Injectable Tumor Necrosis Factor α Inhibitors and Outpatient Antimicrobial Use in Children With Rheumatic Diseases: Analyses of Prescription Claims From a Pharmacy Benefit Manager Database†
The content contained herein is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality.
To estimate the incidence and rate of outpatient antibiotic and antiviral medication use among children receiving methotrexate and/or an injectable tumor necrosis factor α (iTNFα) inhibitor (etanercept and/or adalimumab) and to compare these rates with those of a control population.
Data were obtained from a pharmacy benefit manager (PBM) database. Children were included if they had ≥1 prescription claim for an iTNFα inhibitor or methotrexate prescribed by a pediatric or adult rheumatologist between 2008 and 2010 and if they were age <18 years at the time of the claim. A control cohort of randomly selected children was generated from the PBM database. Poisson regression was used to compare antimicrobial rate ratios (RRs). Incidence rates and RRs were adjusted for age, sex, and prednisone exposure.
In total, 4,312 children were included. The adjusted RRs for antibiotic prescriptions among children receiving methotrexate monotherapy or iTNFα inhibitor and methotrexate combination therapy compared with the control cohort were 2.18 (95% confidence interval [95% CI] 1.92–2.47) and 2.12 (95% CI 1.79–2.50), respectively. The adjusted RRs for antiviral prescriptions among children receiving methotrexate monotherapy or iTNFα inhibitor and methotrexate combination therapy compared with the control cohort were 3.67 (95% CI 1.98–6.78) and 4.34 (95% CI 1.86–10.14), respectively. The RRs for the iTNFα inhibitor group were similar in magnitude. There was no significant difference in RRs between the medication exposure categories for either antibiotic or antiviral prescriptions.
Children receiving methotrexate and/or an iTNFα inhibitor had higher rates of antibiotic and antiviral use compared with the control cohort. Data sets with additional patient-level and disease-specific data are required to assess this association in more detail.
Since etanercept was approved by the US Food and Drug Administration (FDA) in 1999 for children with polyarticular juvenile idiopathic arthritis (JIA), tumor necrosis factor α (TNFα) inhibitors have been used with increasing frequency in children with rheumatic diseases. The subsequent FDA approval of adalimumab in 2008 further increased the availability of injectable TNFα (iTNFα) inhibitors for children. Medication claims data for children with JIA have suggested that the percentage of children receiving a TNFα inhibitor nearly tripled between 2005 and 2008 (an increase from 7–19%) and is approaching the percentage for adult rheumatoid arthritis (RA; 24%) ().
Given the central role of TNFα in innate immunity, concerns that TNFα inhibition increases the risk of infection have accompanied these medications since their initial use in RA. In patients with RA, the hospitalization rates for bacterial infection have been reported to be ∼2-fold higher among those receiving TNFα inhibitors compared with patients receiving nonbiologic disease-modifying antirheumatic drugs, with an ∼4-fold increase during the first 6 months of exposure to the TNFα inhibitor ([2, 3]). Recent US national Medicaid administrative claims data reported that children with JIA had a nearly 3-fold increased risk of hospitalization for severe bacterial infection relative to comparison cohorts of non-JIA patients (). In this same analysis, children with JIA receiving methotrexate and/or TNFα inhibitors did not have an increased risk of infection compared to children with JIA not exposed to these medications, suggesting that the increased infection risk may be more closely associated with the underlying disease rather than specific medication exposure.
Less is known about the risk of outpatient bacterial and viral infections associated with TNFα inhibitors, largely because of the lack of large, observational data sets with sufficient numbers of children receiving these medications. Current estimates of outpatient infections in children with JIA receiving TNFα inhibitors have been derived primarily from clinical trials that contained relatively small numbers of children, were frequently of short duration, and may not be generalizable to children not enrolled in clinical trials. Outpatient infections, particularly those requiring treatment with systemic antimicrobial therapy, may be important contributors to the burden of disease for children with JIA due to their associated morbidity and their impact on both direct and indirect health care costs through added clinic visits, days missed from school, and parent days missed from work. Therefore, the objective of this analysis was to use prescription claims data from a large pharmacy benefit manager (PBM) database to estimate the rates of outpatient antibiotic and antiviral use in children treated with an iTNFα inhibitor and/or methotrexate.
Box 1. Significance & Innovations
- These analyses are the first to describe rates of antibiotic and antiviral medication use in children receiving methotrexate and/or an injectable tumor necrosis factor α inhibitor.
- Children exposed to these medications had an ∼2-fold increased rate of antibiotic use and a 3–4-fold increased rate of antiviral use when compared to a control cohort.
Patients and methods
The cohort was derived from medication claims and patient enrollment data provided by CVS Caremark, a large PBM company with ∼45 million beneficiaries in the US. Children were included in the cohort if they had at least 1 prescription claim in the CVS Caremark database between January 1, 2008 and December 31, 2010 for etanercept, adalimumab, or methotrexate that was prescribed by a pediatric or adult rheumatologist and if they were age <18 years at the time of the claim. Children were also required to have a minimum of 1 year of eligibility within the PBM database in order to be included in the analysis. Because PBM databases do not universally collect International Classification of Diseases, Ninth Revision diagnostic code information or medical claims data associated with prescription claims, we anticipated that these criteria would limit the cohort to children receiving these medications for a pediatric rheumatic disease, primarily JIA.
Data from children were excluded if their date of birth could not be verified or if they had been exposed to mycophenolate mofetil, azathioprine, anakinra, leflunomide, and cyclosporine within the month prior to their initial methotrexate or iTNFα inhibitor claim because these medications could have confounded the infection risk. Children who received cyclophosphamide at any time during the study period were also excluded. Data from patients with a medical claim for an infused biologic agent (infliximab, abatacept, and rituximab) were not included because there were very small numbers of children with available medical claims and an even smaller number of children receiving these medications.
Children were categorized into the following 3 medication exposure categories: 1) iTNFα inhibitor monotherapy (a claim for etanercept or adalimumab and no claims for methotrexate), 2) methotrexate monotherapy (a claim for methotrexate and no claims for an iTNFα inhibitor), and 3) combination therapy (claims for methotrexate and an iTNFα inhibitor that overlapped for ≥30 days). Each group was further dichotomized by whether they had a claim for prednisone in the 60 days prior to their antimicrobial claim.
A control cohort of children was randomly selected from the CVS Caremark database. The control cohort was matched to the treatment cohort in a 2:1 ratio based on age, sex, insurance status, and geographic region. Each child in the control population had at least 1 year of eligibility in the PBM database between 2008 and 2010, during which they may or may not have had a medication claim. Children were excluded from the control group if they had missing date of birth and/or a medication claim from an adult or pediatric rheumatologist. The observation period for the control children included the total duration of their PBM database eligibility between 2008 and 2010 (e.g., a child could contribute a maximum of 3 years of observation if they were eligible for the entire study duration). The approval for this study was obtained from the Seattle Children's Hospital Institutional Review Board. All individually identifiable health information was protected in accordance with federal and state laws.
Descriptive statistics were generated for the children in this cohort by each treatment group, as specified above. The infection rates were calculated based on the numbers of claims for commonly used antibiotic and antiviral medications. Antimicrobial claims that overlapped or occurred up to 30 days after a claim for the medications of interest were included. To estimate the rate of outpatient infections per person-year of exposure for each medication category, the total number of days receiving therapy was estimated and summed for children in each of the treatment groups above. This period was defined as beginning with the first claim for the medication of interest (e.g., methotrexate) through the last claim. Patients were considered discontinued from the medications of interest if they had no additional claims for the medication and/or a ≥60-day gap in claims for the medication of interest (). Patients were censored from the analysis at the time of their last claim. In the case of combination therapy, the duration of exposure was calculated based on the medication the patients received for the longest duration. In the control cohort, exposure time was equivalent to the duration of eligibility.
The total number of antibiotic and/or antiviral claims for children within each exposure category was divided by the total years of medication exposure within each category to generate the infection rate per 100 person-years of exposure. These estimates were adjusted for the patients' age, sex, and prednisone exposure (yes/no; defined as a claim for prednisone within 60 days of the antimicrobial claim). Rate ratios (RRs) were calculated using Poisson regression and adjusted for age, sex, and prednisone exposure (yes/no), with robust SE estimates. The analyses were performed using SAS, version 9.2.
In total, 4,312 children with a claim for methotrexate and/or an iTNFα inhibitor and 8,624 controls were included in the analyses (Table 1). Most children (80%) had private insurance, which is consistent with the overall CVS Caremark population. In the methotrexate monotherapy group, 42% of children had a claim for a systemic corticosteroid (e.g., prednisone, prednisolone, or methylprednisolone), as did 44% of children in the methotrexate and iTNFα inhibitor combination therapy group and 22% of children in the iTNFα inhibitor monotherapy group. The overall prevalence rate of corticosteroid use for the overall CVS Caremark cohort was 6.5%. There were no children identified with claims for isoniazid or rifampin, indicating no episodes of an active or latent Mycobacterium tuberculosis infection in this cohort.
Table 1. Cohort characteristics*
|Age, years|| || || || |
|0–8||145 (20)||934 (33)||211 (27)||1,290 (30)|
|9–12||195 (27)||847 (30)||209 (27)||1,251 (29)|
|13–15||244 (34)||696 (25)||241 (31)||1,181 (27)|
|≥16||135 (19)||332 (12)||123 (16)||590 (14)|
|Female||455 (63)||2,002 (71)||566 (72)||3,023 (70)|
|Pediatric rheumatologistc||591 (82)||2,432 (87)||667 (85)||3,690 (86)|
|Adult rheumatologistd||616 (86)||2,192 (78)||659 (84)||3,467 (80)|
Incidence of antibiotic and antiviral claims
Children receiving methotrexate monotherapy composed the largest medication exposure group (n = 2,809), with 2,032 person-years of exposure. The iTNFα inhibitor monotherapy and iTNFα inhibitor and methotrexate combination therapy groups had 593 and 1,114 person-years of exposure, respectively. In the control group, 35% of children had ≥1 claim for an antibiotic medication, while 32% of children in the methotrexate monotherapy group, 36% of children in the iTNFα inhibitor monotherapy group, and 48% of children in the iTNFα inhibitor and methotrexate combination therapy group had ≥1 antibiotic claim. The unadjusted and adjusted incidences of antibiotic claims for each medication group are shown in Table 2.
Table 2. Incidence rate of antibiotic and antiviral prescriptions among children treated with methotrexate and/or an iTNFα inhibitor and a control cohort*
|Antibiotic|| || |
|Methotrexate and iTNFα||0.89 (0.74–1.05)||0.87 (0.72–1.04)|
|iTNFα||0.97 (0.83–1.14)||1.05 (0.90–1.23)|
|Methotrexate||0.95 (0.84–1.07)||0.89 (0.78–1.01)|
|Control||0.36 (0.35–0.38)||0.41 (0.37–0.45)|
|Antiviral|| || |
|Methotrexate and iTNFα||0.07 (0.04–0.12)||0.06 (0.03–0.11)|
|iTNFα||0.05 (0.03–0.14)||0.05 (0.02–0.10)|
|Methotrexate||0.06 (0.03–0.10)||0.05 (0.03–0.08)|
|Control||0.01 (0.01–0.01)||0.01 (0.01–0.02)|
The overall incidence rate of antiviral claims was lower than that for antibiotic claims. The percentage of children with ≥1 antiviral claim was 1% in the control group, 1.7% in the methotrexate monotherapy group, 2.2% in the iTNFα inhibitor monotherapy group, and 2.8% in the iTNFα inhibitor and methotrexate combination therapy group. The unadjusted and adjusted incidences of antiviral claims for each medication group are shown in Table 2.
Comparison of antibiotic and antiviral claims RRs
There was an ∼2-fold increase in antibiotic claims rates between each medication group relative to the control group (Table 3). The RRs for antibiotic claims were significantly higher for each of the medication exposure categories compared with the control group (P < 0.0001 for each). However, there were no significant differences between the RRs for either the iTNFα inhibitor monotherapy group or the iTNFα inhibitor and methotrexate combination therapy group compared with the methotrexate monotherapy group.
Table 3. Rate ratios for antibiotic and antiviral prescriptions among children treated with methotrexate and/or an iTNFα inhibitor*
|Antibiotic|| || |
|Methotrexate and iTNFα||2.44 (2.04–2.93)||2.12 (1.79–2.50)|
|iTNFα||2.7 (2.3–3.17)||2.58 (2.17–3.06)|
|Methotrexate||2.62 (2.30–2.99)||2.18 (1.92–2.47)|
|Antiviral|| || |
|Methotrexate and iTNFα||7.1 (3.69–13.64)||4.34 (1.86–10.14)|
|iTNFα||5.27 (2.38–11.67)||3.63 (1.62–8.12)|
|Methotrexate||5.69 (2.96–10.93)||3.67 (1.98–6.78)|
The RRs for antiviral claims followed a similar trend to the antibiotic claims (Table 3). There was a 3–4-fold increase in antiviral claims rates for the medication groups when compared with the control cohort. Each medication exposure category had a significantly higher RR than the control group (P < 0.002 for each comparison), and the RRs for the iTNFα inhibitor monotherapy group and the iTNFα inhibitor and methotrexate combination therapy group were not significantly different from the methotrexate monotherapy group.
These analyses suggest that children receiving methotrexate and/or iTNFα inhibitors have an ∼2-fold increased rate of outpatient antibiotic prescriptions and an ∼3–4-fold increased rate of outpatient antiviral therapy compared with the control cohort. This risk was not significantly different between children receiving either of these medications alone or in combination, and adjustment for prednisone exposure, age, and sex did not significantly change this risk. To our knowledge, these results represent the first estimates of the rates of outpatient systemic antimicrobial use in children receiving these medications.
While the overall risk of outpatient infections requiring systemic antimicrobial therapy has not been well described for children receiving iTNFα inhibitors alone or in combination with methotrexate, this risk has been explored in more detail among patients with RA. Prior investigations of the risk of mild infections associated with methotrexate monotherapy have been equivocal, with some studies indicating no increase in risk and others indicating a small increase in risk ([6, 7]). An analysis of patients with RA included in the Consortium of Rheumatology Researchers of North America (CORRONA) registry reported an overall outpatient infection rate of 31.2 events per 100 person-years. However, infections in this cohort were identified by a physician's report and limited chart review and were not limited to infections requiring antimicrobial therapy. Upper respiratory infections and pneumonias composed the largest proportion of these infections (13.6 events per 100 person-years) (). The incidence rates calculated in our current study were much smaller than those reported from the CORRONA registry, which was likely due to the restriction of the analyses to events requiring antimicrobial therapy, suggesting that the majority of infections in patients receiving these medications do not require antimicrobial therapy. This same CORRONA study also reported that daily oral prednisone doses <10 mg were not associated with an increased risk of infection. These results may help explain why adjustment for prednisone use in the current study did not significantly impact antimicrobial prescription rates, given our hypothesis that the majority of children would be receiving relatively low doses of prednisone.
A recent meta-analysis assessing the risk of nonserious infections (defined as infections not requiring intravenous antibiotics) in patients with RA estimated RRs of 1.194 (95% confidence interval [95% CI] 0.909–1.162) and 1.028 (95% CI 1.014–1.406) for adalimumab and etanercept, respectively, compared to persons with RA not receiving these medications (). The RRs calculated for our PBM cohort were higher, likely reflecting our use of a healthy cohort as the control population.
Although the CORRONA studies described above did not estimate separate risks for bacterial and viral infections, studies have been equivocal regarding the risk of zoster infections in RA patients receiving TNFα inhibitors ([9-12]). A study from the German biologics registry for RA found an ∼2-fold elevated risk of herpes zoster among patients receiving monoclonal TNFα inhibitors (adalimumab and infliximab; hazard ratio 1.82 [95% CI 1.05–3.15]) but no increased risk associated with etanercept exposure after controlling for disease severity, age, and glucocorticoid use (). Another recent analysis of claims data reported no difference in the risk of zoster infections between RA patients receiving and not receiving TNFα inhibitors (adjusted hazard ratio 1.00 [95% CI 0.77–1.29]) and no difference in rates between patients exposed to different TNFα inhibitors (). While our current PBM cohort had an ∼3–4-fold increase in antiviral use compared with controls, we were unable to investigate the types of infections or suspected infections for which the antivirals were prescribed because of limitations in the PBM data.
While this study represents the first estimates of the rate of outpatient systemic antimicrobial use in children exposed to methotrexate and/or iTNFα inhibitors, the use of PBM data has several limitations. Specifically, we were not able to determine the specific type of infection being treated, whether the prescription was provided for a documented or suspected infection, and/or whether there was a separate indication for the medication, such as prophylaxis for a dental procedure or surgery. We were also unable to determine whether the child subsequently took any of the medications of interest after the prescription was filled. Because data regarding the primary diagnosis were not available, we were also unable to determine whether there were additional disease or patient-specific characteristics that contributed to the risk of receiving an antibiotic or antiviral medication, although we hypothesized that most children in this cohort had JIA because this is the most common pediatric rheumatic disease and the most common indication for methotrexate and iTNFα inhibitors. Although the abovementioned study from CORRONA also reported a significant association between RA disease activity and infection risk, with patients with higher disease activity having a higher risk of infection, we were not able to assess this relationship or confounding by indication in our current analyses due to limitations in the data set (). Last, our current study did not assess infections that did not require antimicrobial use, which likely also significantly contribute to the burden of disease in these children.
The reasons for the increase in antibiotic and antiviral medication claims noted in the current study will require additional exploration in data sets that include more disease-specific and patient-level data to determine whether these differences reflect true infections versus physicians tending to prescribe antimicrobials more commonly to children receiving immunosuppressants and to differentiate the infection risk associated with medication exposure from the risk associated with the underlying disease.
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. Ringold 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. Ringold, Girdish, Wallace, Sullivan.
Acquisition of data. Ringold, Girdish.
Analysis and interpretation of data. Ringold, Grant, Girdish, Wallace, Sullivan.
Ms Grant is an employee of Axio Research. Ms Girdish is an employee of CVS Caremark.