Juvenile idiopathic arthritis and risk of cancer: A nationwide cohort study

Authors

  • J. F. Simard,

    Corresponding author
    1. Karolinska Institutet, Stockholm, Sweden
    • Clinical Epidemiology Unit, T2, Karolinska University Hospital, 171 76 Stockholm, Sweden

    Search for more papers by this author
  • M. Neovius,

    1. Karolinska Institutet, Stockholm, Sweden
    Search for more papers by this author
  • S. Hagelberg,

    1. Astrid Lindgren Children's Hospital at Karolinska University Hospital, Stockholm, Sweden
    Search for more papers by this author
    • Dr. Hagelberg has received consulting fees from Abbott Scandinavia (less than $10,000).

  • J. Askling

    1. Karolinska Institutet and Karolinska University Hospital, Stockholm, Sweden
    Search for more papers by this author
    • Dr. Askling has received honoraria (less than $10,000) for presentations of results from safety studies, based on data from the Swedish Biologics Register, at scientific meetings that were funded or organized by Bristol-Myers Squibb or Wyeth. The Swedish Biologics Register ARTIS is run by the Swedish Society for Rheumatology, and for the maintenance of this register, the Swedish Society for Rheumatology has received funding, independent of the conduct of these scientific analyses, from Schering-Plough, Bristol-Myers Squibb, Wyeth, Abbott Laboratories, and Roche.


Abstract

Objective

Reports of therapy-related adverse events suggest an elevated rate of malignancy in patients with juvenile idiopathic arthritis (JIA) treated with biologic therapies. However, the scarcity of data on the underlying risk of malignancy in JIA hampers interpretation of these signals. Therefore, the aim of this study was to determine the risk of cancer in patients with JIA as compared with that in the general population.

Methods

Through linkage with a national database, the Swedish Patient Register (comprising inpatient discharges in 1969–2007 and specialist outpatient visits in 2001–2007 in Sweden), a national JIA cohort (n = 9,027) was identified, and each JIA case was matched with 5 general population comparators. Using data from the Swedish Cancer, Census, Death, and Biologics Registers, the occurrence of cancer, vital status, and start of a biologic therapy were identified. The relative risk (RR) of first occurrence of a primary cancer in patients who had not been treated with biologics (biologics-naive patients with JIA) was estimated using Poisson regression, stratified a priori by year of earliest identification of JIA (before 1987 versus 1987 and thereafter). In sensitivity analyses, the data were followed up to 1999, when biologics first became available.

Results

In this biologics-naive JIA cohort, 60 malignancies were observed during 131,144 person-years of followup, compared with 266 cancers observed during 661,758 person-years in the general population comparator (0.46 cases/1,000 person-years versus 0.40 cases/1,000 person-years; RR 1.1, 95% confidence interval [95% CI] 0.9–1.5). Patients with JIA identified before 1987 were not at increased risk of cancer, whereas JIA identified in 1987 and thereafter was significantly associated with incident lymphoproliferative malignancies (RR 4.2, 95% CI 1.7–10.7) and cancers overall (RR 2.3, 95% CI 1.2–4.4). Sensitivity analyses did not reveal any ready explanation for this heterogeneity.

Conclusion

Although absolute risks were low, an elevated risk of malignancy was observed among biologics-naive patients in whom the diagnosis of JIA was made in the past 20 years, which may have implications for the interpretation of cancer signals in patients with JIA treated with newer therapies.

Recent reports have suggested an elevated rate of malignancy, particularly lymphoproliferative cancers, in patients with juvenile idiopathic arthritis (JIA) treated with tumor necrosis factor α (TNFα) inhibitors, leading to a recent black box warning for these drugs (1). Between 2001 and 2008, the US Food and Drug Administration (FDA) received reports of 48 malignancies occurring in children and adolescents exposed to TNFα inhibitors, of which 15 occurred in patients with JIA who were <18 years of age (2).

The interpretation of this signal is, however, hampered by the scarcity of data on the risks of cancer in JIA in the absence of biologic therapies. In August 2009, the FDA concluded that “the background incidence of malignancy in children with JIA is not well defined” (2). Indeed, it is far from certain whether the increased risks of specific cancers reported among adult patients with rheumatoid arthritis (RA) who had not been treated with biologic therapy (3) can be extrapolated to the pediatric setting, in which the background rate of cancer, and the distribution of cancer types, is markedly different from that in adults.

Therefore, the aim of this study was to assess the occurrence and risk of cancer in patients with JIA who had not been treated with biologics (biologics-naive JIA) as compared with that in the general population. By cross-referencing data from Swedish population-based registers, we assembled a national cohort of 9,027 patients with JIA whose diagnosis was identified from 1969 through 2007, a matched general population comparator, and data on cancer occurrence and vital status through 2007.

PATIENTS AND METHODS

Setting and data sources.

Swedish healthcare is public and tax-funded. Pediatric rheumatology today is managed by pediatricians with a subspecialty or special interest in rheumatology. However, prior to the 1990s, the care was jointly managed by internist-rheumatologists and pediatricians. As in adult rheumatologic care, biologic therapies for JIA were introduced in 1999.

The national registration number is a unique 10-digit number issued to all Swedish residents alive in, or born after, 1947. This number is recorded in health registers and facilitates register linkage. In this study, we used the following nationwide population-based registers, which have all been used for similar studies in the adult setting (4). The Swedish Population Register hosts information on sex, year of birth, residency, immigration, and emigration of Swedish residents. The Swedish Patient Register contains information on Swedish inpatient care since 1964 (nationwide since 1987). For each visit, primary and contributory discharge diagnoses, as assigned by the discharging physician, are coded according to the calendar year–specific International Classification of Diseases (ICD) seventh through tenth revision codes, together with dates of admission and discharge. The completeness of this register is more than 99% (5). In 2001, nationwide data from specialist outpatient care (e.g., pediatric outpatient clinics and adult rheumatology outpatient clinics) were added to the Patient Register, and the same coding scheme as that used for the inpatient component is used (5, 6).

The Swedish Cancer Register includes information on incident malignancies since 1958 in Sweden. Reporting is mandatory for treating clinicians and pathologists, resulting in a nationally high rate of coverage. The reports include cancers diagnosed clinically, morphologically, via laboratory examination, and at autopsy. To facilitate studies spanning long study periods of time, each case is coded with an ICD-7 code in addition to calendar year–specific codes (e.g., ICD-10 after 1997). The Swedish Causes of Death Register contains information on >99% of all deaths since 1952 in Sweden, including that reported on death certificates (7).

The Swedish Prescribed Drug Register was started on July 1, 2005 and retains information on dates and drugs for all pharmacy dispensations from the Swedish governmental pharmacy monopoly during the study period (no other sources for drug dispensing exist). Drugs administered by infusions provided in hospitals, such as infliximab, are not covered on a patient-level basis or (as with infliximab) are covered to a lesser extent (in 2009, 18% of infliximab was used in ambulatory care in Sweden) (8). The Swedish Biologics Register ARTIS (Anti-Rheumatic Therapy in Sweden) is a collection of data on adult patients who have been prescribed biologic therapies for the treatment of rheumatic disease in Sweden since 1999. In this register, patients whose data are entered on the basis of a diagnosis of JIA and who are starting a biologic therapy in adult care can be identified, along with data on the date of and age at diagnosis and disease duration (9).

Design and participants.

We identified 5,748 individuals with inpatient hospital discharges (1969–2007) and 4,353 individuals with outpatient visits (2001–2007) for whom a diagnostic code of JIA (712.0 on ICD-8, 714D on ICD-9, and M08 and M09 on ICD-10) was assigned before the age of 16 years (total n = 9,027) (Table 1). For each patient with JIA, 5 age-, sex-, and geographic region–matched general population comparators who were alive at the time of the date of first identification of JIA (index date) in the index case were identified through the Swedish Population Register (n = 44,900). Most cases, but not all, were successfully matched to 5 population-based comparators.

Table 1. Characteristics of the patients with juvenile idiopathic arthritis (JIA) over the entire followup period, stratified by calendar period of initial identification of JIA in the Swedish Patient Register
 All JIAJIA identified 1969–1986JIA identified 1987–2007
No. of participants9,0203,7245,296
Female sex, no. (%)5,016 (55.6)1,706 (45.8)3,310 (62.5)
Age, mean/median years   
 Start of followup9/98/710/10
 End of followup23/2135/3515/15

Followup for incident malignancies, and statistical analysis.

Through linkage to the Swedish Cancer Register, all registered cancers (including malignant nonmelanoma skin cancers but excluding basal cell carcinomas) up through the end of 2007 in the study population were identified. Lymphoproliferative malignancies were defined as Hodgkin's lymphoma and non-Hodgkin's lymphoma, acute lymphocytic leukemia, chronic lymphocytic leukemia, and multiple myeloma. The followup period (expressed in person-time units) was the amount of time accrued from the index date until the first occurrence of the following events: emigration, death, first cancer, end of followup (December 31, 2007), or (if present) first recorded date of the start of biologic therapy. Data on the start of biologic therapy among the members of the JIA cohort were obtained from ARTIS and the Swedish Prescribed Drug Register.

Individuals with a history of cancer by the index date were excluded (7 patients with JIA, 42 general population comparators), leaving 9,020 patients with JIA and 44,858 subjects in the general population comparator cohort. For each individual, a maximum of 1 cancer (i.e., his or her first-ever invasive cancer) was considered. To reduce the risk that an undiagnosed malignancy might be initially masquerading as JIA, we excluded all person-time data and all events during the first year of followup. Sensitivity analyses were performed with inclusion of the first year of followup, and the results did not change appreciably. In additional sensitivity analyses, the definition of JIA was refined by first, removing the ICD-10 code M09 from the list of possible diagnoses (leaving 8,976 JIA cases), and second, requiring that at least 2 visits be recorded in the Swedish Patient Register to be defined as a case (n = 5,123). Outcomes assessed included cancer overall, lymphoproliferative malignancies, and all cancers other than lymphoproliferative malignancies.

The relative risk (RR) of cancer in the JIA cohort, in comparison with the general population comparator, was estimated using Poisson regression, adjusting for age (in 5-year age categories until age 85 years, then 85+ years), sex, followup time (<1 year, 1–2 years, >2–5 years, >5–10 years, >10–20 years, and >20–50 years), and calendar period (1969–1979, 1980s, 1990s, 2000–2007). Analyses that compared the JIA cohort with the general population using national age-, sex-, and calendar period–specific rates through standardized incidence ratios were also performed, and similar RRs were obtained (data not shown).

Analyses were stratified a priori by calendar year of entry into the JIA cohort (3,724 patients first entering with a diagnosis of JIA before 1987, and 5,296 patients first entering with JIA beginning in 1987, when the Swedish Patient Register included all counties throughout Sweden and when the ICD-9 was implemented). Further sensitivity analyses assessed cancer risks in patients up to the age of 18 years only (among the cohort, the age at cancer incidence ranged from 3 years to 51 years, with a mean age of 29 years [SD 11 years]). Moreover, to account for the possibility of missing data on the start of biologic therapy in the JIA cohort in 1999 and onward, the followup was ended in 1999, when biologics first became available.

RESULTS

Based on the JIA cohort assembled, the prevalence of JIA in the Swedish population up to age 16 years in 2007 was estimated to be 0.16% (0.12% in male subjects and 0.20% in female subjects). The exposure to biologics therapy at age 16 years in this cohort was estimated to be 9% in 2007.

In the JIA cohort, 60 incident malignancies were observed during 131,144 person-years of followup (incidence rate of 0.46 per 1,000 person-years) from 1969 through 2007. In the matched general population comparator, 266 incident malignancies were identified during 661,758 person-years of followup (incidence rate of 0.40 per 1,000 person-years), corresponding to an absolute excess risk of 0.1 cancers per 1,000 person-years and an RR for all-site cancers of 1.1 (95% confidence interval [95% CI] 0.9–1.5), but without any increased risk of lymphoproliferative malignancies (Table 2). Sensitivity analyses in which the criteria for the definition of JIA were altered did not appreciably change the results. In analyses using the first alternate definition, in which ICD-10 M09 was removed as a diagnostic code for JIA, results were nearly identical to those in the unaltered analyses (results not shown). When at least 2 visits in the inpatient component, outpatient component, or both components of the Swedish Patient Register were required to define JIA, the results were similar (Table 3).

Table 2. Risk of cancer in the cohort of patients with JIA (n = 9,020) as compared with the general population comparator group (n = 44,858) by calendar period of first identification of JIA versus all JIA, stratified by age at time of cancer diagnosis*
 JIA by year of identificationAll years
1969–19861987–2007
  • *

    JIA = juvenile idiopathic arthritis; RR = relative risk; 95% CI = 95% confidence interval.

All ages
 All cancers
  RR (95% CI)1.0 (0.7–1.4)2.3 (1.2–4.4)1.1 (0.9–1.5)
  No. of cancers, JIA vs. comparator47 vs. 23613 vs. 3060 vs. 266
 Lymphoproliferative cancers
  RR (95% CI)0.3 (0.1–1.1)4.2 (1.7–10.7)1.0 (0.5–1.9)
  No. of cancers, JIA vs. comparator3 vs. 458 vs. 1011 vs. 55
 Other cancers
  RR (95% CI)1.2 (0.8–1.6)1.3 (0.5–3.6)1.2 (0.9–1.6)
  No. of cancers, JIA vs. comparator44 vs. 1915 vs. 2049 vs. 211
Age 0–17 years
 All cancers
  RR (95% CI)0.5 (0.1–1.5)3.9 (1.6–9.3)1.4 (0.7–2.6)
  No. of cancers, JIA vs. comparator3 vs. 329 vs. 1212 vs. 44
 Lymphoproliferative cancers
  RR (95% CI)0.0 (0.0–3.7)5.2 (1.7–16.2)1.3 (0.5–3.2)
  No. of cancers, JIA vs. comparator0 vs. 176 vs. 66 vs. 23
 Other cancers
  RR (95% CI)1.0 (0.3–3.4)2.6 (0.7–10.4)1.4 (0.6–3.6)
  No. of cancers, JIA vs. comparator3 vs. 153 vs. 66 vs. 21
Age 18+ years
 All cancers
  RR (95% CI)1.1 (0.8–1.6)1.2 (0.4–3.6)1.1 (0.8–1.5)
  No. of cancers, JIA vs. comparator44 vs. 2044 vs. 1848 vs. 222
 Lymphoproliferative cancers
  RR (95% CI)0.5 (0.2–1.8)2.7 (0.5–14.7)0.8 (0.3–2.0)
  No. of cancers, JIA vs. comparator3 vs. 282 vs. 45 vs. 32
 Other cancers
  RR (95% CI)1.2 (0.8–1.7)0.8 (0.2–3.4)1.2 (0.8–1.6)
  No. of cancers, JIA vs. comparator41 vs. 1762 vs. 1443 vs. 190
Table 3. Risk of cancer in a cohort of patients identified as having JIA according to an alternate definition (n = 5,123) as compared with a general population comparator group (n = 25,522) by calendar period of first identification of JIA versus all JIA, stratified by age at time of cancer diagnosis*
 JIA by year of identification 
1969–19861987–2007All years
  • *

    The alternate definition of juvenile idiopathic arthritis (JIA) for this analysis required that at least 2 visits be recorded in the Swedish Patient Register to be defined as a case. RR = relative risk; 95% CI = 95% confidence interval.

All ages
 All cancers
  RR (95% CI)1.4 (0.9–2.3)2.7 (1.3–5.7)1.7 (1.1–2.5)
  No. of cancers, JIA vs. comparator22 vs. 8111 vs. 2222 vs. 103
 Lymphoproliferative cancers
  RR (95% CI)0.5 (0.1–2.2)4.7 (1.7–13)1.6 (0.8–3.5)
  No. of cancers, JIA vs. comparator2 vs. 207 vs. 89 vs. 28
 Other cancers
  RR (95% CI)1.8 (1.1–2.9)1.6 (0.5–4.9)1.7 (1.1–2.7)
  No. of cancers, JIA vs. comparator20 vs. 614 vs. 1424 vs. 75
Age 0–17 years
 All cancers
  RR (95% CI)0.3 (0.04–2.5)4.7 (1.8–12.1)1.9 (0.9–4.1)
  No. of cancers, JIA vs. comparator1 vs. 158 vs. 99 vs. 24
 Lymphoproliferative cancers
  RR (95% CI)0.0 (0.0–3.7)5.3 (1.7–16.3)2.2 (0.8–5.7)
  No. of cancers, JIA vs. comparator0 vs. 86 vs. 66 vs. 14
 Other cancers
  RR (95% CI)0.7 (0.1–5.7)3.5 (0.6–21)1.5 (0.4–5.6)
  No. of cancers, JIA vs. comparator1 vs. 72 vs. 33 vs. 10
Age 18+ years
 All cancers
  RR (95% CI)1.7 (1.0–2.8)1.3 (0.4–4.7)1.6 (1.0–2.6)
  No. of cancers, JIA vs. comparator21 vs. 663 vs. 1324 vs. 79
 Lymphoproliferative cancers
  RR (95% CI)0.9 (0.2–3.9)2.9 (0.3–31.7)1.1 (0.3–3.9)
  No. of cancers, JIA vs. comparator2 vs. 121 vs. 23 vs. 14
 Other cancers
  RR (95% CI)1.9 (1.1–3.2)1.0 (0.2–4.7)1.7 (1.1–2.9)
  No. of cancers, JIA vs. comparator19 vs. 542 vs. 1121 vs. 65

When the analyses were stratified by calendar period of first identification of JIA, patients with JIA first identified before 1987 did not appear to be at either increased risk or decreased risk of cancer overall compared with their matched controls (RR 1.0, 95% CI 0.7–1.4 for 47 cancers versus 236 cancers). In contrast, patients with JIA first identified in 1987 or later were at increased risk of cancer overall (RR 2.3, 95% CI 1.2–4.4 for 13 cancers versus 30 cancers). The risk estimates were significantly different between these 2 groups (JIA diagnosis before 1987 versus 1987 and thereafter; P = 0.02). This was most pronounced for lymphoproliferative malignancies (RR 0.3 in those identified as having JIA before 1987 versus RR 4.2 in those identified as having JIA in 1987 and thereafter; P < 0.001), although the numbers of cancers were smaller in the earlier diagnosis period, at a time when the Swedish Patient Register did not have full national coverage.

Among the 8 cases of lymphoproliferative malignancy occurring in patients with JIA identified in the later period, the mean age at JIA presentation was 9.5 years, and the mean age at cancer diagnosis was 14 years. Four of these patients were diagnosed as having Hodgkin's lymphoma, 3 as having non-Hodgkin's lymphoma, and 1 as having acute lymphocytic leukemia. In contrast, the mean age at lymphoproliferative cancer diagnosis was 17 years in the comparator group; 6 of these subjects were diagnosed as having Hodgkin's lymphoma, 3 as having acute lymphocytic leukemia, and 1 as having non-Hodgkin's lymphoma. Nonlymphoproliferative cancers were not significantly associated with JIA, neither overall nor in analyses by calendar period of identification of JIA (Table 2).

When the analyses were stratified by sex, we found an overall null association with cancer both among male patients (RR 1.0, 95% CI 0.6–1.5 for 21 cancers versus 111 cancers in the comparator group) and among female patients (RR 1.3, 95% CI 0.9–1.8 for 39 cancers versus 155 cancers in the comparator group). Moreover, the result in female patients was not significantly different from the estimate in male patients (P = 0.31).

The overall pattern of RRs for cancer in the JIA cohort remained largely the same when followup was truncated at 18 years of age (Table 2). When analyses were restricted to followup at 18 years of age or older, the pattern was similar to that in the total group, with an elevated risk of lymphoproliferative cancers observed in patients with JIA identified in 1987–2007, although the risk was not statistically significantly elevated, despite the larger numbers of events overall (Table 2).

In further sensitivity analyses in which the followup ended on December 31, 1998, i.e., before biologics first became available, a nonelevated risk of cancer overall in the JIA cohort was observed (RR 0.7, 95% CI 0.4–1.2 for 19 cancers versus 115 cancers in the comparator group). When this analysis was further stratified by calendar period of JIA identification, the RR for cancer in patients with JIA identified in 1969–1986 with followup until 1999 was 0.6 (95% CI 0.4–1.1 for 16 cancers versus 109 cancers in the comparator group), whereas the RR for cancer in patients with JIA identified in 1987–1998 with followup until 1999 was 2.5 (95% CI 0.6–10 for 3 cancers versus 6 cancers in the comparator group).

When we further subdivided the nonlymphoproliferative cancers into 7 groups (respiratory, gastrointestinal, urogenital, nervous system, cutaneous, hematologic, and other sites), we found no statistically significant overall association between JIA and these cancer types. A borderline significant association with gastrointestinal cancers was observed (RR 2.0, 95% CI 1.0–4.3 for 10 cancers versus 25 cancers in the comparator group) (Table 4).

Table 4. Association between JIA and site-specific nonlymphoproliferative cancers, relative to a general population comparator group*
Cancer siteNJIANGenPopRR (95% CI)
  • *

    The relative risk (RR) with 95% confidence interval (95% CI) was adjusted for age and sex. NJIA = number of patients with juvenile idiopathic arthritis who had cancer at the specified site; NGenPop = number of general population comparator subjects who had cancer at the specified site.

Respiratory040.0 (0.0–3.7)
Gastrointestinal10252.0 (1.0–4.3)
Urogenital18801.2 (0.7–1.9)
Nervous system3290.5 (0.2–1.7)
Cutaneous10341.5 (0.7–3.1)
Hematologic3131.2 (0.3–4.1)
Other sites5261.0 (0.4–2.6)

Because of the differences in calendar year of JIA identification from the inpatient and outpatient parts of the Swedish Patient Register, we performed an analysis in which we restricted the period to 2001–2007, when both inpatient and outpatient components of the Patient Register were available. During this short followup period, we found an elevated risk of overall cancer associated with JIA regardless of the source cohort (results not shown). The power to detect an association was, however, limited; only 2 cancers were observed when only the data from the inpatient register for identification of JIA were used (1 cancer in the JIA group versus 1 cancer in the general population comparator).

DISCUSSION

In this study, we found that biologics-naive patients from Sweden whose diagnosis of JIA was identified and followed up during the last 40 years did not have an increased risk of cancer. Importantly, however, there appeared to be heterogeneity in this risk, in that patients identified as having JIA beginning 40 years ago up to 20 years ago did not have an elevated risk of cancer during followup, whereas those patients first identified as having JIA during the past 20 years appeared to have a significantly increased risk of lymphoproliferative cancers. Despite conducting several sensitivity analyses, we could not find any immediate explanation for this observed difference in cancer risk between the 2 calendar periods.

JIA is an inflammatory disease that is most often (although not always) chronic in nature. As such, it may have significant effects on the priming or function of the immune system during childhood and adolescence, although the significance of any such effects on the risk of cancer is not well understood. Treatment of JIA may include nonbiologic or biologic disease-modifying antirheumatic drugs; little is known about their effects on immune function with respect to cancer surveillance in the pediatric population.

A recent study, the results of which have been presented in abstract form, showed no elevated risks of malignancy associated with JIA between 1974 and 2006 in a Canadian cohort; in that study, no cancer cases were observed among 1,168 patients with JIA followed up for 16,396 patient-years, although ∼6 cases were expected (10). A second study of biologics-naive patients with JIA using US insurance claims data, the results of which have also been presented in abstract form, found between a 3-fold and a 4-fold increase in malignancy as compared with either matched non-JIA claims-based comparators or when using reference incidence rates from the US Surveillance, Epidemiology, and End-Results data (11). An earlier study that involved patients with various rheumatic diseases found that patients with juvenile chronic arthritis were not at significantly elevated risk of malignancy compared with the general population during followup from 1981 to 1996 (12).

Consistent with the study by Thomas et al (12), we showed no significantly increased risk of malignancy in the patients with JIA identified in the earlier period and also no significantly increased risk overall. Although we did not observe any overall increased risk of cancer in biologics-naive patients with JIA, we found an increased risk of malignancies, particularly lymphoproliferative, in patients presenting with JIA during the last 20 years. This increased risk was not explained by the introduction of biologic therapies, since the association was similar in analyses ending in 1999, although the increase was not statistically significant. Besides attributing the differences to chance, tentative explanations for the heterogeneity in RRs by calendar period include differences in the definition of JIA, changes in standard treatment exposures in the JIA cohort, such as treatment with methotrexate, and possible underascertainment of JIA during the earlier period, at a time before the national Patient Register covered all Swedish counties. Among solid organ–specific cancer types, we found no association between JIA and cancer.

The substantial heterogeneity of JIA and the varying classification schemes used over time make longitudinal studies of JIA, and comparisons across studies, difficult. In our study, the JIA cohort was defined from 2 highly overlapping data sources. The calculated cumulative prevalence of 1.6 JIA cases per 1,000 minors (age <16 years) is consistent with the data reported by others (13). In an earlier regional Swedish study, investigators found a lower prevalence. That study, however, measured point prevalence and did not include patients with disease in remission at study start (14). In contrast, the present study used the Swedish Patient Register to identify patients ever having had a hospitalization or specialist outpatient visit with a diagnosis of JIA at discharge or visit through 2007.

The possibility of misclassification of the JIA diagnosis cannot be excluded. We would expect that such misclassification, if present, would yield a conservative bias (i.e., toward no association). Therefore, we cannot exclude the possibility that the observed RRs may actually be underestimates of the true associations between JIA and malignancy. When 2 alternative definitions of JIA were evaluated, the findings were similar. For the cohort with JIA identified in the earlier time period, the existence of only the inpatient component of the Swedish Patient Register could suggest that there was an underascertainment of cases. Furthermore, it is important to mention that we cannot exclude the possibility of misclassification of malignancy, although the Cancer Register has been shown to be reliable and valid, particularly with regard to lymphoma registration classification (15). In addition, we excluded the first year of followup after initial JIA presentation to reduce the risk that an undiagnosed malignancy could be masquerading as JIA.

Although data were available to identify the initiation of biologics therapy using the Prescribed Drug Register, these data only span July 2005 through December 2007 for the present study. For patients older than 18 years of age and entered into ARTIS with a biologic treatment for JIA, information on earlier start dates were available in some cases; however, because JIA was not one of the initial ARTIS indications, some of the data on biologics therapy use may be missing. The biologics experience (estimated exposure rate of 9%) in the present study population may be an underestimate, since few patients with JIA were included in ARTIS, and prescriptions of infliximab are less often included in the Prescribed Drug Register, because they are typically not dispensed through the pharmacy. Being fairly small, the biologics exposure in this study did not permit any precise comparisons of cancer risks during childhood and adolescence in patients with JIA treated with biologics. Such analyses, instead, need to be performed in a wider population or setting, perhaps extending international pooling efforts beyond those recently presented at the European League against Rheumatism annual meeting (16).

While there are limitations to this study, there are also several strengths. We have amassed a population-based cohort of >9,000 patients identified as having JIA, along with multiple matched population-based comparators for each JIA case. Linkage with a high-coverage national cancer register provided an independent and virtually complete source of outcome identification with high diagnostic accuracy. A number of sensitivity analyses demonstrated the robustness of our findings.

Thus, although the absolute risks of cancer were low and the overall relative risk of cancer was not elevated, we found a significantly higher risk of malignancy among patients with JIA identified during the last 20 years compared with the general population. Whether the observed heterogeneity of cancer risk reflects changes in the definition and selection of JIA, its treatment, or methods of cancer detection remains unknown. Nevertheless, these risks observed in patients with JIA who have had no exposure to biologics therapy have implications for the interpretation of cancer signals in patients with JIA treated with newer therapies.

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. Simard 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. Simard, Hagelberg, Askling.

Acquisition of data. Simard, Neovius, Askling.

Analysis and interpretation of data. Simard, Askling.

ROLE OF THE STUDY SPONSOR

Wyeth provided funding for the study but had no influence on the study design, data acquisition and analysis, interpretation of the results, or manuscript preparation. Wyeth was allowed to comment on the findings prior to submission, although all final decisions resided with the investigators and publication of this article was not contingent upon approval by Wyeth.

Acknowledgements

We would like to thank all of the clinicians who registered their patients, and the members of the ARTIS Study Group (all in Sweden), as follows: Eva Baecklund (Uppsala University), Lars Cöster (Linköping University), Christina Dackhammar (Sahlgrenska Academy), Nils Feltelius (Swedish Medical Products Agency), Pierre Geborek (Lund University), Lennart Jacobsson (Lund University), Lars Klareskog (Karolinska Institutet), Staffan Lindblad (Karolinska Institutet), Solbritt Rantapää-Dahlqvist (Umeå University), Tore Saxne (Lund University), and Ronald van Vollenhoven (Karolinska Institutet).

Ancillary