RA ([21]) and were clinically evaluated as RA patients through the medical records. In a third approach, we excluded the 3 patients misclassified as having PsA and lymphoma from the original numerator and estimated the crude incidence based on all TNFi-treated PsA patients who also had a PsA diagnosis in the biologics-module of the Swedish Rheumatology Quality Register, i.e., in ARTIS (n = 2,125). In this analysis, the crude incidence of lymphoma following TNFi treatment of PsA was 64 per 100,000 (95% CI 21–151).

Finally, the revalidation of the lymphoma specimen of the remaining 5 patients with PsA and lymphoma revealed that 1 of these patients did not have a lymphoma diagnosis, such that of the original 8 patients with PsA and lymphoma, 4 actually had PsA and lymphoma. When this patient was excluded from the analysis, this resulted in a lymphoma incidence rate of 52 per 100,000 (95% CI 14–131).

Since in clinical practice in Sweden, a previous cancer diagnosis has generally represented a contraindication for TNFi therapy, we performed separate analyses excluding all patients with a previous cancer diagnosis. This exclusion did not significantly change the results (data not shown). Because of the small numbers, we did not perform more detailed assessments of HRs with TNFi therapy.

Distribution of lymphoma subtypes

When categorizing the observed lymphomas into different subtypes according to ICD-10 codes in the Cancer Register, we did not find any markedly increased proportion of any specific lymphoma subtype, apart from a slightly higher proportion of DLBCL in the AS patients compared with their comparator subjects, although the numbers of lymphomas per each lymphoma subtype in the AS cohort were small. There was no increased proportion of T cell lymphoma in PsA patients as compared with the general population (Table 6).

Table 6. Distribution of lymphoma subtypes in patients with AS and patients with PsA compared with their comparator subjects*
Lymphoma subtypesAS patients with lymphoma (n = 14)General population with lymphoma (n = 75)PsA patients with lymphoma (n = 45)General population with lymphoma (n = 175)
  1. Lymphoma subtypes were identified based on International Statistical Classification of Diseases and Related Health Problems, Tenth Revision diagnosis codes as registered in the Swedish Cancer Register. Values are the no. (%) of patients with lymphomas. AS = ankylosing spondylitis; PsA = psoriatic arthritis; DLBCL = diffuse large B cell lymphoma; NHL = non-Hodgkin's lymphoma; CLL = chronic lymphocytic leukemia.
B cell lymphoma    
DLBCL7 (50)22 (29)19 (42)59 (34)
Follicular lymphoma0 (0)11 (15)3 (7)24 (14)
Other B cell lymphoma1 (7)7 (9)5 (11)19 (11)
T cell lymphoma1 (7)3 (4)1 (2)9 (5)
NHL, unspecified0 (0)6 (8)7 (16)13 (7)
Hodgkin's lymphoma0 (0)6 (8)5 (11)5 (3)
CLL4 (29)20 (27)5 (11)44 (25)
Unspecified lymphoma1 (7)0 (0)0 (0)2 (1)

In the validation of tissue specimens for lymphoma subtype occurring in the TNFi-exposed patients (both AS and PsA) by one of the authors (CS), 1 lymphoma subtype diagnosis was changed from the original pathology report (from CLL to unspecified B cell lymphoma in a PsA patient). The remaining lymphomas were distributed as follows: 2 DLBCLs (1 in an AS patient and 1 in a PsA patient), 1 follicular lymphoma (in a PsA patient), 1 mantle cell lymphoma (in an AS patient), and 1 Hodgkin's lymphoma (in a PsA patient).

DISCUSSION

In this study, we made several key observations. First, we found no increased risk of lymphoma in patients with AS compared with the general population. For patients with PsA, there was a borderline significantly increased risk of 20%. However, when all patients in the PsA cohort who had ever had a diagnosis code of RA were excluded, this risk increase disappeared. Second, the distribution of lymphoma subtypes did not differ substantially between patients with AS or PsA and the general population. Third, treatment with methotrexate and/or sulfasalazine was associated with an increased risk of lymphoma in patients with PsA compared with the comparator subjects, which persisted when patients with a diagnosis of RA were excluded, but was not detected in patients with AS. Fourth, the crude incidence of lymphoma in AS or PsA patients treated with TNFi was not substantially different from those of TNFi-naive AS and PsA patients, although the numbers were small.

For AS, our observation of an, on average, nonincreased risk of lymphoma is consistent with previous data ([8, 11, 22-24]), with the exception of one retrospective cohort study (published as an abstract only) that identified an incidence risk ratio of lymphoma of 2.8 among AS patients ([25]).

Corresponding data for PsA are scarce. Our literature search resulted in the identification of one published observational study, a Canadian cohort study by Rohekar et al ([26, 27]) that assessed cancer risks in PsA patients followed up prospectively from 1978 to 2004 and found a standardized incidence ratio for all hematopoietic malignancies combined of 0.7 (95% CI 0.3–1.8). Our study may therefore be the first published study to assess lymphoma risk separately in PsA ([27]).

Conversely, in cutaneous psoriasis, study results with regard to lymphoma risk are mixed. Some have reported increased risks ([3, 5, 6]), while others have not ([9-11]). As with RA, there is some evidence of a link to disease severity ([4, 7]). From a clinical perspective, it is sometimes difficult to distinguish between RA and PsA, at least the PsA phenotype characterized by small-joint inflammation, and occasionally, the two conditions are confusingly similar (as illustrated by the clear diagnostic ambiguity in our validation of medical records). A significant proportion of patients with PsA in this study (17%) had thus at some point (correctly or incorrectly) received a diagnosis code for RA. Although the average relative risk of lymphoma in PsA approximated 1 when these patients were excluded, a truly increased lymphoma risk in a particular subset of PsA patients (i.e., those with RA-like disease) is difficult to rule out, as is a risk increase in those patients with the most severe PsA. In this study, we were not able to adjust for inflammatory activity or disease severity. Interestingly, however, a significant risk increase was observed among PsA patients who were treated with methotrexate and/or sulfasalazine (which remained after exclusion of individuals who also received an RA diagnosis code at least once), possibly reflective of a patient group in which systemic treatment is required, and which presumably has more severe disease.

The discrepancies between lymphoma risk in RA and that in AS and PsA may have several explanations. RA is highly characterized by activated autoimmune B cells (with the presence of rheumatoid factor and anti–cyclic citrullinated peptide). B cell activation (and in turn, overstimulation and defective apoptosis of B cells), as well as systemic inflammation, gives rise to a state of chronic immune stimulation, which is one of several possible biologic mechanisms linking autoimmunity and RA to lymphoma development ([1, 28, 29]). Further, the strongest association with RA has been shown for a particular lymphoma subtype, i.e., DLBCL ([2]), whereas T cell lymphomas, for example, have been linked to other autoimmune conditions (such as celiac disease and psoriasis) ([30-34]), pointing to different underlying pathways in different inflammatory diseases that are of importance both for the inflammatory disease and for the lymphoma development. In this study, there was a slightly higher proportion of DLBCL in AS patients than expected, based on small numbers. In contrast, there was no increased proportion of T cell lymphoma in the groups of PsA or AS patients. Nor did we observe any hepatosplenic T cell lymphomas, for which there has been a particular concern in patients with IBD treated with a combination of TNFi and azathioprine ([12]).

Treatment of inflammatory diseases has repeatedly been linked to lymphoma risks. With regard to methotrexate, an increased lymphoma risk due to treatment, especially in RA, has been reported ([1]), although larger, observational population-based studies have produced little evidence to support the notion of excess risks associated with methotrexate per se ([2, 35, 36]). With respect to sulfasalazine, data are limited, without evidence of increased risks ([2]). Assessing the link between therapy and lymphoma risk is complex, considering potential preexisting variations in risk, such as that due to disease severity, and channeling bias. Furthermore, the choice of DMARDs does not differ to any great extent in RA versus AS (at least in the subset of patients with peripheral polyarthritis) or PsA, even if the proportion of patients with each diagnosis who are treated with DMARDs may differ. Thus, it is less likely that the differences in lymphoma risk between RA and AS or PsA are a consequence of an intrinsic effect of DMARDs.

The safety profile of TNFi therapy with respect to risk of lymphoma has been a concern. The majority of studies addressing this issue have focused on the risk of lymphoma following exposure in RA ([13, 14, 37]). In AS and PsA, data so far have been hampered by small numbers of events, making risk estimation difficult ([15, 38-40]). In the present study, the crude incidences of lymphoma in TNFi-exposed versus TNFi-naive patients were largely similar in AS and in PsA, after correction for misclassification of RA and when applying a stricter case definition of PsA. These results are generally consistent with the conclusions from previous assessments based on 1 TNFi-exposed AS patient with lymphoma and 2 TNFi-exposed PsA patients with lymphomas ([38]) and on 4 TNFi-exposed AS patients with lymphomas and 3 TNFi-exposed PsA patients with lymphomas ([39]), making the present (albeit small) study the largest to date to evaluate the association between lymphoma and TNFi exposure in patients with AS and PsA.

Some further strengths and limitations of the present study need to be addressed. Linkage to the Swedish Cancer Register allowed the capture of lymphomas independently of registration in ARTIS and in the National Patient Register, resulting in an independent and prospective registration of exposure and outcome, minimizing bias from selective recall or reporting related to knowledge of treatment status. By using matched population comparator subjects, we were also able to assess the underlying risk of lymphoma in our AS and PsA cohorts and compare that with the lymphoma incidences in TNFi-treated and TNFi-naive patients from the same cohorts. However, we were unable to fully control for potential confounding from factors influencing treatment decisions, such as comorbidity (other than a previous cancer) and disease severity, when comparing TNFi-treated versus non–TNFi-treated patients. Finally, despite our efforts to assemble sufficiently large patient cohorts to allow estimation of lymphoma risk in different subsets of patients and in subsets of lymphomas, some of our sensitivity analyses were limited by small numbers and low precision, which is why we did not estimate adjusted HRs, e.g., for TNFi treatment.

To conclude, the average risk of lymphoma in AS and PsA is no different from that in the general population, although an increased risk in a subpopulation of patients with PsA cannot be excluded. Overall, the distribution of lymphoma subtypes was no different in AS or PsA versus the general population. Reassuringly, the numbers and crude incidences of lymphoma in AS and PsA patients treated with TNFi were not increased compared to those in non–TNFi-treated patients.

AUTHOR CONTRIBUTIONS

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. Hellgren 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. Hellgren, Backlin, Feltelius, Baecklund, Askling.

Acquisition of data. Hellgren, Backlin, Sundstrom, Feltelius, Eriksson, Baecklund, Askling.

Analysis and interpretation of data. Hellgren, Smedby, Backlin, Feltelius, Eriksson, Baecklund, Askling.

Acknowledgments

The authors acknowledge the ARTIS Study Group, which includes Johan Askling, Eva Baecklund, Lars Coster, Helena Forsblad, Nils Feltelius, Pierre Geboreck, Lennart Jacobsson, Lars Klareskog, Lars-Erik Kristensen, Staffan Lindblad, Solbritt Rantapaa-Dahlqvist, and Ronald van Vollenhoven.

Ancillary