Cost-effectiveness of biologics in polyarticular-course juvenile idiopathic arthritis patients unresponsive to disease-modifying antirheumatic drugs


  • The opinions, results, and conclusions reported in this paper are those of the authors and are independent from the funding sources. No endorsement by the Ontario Ministry of Health and Long-Term Care is intended or should be inferred.



Juvenile idiopathic arthritis (JIA) is the most common chronic pediatric rheumatic disease and can have long-term effects leading to disability in adulthood. Biologics are a new class of drugs increasingly used to treat JIA. The primary study objective was to determine the incremental costs of biologics per additional responder compared to conventional treatment (methotrexate).


A separate decision model was created for etanercept, infliximab, adalimumab, and abatacept. The study population consisted of polyarticular-course JIA patients with a prior inadequate response or intolerance to disease-modifying antirheumatic drugs (DMARDs). The effectiveness measure was the proportion of patients who had a treatment response at 1 year according to the American College of Rheumatology (ACR) Pediatric 30 (Pedi 30) improvement criteria. Direct and indirect costs were calculated in 2008 Canadian dollars. Incremental cost-effectiveness ratios and 95% confidence intervals (95% CIs) were calculated for each biologic agent using probabilistic sensitivity analyses.


The additional costs per additional ACR Pedi 30 responder at 1 year were $26,061 (95% CI $17,070, $41,834), $46,711 (95% CI $30,042, $75,787), $16,204 (95% CI $11,393, $22,608), and $31,209 (95% CI $16,659, $66,220) for etanercept, adalimumab, abatacept, and infliximab, respectively.


Biologics are more effective than methotrexate in achieving a short-term response in JIA patients with prior inadequate responses to DMARDs; however, this comes at a high annual cost. Adequate long-term data with respect to both safety and effectiveness are not currently available, nor are utility estimates. Such data will be important to estimate value for money for treating JIA with biologic drugs over the long term.


Juvenile idiopathic arthritis (JIA) is the most common chronic pediatric rheumatic disease (1). Prevalence estimates vary from 7–400 per 100,000 children (2). Severe disease is characterized by pain and stiffness, joint damage, growth abnormalities, and functional disability (1). Polyarticular JIA is one of the more severe subtypes, where 5 or more joints are affected within the first 6 months of illness (3). In approximately 50% of cases, inflammation and disability continue into adulthood (4). Affected children may be limited in activities and miss school due to symptoms or medical appointments (5). Impacts on other family members include lost work time by parents or the need to make significant changes in work status to be available to provide care (6).

Treatment of JIA includes pharmacotherapy, physical and occupational therapy, and psychosocial support. Medications include nonsteroidal antiinflammatory drugs (NSAIDs), glucocorticoids, and disease-modifying antirheumatic drugs (DMARDs). Long-term studies with the DMARD methotrexate (MTX) have shown that it is effective in 60–70% of JIA patients, with effectiveness measured by the American College of Rheumatology (ACR) Pediatric 30 (Pedi 30) improvement criteria (7, 8). The ACR Pedi 30 criteria are defined as an improvement of ≥30% in at least 3 of 6 core variables and worsening of ≤30% in not more than one variable (core variables include physician global severity assessment, parent or patient global assessment of overall well-being, number of active joints, number of joints with motion limitation, erythrocyte sedimentation rate/C-reactive protein level, and functional assessment) (8). The ACR Pedi 30 criteria are a common efficacy measure in clinical trials. Other outcomes include rates of disease flare and remission (9).

Complete disease control with medications is elusive for many JIA patients; MTX may not be effective in achieving control in some patients, even at high doses (7). In a study of polyarthritis patients, disease remained active over 63% of a 5-year period (10). Growth retardation and osteoporosis may also occur secondary to JIA or treatment with glucocorticoids (1, 11).

Biologics are newer DMARDs that have been developed based on an enhanced understanding of inflammation. Their introduction has resulted in improved outcomes, permitting better control of disease in patients refractory to previous medications (12). The most common adverse events reported with biologics are injection site reactions and infections (13).

While biologics may improve outcomes for JIA patients, they are expensive. Economic evaluations weigh the additional costs of new interventions against their added health benefits compared to standard care. As children receive medication during physiologic and psychological development, and because they may need treatment for long periods, it is important to evaluate the use of biologics in children. In a climate of limited health budgets, health economic evidence can inform decision making. The study objective was to determine the incremental costs of treatment with the biologic drugs most commonly indicated for JIA (etanercept, adalimumab, abatacept, and infliximab) compared to continuing conventional treatment (MTX) per responder gained in patients with a prior inadequate response or intolerance to MTX or other DMARDs.


Study design.

A cost-effectiveness analysis was conducted to weigh the additional costs of treatment with a biologic drug compared to MTX against additional effectiveness. Separate analyses were conducted for etanercept, adalimumab, abatacept, and infliximab. A systematic review of the peer-reviewed (PubMed, EMBase, Cochrane) and grey literature was conducted to identify randomized controlled trials (RCTs), observational studies, and systematic reviews that evaluated biologic drugs for the treatment of polyarticular-course JIA or juvenile rheumatoid arthritis (RA) with any onset type in rheumatoid factor–negative or –positive patients. Studies were reviewed if they evaluated more than 20 patients, evaluated a single biologic, and reported disease improvement according to the ACR Pedi 30 criteria (8). Studies of biologics in systemic JIA were excluded because these patients may have different outcomes. Only studies relevant to the cost-effectiveness analysis were included in the analytic model.

The economic evaluation was conducted from the societal perspective (where applicable) with a time horizon of 2 consecutive 6-month intervals, as data for longer-term models were limited. An initial cost analysis compared annual treatment costs between biologic agents. This analysis included direct health care costs associated with resources consumed in drug administration and routine patient monitoring and non–health care costs consisting of parent/caregiver productivity losses associated with their children's care. In addition to these costs, the cost-effectiveness analysis included costs associated with serious adverse events and safety monitoring.

Cost identification, measurement, and evaluation.

The annual cost of treatment with each biologic was calculated in 2008 Canadian dollars (2008 annual average Canadian $1 = US $0.938). In the base-case analysis, a 40 kg patient was assumed, approximating the mean weight in the 2 pediatric RCTs that reported weight (14, 15). The direct medical cost items included biologic and MTX acquisition costs; concomitant drug costs (hydrocortisone, acetaminophen, folic acid, and diphenhydramine); drug administration materials; nursing time before, during, and after drug administration and in monitoring; dispensing fees; physician assessments; and laboratory tests, including tuberculosis screening (chest radiograph and skin test) and blood tests. For abatacept and infliximab, which are administered intravenously in a hospital-based clinic or infusion center, the costs of premedications (acetaminophen, diphenhydramine, and hydrocortisone), pharmacy preparation, intravenous bags and solutions, and parental/caregiver productivity time losses were also included. Cost calculations assumed no vial reuse. Productivity costs were assumed to be zero for medications that could be administered at home.

Unit prices of health resources were gathered from public sources, including Quebec and Ontario provincial drug plan formularies (medications) and Ontario Ministry of Health and Long-Term Care fee schedules (laboratory tests and physician fees). The chest radiograph cost was obtained from a previous report on biologics (16). The cost of the tuberculin test was obtained from a previous publication (17). Productivity costs were estimated by applying average wages obtained from Statistics Canada to the estimates for parent/caregiver time spent in the hospital (18). Detailed information on unit prices is available from the corresponding author.

Outcome measurement.

The effectiveness measure used in this analysis was the proportion of patients who had a reduction in symptoms at 1 year according to the ACR Pedi 30 criteria, as this was the most commonly available and consistently used outcome measure in the RCTs. Response rates for each biologic and comparator were extracted from multiple sources, including individual RCTs (14, 19–21) and prospective observational studies (22–25), to reduce sampling uncertainty. All of the RCTs except infliximab had a withdrawal design and were divided into 3 phases. In the open-label phase 1, the biologic drug ± MTX was administered to all eligible patients. Patients who achieved an ACR Pedi 30 response were then randomized into the double-blind phase 2 to receive either the active drug ± MTX or its matching placebo ± MTX for 4 to 8 months, depending on the study. Phase 2 was followed by an open-label noncomparative extension phase 3, where the biologic was administered to patients enrolled in the double-blind phase. Because only treatment responders were carried forward into phase 2, absolute effectiveness rates from phase 1 of these studies were used in the analyses to represent a more generalizable patient population. In the infliximab RCT, patients were initially randomized to receive infliximab 3 mg/kg + MTX or matching placebo + MTX for 14 weeks. After this period, patients received infliximab 3 or 6 mg/kg + MTX until week 52, and could then continue to an open-label extension phase. The absolute effectiveness rate from phase 1 was used for the infliximab analysis.

The randomized withdrawal design precluded the use of effectiveness estimates for the comparator (MTX) arm, which were not available in phase 1 (except for infliximab). In the infliximab RCT, patients were randomized to a control (MTX alone) arm in phase 1 (20). The observed treatment response rate of 49.2% (ACR Pedi 30) at 14 weeks was higher than the rate of 20–30% expected by the study investigators (20). A previous meta-analysis that pooled the results of the placebo groups of JIA RCTs yielded a 28.5% ACR Pedi 30 response rate (95% confidence interval [95% CI] 24.0%, 34.2%) at 6 months (26). For the base-case model, it was therefore assumed that in patients with optimized doses of nonbiologic DMARDs, 30% of patients would have a treatment response for a period of 6 months. In the absence of pediatric data beyond this point, it was assumed that the rate of responders would remain stable for the next 6 months. This assumption was based on RCTs of biologics in adults, which showed that the response rate in MTX-treated patients remained stable during the first year of treatment (27–29). Rates of serious infections and discontinuations due to adverse events were taken from the pediatric RCTs and from previous meta-analyses and systematic reviews in adults (30–33).

Decision model.

A separate decision analysis model was created for etanercept, adalimumab, abatacept, and infliximab. A general representation is displayed in Figure 1. The 1-year model consists of 2 consecutive 6-month intervals and incorporated probabilities that patients would, based on their response at 6 months, either continue with the same treatment or switch. The model incorporated the costs described above as well as the cost of treating serious infections. The value of $6,065 (range $1,814–11,277) representing average inpatient costs for infections in children ages <1 to 14 years was obtained from the Canadian Institute for Health Information (34).

Figure 1.

Schematic of the decision model used in the cost-effectiveness analysis. JIA = juvenile idiopathic arthritis; DMARDs = disease-modifying antirheumatic drugs.

To derive 6-month response rates for each biologic, data from the pediatric RCTs (14, 19–21) were pooled with data from the registry and observational studies (22–25, 35) in a meta-analysis. The inverse variance method, which uses the variance as a weight, was used to calculate a weighted average response rate with 95% CIs. At 6 months, patients treated with a biologic could stay on treatment or could switch due to lack of response, intolerance to therapy, or adverse events. Patients who switched were assumed to receive treatment with a second biologic for the ensuing 6 months. Probabilities for switching due to nonresponse or adverse events were derived from the RCTs (14, 19–21) and observational studies (22–25) corresponding to each biologic, with probabilities pooled as described above. For biologics where these data were not available in pediatric studies, the relative risk of switching from biologics due to nonresponse or adverse events was extrapolated from adult studies (27, 36, 37). These pediatric and adult sources were used to derive rates of serious infections.

For the MTX comparator, the same response rate of 30% was used for each model, as described above. Due to differences in the duration of followup in the various pediatric studies, the relative risk from adult meta-analyses comparing each biologic to MTX was used to represent the probability of switches due to inadequate response or adverse events (30–33). Patients who switched from MTX were assumed to receive a biologic for the next 6 months, and the cost of the biologic was represented by the average cost of all of the biologics, regardless of the initial therapy.

Incremental cost-effectiveness ratios (ICERs) and 95% CIs were calculated for each model using probabilistic sensitivity analysis (PSA) with 10,000 Monte Carlo simulations. The PSA simultaneously incorporates the uncertainty in estimates of different parameters into the distribution of ICERs from the simulations. This provides an estimate of the variation of the expected cost-effectiveness of each drug. To facilitate comparisons, treatment costs for a 40 kg/1.3 m2 child were used in each model. Due to uncertainty in the effectiveness estimates, additional PSAs were conducted under extreme contrast scenarios with fixed-effectiveness estimates, where the lowest effectiveness estimates for biologics were compared to the highest effectiveness estimates for DMARDs and vice versa. High and low effectiveness estimates for MTX were derived from an RCT (20) and for the biologics, from observational data with etanercept (35). The same relative decrease and increase in base-case effectiveness was applied to the other biologics.

Cost-effectiveness acceptability curves (CEACs) were calculated for each analysis. CEACs show the probability that a biologic is cost-effective at different willingness-to- pay thresholds. At the point where the curve crosses the 50% probability mark, there is an equal probability that either treatment (biologic or MTX comparator) is cost effective, i.e., the benefit (in monetary terms) exceeds the cost.



The RCTs for etanercept (21), adalimumab (14), and infliximab (20) were conducted in children with polyarticular-course juvenile RA, whereas the abatacept trial was conducted in children with JIA (19). All 4 were conducted in children with an inadequate response or intolerance to MTX or other DMARDs (14, 19–21). In the abatacept trial, treatment failure or intolerance to an anti–tumor necrosis factor α drug also qualified the patient for inclusion, whereas in the trials of etanercept, adalimumab, and infliximab, prior use of a biologic was an exclusion criterion. Table 1 shows the findings for the etanercept, adalimumab, and abatacept trials, which shared a similar design. During the open-label phase (phase 1) of the RCTs, ACR Pedi 30 criteria were met by two-thirds or more of children who received etanercept, adalimumab ± MTX, and abatacept ± MTX. The more stringent ACR Pedi 70 criteria were met by 36%, 59%, and 28% of patients in the etanercept (21), adalimumab (14), and abatacept (19) studies, respectively. Discontinuation rates varied between 24% and 36% during the open-label phase. Among children who completed a 4–8-month double-blind phase, the majority of children receiving biologics were without a disease flare. During the double-blind phase, discontinuation rates with biologics varied between 6% and 24%.

Table 1. Summary of the results of randomized controlled trials of biologics in juvenile idiopathic arthritis*
  • *

    Values are the percentage unless otherwise indicated. ACR Pedi 30 = American College of Rheumatology Pediatric 30 improvement criteria; MTX = methotrexate.

  • In the etanercept trial, the comparison was to placebo alone. Some patients discontinued due to adverse events rather than nonresponse; therefore, the proportion of ACR Pedi 30 responders and the rates of lead-in discontinuation may not always sum to 100%.

Sample size at the beginning of the lead-in phase69171190
ACR Pedi 30 response rate in the lead-in phase748465
ACR Pedi 70 response rate in the lead-in phase365928
Rate of drug discontinuation in the lead-in phase262236
Rate of drug discontinuation in the double-blind phase24618
Rates of patients without a disease flare at the end of the double-blind phase, biologic ± MTX vs. placebo ± MTX alone72 vs. 1960 vs. 3280 vs. 47

In the infliximab study (n = 60 at the start of the lead-in phase), the difference in the percentage of ACR Pedi 30 responders treated with infliximab 3 mg/kg + MTX compared to placebo + MTX was not statistically significant at 14 weeks (64% and 49%, respectively). After 14 weeks, all of the patients received infliximab 3 or 6 mg/kg + MTX. At the end of 52 weeks, approximately 75% of the patients were ACR Pedi 30 responders (20). In the infliximab study, 11% withdrew from the study between weeks 6 and 52, mainly due to lack of efficacy, adverse events, withdrawal of consent, or the start of alternative therapy.

Cost analysis.

Annual treatment costs with each biologic are shown in Table 2 (2008 Canadian dollars). The table shows the impact on the total as the costs related to various components are added. The total costs for abatacept and infliximab included parental time losses. Similar total annual treatment costs were observed with the biologics evaluated, ranging from $16,608 for the intravenous medication abatacept to $18,966 for the subcutaneous medication etanercept. Costs were based on a 40 kg child and may differ according to child weight. Annual costs with MTX were estimated at $952, including concomitant medications and monitoring.

Table 2. Summary of treatment costs*
MedicationMedication costsDrug costs plus preparation and administration costsTotal costs, including concomitant medications and monitoring
Per infusionPer yearPer infusionPer yearPer yearTotal costs, including productivity costs
  • *

    All costs are in 2008 Canadian dollars. Costs are based on treatment of a 40 kg child. N/A = not applicable.

Biologics administered intravenously in the hospital      
 Infliximab 3–5 mg/kg$1,880$15,040$2,034$16,274$17,259$18,330
Biologics administered subcutaneously at home      

Economic evaluation.

For the purpose of the cost-effectiveness analysis, data on effectiveness from 4 pediatric RCTs, 5 pediatric observational studies, and 1 meta-analysis were extracted and synthesized for each model as described above. Table 3 lists the variables and ranges used in the PSA. Incremental costs, effectiveness, and ratios from these analyses are shown in Table 4. In the base-case scenarios, each biologic was more expensive but also more effective than the MTX comparator. The mean incremental costs per additional ACR Pedi 30 responder at 1 year were $26,061 (95% CI $17,070, $41,834), $46,711 (95% CI $30,042, $75,787), $16,204 (95% CI $11,393, $22,608), and $31,209 (95% CI $16,659, $66,220) for etanercept, adalimumab, abatacept, and infliximab, respectively. Since head-to-head trials are not available with the different biologics, and because study populations differed by onset type, which may affect disease progression and prognosis, ICERs among biologics should not be compared. In the model where maximal effectiveness of each biologic was contrasted with the lowest-reported effectiveness for MTX, the incremental cost per additional responder gained decreased by 33–37% (depending on the drug), rendering the biologic more attractive. In the opposing model where the lowest-reported effectiveness for each biologic was contrasted with the highest effectiveness reported for MTX, only a very small proportion of simulations (7–16%) demonstrated a lower effectiveness for etanercept, abatacept, and infliximab. Therefore, even in these extreme scenarios the biologics were more effective, but also more costly. The exception was adalimumab, which demonstrated a lower effectiveness than the comparator in 58% of the simulations.

Table 3. Variables and ranges used in the probabilistic sensitivity analysis*
  • *

    Values are the mean ± SD unless otherwise indicated. Model assumes a 40 kg child. All variables were assigned a beta distribution except those designated with the † footnote. MTX = methotrexate; ACR Pedi 30 = American College of Rheumatology Pediatric 30 improvement criteria; RR = rate ratio.

  • Log-normal distribution.

  • The same relative decrease in estimates from observational studies of etanercept (35) was applied to other biologics due to a lack of data.

Patients achieving ACR Pedi 30 at 6 months, base-case %0.79 ± 0.100.80 ± 0.050.82 ± 0.050.80 ± 0.100.30 ± 0.03
Patients achieving ACR Pedi 30 at 12 months, base-case %0.79 ± 0.080.63 ± 0.060.82 ± 0.050.79 ± 0.050.30 ± 0.03
Patients with serious infections, %0.04 ± 0.020.04 ± 0.020.02 ± 0.030.08 ± 0.04RR 0.49–1.43
Patients switching due to adverse events, %0.03 ± 0.050.07 ± 0.020.04 ± 0.030.08 ± 0.11RR 0.68–1.23
Efficacy (low)0.61 ± 0.070.62 ± 0.040.63 ± 0.040.62 ± 0.070.20 ± 0.05
Efficacy (high)0.84 ± 0.030.85 ± 0.050.87 ± 0.050.85 ± 0.100.49 ± 0.06
Table 4. Incremental costs, effectiveness, and cost-effectiveness ratios*
Scenarios/outcomesEtanercept vs. MTX (95% CI)Adalimumab vs. MTX (95% CI)Abatacept vs. MTX (95% CI)Infliximab vs. MTX (95% CI)
  • *

    All costs are in 2008 Canadian dollars. A mean ICER could not be calculated where dominance was present in some of the simulations. MTX = methotrexate; 95% CI = 95% confidence interval; ICER = incremental cost-effectiveness ratio; N/A = not applicable.

Base case    
 Incremental costs, dollars11,090 (10,261, 11,863)13,107 (10,818, 15,491)7,873 (6,226, 9,419)12,167 (8,959, 12,550)
 Incremental effectiveness, %47.6 (26.7, 63.6)29.4 (17.3, 41.0)49.4 (38.1, 59.3)43.2 (18.2, 61.1)
 ICER26,061 (17,070, 41,834)46,711 (30,042, 75,787)16,204 (11,393, 22,608)31,209 (16,659, 66,220)
Extreme efficacy (biologic high, MTX low)    
 Incremental costs, dollars10,191 (9,121, 11,350)12,252 (9,297, 15,249)6,792 (4,907, 8,610)11,297 (7,897, 15,798)
 Incremental effectiveness, %62.1 (39.0, 79.7)43.4 (29.5, 58.0)63.9 (50.2, 75.5)57.4 (33.0, 71.9)
 ICER17,062 (11,914, 27,026)29,298 (18,071, 46,412)10,822 (6,964, 15,890)20,688 (12,121, 36,034)
Extreme efficacy (biologic low, MTX high)    
 Incremental costs, dollars12,833 (11,478, 14,145)12,647 (11,747, 13,477)9,978 (8,154, 11,751)13,951 (10,157, 18,969)
 Incremental effectiveness, %12.0 (−9.7, 32.7)−1.6 (−17.2, 14.3)12.7 (−4.1, 29.5)8.1 (−8.4, 23.8)
 ICERN/A (14% of simulations had lower biologic efficacy)N/A (58% of simulations had lower biologic efficacy)N/A (7% of simulations had lower biologic efficacy)N/A (16% of simulations had lower biologic efficacy)

The CEAC for etanercept is shown in Figure 2, based on the incremental cost per additional ACR Pedi 30 responder at 1 year. Only one figure is shown because the results were similar for all biologics. In the etanercept base-case scenario, the curve crosses 50% at $23,000. This indicates that if the maximum a decision maker was willing to pay to gain one additional ACR Pedi 30 responder was $23,000, both etanercept and MTX would have an equal probability of being cost effective, where the monetary value of health improvement exceeds the cost compared to standard care with MTX. If a decision maker was willing to pay no more than $30,000 to gain a responder, then the probability that etanercept would demonstrate a net economic benefit would be 95%. The willingness-to-pay points at which the biologic had a 50% probability of cost-effectiveness were $45,000, $17,000, and $27,500 for adalimumab, abatacept, and infliximab, respectively.

Figure 2.

Cost-effectiveness acceptability curve for etanercept. In the etanercept base-case scenario, the curve crosses the 50% mark at $23,000. This indicates that if the maximum a decision maker was willing to pay to gain one additional American College of Rheumatology Pediatric 30 improvement criteria responder was $23,000, both etanercept and the methotrexate comparator would have an equal probability of being cost effective, whereby the dollar value of the health benefit exceeds the cost.


This cost-effectiveness analysis demonstrated that JIA patients with a prior suboptimal response or intolerance to MTX may benefit from treatment with biologics for at least 1 year. The annual incremental costs of biologics per additional ACR Pedi 30 responder gained are high, ranging from $16,204 for abatacept to $46,711 for adalimumab from a societal perspective.

Studies of JIA patients have shown that the use of etanercept, adalimumab, abatacept, or infliximab may result in short-term disease improvement, according to the ACR Pedi 30, in approximately 80% of patients with active disease following a nonoptimal response to treatment with nonbiologic DMARDs (14, 19–21). The studies found, however, that up to one-third of patients may need to discontinue the biologic in the first 3–4 months of treatment due to either lack of efficacy or intolerance. The study with the longest followup (8 years) reported a 66% discontinuation rate from etanercept (excluding disease remissions) (38). The long-term results currently available (up to 8 years) suggest that biologics may remain effective for many years in those who tolerate them.

Several health technology assessments (HTAs) of biologics in JIA have been conducted. In a 2006 HTA report from Hungary, the authors concluded that “etanercept can improve the symptoms of JIA”; however, ongoing safety monitoring was suggested (39). The incremental cost per quality-adjusted life year (QALY) gained was estimated as €36,600 (Canadian $57,800) with etanercept compared to MTX (39). Brief HTAs from the National Horizon Scanning Centre at the University of Birmingham on adalimumab (2007) (40), abatacept (2007) (41), and tocilizumab (2006) (42) were identified. Based on JIA RCTs, the authors believed that adalimumab and abatacept may be effective in reducing morbidity and improving the patients' quality of life (40, 41). The authors could not determine the clinical benefits of tocilizumab in JIA due to lack of data (42).

A 2002 UK HTA evaluated the effectiveness and costs of etanercept in JIA (43). The report was based on the etanercept RCT data used in this study (21) and included data up to the second year of the trial. The authors concluded that etanercept is an effective treatment in JIA patients and that “the safety profile of etanercept is acceptable at present despite some reports of blood dyscrasias,” but emphasized the need to continue to monitor safety (43). The annual drug cost of etanercept in that HTA assuming no vial reuse was estimated at £8,996 (Canadian $16,000), based on a dosage of 0.4 mg/kg (maximum 25 mg) twice weekly. If a multiple-use vial was available, the annual cost was estimated as £2,407 (Canadian $4,280) for a 4-year-old (6.7 mg/dose) to £8,996 (Canadian $16,000) for an 18-year-old (25 mg/dose). The authors assumed no additional costs for support services, clinic visits, and monitoring (43). The present analysis also assumed no vial reuse. Because children generally receive lower doses than adults, a complete vial may not be needed. If vials could be reused, the cost-effectiveness of the medication would appear more favorable. At present, it is difficult to estimate the prevalence of vial reuse because this may vary across institutions.

Although biologic drugs represent significant improvements in the treatment of JIA, long-term safety needs to be established. Safety concerns raised by health authorities include the development of malignancies, demyelinating diseases, autoimmune disorders, and an increased risk of opportunistic infections (44–49). Furthermore, the long-term effects of biologics compared to nonbiologic DMARDs on functional disability and quality of life have not been investigated.

The conventional outcome for cost-effectiveness analysis considers quality of life through determination of the QALY, which requires measurements of utilities, or preferences, for health states. Valid utility estimates are difficult to obtain in young children and were not available for the analysis. While the present economic results based on the incremental cost per additional treatment responder may pose a challenge in the interpretation of and comparison to other studies, they may nevertheless be meaningful to clinicians and institutions where biologics are purchased and administered.

Despite evidence of short-term effectiveness, biologic agents for JIA are expensive and may pose significant costs to payers. Payers may vary by jurisdiction and may include the hospital, a government agency or other publicly-funded program, private insurance plans, or the patient's family. Given the potential budget impact of these agents as well as the potential for improvement in long-term patient outcomes, more comprehensive economic analyses should be undertaken once long-term data on safety and effectiveness are available.

Caution is required in interpreting the findings. The biologic studies identified through the systematic literature review were conducted in JIA patients with a suboptimal response or intolerance to MTX and other DMARDs. With the exception of the infliximab RCT, the design of the other studies did not permit a rigorous controlled comparison of the biologics with MTX/DMARDs since the randomized phase was restricted to responders. In addition, the outcome used was the ACR Pedi 30, as there were insufficient data on other ACR outcomes, such as the ACR Pedi 70, for the 1-year model. Although often a primary end point in RCTs, this outcome may represent only the minimal clinically important improvement by practitioners. With the more rigorous ACR Pedi 70, patients treated with biologics or with DMARDs would be expected to demonstrate lower response rates. As the economic evaluation represents an incremental analysis, the results would be changed only if the differences in ACR Pedi 70 response rates between the biologic and MTX groups were greater or less than those found in the present evaluation. The use of extreme contrasts in the present study was thus useful in revealing the impact on results when differences in effectiveness between groups were increased or decreased. It would be of value to incorporate more clinically relevant measures of effectiveness as they become available.

The economic models were based on the best evidence available; however, limitations existed. While each model generated a unique ICER for each biologic, these ICERs cannot be compared due to the absence of head-to-head trials among these drugs. One might attempt to compare the costs and outcomes of the drugs through statistical indirect comparisons. Unfortunately, differences in study populations across the trials precluded the conduct of indirect comparisons. Specifically, the proportion of JIA patients with active polyarticular-course disease who were diagnosed with systemic JIA at the onset of disease varied between studies (17–32%). Since onset type may affect disease progression and response to therapy, it was not possible to conduct indirect comparisons between drugs. Instead, extensive sensitivity analyses were conducted to measure the impact of uncertainty in effectiveness measures and other parameters. An important limitation was the use of a short time horizon of 1 year. The uncertainty in parameter estimates beyond this point was too great to allow for meaningful extrapolation. The costs of uncontrolled arthritis may have been underestimated, as other costs such as physical therapy were not included. The costs of NSAIDs and other pain medications were also not included, as it was difficult to predict their use accurately. Also, productivity losses incurred by parents who must miss work were included only for clinic appointments but not for the care of children with prolonged uncontrolled disease. Although costs were estimated in Canadian dollars, cost components were disaggregated to allow diverse payers to judge the relevance of costs presented for their jurisdictions. While significant differences in financing of the Canadian and US health care systems exist, treatment practices and patterns of health services use are similar.

The current evidence shows a short-term improvement in disease status following treatment with biologics in patients with polyarticular JIA who previously had an inadequate response to conventional treatment, albeit at a high annual cost. It is believed that better control of the disease may result in improvement in important long-term clinical outcomes such as functional disability, which may affect social functioning, employment, and quality of life. Longer-term data from disease registries and observational studies may allow for more comprehensive economic evaluation in the future.


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 submitted for publication. Dr. Ungar 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. Ungar, Costa, Laxer.

Acquisition of data. Costa, Laxer.

Analysis and interpretation of data. Ungar, Costa, Hancock-Howard, Feldman, Laxer.


We thank Dr. Timothy Beukelman for valuable comments on our analysis. We are grateful to the following individuals who provided assistance in the process of this research: Dr. Shirley Tse, Ms Karen Queffelec, and Ms Miranda Vermeer.