SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. ADDITIONAL DISCLOSURES
  10. REFERENCES

Objective

To estimate the relative risk of incident cancer diagnosis among patients with juvenile idiopathic arthritis (JIA) compared to patients without JIA.

Methods

A cohort of biologics-naive patients diagnosed with JIA between 1998 and 2007 and a matched cohort of comparators without JIA were assembled from the PharMetrics Patient-Centric Database. The primary outcome was any incident malignancy, excluding nonmelanoma skin cancer and carcinoma in situ. Claims profiles of patients with any cancer-related diagnosis codes were reviewed to determine outcomes. Incidence rates and 95% confidence intervals (95% CIs) of cancer were calculated and compared between cohorts using Cox proportional hazards regression. Standardized incidence ratios (SIRs) for each cohort compared to the general population were calculated using reference rates from the US Surveillance, Epidemiology, and End-Results (SEER) program.

Results

The JIA and non-JIA cohorts included 3,605 and 37,689 patients, respectively, with a mean age of 11 years. The incidence rates of cancer were 67.0 (95% CI 1.3–132.5) cases/100,000 person-years (PY) for JIA and 23.2 (95% CI 12.2–34.2) cases/100,000 PY for non-JIA. The risk of cancer associated with biologics-naive JIA was elevated (hazard ratio 2.8, 95% CI 0.9–8.3). The JIA cohort had a significantly elevated SIR of 4.0 (95% CI 2.6–6.0); the non-JIA cohort SIR was not significantly above SEER rates (SIR 1.4, 95% CI 0.6–2.6).

Conclusion

We found a nearly 3-fold increased risk of cancer in biologics-naive JIA patients, which approached significance despite the small number of outcomes. This finding suggests an elevated underlying risk of cancer in this disease population.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. ADDITIONAL DISCLOSURES
  10. REFERENCES

Juvenile idiopathic arthritis (JIA) is the most common pediatric rheumatic disease and is among the most prevalent chronic childhood illnesses. The prevalence of JIA has been estimated to be between 7 and 400 per 100,000 children (1). The International League of Associations for Rheumatology classification criteria for JIA require that signs and symptoms be present before age 16 years and persist for at least 6 weeks, and that other conditions are excluded (2). Besides chronicity and age at onset, the heterogeneous conditions included in JIA are unified only by being immune mediated. As such, the treatment for these conditions, especially in more severe cases, is often similar despite the differences in presentation, clinical course, and outcomes (3).

It is well recognized that the risk of some cancers, specifically lymphoma, is increased in adult rheumatoid arthritis (RA) patients, and an association between the use of tumor necrosis factor (TNF) antagonists and malignancy in adult RA has long been suspected (4). Recently, concerns have been raised about the possibility of an increased risk of cancer in JIA patients treated with such agents (5–9). The increased cancer risk in adult RA is believed to be partly related to the immune dysregulation that characterizes the disease. Additional evidence suggests that the greater the inflammatory burden of RA, the greater the risk of cancer (10, 11). Because the patients with the most severe RA require the most aggressive interventions, it is often unclear whether the increased risk results from the disease itself or its treatment. Regardless, any evidence of risk associated with RA treatment is properly interpreted in the context of this underlying risk (12).

JIA is also characterized by immune system dysfunction; therefore, a similarly increased underlying risk of malignancy is plausible. Whereas several studies have demonstrated an increased risk of cancer in the biologics-naive RA population, few have examined the risk of malignancy in patients with JIA who have never received anti-TNF therapy. A recent population-based study from Sweden found an increase in the risk of cancer overall and of lymphoproliferative malignancies among patients with JIA diagnosed between 1987 and 2007, but not earlier; the reasons for the different results across time were not clear (13). To investigate cancer risk among JIA patients in the absence of TNF inhibitors, we conducted a retrospective cohort study to quantify the relative risk of developing malignancy among biologics-naive patients with JIA in the US compared to matched patients without JIA.

Significance & Innovations

  • Juvenile idiopathic arthritis (JIA) appears to be associated with an increased risk of cancer.

  • The elevated risk of cancer was found primarily in JIA patients treated with methotrexate, which could indicate a link between methotrexate and cancer or severity of JIA and cancer.

  • The risk of cancer found in JIA patients treated with biologic agents should be evaluated in the context of the elevated risk associated with JIA itself.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. ADDITIONAL DISCLOSURES
  10. REFERENCES

Data source.

All data were obtained from the PharMetrics Patient-Centric Database. At the time of the study, this database contained fully adjudicated medical and pharmaceutical claims for >60 million commercially insured patients from >95 different health plans across the US from 1995–2009. The database includes the inpatient and outpatient diagnoses (International Classification of Diseases, Ninth Revision [ICD-9]), procedures (Current Procedural Terminology, fourth edition and Centers for Medicare and Medicaid Services Healthcare Common Procedure Coding System), and dispensed prescription drugs. Data on patient demographics (e.g., age, sex, US geographic region), insurance plan characteristics, and dates of enrollment are also available. Data for each patient are maintained longitudinally, with an average enrollment duration of ∼2 years.

Cohort definitions.

We assembled separate JIA and non-JIA cohorts from the PharMetrics Patient-Centric Database. The JIA cohort included both incident and prevalent cases of JIA.

The classification criteria for JIA require at least 6 weeks of active disease for an accurate diagnosis. Therefore, to qualify for the JIA cohort, patients were required to have at least 3 claims with any ICD-9 code 714 (RA and other inflammatory polyarthropathies), where the first 2 occurred at least 6 weeks apart or at least 2 such claims occurred at least 6 weeks apart in addition to at least 1 claim for a traditional disease-modifying antirheumatic drug, including hydroxychloroquine, sulfasalazine, auranofin, azathioprine, methotrexate, leflunomide, penicillamine, cyclosporine, and cyclophosphamide. Meeting the age criterion required either a claim with the specific diagnosis for JIA (ICD-9 code 714.3) at any age or a first claim for RA (ICD-9 code 714.xx) at <16 years of age.

Non-JIA patients were identified from among those with no claims for RA or JIA at any time. Up to 20 non-JIA patients were matched to each JIA patient on age, sex, US geographic region, insurance plan, duration of enrollment, and cohort entry date.

For JIA patients, cohort entry was defined as the latest of the following: date on which all criteria for the cohort were met, January 1, 1998, or 90 days following enrollment. Cohort entry for non-JIA patients was the same as their corresponding matched JIA patient. All patients were required to have at least 90 days of continuous enrollment both prior to and following cohort entry. The first 90 days after cohort entry were considered part of the baseline period and were not included in the eligible outcome period. Therefore, the followup period extended from 91 days after cohort entry until the earliest of the following: initiation of treatment with a biologic agent, disenrollment from the health plan, or the end of the study data (January 31, 2008). Death could not be identified reliably in this database, but patients who die should be disenrolled from their health plan shortly thereafter, and hence should have followup censored appropriately by the disenrollment criterion. Gaps in enrollment of ≤1 month were permitted during the study period.

We excluded any patient who had any diagnosis of cancer, including nonmelanoma skin cancer (NMSC), before or within 90 days after cohort entry, treatment with a biologic medication used in the treatment of JIA or other rheumatic disease prior to cohort entry, or a diagnosis of genetic syndromes or comorbid conditions with a known increased risk of cancer (i.e., familial polyposis syndrome, albinism, ataxic telangiectasia, Fanconi's syndrome, Bloom's syndrome, Down syndrome, organ transplant, human immunodeficiency virus, or neurofibromatosis).

Outcomes.

The primary outcome for this study was malignant cancer, excluding NMSC and carcinoma in situ. The secondary outcome was all malignancies, including NMSC and carcinoma in situ. For all patients with at least 1 claim indicating any neoplasm other than those noted as benign or of uncertain behavior from 91 days after cohort entry through the end of the followup period, we produced claims profiles listing in chronological order all diagnoses, procedures, service dates, the provider's specialty, and dispensed drugs during the patient's followup period. To blind the cohort assignment, the diagnosis and drug records that were clearly linked to the JIA diagnosis were removed from each profile. Three clinicians blinded to the cohort assignment, including a pediatric oncologist, reviewed the claims profiles to determine which cases were possible cancers and probable or highly probable cancers, based on whether the pattern of care was consistent with a new diagnosis of cancer. The reviewers also coded the cancer type and date of cancer diagnosis and indicated whether the patient received cancer surgery, radiation, or chemotherapy. Some profiles showed clear indications that the patient should have been disqualified from the cohort, either because of an error in the recorded age (where the profile was clearly that of an elderly person and not a child or adolescent as indicated by the year of birth) or because the patient's pattern of care indicated an evaluation for suspected cancer that began prior to cohort entry and ended in a cancer diagnosis early during the followup period. Such patients were flagged during the profile review and removed from the cohort. The reviewers resolved all disagreements through discussion.

Statistical analysis.

All data management and analyses were conducted using SAS, version 9.1. Baseline characteristics were compared between cohorts using chi-square statistics. For each cohort, we tabulated the number of all probable or highly probable cases of cancer, as well as specific cancer types, from which we calculated incidence rates and 95% confidence intervals (95% CIs). We reported the rates of specific cancer types only when at least 2 cases were found in at least 1 of the cohorts. In addition to internal comparisons (described below), we also determined whether the observed number of cases were consistent with expectations, using background rates from the US Surveillance, Epidemiology, and End-Results (SEER) program. Standardized incidence ratios (SIRs) with 95% CIs for all cancers and lymphomas were calculated for the JIA and non-JIA cohorts, adjusting for their age and sex composition.

The cancer risk in JIA and non-JIA patients was compared using Cox proportional hazards regression. The covariates in this analysis included age at cohort entry, sex, US geographic region, year of cohort entry, and duration of enrollment prior to cohort entry.

Because specific treatments for JIA were expected to show a strong association with disease status (JIA versus no JIA), the models did not include the treatment variables as covariates. Adjusting for the treatment associated almost exclusively with JIA may remove some of the effect actually attributable to JIA. For example, the use of methotrexate likely would indicate patients with greater disease severity; any effect of this variable may therefore derive from either drug exposure or more severe JIA. A second model was constructed with the JIA cohort separated into patients with versus patients without baseline methotrexate use. For each model, each of the covariates was first examined in a univariate analysis. The final model was adjusted for any covariates that modified the hazard ratio (HR) for JIA by at least 10%.

For the secondary outcome of all malignancies, including NMSC and carcinoma in situ, we found the number of cases in total and by type and calculated incidence rates and 95% CIs for all cancers and for each type with at least 2 cases found within a cohort. Cox modeling was performed in a parallel manner to the primary outcome analysis.

We conducted 2 additional sensitivity analyses. The first, in an effort to further remove any patients with prevalent cancer, excluded patients with <6 months of continuous enrollment before cohort entry. The second included all cases of cancer that the reviewers categorized as possible, as well as probable or highly probable.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. ADDITIONAL DISCLOSURES
  10. REFERENCES

A total of 6,631 patients with at least 1 diagnosis of JIA and 131,731 patients without JIA were initially identified from the PharMetrics Patient-Centric Database (Table 1). JIA patients were most often excluded for having insufficient followup time after the JIA diagnosis; the relatively high turnover from health plan enrollment might reflect changes in health insurance prompted by this new diagnosis. Non-JIA patients were most often excluded when their matched JIA patient was excluded or because of differing enrollment periods from the matched patient. The final cohort included 3,605 JIA patients and 37,689 patients without JIA, for an average of 10.5 comparators for each JIA patient.

Table 1. Patient selection*
Reason for exclusion (all patients)JIA (n = 6,631)Non-JIA (n = 131,731)
  • *

    Values are the number (percentage). JIA = juvenile idiopathic arthritis; RA = rheumatoid arthritis; NA = not applicable.

<2 RA/JIA diagnoses214 (3.2)NA
Age at first RA diagnosis ≥16 years with no JIA claim306 (4.6)NA
<90 days continuous enrollment prior to index date225 (3.4)261 (0.2)
<90 days continuous enrollment after index date1,126 (17.0)247 (0.2)
Cancer diagnosed prior to index date100 (1.5)239 (0.2)
Cancer diagnosed within 90 days after index date25 (0.4)60 (0.0)
Prior biologic RA/JIA treatment471 (7.1)4 (0.0)
Diagnosis of genetic syndrome associated with increased cancer risk76 (1.1)218 (0.2)
Prior transplant4 (0.1)12 (0.0)
Cancer patient excluded during profile review1 (0.0)5 (0.0)
Non-JIA match for excluded JIA patientNA50,646 (38.4)
Non-JIA patient enrollment period differs from matched JIA patientNA42,350 (32.1)
JIA patient with no matches after exclusions478 (7.2)NA
Patients included3,605 (54.4)37,689 (28.6)

Characteristics of JIA patients and comparators.

Most patients were between the ages of 5 and 17 years; the mean ± SD age at cohort entry was 11.1 ± 4.7 years for JIA patients and 11.4 ± 4.5 years for those without JIA (Table 2). The cohorts were predominantly female, with the greatest proportion of patients from the Midwest, followed by the Northeast. Most patients were covered by a preferred provider organization or health maintenance organization. More than 60% of the patients had a year or more of enrollment in the database before cohort entry, although the JIA patients were more likely than comparators to have only 3–6 months of prior enrollment and less likely to be enrolled for >2 years.

Table 2. Baseline characteristics of JIA and non-JIA patients*
Patient characteristicsJIA (n = 3,605)Non-JIA (n = 37,689)
  • *

    Values are the number (percentage). JIA = juvenile idiopathic arthritis; IV = intravenous; IM = intramuscular.

  • Followup period began 91 days after cohort entry.

Age at cohort entry, years  
 0–4405 (11.23)3,320 (8.81)
 5–111,412 (39.17)14,494 (38.46)
 12–171,528 (42.39)17,082 (45.32)
 ≥18260 (7.21)2,793 (7.41)
Sex  
 Male1,027 (28.49)10,669 (28.31)
 Female2,578 (71.51)27,020 (71.69)
Geographic region  
 Northeast885 (24.55)10,330 (27.41)
 South586 (16.26)6,117 (16.23)
 Midwest1,565 (43.41)15,607 (41.41)
 West569 (15.78)5,635 (14.95)
Duration of enrollment prior to cohort entry, months  
 3 to <6546 (15.15)4,781 (12.69)
 6 to <9479 (13.29)4,817 (12.78)
 9 to <12431 (11.96)4,504 (11.95)
 12 to <18641 (17.78)6,520 (17.30)
 18 to <24432 (11.98)4,530 (12.02)
 ≥241,076 (29.85)12,537 (33.26)
Alkylating agent treatment prior to followup1 (0.03)2 (0.01)
Methotrexate treatment prior to followup1,009 (27.99)5 (0.01)
IV/IM or extended use corticosteroid treatment prior to followup87 (2.41)24 (0.06)
Alkylating agent treatment during followup1 (0.03)2 (0.01)
Methotrexate treatment during followup881 (24.44)7 (0.02)
IV/IM or extended use corticosteroid treatment during followup78 (2.16)22 (0.06)

Treatment with alkylating agents was rare in both cohorts; in total, 3 patients received such medications at baseline and during followup (prior to a cancer event for patients judged to have cancer). Methotrexate use occurred almost exclusively in the JIA cohort; 28.0% of JIA patients received methotrexate during the baseline period and 24.4% during followup versus only 0.01% of comparator patients at baseline and 0.02% during followup. Parenteral or extended-use corticosteroids were also more common in the JIA cohort, although they were still seen with some frequency in comparators, with ∼2% of JIA patients and ∼0.1% of non-JIA patients having this exposure in both the baseline and followup periods. The total duration of followup was slightly longer for the non-JIA cohort versus the JIA cohort, with medians of 19 and 16 months, respectively.

Cancer outcomes.

A total of 200 patients, 32 (0.9%) in the JIA cohort and 168 (0.5%) in the comparator cohort, had an ICD-9 code suggesting a potential cancer outcome. The reviewers determined that 3 cases of cancer (all in the non-JIA cohort, 1.8% of the reviewed cases in that cohort) existed before cohort entry, evidenced by benign tumor codes linked to a cancer evaluation at baseline followed by malignant codes after cohort entry, and therefore these patients were excluded from the cohorts. The reviewers judged 3 additional patients to be adults (1 JIA [3.1% of reviewed cases] and 2 non-JIA [1.2% of reviewed cases]) based on several claims for adult diagnoses; they were also excluded. Across both cohorts, 22 patients were judged to have highly probable cancers, including 4 cervical carcinomas in situ and 1 NMSC. Probable cancers were seen in 6 patients, including 2 NMSCs, and possible cancers were seen in 34 patients, of which 1 was a cervical carcinoma in situ and 25 were NMSC. The remaining 132 patients were judged as unlikely to have had cancer.

Therefore, a total of 21 patients met our primary outcome definition (Table 3), 4 (0.1%) in the JIA cohort and 17 (0.05%) among the non-JIA comparator cohort. During a total of 5,974 person-years in the JIA cohort and 73,395 person-years in the non-JIA cohort, the overall incidence rates for cancer were 66.95 (95% CI 1.34–132.57) and 23.16 (95% CI 12.15–34.17) events per 100,000 person-years, respectively. Both of these estimates were higher than the SEER age-standardized incidence rate for any cancer of 16.61 (95% CI 16.37–16.86) per 100,000 person-years, although the non-JIA rate was not significantly different from the SEER rate (SIR 1.39, 95% CI 0.61–2.57) (Table 4). The JIA patients, in contrast, showed a statistically significant elevation in rate compared to the SEER rate (SIR 4.03, 95% CI 2.56–5.99).

Table 3. Primary outcome of cancer events during followup, excluding nonmelanoma skin cancer and carcinoma in situ*
Cancer typeJIA (n = 3,605 [5,974 PY])Non-JIA (n = 37,689 [73,395 PY])
No. (%)IR/100,000 PY (95% CI)No. (%)IR/100,000 PY (95% CI)
  • *

    Incidence rates (IRs) are shown only when at least 2 events of a given cancer type were found in the cohort. JIA = juvenile idiopathic arthritis; PY = person-years; 95% CI = 95% confidence interval.

Any cancer  
 Probable or highly probable cancer4 (0.11)66.95 (1.34–132.57)17 (0.05)23.16 (12.15–34.17)
Hematologic cancer    
 Any hematologic cancer2 (0.06)33.48 (0.00–79.87)7 (0.02)9.54 (2.47–16.60)
 Lymphoma2 (0.06)33.48 (0.00–79.87)2 (0.01)2.72 (0.00–6.50)
 Leukemia0 (0.0)5 (0.01)6.81 (0.84–12.78)
Skin cancer    
 Any skin cancer0 (0.0)1 (0.00)
 Melanoma0 (0.0)1 (0.00)
Solid cancer    
 Any solid cancer2 (0.06)33.48 (0.00–79.87)9 (0.02)12.26 (4.25–20.27)
 Brain0 (0.0)3 (0.01)4.09 (0.00–8.71)
 Bone0 (0.0)3 (0.01)4.09 (0.00–8.71)
 Gum1 (0.03)0 (0.0)
 Soft tissue0 (0.0)2 (0.01)2.72 (0.00–6.50)
 Tendon sheath1 (0.03)0 (0.0)
 Thyroid0 (0.0)1 (0.00)
Table 4. SEER incidence rates and SIRs for JIA and non-JIA cohorts compared to SEER rates*
Cancer typeSEER incidence rate/ 100,000 person-years (95% CI)SIR for JIA (95% CI)SIR for non-JIA (95% CI)
  • *

    Surveillance, Epidemiology, and End-Results (SEER) rates were standardized to the age distribution in each group. Standardized incidence ratios (SIRs) were calculated only when at least 2 events of a given cancer type were found in the cohort. JIA = juvenile idiopathic arthritis; 95% CI = 95% confidence interval.

Any cancer   
 Probable or highly probable cancer16.61 (16.37–16.86)4.03 (2.56–5.99)1.39 (0.61–2.57)
Hematologic cancer   
 Lymphoma2.26 (2.17–2.35)14.81 (7.62–25.67)1.21 (0.02–4.45)
 Leukemia4.53 (4.41–4.66)1.50 (0.25–4.03)
Solid cancer   
 Brain2.59 (2.49–2.68)1.58 (0.11–5.06)
 Bone0.92 (0.86–0.98)4.44 (0.31–14.24)
 Soft tissue1.05 (0.99–1.12)2.60 (0.05–9.58)

In general, there were too few cases of any specific cancer type to allow separate analyses. Although the numbers were limited, the SIR for JIA was significantly elevated for lymphoma at 14.81 (95% CI 7.62–25.67) compared to the SEER rate, but not in the non-JIA cohort, suggesting that lymphoma may be among the primary drivers of any increased risk of cancer in JIA.

All but one of the cases of cancer judged probable or highly probable were associated with additional claims for chemotherapy, cancer-related surgery, and/or radiation for treatment. The single exception was a soft tissue cancer that was deemed probable. No claims were found following the initial diagnosis, suggesting that the patient disenrolled from the health insurance plan shortly afterward.

In the primary univariate Cox models (Table 5), JIA was associated with a nearly 3-fold increased risk of cancer (HR 2.81, 95% CI 0.94–8.34). Patients ages 12–17 years were at decreased risk compared to patients ages ≥18 years (HR 0.18, 95% CI 0.05–0.68). None of the covariates tested changed the HR for JIA by at least 10%, so the multivariate model was the same as the univariate model for JIA.

Table 5. Cox proportional hazards models of any cancer (probable or highly probable, primary outcome)*
PredictorUnivariate models hazard ratio (95% CI)
  • *

    No covariates were found that modified the hazard ratio for juvenile idiopathic arthritis (JIA) by at least 10%, so the final model contained only JIA. 95% CI = 95% confidence interval.

Cohort 
 JIA2.81 (0.94–8.34)
 Non-JIARef.
Age, years 
 0–40.34 (0.06–1.88)
 5–110.40 (0.13–1.28)
 12–170.18 (0.05–0.68)
 ≥18Ref.
Sex 
 MaleRef.
 Female1.24 (0.45–3.38)
Region 
 Northeast1.21 (0.35–4.15)
 South0.79 (0.18–3.52)
 Midwest0.63 (0.18–2.15)
 WestRef.
Enrollment duration prior to cohort entry, months 
 3 to <6Ref.
 6 to <91.01 (0.20–4.98)
 9 to <120.67 (0.11–4.04)
 12 to <180.75 (0.15–3.73)
 18 to <240.39 (0.04–3.77)
 ≥241.63 (0.44–6.06)

Of the 4 JIA patients who developed cancer during followup, 2 received methotrexate during baseline. The model including separate indicators for JIA with and without methotrexate at baseline obtained an HR for JIA without methotrexate that was lower and further from significance than all JIA combined (HR 1.89, 95% CI 0.44–8.17). In contrast, the HR for JIA with methotrexate compared to patients without JIA (HR 5.47, 95% CI 1.26–23.70) was considerably higher than all JIA combined and was significant at P < 0.05.

Including NMSC and carcinoma in situ as defined by our secondary outcome added no additional cases in the JIA cohort and 7 cases (3 NMSCs and 4 cervical cancers in situ) in the comparator cohort. The Cox models of this outcome showed a weaker effect of JIA compared to the primary outcome, with an HR of 2.02 (95% CI 0.70–5.82) for the model of all JIA patients combined.

The sensitivity analysis adding possible cancers to the primary outcome showed a similar pattern in the comparison of JIA to non-JIA, although the total number of events in each group increased (1 additional case in JIA and 7 in non-JIA). When these possible events were added, even the SIR for non-JIA was significantly elevated compared to the SEER rate (SIR 1.97, 95% CI 1.00–3.33). The Cox model of all JIA patients combined estimated the HR for JIA to be 2.48 (95% CI 0.95–6.51).

Similarly, rerunning the analysis including only patients with at least 6 months of baseline enrollment instead of a minimum of 3 months found comparable results to the primary analyses. Excluding patients with <6 months of enrollment dropped 1 event from the JIA cohort and 2 from the comparator cohort.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. ADDITIONAL DISCLOSURES
  10. REFERENCES

This study of cancer among biologics-naive patients with JIA found a 4-fold increased rate of malignancy compared to general population rates obtained from the SEER program. In contrast, patients without JIA had a similar rate to the SEER rate, suggesting that the cancer events identified through claims data may have good validity. Based on our understanding of the role of immunosurveillance and disrupted immune function in cancer development, along with the findings from studies of an increased risk of some cancers in adult RA (12, 14, 15), an increased background risk of cancer in JIA can be expected. One prior study examining the incidence of cancer in various rheumatic conditions found no significant increase in risk with JIA, although the number of juvenile patients enrolled was small (15).

When the cancer events judged as only possible were included in the outcome, the rate in patients without JIA was elevated compared to the SEER rate, suggesting that many of these events were not actual malignancies. This supports our inclusion of the probable and highly probable cancers, but not possible cancers, in the primary outcome. Also, the similar findings between the 3- and 6-month baseline enrollments suggest that the 3-month baseline period was adequate to ascertain new-onset cancers rather than prevalent cancers. Children and adolescents with cancer typically will not disappear from the health care system for longer than 3 months.

In the full study population, biologics-naive JIA was associated with more than double the risk of cancer compared to matched general population comparators, with confidence limits falling close to statistical significance at P < 0.05. A Swedish cohort study demonstrated similar findings, with an estimated relative risk of 2.3 for JIA compared to the general population over a recent 20-year period (13). However, the analysis separating the JIA patients into those with and without baseline methotrexate use showed a significant increase in cancer risk only with methotrexate. Methotrexate was used almost exclusively in the JIA cohort and may be more likely given for severe cases of JIA. The different results by methotrexate exposure might therefore indicate greater cancer risk with more severe JIA and/or use of methotrexate. Interestingly, the Swedish study that found an increased risk of cancer in JIA from 1987–2007 found no significant increase in the prior 20 years, before methotrexate was available. That study did not examine the association between methotrexate exposure and cancer risk, but concluded that such an association cannot be ruled out (13).

Although the present study cannot separate the risk of cancer associated with JIA itself from any increased risk from methotrexate, especially given the restriction in the analysis to baseline methotrexate use, it was designed to evaluate the risk of cancer among patients with JIA who may be eligible for anti-TNF therapies. Anti-TNF therapies are typically used when conventional treatments, especially methotrexate, have failed. Therefore, it is reasonable to assess the cancer risk in methotrexate-treated JIA patients as a baseline for understanding any excess risk conferred by an anti-TNF agent. Since anti-TNF medications have been suspected of increasing the risk of cancer, the possible increase in risk associated with methotrexate use might lead some practitioners to avoid giving combination methotrexate and anti-TNF treatment to patients with JIA. However, considering the low risk of cancer in this population even with prior methotrexate treatment, the high efficacy of methotrexate and anti-TNF combination therapy in treating JIA may outweigh the potential risks (16–18).

The limitations of this study include the use of claims data only, with no verification of JIA status or cancer outcomes from other sources such as medical records, although our restriction to patients with multiple evaluation and management claims, as opposed to claims for tests that might rule out the diagnosis entered, should have considerably increased the validity. We expect that the use of clinical reviewers, including a pediatric oncologist, to determine case status also reduced false-positives, although we cannot exclude the possibility of some outcome misclassification. Additionally, outcome determination was based on the clinical reviewers' expert judgment rather than explicit rules that might be used and tested by other reviewers in future studies. Information on race is unavailable in claims data; the inability to standardize based on race may have biased some of the SIR calculations to the extent that our patient populations did not have the same racial balance as the SEER populations.

The timing of new-onset cancer is difficult to determine; the present analysis restricted outcomes to those diagnosed more than 90 days after cohort entry to avoid considering a case of cancer diagnosed very shortly after the JIA diagnosis as associated with JIA. Still, for some cases, either a longer or shorter interval may have been more appropriate, as evidenced by the 3 non-JIA cases with cancer evaluations that crossed over the cohort entry date, leading to their exclusion due to having prevalent cancer only after examination of their detailed profiles. Patients with JIA are more likely to receive closer clinical monitoring than those without JIA, which may speed identification of new cancer, resulting in an overestimation of the relative risk for JIA. The post hoc exclusion of 3 patients from the cohort based on their apparent status as an older adult rather than a child or adolescent that was discovered only through examination of detailed claims profiles points to a potential source of bias in the study: the possibility of other adults in the denominator of each cohort with erroneous age data. However, the small proportions of reviewed cases affected by this issue and the lack of a clear imbalance in those proportions between groups (3.1% and 1.2% for JIA and non-JIA, respectively) suggest that any resulting bias should be relatively small. Finally, given the exclusion of nearly one-half of the JIA patients initially identified from the cohort, these results may not generalize to the full population of commercially insured JIA patients.

The present study is among the first to our knowledge to examine malignancy risk in biologics-naive JIA patients from a natural history of disease standpoint rather than looking for cancer as a potential adverse event associated with the treatment of JIA. The study sample was drawn from a very large patient population; yet, given the infrequency of JIA and the rarity of cancer in children, it would be difficult to find any data source large enough to make any definitive conclusions. We propose that the interpretation of the risk of cancer in JIA patients treated with biologic agents should be made with consideration for the potential increased risk associated with JIA and conventional treatment.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. ADDITIONAL DISCLOSURES
  10. REFERENCES

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. Nordstrom 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. Nordstrom, Mines, Gu, Mercaldi, Aquino, Harrison.

Acquisition of data. Gu, Mercaldi, Aquino.

Analysis and interpretation of data. Nordstrom, Mines, Gu, Mercaldi, Aquino, Harrison.

ROLE OF THE STUDY SPONSOR

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. ADDITIONAL DISCLOSURES
  10. REFERENCES

Wyeth/Pfizer participated in the study design, acquisition of data, analysis of data, interpretation of the data, the writing of the manuscript, and the decision to submit the manuscript for publication through the 4 coauthors employed by Wyeth/Pfizer. Wyeth/Pfizer approved the manuscript prior to submission.

ADDITIONAL DISCLOSURES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. ADDITIONAL DISCLOSURES
  10. REFERENCES

Drs. Nordstrom and Mercaldi are employees of United BioSource Corporation, which was contracted by and received funding from Wyeth/Pfizer for this study. Dr. Mines is an employee of HealthCore.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. ADDITIONAL DISCLOSURES
  10. REFERENCES