The clinical effectiveness and cost-effectiveness of abatacept, adalimumab, etanercept and tocilizumab for treating juvenile idiopathic arthritis: a systematic review and economic evaluation

Abatacept

Abatacept in combination with methotrexate is indicated for the treatment of moderate to severe active polyarticular JIA in paediatric patients aged ≥ 6 years, who have had an insufficient response to other DMARDs including at least one TNF inhibitor. 45

Abatacept is a fusion protein produced by recombinant deoxyribonucleic acid (DNA) technology in Chinese hamster ovary cells. It inhibits T-cell activation by specifically binding to cluster differentiation (CD)80 and CD86, thereby selectively inhibiting a costimulatory pathway that is required for full activation of T lymphocytes. 45,49 Through this mechanism, abatacept modulates the downstream T lymphocyte-dependent antibody responses and inflammation that cause the symptoms of JIA.

Treatment should be initiated and supervised by specialist physicians experienced in the diagnosis and treatment of JIA at the appropriate dosage as indicated in Table 5. Abatacept is not recommended in combination with TNF inhibitors. 45

Adalimumab

Adalimumab in combination with methotrexate is indicated for the treatment of active polyarticular JIA in patients from the age of 2 years who have had an inadequate response to one or more DMARDs. Adalimumab can be given as monotherapy in the case of intolerance to methotrexate, or when continued methotrexate treatment is inappropriate. Adalimumab is also indicated for the treatment of active ERA in patients aged ≥ 6 years, who have had an inadequate response to, or who are intolerant of, conventional therapy. 46

Adalimumab is a fully human monoclonal antibody drug initially tested as a treatment for rheumatoid arthritis (hence the trade name Humira – HUman Monoclonal antibody In Rheumatoid Arthritis). It binds specifically to the inflammatory cytokine TNF, thereby neutralising its biological function46 and modifying the inflammatory disease process. The European Medicines Agency (EMA) therapeutic indication for adalimumab was extended to the treatment of JIA in July 2008.

Treatment should be initiated and supervised by specialist physicians experienced in the diagnosis and treatment of JIA at the appropriate dosage as indicated in Table 6. 46 The concomitant administration of adalimumab with other biologic DMARDs [e.g. anakinra (Kineret^®, Swedish Orphan Biovitrum) and abatacept] or other TNF antagonists is not recommended. 46

Etanercept

Etanercept (Enbrel^®, Pfizer) is a fully humanised soluble TNF receptor fusion protein produced by recombinant DNA technology in Chinese hamster ovary cells. It is a dimer with two copies of the extracellular domain of the TNF receptor (p75) linked with the Fc component of human immunoglobulin 1, binding to TNF-α. 50 The mechanism of action of etanercept is thought to be its competitive inhibition of TNF binding to cell-surface TNF receptor, preventing TNF-mediated cellular responses by rendering TNF biologically inactive. Etanercept may also modulate biologic responses controlled by additional downstream molecules (e.g. cytokines, adhesion molecules or proteinases) that are induced or regulated by TNF. 47 The EMA therapeutic indication for etanercept in the treatment of JIA was extended in July 2012 to include:

• treatment of polyarthritis (RF+ve or RF–ve) and EO in children and adolescents aged ≥ 2 years who have had an inadequate response to, or who have proved intolerant of, methotrexate

• treatment of PA in adolescents aged ≥ 12 years who have had an inadequate response to, or who have proved intolerant of, methotrexate

• treatment of ERA in adolescents aged ≥ 12 years who have had an inadequate response to, or who have proved intolerant of, conventional therapy.

The age for treating polyarticular disease has been reduced from 4 to 2 years of age and the upper age limit of 17 years has been removed.

Treatment should be initiated and supervised by specialist physicians experienced in the diagnosis and treatment of JIA at the appropriate dosage as indicated in Table 7. The combined use of etanercept and anakinra or etanercept and abatacept is not recommended.

Tocilizumab

Tocilizumab (RoActemra^®, Roche) in combination with methotrexate is indicated for the treatment of juvenile idiopathic polyarthritis (RF+ve or RF–ve and EO) in patients ≥ 2 years of age who have responded inadequately to previous therapy with methotrexate. When the patient is intolerant to methotrexate or where continued treatment with methotrexate is inappropriate, tocilizumab can be given as monotherapy. 48 Tocilizumab is also indicated for the treatment of active systemic JIA but this indication is not included within the current NICE appraisal.

Tocilizumab is a humanised, monoclonal, antihuman interleukin-6 (IL-6) receptor (IL-6R) antibody that binds to membrane and soluble IL-6R, inhibiting IL-6-mediated signalling – a key cytokine in rheumatoid arthritis pathogenesis. 51 IL-6 is involved in causing inflammation and is found at high levels in patients with rheumatoid arthritis, systemic JIA and polyarticular JIA. By preventing IL-6 from attaching to its receptors, tocilizumab reduces the inflammation and other symptoms of these diseases. 48 The EMA was granted a licence for tocilizumab in the treatment of JIA in May 2011.

Treatment should be initiated by health-care professionals experienced in the diagnosis and treatment of JIA at the appropriate dosage as indicated in Table 8.

Reference screening

All studies were selected for inclusion through a two-stage process. Titles and abstracts were screened independently by two reviewers for potential eligibility, using a standardised and piloted eligibility selection worksheet (see Appendix 2) containing the inclusion/exclusion criteria detailed above.

Full-paper screening

Full texts for potentially relevant studies were obtained and screened using a standardised and piloted eligibility section worksheet (see Appendix 3) by one reviewer and checked by a second reviewer, and a final decision regarding inclusion was agreed. At each stage, any disagreements were resolved by discussion or with the involvement of a third reviewer when necessary.

Quantity and quality of available research

Titles and, where available, abstracts of a total of 2651 references identified by searches (after deduplication) were screened and full copies of 60 references were retrieved. Of these, 29 articles were excluded after inspection of the full text as shown in Figure 1 and these are listed in Appendix 4. The most common reason for exclusion of a reference was an irrelevant study design (e.g. systematic reviews, which were used as a source of references, commentaries). One full text56 was of unclear relevance to the review because the type of JIA was not stated and it was not clear whether or not participants met the licensed indication for etanercept therapy in respect of having an inadequate response or intolerance to methotrexate. One full paper and eight conference abstracts relating to four ongoing studies that seemed to be relevant were tagged for inclusion in Ongoing trials (note that a further three ongoing studies were identified from a separate search specifically undertaken for ongoing studies, which is not represented in Figure 1; therefore, a total of seven ongoing studies are summarised in Ongoing trials).

FIGURE 1.

Flow chart for the identification of studies.

Nine full texts and 12 conference abstracts described four RCTs (each of which was described by at least one full paper) that met the inclusion criteria of the review (see Figure 1). As the full texts provided the most complete data, these were the primary source of information for this review.

One of the RCTs evaluated abatacept57–60 [the AWAKEN (Abatacept Withdrawal study to Assess efficacy and safety in Key Endpoints) trial], one RCT, by Lovell and colleagues, evaluated adalimumab,61–64 one RCT evaluated etanercept (Lovell et al. 42,65–67) and one RCT evaluated tocilizumab68–76 (the CHERISH trial). For the sake of brevity, generally only the key reference for each RCT is cited in the main text of this report. All four RCTs used placebo as the comparator; however, with the exception of the etanercept trial, the majority of the patients in the trials received methotrexate in addition to the biologic DMARD or placebo. The key characteristics of the trials are presented in Table 9, with the primary and secondary outcomes measured in trials summarised in Table 10. All studies were multicentre RCTs, with the number of centres ranging from 9 in the etanercept study42 to 58 in the tocilizumab study. 68 Locations of the studies included the USA (all four studies), Canada (one study42), Europe (three studies,57,61,68 with only the tocilizumab study68 including UK centres), Latin America (two studies57,68), Australia (one study68) and the Russian Federation (one study68). In each study, participants were initially treated in an open-label phase with the biologic DMARD under investigation and had to achieve at least an ACR Pedi-30 response to the biologic DMARD to be eligible for entry to the randomised double-blind withdrawal phase, with the number of participants randomised ranging from 51 in the etanercept study to 166 in the tocilizumab study. As each study investigated a different biologic DMARD, study-specific details are provided below by study drug.

Assessment of clinical effectiveness: biologic disease-modifying antirheumatic drugs versus placebo (with methotrexate where permitted)

Assessment of clinical effectiveness: biologic disease-modifying antirheumatic drugs versus each other (with methotrexate where permitted)

Background

None of the RCTs included in the systematic review of clinical effectiveness directly compared any of the biologic DMARDs with each other. It was therefore necessary to undertake an indirect comparison of the drugs to inform the assessment of comparative clinical effectiveness. One published indirect comparison was identified through literature searching, by Otten and colleagues. 79 This was a systematic review of RCTs that constructed two separate evidence networks: polyarticular-course JIA and systemic JIA. For each network, a series of pairwise indirect comparisons was conducted, with placebo as a common comparator, using the method described by Bucher and colleagues. 53

Three RCTs were included in Otten and colleagues’79 polyarticular-course JIA network,42,57,61 all of which have been included in this assessment report. However, this network did not include tocilizumab, as at that time no RCT evidence for that drug in polyarticular-course JIA was published. The network therefore included only comparisons of three of the four biologic DMARDs of relevance to the scope of this assessment (abatacept, adalimumab and etanercept). We have conducted a similar adjusted indirect comparison to Otten and colleagues79 including the recently published tocilizumab RCT by Brunner and colleagues (the CHERISH trial). 68 Figure 2 illustrates the design of the analysis, representing what is termed a star network. 80

FIGURE 2.

Indirect comparison of biologic DMARDs.

Otten and colleagues79 indirectly compared the drugs in relative risk (RR) of disease flare. We have similarly included disease flare as an outcome and, in addition, have chosen ACR Pedi-50 and -70 responses as an outcome. ACR Pedi-50 and -70 were chosen as opposed to ACR Pedi-30, as it was considered that a higher level would be a more clinically relevant level of treatment response. Furthermore, owing to the design of the RCTs, all patients who were randomised had achieved an ACR Pedi-30 response at the end of the open-label lead-in phase.

The adjusted indirect comparison should be considered to be exploratory rather than definitive owing to limitations in the evidence base and heterogeneity between the included trials.

• There is only one trial available for each drug. Although the trials were considered to be of generally good methodological quality and low risk of bias, evidence networks are considered weaker if informed by small numbers of studies and small numbers of participants. 81 The number of patients in the trials was also relatively low (ranging from 51 to 163).

• There is some variation in the proportion of subtypes of JIA in the included trials. Although the network is considered to be most applicable to polyarticular-course JIA, in one trial (etanercept42) around one-third of patients were classified as having systemic JIA with apparent systemic manifestations (which is outside the scope of the current appraisal). There was insufficient evidence from RCTs to construct a network for PA, or ERA because outcome data for these subtypes of JIA were not reported separately in trials, even though some cases with these subtypes may have been included.

• The duration of JIA ranged from just under 4 years57 to approximately 6 years across the trials. 42 A study by the BSPAR etanercept cohort found that shorter disease duration was an independent predictor of achieving an excellent response to treatment (ACR Pedi-90) at one year. 82 Disease duration has also been found to be a predictor of response to etanercept83 and to methotrexate84 among patients from the German Biologika in der Kinderrheumatologie (BIKER) registry. Differences between the trials in disease duration may therefore potentially confound results.

• Three of the four trials permitted patients to take methotrexate in addition to the biologic DMARD (in proportions of patients varying from 74% to 100%), whereas the fourth trial (etanercept)83 did not permit use of methotrexate.

• Previous therapy with biologic DMARDs had been received by approximately one-third of participants who entered the initial open-label run-in of two trials (abatacept57 and tocilizumab68). Prior therapy with another biologic DMARD was an exclusion criterion for the adalimumab trial61 and was not mentioned for the earliest trial (etanercept42), presumably because no other biological therapies were available at the time. Currently, it is unclear whether or not prior biologic DMARD treatment influences the effectiveness of subsequent biologic treatment.

• The mean age of patients across the trials varied from around 7.5 years to 13 years. Part of this variation may reflect the age ranges specified in the inclusion criteria of the trials and potentially the mix of JIA subtypes in the trials, which have a different mean age at onset. Age could be an effect modifier given the progressive nature of JIA.

• The duration of the double-blind randomised treatment phase of the trials varied from 4 months42 to 8 months. 61 Treatment duration may affect outcomes that are time dependent, such as disease flare.

Results

Figure 3 illustrates the results of the four included RCTs comparing the biologic DMARDs with placebo (with background methotrexate where permitted) on the outcome of disease flare (the adalimumab trial61 stratified results according to whether or not patients received methotrexate background therapy, and we have included data only for patients who did receive methotrexate, in accordance with the licensed indication – this applies to disease flare and to ACR Pedi-50/-70). Treatment with each of the four DMARDs resulted in a statistically significant reduction in the RR of a disease flare, ranging from 0.38 to 0.57. (We have not presented a pooled RR given differences between the DMARDs and also heterogeneity between the trials.)

FIGURE 3.

Summary forest plot of biologic DMARDs vs. placebo: disease flare. M–H, Mantel–Haenszel.

Table 25 reports adjusted pairwise indirect comparisons for the four biologic DMARDs for the outcome of disease flare. The point estimate for risk of flare was lower for etanercept than the other three comparators. Abatacept had a lower risk of flare than adalimumab and tocilizumab. Tocilizumab had a lower risk of flare than adalimumab only. Adalimumab was associated with a higher risk of disease flare than the other three comparators. The ranking of treatments in terms of risk of flare was therefore etanercept, abatacept, tocilizumab and adalimumab. However, none of the comparisons demonstrated a statistically significant difference between the treatments being compared, with CIs crossing 1 in every case. The results of our analysis match those of Otten and colleagues,79 with the exception of the comparison with tocilizumab, which was not included in their polyarticular-course JIA trial network (as discussed above).

TABLE 25 - Indirect comparisons of biologic DMARDs: disease flare

Comparison	RR (95% CI)
Etanercept vs. adalimumab	0.61 (0.27 to 1.38)
Etanercept vs. abatacept	0.92 (0.39 to 2.18)
Etanercept vs. tocilizumab	0.65 (0.30 to 1.43)
Adalimumab vs. abatacept	1.51 (0.72 to 3.15)
Adalimumab vs. tocilizumab	1.07 (0.56 to 2.04)
Abatacept vs. tocilizumab	0.71 (0.35 to 1.43)

Figure 4 illustrates the results of the four included RCTs comparing the biologic DMARDs with placebo (with background methotrexate where permitted) on the outcome of ACR Pedi-50 response. Treatment with each of the four DMARDs led to a statistically significant greater proportion of participants with ACR Pedi-50 response, with the RR ranging from 1.41 to 3.12.

FIGURE 4.

Summary forest plot of biologic DMARDs vs. placebo: ACR Pedi-50 response. M–H, Mantel–Haenszel.

Table 26 reports adjusted pairwise indirect comparisons for the four biologic DMARDs for the ACR Pedi-50 response outcome. Etanercept had a higher RR for treatment response than the other three comparators. Adalimumab had a higher RR for treatment response than abatacept and tocilizumab. Adalimumab had a higher RR for treatment response than tocilizumab. The ranking of treatments in terms of treatment response was therefore etanercept, adalimumab, abatacept and tocilizumab. With the exception of etanercept compared with tocilizumab, none of the comparisons indicated a statistically significant difference between the treatments being compared, with CIs crossing 1.

TABLE 26 - Indirect comparisons of biologic DMARDs: ACR Pedi-50 response

Comparison	RR (95% CI)
Etanercept vs. adalimumab	1.87 (0.77 to 4.53)
Etanercept vs. abatacept	2.10 (0.95 to 4.64)
Etanercept vs. tocilizumab	2.21 (1.01 to 4.84)
Adalimumab vs. abatacept	1.12 (0.65 to 1.96)
Adalimumab vs. tocilizumab	1.18 (0.69 to 2.02)
Abatacept vs. tocilizumab	1.05 (0.72 to 1.53)

Figure 5 illustrates the results of the four included RCTs comparing the biologic DMARDs with placebo (with background methotrexate where permitted) on the ACR Pedi-70 response outcome. Treatment with each of the four DMARDs led to a statistically significant greater proportion of participants with ACR Pedi-70 response, with the RR ranging from 1.54 to 2.34.

FIGURE 5.

Summary forest plot of biologic DMARDs vs. placebo: ACR Pedi-70 response. M–H, Mantel–Haenszel.

Table 27 reports adjusted pairwise indirect comparisons for the four biologic DMARDs for the outcome of ACR Pedi-70 response. Etanercept had a higher RR for treatment response than abatacept and tocilizumab but a slightly lower RR for treatment response than adalimumab. Adalimumab had a higher RR for treatment response than abatacept and tocilizumab. Abatacept had a higher RR for treatment response than tocilizumab. The ranking of treatments in terms of treatment response for ACR Pedi-70 (adalimumab, etanercept, abatacept and tocilizumab) was therefore different from the ACR Pedi-50. None of the comparisons indicated a statistically significant difference between the treatments being compared, with CIs crossing 1.

TABLE 27 - Indirect comparisons of biologic DMARDs: ACR Pedi-70 response

Comparison	RR (95% CI)
Etanercept vs. adalimumab	0.98 (0.33 to 2.87)
Etanercept vs. abatacept	1.31 (0.48 to 3.60)
Etanercept vs. tocilizumab	1.49 (0.57 to 3.85)
Adalimumab vs. abatacept	1.34 (0.65 to 2.79)
Adalimumab vs. tocilizumab	1.52 (0.79 to 2.92)
Abatacept vs. tocilizumab	1.13 (0.66 to 1.93)

The results of this exploratory analysis based on the limited evidence available currently (only one trial for each of the four biologic DMARDs) supports etanercept being more effective than the other three biologic DMARDs in terms of preventing disease flares and achieving a response to treatment based on a composite index (ACR Pedi-50), whereas the ACR Pedi-70 exploratory analysis shows adalimumab with a slight advantage over etanercept (although see comment below about CIs). Abatacept appeared to be superior to tocilizumab for all outcome measures. Adalimumab appeared to be less effective than abatacept and tocilizumab in terms of preventing disease flare but more effective in terms of ACR Pedi-50 and -70 responses. Therefore, there was no consistent ranking of treatment comparisons across these outcome measures. The indirect comparisons were generally not statistically significant and CIs were wide, so caution is advised in the interpretation of these results. Furthermore, the etanercept trial42 appears to have some differences from the other trials which may confound the results, namely the absence of methotrexate background therapy and the longer duration of JIA disease. There was also a noticeably higher rate of flares in the placebo arm of that trial than in the other three trials (81% compared with 48–65%) which may account for the bigger treatment effect seen. Taking the above limitations into account, an overall interpretation of the results of the indirect comparison is that, owing to the absence of statistically significant differences between the biologic DMARDs, they currently appear to be similar in treatment effectiveness. This accords with the conclusion reached by Otten and colleagues79 who suggested that the short-term efficacy of the biologic DMARDs in polyarticular-course JIA seem similar. Furthermore, the clinical advisors to the assessment group felt that these data generally reflect clinical experience in that when used for the same indication in the same population, effectiveness was likely to be similar. However, there was also a recognition that for individual patients, and potentially for particular subgroups of JIA patients, differential effects of each biologic DMARD might be apparent but these differential effects have not yet been captured by current trial data.

Summary of the systematic review of clinical effectiveness

• Four multicentre RCTs, one each evaluating abatacept, adalimumab, etanercept and tocilizumab, met the inclusion criteria of this review. Only the tocilizumab RCT included UK patients. Seven additional RCTs (three for adalimumab and four for etanercept) are described as ongoing (details summarised in Ongoing trials, below).

• Each RCT had three phases, an open-label lead-in period, a randomised withdrawal period and an OLE. The lengths of the lead-in and randomised phases varied between studies (open-label lead-in: 12–16 weeks; randomised double-blind withdrawal phase 16–32 weeks). In each study patients had to achieve an ACR Pedi-XX response during the initial open-label phase in order to be eligible for entry to the randomised double-blind withdrawal phase. Therefore, results are applicable only to patients who have already achieved an initial (low) degree of benefit from a biologic DMARD.

• The quality of the included RCTs was reasonable overall, with a low risk of bias judged for most items, although some aspects were rated as unclear, mainly owing to a lack of reporting.

• Disease flare: this was the primary outcome in all four RCTs, with definitions broadly consistent between the studies. Patients who continued to receive biologic DMARDs during the randomised withdrawal phase of the studies had statistically significantly fewer arthritis flares than those receiving placebo in all four studies, whereas time to disease flare reported in three studies (abatacept, adalimumab and etanercept) was statistically significantly longer in those treated with biologic DMARDs.

• ACR Pedi: a greater proportion of patients receiving biologic DMARDs during the randomised withdrawal phase of the studies achieved ACR Pedi responses of ≥ 30 than placebo-treated patients, with differences statistically significant in all but two instances where p-values were reported. The proportion of biologic DMARD-treated patients with inactive disease was more than twice that of placebo-treated patients in the two studies (abatacept and tocilizumab) reporting this outcome.

• ACR Pedi core variables: in the three studies reporting this outcome (abatacept, etanercept and tocilizumab) results were generally in favour of treatment with biologic DMARDs when compared with placebo.

• Joint-related outcomes: one study (etanercept) reported additional joint outcomes without statistical comparisons. All outcomes favoured etanercept over placebo.

• Pain: three studies reported pain (abatacept, etanercept and tocilizumab). A statistically significant difference in the mean change from baseline favoured the tocilizumab group. Although pain scores were lower for those receiving biologic DMARDs in the remaining two studies, differences between treatment groups were not statistically significant in the abatacept study and no statistical comparison reported in the etanercept study.

• Mortality: no treatment-related deaths were reported in the three studies reporting this outcome (adalimumab, etanercept and tocilizumab).

• Outcomes not reported by the included RCTs for the randomised withdrawal phase of the trials were corticosteroid-reducing regimens, extra-articular manifestations (such as uveitis), height and weight.

• HRQoL: reported by the abatacept study only, with differences between treatment groups for the physical and psychosocial summary scores not statistically significant. Those treated with abatacept had improved scores for 14 of the 15 CHQ subscales compared with 6 out of 15 for placebo-treated patients.

• AEs: during the randomised withdrawal phase of the trials, the proportions of AEs and SAEs were generally fairly similar between the biologic DMARDs and the placebo groups.

• Subgroup analyses: the tocilizumab study reported data for subgroups but no statistical comparisons between treatment groups were reported.

• OLE: all four studies included an OLE phase. There were differences in eligibility criteria between studies and in how data were presented. Only results for ACR Pedi, AEs and growth are included in this report.

• OLE ACR Pedi:

  • Abatacept: the proportion of patients achieving ACR Pedi-30, -50 and -70 scores were similar for those with continuous abatacept therapy and those who received placebo during the double-blind phase, but were greater at achieving ACR Pedi-100 and inactive disease in abatacept-treated patients (ACR Pedi-100 abatacept 39%, placebo 19%; inactive disease abatacept 43%, placebo 23%).

  • Adalimumab: there was no diminution of ACR Pedi responses, and 40% of patients had an ACR Pedi-100 response after open-label treatment in the extension phase, but results included patients not meeting the licensed indication.

  • Etanercept: 26/58 (45%) patients who took part in the OLE entered the eighth year of follow-up. ACR Pedi responses appear to have remained constant over the OLE.

  • Tocilizumab: limited results were based on conference abstracts for 82 patients who received continuous tocilizumab throughout the study. The proportion of patients achieving ACR Pedi-70 and -90 increased since the end of the double-blind phase, with 63% having inactive disease.

• OLE AEs:

  • Abatacept: at 7-year follow-up, 19.6% of patients had SAEs, with similar incidence rates between the OLE phase (5.6 per 100 patient-years) and the 6-month double-blind phase (6.8 per 100 patient-years).

  • Adalimumab: SAEs possibly related to study drug occurred in seven patients during the OLE, but would appear to be in patients not in line with licensed indication (OLE phase ongoing, time period unclear). Three patients discontinued treatment owing to AEs.

  • Etanercept: there were a total of 39 SAEs based on 318 patient-years of etanercept exposure, with 26/69 patients entering their eighth year of etanercept treatment (0.12 events per patient-year). Nine MIIs resulted in the need for intravenous antibiotic therapy or hospitalisation (0.03 events per patient-year).

  • Tocilizumab: AEs rates were 406.5 per 100 patient-years and the SAEs rate 11.1 per 100 patient-years over around 2 years, with the most common AEs and SAEs related to infections (151.4 and 5.2 per 100 patient-years, respectively).

• Growth: limited data reported in abstracts at week 104 for adalimumab and tocilizumab appear to support the positive effect of these drugs on growth, but the use of different outcome measures prevents a comparison between the drugs.

• An exploratory adjusted indirect comparison found that there was a lack of statistically significant differences between the four biologic DMARDs in terms of disease flare and ACR Pedi-50/-70 response, with wide CIs and clinical heterogeneity between the trials.

Review of Bristol-Myers Squibb company evidence submission for abatacept

The company did not report a systematic review of clinical effectiveness of abatacept. 85 There is no indication that any databases were searched and no search strategies were supplied. Furthermore, there is no search or report for any ongoing studies. The majority of the clinical effectiveness information in the CS comes from published papers with a few details that are commercial-in-confidence (CiC) which come from the clinical study reports.

The CS includes one phase-III double-blind randomised withdrawal study, the AWAKEN trial. Although not clearly summarised, the CS draws on two published papers,57,58 one conference presentation86 and the trial clinical study reports. The published papers57,58 met the inclusion criteria of this assessment report. One published paper59 relating to the AWAKEN trial that was identified in this assessment report was not cited by the CS but the data in this appear to have been superseded by the more recent conference presentation. 86 However, the more recent efficacy data are not presented according to the randomised groups in the double-blind period, whereas the safety summary data are. 86 Furthermore, there is limited detail regarding the length of follow-up, which is stated to be ≥ 56 months and up to 7 years of total follow-up. No critical appraisal is reported for any of the studies cited in the CS.

A summary of the AWAKEN trial is provided in the CS which is broadly similar to the information presented in the published papers. 57,58,86 Information from the OLE phase drawn from the conference presentation86 is more recent than the data from the published paper,59 which is included in the assessment report. Furthermore, an analysis was conducted (using Fisher’s exact test) to compare SAEs during the double-blind phases of trials of abatacept, adalimumab etanercept and tocilizumab (CS section 3.4.6). The CS highlights the lack of statistical power attributable to the low numbers of patients and event rates, which should be taken into account in the interpretation of their finding that the incidence of SAEs was likely to be similar between the biologic DMARDs.

The CS focuses on abatacept, with very little information provided regarding the other biologic DMARDs included in this MTA. However, information is presented in the CS on indirect pairwise comparisons for the four biologic DMARDs for the outcome of disease flare. The comparisons for abatacept, adalimumab and etanercept are taken from a published paper by Otten and colleagues79 and this is supplemented by new indirect pairwise comparisons with tocilizumab, taking data from a RCT68 that has been published since the Otten and colleagues study (the CHERISH trial. 68) The results of the indirect comparisons reported in the CS (tables 4 and 5) match those reported in the indirect comparisons conducted for this assessment report [see Assessment of clinical effectiveness: biologic disease-modifying antirheumatic drugs versus each other (with methotrexate where permitted)].

In summary, the CS has not conducted a systematic review of clinical effectiveness but has summarised data from the AWAKEN trial,57 presented indirect pairwise comparisons of the four biologic DMARDs included in this appraisal and conducted an analysis to compare SAEs during the double-blind phases of trials of the four biologic DMARDs. No additional RCTs were included in the CS that would have met the inclusion criteria of this assessment report.

Review of AbbVie company evidence submission for adalimumab

AbbVie submitted a report to NICE on adalimumab as a treatment for JIA. 77 The clinical effectiveness evidence has been briefly appraised.

The company did not conduct a formal systematic review of the clinical effectiveness evidence, but provided what they describe as ‘an iterative literature review’ (CS page 15). The company asserts that all RCTs of adalimumab in the treatment of JIA have been identified (CS page 15). It would appear that RCTs were identified chronologically from an adalimumab trial programme, and there is no mention that any databases were searched and no search strategies were provided. There is no search for or report of any ongoing studies, but the CS does contain information about a trial in progress, the SYCAMORE RCT. 87 This trial is evaluating the clinical effectiveness, safety and cost-effectiveness of adalimumab in combination with methotrexate for the treatment of JIA-associated uveitis (further information on this trial is given Ongoing trials of this assessment report). Data from abstracts/conference proceedings are also presented in the CS.

The submission contains narrative summaries for the pivotal RCT by Lovell and colleagues,61 which formed the basis of the original marketing authorisation in 2008; an ongoing RCT of adalimumab treatment in patients with ERA by Burgos-Vargas and colleagues88–93 (see Ongoing trials of this assessment report for details of this study); two open-label single-arm studies94,95 and supporting data from an ongoing registry (STRIVE) funded by AbbVie. The multinational STRIVE registry is assessing the long-term safety and effectiveness of adalimumab in patients with moderate to severe polyarticular JIA. Some data from this registry are given in the CS, for efficacy outcomes up to 1 year, and safety outcomes for longer (mean duration of drug exposure was 643 days for methotrexate patients and 653 days for adalimumab and methotrexate patients). The registry does not appear to include patients from the UK. Other evidence such as case series, open-label trials, a systematic review96 and data from an Italian registry were included to provide evidence for the effectiveness of adalimumab in JIA-associated uveitis (see JIA-associated uveitis of this assessment report for details of these).

Based on the Lovell and colleagues RCT,61 the key outcomes of disease flares and ACR Pedi responses are the same in the CS and the assessment report.

The CS notes several methodological concerns that prevented the presentation of a network meta-analysis comparing the four biologic DMARDs. An indirect comparison was therefore not presented.

In summary, the CS has not conducted a systematic review of clinical effectiveness but has summarised data separately for RCTs and other non-randomised studies, as well as data from a registry. No indirect comparison of the biologic DMARDs was conducted by the company. No additional RCTs were included in the CS that met the inclusion criteria for systematic review in this assessment report; however, some of the non-randomised study evidence in the CS is presented in this assessment report for patient subgroups where randomised evidence is lacking (i.e. ERA and JIA-associated uveitis) (see Additional supporting evidence).

Review of Pfizer Ltd company evidence submission for etanercept

The company report a systematic review of clinical effectiveness of etanercept (in addition they report a systematic review of observational evidence on etanercept-associated innovation, caregiver burden and treatment adherence, plus a systematic review of HRQoL associated with etanercept). 97 Details of the literature search strategy are provided and the search appears to be comprehensive, up to date and reproducible. A search for ongoing studies was also conducted. A systematic process was followed to screen studies for inclusion, with titles, abstracts and full texts screened independently by two reviewers. The inclusion criteria are in keeping with the scope of the appraisal, with the exception that only studies of etanercept were included, not the other biologic DMARDs. A broad range of study designs were eligible, with the exception of case reports. The majority of the data are in the public domain, although some information is either academic-in-confidence (AiC) or CiC.

The review included 11 publications relating to five primary interventional studies and three extension studies (see CS table 7, CS page 44). It also included 41 observational studies (including registry studies) plus 2 unpublished studies (see CS table 18, CS page 81). Of the five included primary interventional studies, only one meets the inclusion criteria for the systematic review in this current report – Lovell and colleagues. 42 Of the remaining four studies, three were not relevant, as they were single-arm studies, and a fourth was a RCT reported in a conference abstract98 with only limited detail available. However, one of the single-arm studies – the CLIPPER study99 – is noteworthy as it focuses specifically on the JIA subtypes that were absent from the pivotal Lovell and colleagues trial,42 namely EO JIA, ERA and PA. Details are presented in Additional supporting evidence.

Of the three extension studies, only one was relevant to the inclusion criteria of this assessment report, namely the long-term follow-up publications65–67 of the Lovell and colleagues RCT,42 all of which have been included in the data extraction for this study (see Appendix 5). The other two extension studies included a Japanese open-label single-arm multicentre study followed by a double-blind, randomised dose-down extension study (two doses of etanercept – no comparator), and an open-label multicentre Phase-IIIb long-term safety and efficacy study of the CLIPPER study (reported in Additional supporting evidence). 99

A critical appraisal of the interventional studies is provided in CS section 4.7. The company’s appraisal of the Lovell and colleagues RCT42 is provided in CS table 13 (CS page 62). Our critical appraisal differs slightly from the company’s (see Quantity and quality of research available). Specifically, we did not consider that adequate details had been provided of the study’s randomisation method or concealment of allocation. We also note that there was a large imbalance in drop-outs between the randomised groups (see Table 12 and Appendix 5).

A narrative synthesis of the interventional and observational studies is provided in the CS, with detailed tabulation of study characteristics and results. A meta-analysis was not considered feasible or appropriate by the company, and an indirect comparison was not conducted as it was not considered feasible to conduct one owing to differences in respective marketing authorisations across biologic treatments, paucity of data and heterogeneity.

Of note, some of the observational studies of etanercept included in the CS reported (limited) data for outcomes relevant to the scope of the appraisal that were not included in the RCT by Lovell and colleagues,42 namely corticosteroid reduction, growth and disease activity according to the JADAS (see CS section 4.12.5).

In summary, the systematic review of clinical effectiveness reported in the CS appears to be of a good standard and no additional RCTs were included in the CS that met the inclusion criteria for the systematic review in this assessment report.

Review of Roche company evidence submission for tocilizumab

The company did not conduct a formal systematic review of the clinical effectiveness evidence, but provided ‘most relevant literature’ on the use of tocilizumab in patients with polyarticular JIA and EO (CS page 7). There is no evidence that searches were conducted and no search strategies were reported. The CS does state that a systematic literature review was completed for the indirect comparison presented in the submission but provides no further detail. The CS did not report searching conference proceedings or details of any ongoing trials, but data from abstracts/conference proceedings are included in the submission. CiC data are limited to the economic model.

The submission contains narrative summaries of two studies. One of the studies is a RCT comparing tocilizumab with placebo (CHERISH)68 linked to six additional conference publications/abstracts. 69,71–73,76,100 The CHERISH RCT68 met the inclusion criteria of this assessment report and was reported above in Results. Of the six conference publications/abstracts linked to the CHERISH RCT, only two were related to the randomised phase of the trial72,73 and the remaining four were related to the OLE phase. One of these four conference abstracts was not identified by searches for this assessment report, but if it had been, it would not have met the inclusion criteria as none of the outcomes reported was relevant. 100

The other study was a single-arm open-label study of efficacy, pharmacokinetics and safety of adalimumab in Japanese patients with polyarticular JIA. 95

The CS presents all the evidence separately for each study in the form of a narrative summary. Individual tables of baseline patient characteristics, as well as details of methods and design are reported for both the CHERISH trial68 and the open-label Japanese study. 95 No quality assessment of the studies is presented. The CS reports growth data from the OLE phase of the CHERISH trial at week 104, which have been included in this assessment report. The assessment report contains additional data for ACR Pedi-90 responses relative to baseline at week 40 and inactive disease from the CHERISH trial, both of which are not reported in the CS.

The CS includes a hierarchical Bayesian indirect treatment comparison of adalimumab and tocilizumab, conducted in WinBUGS version 1.4 software (Medical Research Council Biostatistics Unit, Cambridge, UK) and using methods described by Dias and colleagues. 101 Limited detail is provided on the specific methods used to conduct the indirect comparison (e.g. which adalimumab trial was used to compare against tocilizumab; this is likely to be Lovell and colleagues61 but is not explicitly stated). An indirect comparison with abatacept was not considered possible owing to the difference in trial design, the fact that it is not approved (appraised) by NICE, and also because of slight differences in licences (i.e. lower age for which treatment is indicated). We note the heterogeneity between the RCTs that increases uncertainties in any indirect comparison, although the fact that abatacept has not been appraised by NICE is not an adequate justification for not performing the comparison. The CS provides an additional analysis which assumes a class effect across antiTNF drugs (based on the indirect comparison with adalimumab which showed ‘overlapping ACR response rates’), permitting a comparison between tocilizumab with etanercept (CS section 5.17). The exploratory pairwise indirect comparisons of all four biologic DMARDS presented in this assessment report showed no statistically significant differences between the drugs [see Assessment of clinical effectiveness: biologic disease-modifying antirheumatic drugs versus each other (with methotrexate where permitted)].

Most of the AE and safety data are presented for the CHERISH RCT. 68 The AE data reported at week 40 (end of the randomised phase of the RCT) in the CS do not include any data for the placebo group, which is presented in this assessment report.

In summary, the CS has not conducted a systematic review of clinical effectiveness but has summarised data from the CHERISH trial and an open-label Japanese study, presented an indirect pairwise comparisons of two of the biologic DMARDs included in this appraisal (tocilizumab and adalimumab), as well as exploratory analysis comparing tocilizumab with etanercept. No additional RCTs were included in the CS that met the inclusion criteria of the assessment report.

Adalimumab ongoing trial 1

This is a Phase-III, multicentre, randomised, double-blind study (NCT01166282) described by six conference abstracts (Burgos-Vargas and colleagues88–93) in children aged ≥ 6 to < 18 years with ERA based on ILAR criteria, with active disease not responsive to ≥ 1 NSAID and ≥ 1 DMARD. No full paper appears to have been published so far, and the six abstracts provide limited information, hence preventing a full assessment of the methodology and trial quality and risk of bias. In addition, baseline characteristics were reported only for the overall trial population and not separately for each randomised group. The estimated study completion date was December 2015; however, the study was published in full after the submission date of this report in July 2015. 102

Forty-six patients were randomised in a 2 : 1 ratio (adalimumab n = 31; placebo n = 15) to receive blinded adalimumab (24 mg/m² body surface area up to 40 mg every other week) or placebo for 12 weeks followed by open-label adalimumab every other week for up to 144 weeks. It is unclear whether or not patients also received methotrexate. A table in one of the abstracts93 shows that 11/15 placebo patients and 21/31 adalimumab patients received DMARDs at baseline.

The primary endpoint of this study was per cent change from baseline in the number of active joints with arthritis (active joint count) at week 12 and secondary variables assessed included enthesitis count, tender and swollen joint counts, and ACR Pedi-30/-50/-70 responses. Active disease was defined as ≥ 3 active joints (swelling or loss of motion and pain/tenderness) and enthesitis in ≥ 1 location (past or present). Safety was assessed in terms of AEs, laboratory values and vital sign measurements. Some interim data were reported for 52 weeks including discontinuation of concomitant medication (at the discretion of the treating physician).

Authors state that no children discontinued the double-blind period, while at the same time reporting that seven children ‘escaped early’ to open-label adalimumab.

Adalimumab ongoing trial 2

The second adalimumab RCT (SYCAMORE, ISRCTN10065623)87 is funded by the NIHR HTA programme and Arthritis Research UK. The study is assessing adalimumab combined with methotrexate compared with placebo combined with methotrexate for JIA-associated uveitis in participants aged between 2 and 18 years. All participants will receive 18 months of treatment with a 3-year follow-up. The study will also include an assessment of cost-effectiveness. Originally expected to report findings in 2020, it has recently been announced that the trial has closed for recruitment early following interim analysis showing a favourable effect for adalimumab. Analysis of the primary outcome is under way and key findings will therefore be available earlier than expected. Collection and analysis of health economic data will continue as planned.

Adalimumab ongoing trial 3

This third adalimumab RCT [Effect of Adalimumab for the Treatment of Uveitis in Juvenile Idiopathic Arthritis (ADJUVITE); NCT01385826] is also currently in progress and is not expected to report findings until June 2016. The study is set in France (seven hospital ophthalmology departments) and assesses the efficacy of 2-month adalimumab treatment versus placebo treatment on reduction of ocular inflammation quantified by laser flare photometry in patients aged ≥ 4 years, with JIA-associated uveitis that is resistant to steroid therapy. The investigators plan to include 40 patients, follow-up appears to be 12 months and the final data collection date for the primary outcome measure is November 2015. The primary endpoint of this study is improvement of uveitis.

Etanercept ongoing trial 1

This is a RCT evaluating the efficacy of etanercept in 124 Chinese JIA patients (no clinical trial registration number has been found for this study). 98 No full paper appears to have been published so far and only one conference abstract was identified, which includes very limited information. The abstract states that a ‘randomised principle was applied’ to divide the JIA patients into a control and a treatment group. Although no baseline characteristics were reported, the authors of the abstract state that there were no significant differences of clinical classification and basic treatment between the groups.

Sixty-two patients in the treatment group (oligoarticular JIA n = 17, polyarticular JIA n = 15 and systemic JIA n = 30) received 0.8 mg/kg per week of subcutaneous etanercept for 6 months. No details for the control group are reported.

American College of Rheumatology Pedi-30, -50 and -70 responses were used to assess the clinical efficacy (primary outcome not stated) and adverse reactions were recorded.

Etanercept ongoing trial 2

The second placebo-controlled etanercept RCT [Remission Induction by Etanercept in Enthesitis related Arthritis JIA-Patients (REMINDER) Study, EudraCT Number: 2010–020423–51] was identified from the search of ongoing clinical trials registers. The trial is set in Germany and has a start date of February 2016. This study has a withdrawal RCT design, with a 12-week open-label treatment phase prior to the controlled randomised double-blind phase The study will assess the safety and effectiveness of etanercept in patients diagnosed with ERA-JIA age ≥ 6 years and < 18 years having met all criteria for eligibility for treatment with etanercept according to SPC and local guidelines, with expectation of the requirement of a minimum of five affected joints. The online record does not provide the treatment time for the double-blind phase or report any follow-up period. The primary endpoint of the study will be inactive disease of ERA-JIA.

Etanercept ongoing trial 3

The third multicentre etanercept RCT was identified from a conference abstract103 which does not provide a clinical trial registration number. This trial appears to be set in Germany, has enrolled patients with ERA and has a withdrawal design, with a 24-week open-label etanercept treatment phase prior to the 24-week placebo-controlled double-blind withdrawal phase. Patients had to achieve at least an ACR Pedi-30 response in order to be randomised to the double-blind phase (terminated in case of a disease flare or at week 48, whichever occurs earlier). No details of the study’s inclusion or exclusion criteria are reported in the abstract.

Etanercept ongoing trial 4

The fourth multicentre etanercept RCT (set in the Netherlands) assessed when and in whom to stop etanercept after successful treatment of JIA (NCT01287715) and was identified from the search of ongoing clinical trials registers. Estimated enrolment was 50, the study completed in September 2013 and final data collection for the primary outcome measure is reported as September 2012 in the online record. No publication of data has been identified. Patients aged 4–17 years and in remission were selected from the National Arthritis and Biologicals in Children (ABC) register, an observational study including all Dutch JIA patients on etanercept therapy. The inclusion criteria states no or low dose of methotrexate and it is therefore unclear if patients were intolerant or had a previous inadequate response to methotrexate. All JIA subtypes were included in the study. Patients were randomised to a stop arm (discontinuation of etanercept – half of the dose for 3 months and discontinuation thereafter) or a control arm (etanercept continued for another 9 months and, if still meeting the eligibility criteria, discontinued thereafter). The primary outcome of the study was flare rate.

Enthesitis-related arthritis and psoriatic arthritis

The most informative study available for these subtypes is the CLIPPER study. 99

This is a single-arm, Phase-IIIb open-label, multicentre interventional study funded by Wyeth (subsequently acquired by Pfizer Inc.). The study was designed to assess the safety and efficacy of etanercept in children and adolescents with three JIA subtypes classified using the ILAR criteria: EO, ERA and PA. There are two parts to the study. Part 1 (which has been published99) has investigated the efficacy and safety of etanercept in the three JIA subtypes over an initial 12-week period with a primary endpoint of the percentage of patients achieving ACR Pedi-30 criteria at week 12. Part 2 of the study is a 96-week OLE, assessing the long-term safety and efficacy of etanercept in JIA subtypes, which is currently published in poster format only. 104 This study formed part of the evidence base supporting the licence extension for etanercept across the JIA subtypes in 2012. The assessment group has extracted 12-week data from the published paper99 on ACR Pedi response rates and inactive disease, change in CHAQ score, PGA of pain, number of active joints, number of joints with LOM and JIA category-specific assessments (data at week 12 compared with historical placebo data, historical active control data and data from a meta-analysis have not been data extracted). Data from the conference poster104 on ACR Pedi response rates and inactive disease have also been extracted and this summary is supplemented with some data presented in the Pfizer CS.

The study included 127 patients with the JIA subtypes of ERA (n = 38, age 12–17 years), EO (n = 60, age 2–17 years) and PA (n = 29, age 12–17 years) who received 0.8 mg/kg of etanercept once weekly (maximum dose 50 mg/week). Key inclusion criteria were ≥ 2 active joints (swollen or LOM accompanied by either pain or tenderness); history of intolerance or unsatisfactory response to at least a 3-month course of ≥ 1 DMARD or, for ERA only, unsatisfactory response to at least a 1-month course of ≥ 1 NSAID (i.e. for ERA, prior methotrexate treatment was not required). A stable dose of concomitant medication (only one DMARD, one oral corticosteroid and one NSAID) was permitted. The inclusion criteria extended below the threshold number of active joints for the classification of polyarticular disease. Key exclusion criteria included other rheumatic diseases, active uveitis within 6 months of baseline and any prior receipt of biologic DMARDs. A total of five patients failed to complete part 1 of the study (completed part 1: EO 97%, ERA 95% and PA 97%) and 13 patients part 2 (completed part 2: EO 90%, ERA 79% and PA 86%).

Key baseline characteristics can be seen in Table 33 (additional baseline characteristics are available in the published paper99).

Patients with EO and PA had a higher number of active joints and number of joints with LOM [EO: mean 7.6 (SD 5.1) and 6.3 (SD 4.4), respectively; PA: mean 7.0 (SD 4.3) and 5.6 (SD 4.1), respectively] at baseline than ERA patients [mean 5.2 (SD 3.6) and 4.8 (SD 4.0), respectively]. The number of painful joints was highest in PA patients [mean 7.8 (SD 7.0)] compared with the other two subgroups, and the number of swollen joints was the lowest in ERA patients [mean 3.8 (SD 2.8)] (Table 34). Mean CHAQ subgroup scores ranged between 0.7 and 0.9, and the parent global assessment of pain VAS ranged between 4.6 and 5.8. Also reported are JIA category-specific assessments [ERA: tender entheseal score, back pain (VAS) and modified Schober’s test; PA: body surface area and PGA of psoriasis] at baseline. Limitations of the study noted by the authors were the difference in concomitant medication use at baseline which may have affected efficacy responses and the lower age limit of 12 years set for inclusion of EO patients (the licensed indication is from 2 years of age).

Juvenile idiopathic arthritis-associated uveitis

As stated earlier, the effects of biologic DMARD on extra-articular manifestations such as uveitis were not assessed by the included RCTs. However, evidence from non-randomised studies is available, as summarised by systematic reviews.

A recently published systematic review by Simonini and colleagues96 assessed the effectiveness of antiTNF drugs for childhood uveitis. To be included, studies had to include patients with autoimmune uveitis refractory to topical and/or systemic steroids and at least one immunosuppressive therapy (e.g. methotrexate). The antiTNFα drugs of relevance to the review were etanercept, infliximab and adalimumab (infliximab is not within the scope of this NICE appraisal). The primary outcome was improvement in intraocular inflammation, with additional outcomes including tapering/stopping systemic steroid administration, improvement in visual acuity and treatment discontinuation among others. A number of bibliographic databases were searched from January 2000 to October 2012.

The review included 23 studies, mainly retrospective chart reviews with very small patient numbers. Of these 23 studies, only 7 were conducted exclusively in JIA uveitis patients (one RCT of etanercept; two retrospective studies of etanercept; two retrospective studies of infliximab; and two retrospective studies of adalimumab). Eleven studies comprised mixed study populations with uveitis associated with a range of conditions, including JIA. The remaining five studies included populations that did not include any children with JIA. It was not possible to analyse results separately by uveitis-associated condition. However, of the 229 children included across all the studies, 152 had chronic uveitis associated with JIA. The results can therefore be interpreted as being generally relevant to JIA uveitis.

A pooled analysis of the observational studies found that adalimumab and infliximab were more efficacious at improving intraocular inflammation than etanercept. The proportion of children with improved intraocular inflammation (responders) was 87% (95% CI 75% to 98%) for adalimumab, 72% (95% CI 64% to 79%) for infliximab, and 33% (95% CI 19% to 47%) for etanercept. There was no statistically significant difference in the proportion of responders between adalimumab and infliximab (p = 0.08), but there was a significant difference for both compared with etanercept (p = 0.001 for both comparisons).

Simonini and colleagues96 did not pool the results of the single RCT identified in the systematic review with the observational studies. 56 This was a small RCT (n = 12 children) of treatment with etanercept. The authors state that this study did not report substantial benefits for the biologic treatment. (Owing to limitations in reporting, this RCT was judged to be unclear for inclusion in our systematic review of clinical effectiveness, as it was not clear whether or not the etanercept was given within its licensed indication). Caution is advised in the interpretation of the findings of the Simonini and colleagues96 systematic review given the weaknesses of the study designs included.

The assessment group are aware of only one other recent systematic review of biologic DMARD treatment of children with uveitis, by Cordero-Coma and colleagues. 105 The most recent search date for literature was October 2011. This review had a broader inclusion criteria than that of Simonini and colleagues96 and a total of 61 studies was included. Again, much of the included evidence was from observational studies. A total of 14 studies assessed adalimumab, 11 assessed etanercept and 50 studies assessed infliximab [studies assessing certolizumab and golimumab were also included]. Of the 1093 patients included across the studies, 316 (30%) were classed as having JIA uveitis. The review does not provide any formal synthesis and quantification of the effectiveness of treatment, but provides a narrative conclusion for each biologic DMARD and a level of evidence. Adalimumab and infliximab were considered by the authors to be effective in autoimmune uveitis, both based on level-2b evidence (individual cohort study or low-quality RCT). Etanercept was judged ineffective, based on level-1b evidence (individual RCTs with narrow CIs).

The Abbvie CS to NICE77 provides narrative summaries of five selected studies published since the Simonini and colleagues96 systematic review. All of them were observational in design (three case series;106–108 one Italian registry-based study;109 one comparative cohort study110), and all assessed treatment with adalimumab (with one also assessing infliximab110 in uveitis patient populations with varying proportions of JIA uveitis. We have not performed an independent critical appraisal of these studies in this assessment report. From the summaries given it appears that adalimumab is associated with improvements in intraocular inflammation and visual acuity, and a decrease in use of corticosteroids. AEs appeared to be minor. The other CSs to NICE did not present much detail of studies of treatment of JIA uveitis with other biologic DMARDs.

In summary, the evidence from observational studies suggests that biologic DMARDs can improve uveitis symptoms in children with JIA, such as intraocular inflammation. Adalimumab and infliximab appear to be more effective than etanercept in improving uveitis. The effects of the treatments in terms of arthritis outcomes in JIA uveitis patients have not been reported. As noted above (see Ongoing trials), the UK-based SYCAMORE RCT87 has investigated adalimumab in the treatment of JIA uveitis patients, and the results of the trial (which will be available sooner than expected, although an exact date has not been specified) will provide more rigorous evidence for effectiveness than that currently available.

Methods for the systematic review

A systematic literature search was undertaken to identify economic evaluations of the biologic DMARDs, within the NICE scope for this appraisal. Studies were included if they were full economic evaluations (cost-effectiveness, cost–utility, cost–consequence or cost–benefit analyses) conducted in children and young people with JIA that compared one or more biologics with a DMARD, such as methotrexate. Studies that were not reported in the English language or that did not provide sufficient information on the model structure, data and results were excluded. This systematic review aimed to summarise the currently available evidence and inform the construction of a de novo model.

Results of the systematic review

Searches for economic evaluations identified 387 potentially relevant references, and a further study was identified through ad hoc searching. The full texts for 17 papers were retrieved for further screening. A summary of the selection process and the reasons for exclusion are presented in Figure 6 and a list of excluded studies can be found in Appendix 6. Although seven studies reported as abstracts appeared to meet the a priori inclusion criteria, they did not contain sufficient information on the methods used and the results to permit formal data extraction or critical appraisal. 111–117 Five studies were found not to be economic evaluations. 118–122 Four studies were included, described in a total of five publications. 123–127 The characteristics of the four included economic evaluations are shown in Table 37. Data extraction forms for the studies can be found in Appendix 7.

FIGURE 6.

Flow chart of identification of studies for inclusion in the review of cost-effectiveness. a, including one study found through hand searching; b, the abstracts provided insufficient details of methods and results to allow inclusion in the systematic review.

Critical appraisal of the studies

The cost-effectiveness studies were assessed using a critical appraisal checklist (Table 38). The checklist assessed study quality and generalisability to the UK. The checklist was adapted by the review authors from checklists by Philips and colleagues,128 Drummond and colleagues129 and methodological requirements stated in the NICE reference case. 130

The Cummins and colleagues study was conducted in the UK,123 whereas the generalisability of the other studies to the NHS is unclear. The other studies were conducted in the Netherlands,124 Russia125 and Canada;127 used appropriate modelling methodology; and included relevant costs.

In terms of the analytical and modelling methodology used, the studies were generally considered appropriate, except for the model reported in Cummins and colleagues,123 which was based upon a number of questionable assumptions as a result of limitations in the data available at the time.

The data inputs for the model were clearly described and justified by two studies,123,127 but the description of some of the data inputs are missing from Simpson and colleagues. 125 Prince and colleagues124 conducted a cost–consequence analysis based on a prospective observational study that collected cost and utility data. The study did not measure quality-adjusted life-years (QALYs).

Two of the studies, Cummins and colleagues123 and Simpson and colleagues,125 used appropriate time horizons, measured the health outcomes in QALYs and discounted costs and outcomes. The model by Ungar and colleagues127 did not use QALYs in the model and used a 1-year time horizon, eliminating the need for discounting.

All three modelling studies analysed results incrementally and assessed uncertainty through sensitivity analyses. 123,125,127

In summary, the cost-effectiveness studies have certain limitations with regard to methodology, reporting of results or generalisability to the UK NHS (see Table 38).

Methods for the systematic review

A systematic literature review was undertaken to assess the HRQoL of people with JIA treated with biologic DMARDs. The aim of the review was to provide data to populate the de novo economic model in this report with health-state utility values to calculate QALYs. The description of the search strategy is shown in Appendix 1. The inclusion criteria included primary studies that investigated HRQoL in people with JIA. To be eligible, the study should report health utility values using any generic preference-based HRQoL measure [e.g. EQ-5D, Short Form questionnaire-6 Dimensions (SF-6D)] or choice-based valuation methods (e.g. time trade-off, standard gamble). Studies that were not reported in English or did not provide sufficient information were excluded. The methodology used for searching and data extraction is outlined in Chapter 3 of this assessment report.

Results of the systematic review

The database searches identified 2249 references, with one further study retrieved by hand searching, making the total number of references identified 2250. Full-text papers for 28 references were retrieved, meeting the a priori inclusion criteria. Figure 7 presents a flow chart of the selection process and the excluded studies, with reasons for exclusion listed in Appendix 8. A total of 6 references were considered to have insufficient information on the study methods, population and results, 9 included an inappropriate population and 10 did not report a relevant outcome measure. Two studies, described in three publications, met the inclusion criteria and the characteristics of these studies are presented in Table 39. Data extraction forms for the included studies can be found in Appendix 9.

FIGURE 7.

Flow chart of identification of studies for inclusion in the review of QoL studies.

The two HRQoL studies are each now described in more detail.

Summary and conclusions of the health-related quality-of-life review

The included studies assessed the HRQoL of children and adolescents with JIA, applying EQ-5D and HUI3 preference-based utility measures. Although both studies reported utility values, they are not directly comparable. The study by Hendry and colleagues136 assessed the effectiveness of a foot care programme in a RCT, whereas Prince and colleagues124,137 conducted an observational study reporting HRQoL and costs from patients in the Dutch ABC registry to assess the effect of treating patients with etanercept. The mean utility values reported for baseline by Hendry and colleagues136 for the intervention and control group (0.57 and 0.58, respectively) are relatively similar to the baseline values reported by the Prince and colleagues study (0.51 and 0.53, respectively). 124,137 The HRQoL values may be higher for Hendry and colleagues136 owing to 23% of patients receiving etanercept, whereas no patients received etanercept at baseline in the Prince and colleagues study. 124,137 The sample size of both cohorts is considered relatively small, but it is reasonable given the population group. The cohort used in the Prince and colleagues study124,137 included patients with systemic JIA, who are outside the scope of the current review; however, the proportion of patients within the group with systemic JIA is relatively small (< 30%). Neither of the studies can be considered fully informative for the de novo economic evaluation in this assessment report. However, estimates provided by Prince and colleagues124,137 are considered reasonably appropriate for use in the economic evaluation, despite not being considered directly generalisable to the UK population.

Review of Bristol-Myers Squibb’s submission to the National Institute for Health and Care Excellence (abatacept)

The company submitted a de novo economic model that included all comparators specified in the NICE scope except for methotrexate monotherapy (i.e. abatacept, etanercept, adalimumab and tocilizumab). 85 The company states that methotrexate monotherapy was not included owing to inconsistency with the clinical effectiveness data (i.e. all patients in the RCTs either did not have sufficient response with methotrexate or were refractory to methotrexate). The scope reflects the licensed indication of abatacept, namely polyarticular JIA patients aged ≥ 6 years who have received at least one TNFα inhibitor (etanercept or adalimumab). All included drugs were assumed to be administered with subcutaneous methotrexate.

Review of the Roche submission to the National Institute for Health and Care Excellence (tocilizumab)

The CS includes an economic model and reports the total costs, the QALYs gained and cost-effectiveness of tocilizumab in the treatment of polyarticular JIA. 78 The model evaluates the lifetime costs and benefits for tocilizumab compared with adalimumab. The perspective of the analysis is that of the NHS and Personal Social Services (PSS).

Review of the AbbVie submission to the National Institute for Health and Care Excellence (adalimumab)

The company did not provide a systematic review of cost-effectiveness studies or an economic evaluation. 77 They discussed the interventions in the NICE scope: adalimumab, etanercept, abatacept, tocilizumab and methotrexate. In sections 5.8–5.10 of the CS, the company provides justifications for not conducting an economic evaluation. Other sections of the CS provide details on what the company would consider important in conducting an economic evaluation in JIA, including an evaluation of the costs associated with surgeries and vision loss.

The company states that an economic evaluation was not conducted owing to a lack of appropriate utility data for HRQoL, heterogeneity in study methods and populations between the interventions that complicated indirect comparisons, and a lack of long-term effectiveness data. The company identified one HRQoL study (Prince and colleagues124), which collected HUI3 utilities in addition to other JIA clinical variables, such as the CHAQ score. The data collected from Prince and colleagues124 were deemed unsuitable to map the CHAQ to HUI3 owing to insufficient sample size, but were considered the most suitable source of utility data by the company. The CS discusses the use of an algorithm by Khan and colleagues141 that mapped the Pediatric Quality of Life Inventory instrument to EQ-5D in secondary school pupils. The company notes the potential limitations in the use of this method in JIA, as Pediatric Quality of Life data were not collected in any of the JIA biologic DMARD RCTs. The company considers that the biologic DMARD trial populations and study methods were not sufficiently similar to allow indirect comparison through network meta-analysis. The CS concluded that using current data and methods would lead to ‘untenable’ uncertainty (CS page 90).

Review of the Pfizer submission to the National Institute for Health and Care Excellence (etanercept)

Pfizer did not submit any cost-effectiveness evidence. 97 The CS notes the limitations raised in the previous submissions for NICE of etanercept TA3543 and tocilizumab TA238. 44 These relate to the limitations in the HRQoL data and the limited evidence on the long-term outcomes and the effectiveness of the treatments. The CS states that any cost-effectiveness evidence would be associated with considerable and unresolvable uncertainty. The company submitted a cost analysis that compared the annual costs for the first year of treatment based on etanercept against adalimumab and tocilizumab in patients with polyarticular JIA. The CS states that the cost analysis showed that etanercept is the biologic DMARD with the lowest acquisition cost compared with list prices for tocilizumab and adalimumab.

Comparison of economic models in company submissions

The CSs differ in the approach to providing economic evidence for biologic DMARDs for JIA. Only one company (Roche) constructed an economic model for tocilizumab that included both costs and outcomes. Two companies (BMS for abatacept and Pfizer for etanercept) submitted cost analyses and assumed that the biologic DMARDs were equivalent, whereas the remaining company (AbbVie for adalimumab) considered there to be too many limitations with any potential analysis and, therefore, did not submit an economic analysis.

Although AbbVie has raised valid concerns about uncertainty in the data available for conducting an economic evaluation, we consider that concerns about uncertainty are an insufficient justification for not building an economic model. A model provides a representation of current knowledge in a subject and uncertainty is part of that current knowledge. A model, even an uncertain one, with limitations noted by the company, gives a more transparent description of available knowledge and enables more informed decision-making than simply presenting clinical trial data from trials that represent only a highly selected subgroup of the drug licences. Modelling also allows the exposure of the most valuable areas for future research enquiries.

Bristol-Myers Squibb and Pfizer consider that all of the treatments are equivalent. It is noted that the available evidence base consists of small trials that lack the statistical power to justify this assumption, and Otten and colleagues79 do not conclude that there is equivalent efficacy between the treatments, but that the treatments are similar.

Briggs and O’Brien142 argue that a cost-minimisation analysis should be conducted only when equivalence of comparators has been statistically demonstrated. Dakin and Wordsworth143 argue that the limitations of a cost-minimisation analysis do not allow for an appropriate assessment of uncertainty or for the value of future research and may lead to biased conclusions. It is also the case that equivalence of one clinical outcome does not mean equivalence of all clinical outcomes. Patients may have the same QoL on treatment but have different adherence and discontinuation or different AEs, for example. For these reasons, a cost-minimisation analysis is generally forgone in favour of a cost-effectiveness analysis and/or a cost–utility analysis.

Roche have provided a cost–utility analysis that compared two of the biologic DMARDs (adalimumab and tocilizumab). The model appears to be a reasonable attempt at modelling JIA, albeit in only two of the biologic DMARDs. However, we have noted errors in the calculation of QALYs in the model, which limit the credibility of the results.

Evaluation of uncertainty

The evaluation of the cost-effectiveness of JIA treatment is based on uncertain information about variables such as clinical effect, HRQoL and resource use. This uncertainty was evaluated using deterministic and PSA. One-way deterministic sensitivity analyses were conducted to evaluate the influence of individual parameters on the model results and to test the robustness of the cost-effectiveness results to variations in the structural assumptions and parameter inputs (see Results of the independent economic analysis). Where possible, the parameters were varied according to the ranges of the CIs of these parameters based on the published estimates. Where these data were not available an alternative range was chosen.

Multiparameter uncertainty in the model was addressed using PSA (Results of the independent economic analysis). 145 In the PSA, probability distributions are assigned to the parameter point estimates used in the base-case analysis. The model is run for 1000 iterations, with a different set of parameter values for each iteration by sampling parameter values at random from their probability distributions.

The uncertainty surrounding the cost-effectiveness of the treatment is represented on a cost-effectiveness acceptability curve according to the probability that the intervention will be cost-effective at a particular willingness-to-pay threshold. Appendix 11 reports the parameters included in the PSA, the form of distribution used for sampling each parameter and the upper and lower limits assumed for each variable.

Model validation

The model was validated by checking the model structure, calculations and data inputs for technical correctness by another researcher. The structure was reviewed by clinical experts from the advisory group for its appropriateness for the disease and its treatment. A senior health economist from the advisory group reviewed the methods and assumptions of the economic evaluation. The robustness of the model to changes in input values was tested using sensitivity analyses to ensure that any changes to the input values produced changes to the results of the expected direction and magnitude.

Data sources

Licensed first-line biologics: adalimumab, etanercept and tocilizumab

The undiscounted summary results of the analyses for adalimumab, etanercept and tocilizumab compared with methotrexate for the treatment effects are shown in Tables 55–57. In the base case, total undiscounted QALYs vary between 14.98 for methotrexate and 17.99 for tocilizumab (see Table 55). Patients on methotrexate have higher QALYs only in the off-biologic-DMARD health state than the patients on biologics, as they spend more time in this health state. The summary results of the undiscounted drug costs are shown in Table 56. The total undiscounted drug acquisition cost of the biologic DMARDs varied between £103,497 and £128,071 for treatment first with etanercept and tocilizumab, respectively, compared with a total undiscounted cost of £7029 for patients treated with methotrexate only. The total patient costs varied between £107,299 and £225,797 for methotrexate only and tocilizumab, respectively. As noted earlier, patients taking etanercept do not receive methotrexate, which partially explains the lower costs for the etanercept regimen.

The base-case discounted cost-effectiveness results are shown in Table 57. Each of the biologic DMARDs is more expensive than methotrexate only, with the incremental cost ranging from £77,513 to £82,995 for etanercept and tocilizumab, respectively. The ICER versus methotrexate only for adalimumab, etanercept and tocilizumab is £38,127, £32,256 and £38,656 per QALY gained, respectively. The results are not presented as an incremental analysis of the biologic DMARDs, as the costs and QALYs for each biologic DMARD are generally similar and it has been previously discussed in Chapter 4 that the biologic DMARDs may be regarded as similar in effectiveness.

Licensed second-line biologic: abatacept

Abatacept is licensed for use after at least one previous antiTNF biologic DMARD. The results are shown for abatacept, adalimumab, etanercept and tocilizumab compared with methotrexate. For each biologic comparator, patients are assumed to have been treated initially with etanercept as the first-line biologic. The summary results of the non-discounted treatment effects are shown in Table 58. In the base case, total undiscounted QALYs vary between 14.98 for methotrexate and 20.07 for abatacept. The summary results of the undiscounted costs are shown in Table 59. The total undiscounted drug acquisition cost of the DMARDs varied between £7029 for methotrexate only to £222,533 for abatacept. The total patient costs varied between £107,299 and £317,097 for methotrexate only and abatacept.

The base-case discounted cost-effectiveness results are shown in Table 60. The costs and QALYs are different from those for the first-line biologic cost-effectiveness analysis (see Table 57) because this analysis includes the costs and QALYs of two lines of biologics. The cost-effectiveness of abatacept compared with methotrexate is £39,536 per QALY. The results are not presented as an incremental analysis, as the costs and QALYs for the biologic DMARDs are similar.

Sensitivity analysis

Scenario analysis

We conducted several scenario analyses to investigate uncertainty for specific aspects of the modelling. The results of these analyses are presented for the first-line biologics.

Probabilistic sensitivity analyses

In the PSA, all parameters were sampled probabilistically from an appropriate distribution using similar ranges as used in the deterministic sensitivity analyses. The parameters sampled were: treatment effectiveness, discontinuation rate, health-state costs, disease flare parameters and HRQoL. The distribution assigned to each variable included in the PSA and the parameters of the distributions are reported in Appendix 11.

First-line biologics

A total of 1000 simulations were run. The PSA results are presented in Table 72 for first-line biologics and show similar results to the deterministic analyses (see Table 58). The cost-effectiveness for biologics versus methotrexate only varied between £32,554 and £38,744 per QALY for tocilizumab.

TABLE 72 - Summary of the probabilistic sensitivity results for first-line biologics vs. methotrexate only

Treatment	Costs, £	QALYs	Incremental costs, £	Incremental QALYs	ICER (£/QALY)
Methotrexate only	67,531	9.38
Adalimumab	145,933	11.43	78,402	2.05	38,181
Etanercept	135,803	11.48	68,272	2.10	32,554
Tocilizumab	151,800	11.55	84,269	2.18	38,744

The cost-effectiveness acceptability curve is shown in Figure 9 and indicates that at the £20,000 and £30,000 willingness-to-pay thresholds methotrexate has the highest probability of being cost-effective, at 0.98 and 0.62, respectively.

FIGURE 9.

Cost-effectiveness acceptability curve from the PSA for first-line biological treatments compared with methotrexate.

Second-line biologics

The PSA results are presented in Table 73 for second-line biologics and show similar results to the deterministic analyses (see Table 61). The cost-effectiveness of abatacept in the PSA is £39,608 per QALY.

TABLE 73 - Summary of the probabilistic sensitivity results for second-line biologics vs. methotrexate only

Treatment	Costs, £	QALYs	Incremental costs, £	Incremental QALYs	ICER (£/QALY)
Methotrexate only	67,168	9.35
Abatacept	203,396	12.81	136,041	3.43	39,608
Adalimumab	183,563	12.66	116,208	3.29	35,366
Etanercept	179,807	12.67	112,452	3.30	34,053
Tocilizumab	194,464	12.77	127,109	3.39	37,443

The cost-effectiveness acceptability curve is shown in Figure 10 and indicates that at the £20,000 and £30,000 willingness-to-pay thresholds methotrexate has the highest probability of being cost-effective, at 0.99 and 0.71, respectively.

FIGURE 10.

Cost-effectiveness acceptability curve from the PSA for second-line biological treatments compared with methotrexate.

Subgroups

There are a number of potential subgroups that were within the NICE scope, including the subtypes of JIA (EO, polyarticular arthritis, ERA and PA) and patients with extra-articular manifestations such as uveitis. As stated earlier, subgroup analyses by subtype of JIA was not possible owing to insufficient evidence for input parameters to support modelling. The modelled patient population is therefore people with JIA, with the results of particular relevance to those with polyarticular-course JIA (EO, and RF+ve and RF–ve polyarthritis).

In considering the potential for modelling the clinical effectiveness and cost-effectiveness of biologic DMARDs in patients with JIA-associated uveitis, a draft NHS clinical commissioning policy on the use of antiTNFα agents in paediatric patients with severe refractory uveitis was consulted. 29 The policy discusses the cost-effectiveness of treatment and the elements of an economic evaluation are given, although the full results from such an economic evaluation have not been reported.

The report states that infliximab and adalimumab in combination with methotrexate are widely used worldwide for the treatment of refractory uveitis, and that etanercept is not recommended for use in this patient group. The report also cites evidence from a systematic review by Simonini and colleagues96 (see Chapter 4, Juvenile idiopathic arthritis-associated uveitis), which shows that, based on a pooled analysis of observational studies, the proportion of children with improved intraocular inflammation (responders) was 87% for adalimumab, 72% for infliximab and 33% for etanercept. Potential modelling of the clinical effectiveness and cost-effectiveness in JIA-associated uveitis in this report would therefore apply only to adalimumab, as this is the only one of the four biologic DMARDs within the scope of the appraisal recommended for treating this patient subgroup.

With regard to QoL, the clinical commissioning policy assumes that loss of vision causes detrimental effects on utility based on the results of a study of age-related macular degeneration by Reeves and colleagues. 165 This study measured HRQoL changes associated with loss of vision using data from the SF-6D and best corrected visual acuity. This population is quite different from JIA-associated uveitis, and the data do not capture aspects of JIA related to arthritic joints.

In our model, we have used utility data derived using the HUI3 generic preference instrument from the study by Prince and colleagues. 124 This instrument is appropriate for conditions that involve vision impairment, as it includes a domain for vision. HUI3 is not compatible with the SF-6D and the instruments will produce different QoL estimates. Attempting to combine data from Reeves and colleagues165 and Prince and colleagues124 would be inappropriate owing to the differences in the populations of the studies and the incompatibility of SF-6D and HUI3. Moreover, if it is assumed that adding vision loss to the other QoL decrements owing to advancing JIA even partially decreases patient QoL, then it follows that adalimumab will be more cost-effective in JIA patients with uveitis and joint inflammation than it is in JIA patients without uveitis.

Likewise, if most of the costs related to uveitis relate to the management of vision loss, as stated in the clinical commissioning policy,29 then any reduction of these costs attributable to improving vision would increase cost-effectiveness in the subgroup of JIA patients with uveitis. Any additional analysis of cost-effectiveness in a JIA uveitis population that is refractory to methotrexate, as is indicated in the licensing for adalimumab, is therefore likely to have predictable results.

As discussed Chapter 4, the SYCAMORE trial of adalimumab and methotrexate in JIA-associated uveitis patients has recently closed early following an interim analysis that showed a favourable effect for treatment. 87 The trial also includes a cost-effectiveness analysis, the results of which would be likely to concur with the logical implications discussed above.

Clinical effectiveness

The systematic review of clinical effectiveness conducted for this report found that biologic DMARDs are superior to placebo (with methotrexate where permitted) across a number of outcome measures in children with JIA (predominantly polyarticular course) and an insufficient response to previous DMARD treatment. With the exception of the etanercept trial, the majority of patients in the trials received methotrexate in addition to the biologic DMARD/placebo. Biologic DMARD-treated patients had fewer arthritis flares, a longer time to disease flare (applicable to abatacept, adalimumab and etanercept), were more likely to achieve a treatment response as defined by the ACR Pedi criteria and were more likely to have inactive disease (measured only in the abatacept and tocilizumab trials). This last outcome can be considered to be the most clinically significant, as absence of disease activity (e.g. no joints with active arthritis; PGA indicates no disease activity, etc.) is a key treatment goal. Treatment was associated with reduced pain scores, although this was reported as statistically significant in only one study (tocilizumab). HRQoL as measured by the CHAQ appeared to be higher for treated patients, although this was not always statistically significant.

The percentage of patients achieving ACR Pedi-30 in the open-label lead-in phases of the RCTs ranged from 65% to 94% across the trials. It should be acknowledged that owing to the withdrawal design of the RCTs, in which only patients achieving an ACR Pedi-30 response during the open-label lead-in phase are eligible for randomisation, the results of the double-blind randomised phase of the trials are therefore applicable only to patients who have achieved an initial degree of treatment benefit. The effects seen during the double-blind period in the placebo group may not necessarily be the same for a placebo group who had not received a biologic DMARD prior to randomisation. However, expert clinical opinion suggests that ACR Pedi-30 can be considered an inadequate or partial response threshold, and higher rates, such as ACR Pedi-70 or above, are considered more clinically significant. In this respect, the patients responding to ACR Pedi-30 in the open-label lead-in phase (and eligible to be randomised) may not necessarily be considered atypical of patients eligible for treatment in clinical practice, as both would have active disease.

The clinical significance of the ACR Pedi-30 results of the randomised phases of the trials may also be questioned. ACR Pedi-30 response rates varied from 63% to 80% across the trials and declined with increasing response thresholds. Nonetheless, at ACR Pedi-70 (the highest threshold for which data were available across all four RCTs), the response rate varied from 44% to 65% and remained higher in biologic DMARD-treated patients than placebo patients in all trials. Research is under way to further develop the JADAS tool as a clinically useful measurement tool,35–37 although clinical trials are continuing to use the ACR Pedi criteria, albeit with effectiveness judged at thresholds higher than ACR Pedi-30.

In the longer term, treatment effectiveness, in terms of ACR Pedi response, appears to be sustained, as reported in the observational OLE studies for all four included RCTs. The longest follow-up available is for etanercept, where ACR Pedi responses were maintained up to 8 years after treatment.

The occurrence of AEs was generally similar between biologic DMARD- and placebo-treated patients, based on reported non-statistically significant differences. A range of AEs were reported, including viral and upper respiratory tract infections, injection-site reactions and nasopharyngitis. Serious AEs were uncommon. Discontinuations attributable to AEs were also uncommon (< 3% patients). In the lead-in phase of the RCTs, discontinuations attributable to AEs were low, ranging from 0.5% to 1.8%. The incidence of AEs and SAEs during open-label long-term follow-up did not appear to be excessive. The safety profile of the biologic DMARDs, therefore, appears to be relatively favourable.

Subgroup analyses were reported in only one of the included RCTs (tocilizumab). 68 Patients receiving methotrexate background therapy had higher ACR Pedi response rates than those not receiving it, as did patients receiving background glucocorticoids. Patients who had received previous treatment with a biologic agent had lower ACR Pedi responses than those who were naive to biologic DMARDs. It is not clear whether these subgroup analyses were pre-planned or post hoc, so caution is advised in their interpretation.

Two recently published systematic reviews of the effectiveness of biologic DMARDs were identified during the production of this report. 167,168 Both of these included a range of biologic DMARDs, including the four relevant to the scope of this assessment. However, none of these reviews identified any additional RCT evidence to this assessment report. The only other relevant published systematic review of biologic DMARDs that we are aware of is by Otten and colleagues,79 (most recent search date January 2012). As discussed earlier in this report [see Chapter 4, Assessment of clinical effectiveness: biologic disease-modifying antirheumatic drugs versus each other (with methotrexate where permitted)], Otten and colleagues79 conducted an adjusted indirect comparison of adalimumab, abatacept and etanercept, using the same RCTs as included in this assessment report. We replicated the indirect comparison, extending it to include the tocilizumab RCT,68 which was not published during the timescale of the Otten and colleagues79 review. Our results and conclusions match those of Otten and colleagues,79 namely that the biologic DMARDs appear similar in effectiveness in polyarticular-course JIA, in terms of ACR Pedi response and preventing disease flares. Otten and colleagues79 also share some of the caveats made in this assessment report about the limitations of the data included in the indirect comparison, namely, the small number of trials (and patient numbers), and differences between the trials in key patient characteristics and in treatment duration.

The conclusion that biologic DMARDs may be similar in clinical effectiveness was supported by the expert advisers to this assessment report. In their experience, there is similarity in effects between the drugs at a population level. However, it is noted that interpatient variation in effects may occur, and comparative effectiveness of the biologic DMARDs may potentially vary between JIA subtypes. Currently, there are a lack of clinical trial data to confirm this. Experts suggested that future trials of biologic DMARDs should stratify by disease phenotype to assess the differential effects of each treatment.

As noted earlier in this report, the RCTs of the biologic DMARDs included a mixture of JIA subtypes, broadly under the classification of polyarticular-course JIA (including EO). The trials did not appear to include patients with ERA or PA; thus, we reviewed available evidence from trials in progress (see Chapter 4, Ongoing trials) and from non-randomised studies (see Chapter 4, Additional supporting evidence) to gauge the effectiveness of biologic DMARD treatment in these groups. Much of the evidence is for etanercept (licensed for ERA and PA) with some available for adalimumab (licensed for PA). A broad comparison of the results of these studies with those of the RCTs included in this assessment report suggests that effectiveness is generally similar between these JIA subtypes. For example, ACR Pedi-70 response rates for biologic DMARDs were in the range of 44–65% across the RCTs (see Table 14), compared with around (CiC information has been removed) across the JIA subtypes in the CLIPPER study of etanercept99 (see Table 34) (notwithstanding differences in study variables such as length of follow-up). Evidence from trials in progress will provide greater clarity regarding the efficacy and safety of biologic DMARDs in these JIA subtypes. At present, there do not appear to be any studies of the comparative effectiveness of biologic DMARDs in these subtypes (e.g. adalimumab vs. etanercept).

All of the RCTs were multinational, with only one specifying that it included patients from the UK. The distribution of JIA subtypes within the trials, as far as reported (see Table 11), appears reasonably similar to that seen in UK registry studies (see Table 2), although this comparison may be limited by different reporting classifications used between studies. In addition, clinical practice in the RCTs (e.g. the oldest one published in 200042) may not necessarily reflect current NHS care. The generalisability of the RCTs to the NHS is considered uncertain.

Cost-effectiveness

A systematic search of the literature found four economic evaluations of biologic DMARDs for patients with JIA. Two of the studies were presented as cost–utility studies, one was a cost-effectiveness study and the other was a cost–consequence study. The evaluations were published between 2002 and 2012 in the UK, the Netherlands, Russia and Canada. One of the studies was the assessment report which informed the previous NICE appraisal of etanercept (NICE TA35). 123 The studies varied in design and structure, time horizons and the comparators included. There were limitations in the methodological quality in all the studies identified, and limited reporting of model parameters and assumptions.

A systematic search of the literature found two HRQoL studies in children and adolescents with JIA. One study assessed the effectiveness of a foot care programme in a RCT setting, whereas the other evaluated the QoL in a cohort of patients from the Dutch ABC Registry before and after treatment with etanercept.

Four drug companies submitted evidence to be considered as part of the NICE appraisal. Only one company (Roche) constructed a cost–utility analysis that included both costs and outcomes. Two companies (BMS and Pfizer) submitted cost analyses and assumed that the biologic DMARDs were equivalent in effectiveness, whereas another company (AbbVie) did not submit an economic analysis owing to suggested methodological limitations with any potential analysis. Roche submitted a Markov state-transition model with health states for uncontrolled/off treatment, on treatment and dead. The model compared treatment with adalimumab to that with tocilizumab. The base-case results from the submission conclude that tocilizumab is of similar effectiveness and is less expensive than adalimumab.

We developed an independent cost–utility model comparing the biologic DMARDs to methotrexate only. From the model, the incremental cost-effectiveness versus methotrexate only for adalimumab, etanercept and tocilizumab is estimated at £38,127, £32,526 and £38,656 per QALY gained, respectively. The incremental cost-effectiveness for abatacept as a second-line biologic was £39,536 per QALY gained. The model results are most sensitive to changes to the HRQoL utility values.

The cost-effectiveness of biologic DMARDs estimated in this report is associated with some uncertainty owing to the limitations of the evidence base. For this reason, assumptions have had to be made to simplify the modelling. There was limited evidence on HRQoL, in particular with regard to disease progression. The HRQoL utility values were taken from a small Dutch registry study of patients receiving etanercept. The HRQoL values for patients treated with methotrexate were assumed to be constant over time. Patients with JIA who do not receive a biologic will experience disease progression and, therefore, their HRQoL will decline over time. In the model, we have assumed a constant HRQoL utility value for patients receiving methotrexate only and so the biologic DMARDs would be more cost-effective than estimated by the economic model.

The model has not considered the underlying disease progression in terms of joint damage for patients with JIA. These patients may have sustained permanent damage to one or more joints, thereby affecting their physical function and HRQoL into adulthood and potentially requiring joint surgery. The model has not considered the cost of this surgery and this assumption implies that biologic DMARDs have no impact on long-term disease progression in terms of joint damage. However, a prospective registry-based cohort study by Minden and colleagues169 showed improved long-term prognosis for adult JIA patients who received etanercept during childhood. Furthermore, the AbbVie CS suggests that the reduction in orthopaedic surgery in JIA patients has been attributable to the increase in the use of immunomodulatory agents among children in recent decades and so DMARDs and biologic agents may have successfully prevented end-stage joint damage, based upon historical data that have shown a reduction. Therefore, biologic DMARDs are likely to reduce long-term damage compared with treatment with methotrexate, and they would potentially be more cost-effective than estimated by the independent economic model.

The cost-effectiveness of biologic DMARD treatment of patients with JIA-associated uveitis has not been formally estimated in this economic evaluation, owing to a lack of suitable input parameter data. The current evidence base comprises mainly small retrospective observational studies and suggests that adalimumab and infliximab are clinically effective in terms of improving intraocular inflammation and vision impairment. 96 A US cohort study of children with JIA compared those with JIA-associated uveitis with those without JIA-associated uveitis. 170 It reported that vision-related HRQoL was worse in uveitis patients, but general HRQoL was similar to that of JIA patients without uveitis. 170 It can be assumed that biologic DMARD treatment in JIA-associated uveitis patients will result in bigger overall HRQoL improvement (including vision-related HRQoL) and therefore would be more cost-effective in this group than in JIA patients without uveitis.

It was also reported that significant predictors of uveitis were persistent oligoarthritis and younger age at JIA diagnosis. 170 As discussed in Chapter 1, persistent oligoarthritis accounts for up to 48% of JIA cases in the UK and is regarded as a milder form of JIA. In contrast, EO accounts for between 6% and 17% of JIA cases in the UK and results in more severe symptoms and disease progression. Only EO was explicitly included in the NICE scope for this appraisal, and, therefore, it can be considered that uveitis is less likely to affect the patient subtypes that are relevant to the appraisal.

The economic model does not include the wider societal costs associated with JIA, which are described in more detail in Chapter 6. In the base-case analysis we have not included caregiver benefits associated with biologic DMARD treatment. A scenario analysis showed an improvement in cost-effectiveness for the biologic DMARDs when incorporating a utility disutility for patient caregivers.

The base-case analysis includes only one line of biologic DMARD treatment; however, in clinical practice some patients may switch to second- or third-line DMARDs. A scenario analysis that included a sequence of biologic treatments that most resembles current clinical practice was performed. The cost-effectiveness of multiple lines of biologic therapy is similar to that seen in the base-case analysis for one line of biologic therapy. There are many other possible treatment sequences but these have not been modelled, as they were considered to be less likely to occur in clinical practice and the results for these sequences are similar to those presented. In clinical practice, infliximab is often used but this has not been included as a treatment in the economic model, as it is licensed for this indication.

The cost-effectiveness results in this report are consistent with those from an earlier NICE technology appraisal for etanercept for patients with JIA (NICE TA3543). The previous appraisal used a discount rate of 6% for costs and 1% for benefits. 123 We ran the analysis for etanercept using these discount rates and the cost-effectiveness of etanercept improved to £21,718 per QALY. Using these discount rates, etanercept would be cost-effective at a willingness-to-pay threshold of £30,000 per QALY.