Omalizumab for the treatment of severe persistent allergic asthma: a systematic review and economic evaluation

Significance for patients in terms of ill health (burden of disease)

Asthma affects the patients and their families, and also society in terms of days lost from work and school, reduced QoL, and avoidable health-care visits, hospitalisations and deaths. 11 Although severe uncontrolled asthma affects only a relatively small population, it accounts for a significant proportion of health-care resource use. 12 This group of patients remain at high risk of exacerbations that require additional treatment, health-care consultations and often hospitalisations. Severe exacerbations are also potentially life-threatening. 11

Psychological conditions such as anxiety and depression may be up to six times more common in people with asthma than in the general population. Depression may be present in between 14% and 41% of those with asthma. 13 It is particularly common in people with severe and difficult-to-control asthma, and this is emphasised in the British Asthma Guidelines. 13 Those with asthma who also have depression or anxiety experience more asthma symptoms and have worse outcomes in terms of higher use of health-care resources, increased health-care costs, less successful emergency treatment and increased hospitalisation. 10

Significance for the NHS

The costs of asthma are substantial and include both direct health costs (hospital admissions and cost of treatment) and indirect, non-medical costs (time lost from work, premature death). 1

Search strategy

Studies relevant to an assessment of the therapeutic effect of omalizumab were identified by searching the following databases: MEDLINE, MEDLINE In-Process & Other Non-Indexed Citations, EMBASE, Cochrane Database of Systematic Reviews (CDSR), Cochrane Central Register of Controlled Trials (CENTRAL), Database of Abstracts of Reviews of Effects (DARE) and Health Technology Assessment (HTA) Database, National Institutes for Health (NIH) ClinicalTrials.gov Register, Current Controlled Trials, Conference Proceedings Citation Index (CPCI-S), and EconLit. Searches were run in September 2011 and rerun in October 2011 following the identification of an additional search term at the screening stage of the review. Full details of the search strategy are provided in Appendix 1. Additional searches of trial registers, journals and reference lists of relevant published systematic reviews were conducted to identify any further studies of relevance. No limits on date, language or study design were applied. EndNote software (Thomson Reuters, CA, USA) was used to download and import references and remove duplicates. The submissions were provided to NICE by Novartis Pharmaceuticals and the associated documents were also used as sources of relevant studies for the review.

Study selection

Abstracts of identified studies and potentially relevant full papers were independently assessed for inclusion in the review by two reviewers using the criteria outlined below. Disagreements were resolved through discussion and, where necessary, by consultation with a third reviewer. In cases where it was unclear if a study met inclusion criteria, attempts were made to contact authors for further information.

Data extraction

Data relating to both study design and quality were extracted by one reviewer using a standardised data extraction form and independently checked for accuracy by a second reviewer. Disagreements were resolved through consensus, and, if necessary, a third reviewer was consulted. Attempts were made, where possible, to contact authors and study sponsors when a potentially relevant study was reported in abstract form only or could not be fully identified. Authors were also contacted when it was unclear whether multiple publications referred to the same study. Data from studies with multiple publications were extracted and reported as a single study. Additional data were also extracted from the manufacturer's submission (MS); where this is the case the fact is noted and the trial publications are not referenced.

Intention-to-treat (ITT) data were used where possible; where this was not possible, the fact was noted. Rate ratios were reported for the outcomes of exacerbations and hospitalisation and other unscheduled care use. Mean differences were reported where possible for outcomes of QoL and asthma symptoms.

Where rate ratios or other summary measures were reported in the source documents, these were used. The majority of rate ratios reported in the published papers and in the MS were calculated using a Poisson regression. Where reported, this included factors such as dosing schedule, country grouping and asthma medication strata. Where these summary measures were not reported, they were calculated by the review team using numbers of exacerbations and patient time, where possible, and this is noted. When only very limited data were reported which precluded calculation of a summary measure, or where only the result of a statistical test (with or without a p-value) were reported, then these were presented.

Quality assessment

In all cases, quality assessment was performed by one reviewer, and independently checked by a second. All disagreements were resolved through consensus, and, if necessary, a third reviewer was consulted.

The quality of RCTs was assessed using standard checklists following the principles of CRD. 23 The following criteria were assessed: randomisation (were details of an appropriate method reported?); allocation concealment (were details of an appropriate method reported?); blinding of outcome assessors; baseline comparability of groups and equal treatment of groups (with the exception of the intervention); use of appropriate analysis including use of a sample size calculation and selective outcome reporting; treatment of withdrawals and dropouts. Full details are given in Appendix 7.

The original protocol was amended to also include the assessment of risk of bias following the principles of the Cochrane Collaboration. 25 This assessment was conducted by using the answers to the original quality assessment to give a risk of bias score. The overall risk of bias was considered to be the highest risk scored for any single criterion. For example, if a trial was considered at low risk of bias on all criteria except one where the risk was unclear, then the overall risk of bias was recorded as unclear; where the risk was low or unclear on all criteria except one which was scored as high then the overall risk of bias was recorded as high. All outcomes were considered to be subjective. Because there was therefore no reason to believe risk of bias differed between outcomes, a risk of bias was calculated for the study as a whole rather than for each outcome. This is because even seemingly objective outcomes, such as exacerbations, are influenced by (for example) compliance with concomitant medication and symptom perception. Criteria not included in the risk of bias assessment (such as reporting of a power calculation) did not contribute to the overall risk of bias for the study. Criteria such as comparability of baseline characteristics and use of an appropriate analysis were independently assessed by the review team; if authors reported tests of statistical significance for potential baseline imbalances, these were used to inform the assessment.

For non-randomised studies, tools based on CRD guidance23 were used. The criteria assessed were: recruitment of a representative sample from a relevant population; use of explicit criteria for inclusion; baseline comparability of groups (where applicable), blinding of outcome assessors (where applicable); completeness of follow-up; adequate reporting of outcomes, including sufficient follow-up. No overall quality score was calculated but aspects of the quality assessment were used to inform the consideration of observational study results.

Details of the methods used in the quality assessment are provided in Appendix 7.

Data analysis

Approach to synthesis

Although a statistical synthesis (meta-analysis) of the results of the identified RCT was planned, in practice this was not appropriate for any analysis. The trials of adult patients were subject to significant methodological and, to a lesser degree, clinical, heterogeneity identified by the quality assessment (see Validity assessment and risk of bias of RCTs). There was clear clinical heterogeneity between trials in the licensed population and those included as supportive evidence. Methodological heterogeneity prevented meaningful pooling of the trials in the adult licensed population. In the case of children there was only one trial in which a defined subgroup met the licence criteria and one further trial was included as supportive evidence.

Search strategy

In addition to the searches conducted for the review of the efficacy of omalizumab (see Methods for reviewing the efficacy of omalizumab (including long-term outcomes and steroid-sparing), Search strategy), information on adverse events of omalizumab was identified from searching resources of the US and European drug regulatory agencies (FDA and EMA). No language or date restrictions were applied to the search strategy. In addition, reference lists of all included studies and industry submissions made to NICE in this and previous appraisals (TA133 and TA201) were hand-searched to identify further relevant studies.

Inclusion and exclusion criteria

The FDA, EMA and NICE documents and efficacy studies that reported information on the adverse effects of omalizumab were relevant for the review. The lists of titles/abstracts generated by the electronic searches and all full paper manuscripts and documents of possible relevance to the review of safety of omalizumab were obtained, where possible, and the relevance of each study was assessed by two reviewers; any discrepancies were resolved by consensus. Potentially relevant studies that did not meet all of the criteria were excluded and their bibliographic details listed with reasons for exclusion.

Data extraction, quality assessment and data analysis

Data relating to adverse and serious adverse events were extracted using a standardised data extraction form and the quality of RCTs and other study designs were assessed using standard checklists as detailed in Methods for reviewing the efficacy of omalizumab (including long-term outcomes and steroid-sparing), Quality assessment. Reviews and regulatory documents were not formally quality assessed, but the reliability of the data were discussed where relevant. Data extraction and quality assessment was performed by one reviewer and independently checked for accuracy by a second reviewer. Disagreements were resolved through consensus. No formal analysis of the data was performed; the adverse effects of omalizumab were presented as a narrative synthesis.

Search strategy

The review team were given access to an existing internal CRD database of systematic reviews of adverse events as previously used by Golder et al. 27 This database was searched using the terms steroid, corticosteroid, glucocorticoid and all individual steroid names (see Appendix 1). This search was supplemented by a search of The Cochrane library and DARE using terms for steroids coupled with terms for asthma. A further supplementary search was conducted on PubMed to try to identify any very recent relevant systematic reviews (SR).

As for the other review questions, information provided in the MS to NICE was also considered

Inclusion and exclusion criteria

Any systematic review of the adverse effects of OCSs was considered for inclusion in the review. The steroid-related adverse events of particular interest included: bone outcomes (such as fracture), incidence of infectious disease, hypertension, ocular outcomes including cataracts and glaucoma and, in children and adolescents, growth retardation.

Data extraction, quality assessment and data analysis

Relevant data were extracted by one reviewer and checked by a second. The quality of the included reviews was discussed in terms of accepted criteria for systematic reviews as used for the DARE database (clear review question and inclusion criteria, adequate literature searches, quality assessment of included studies, appropriate synthesis), but a formal checklist was not used. The findings of the included reviews were combined in a narrative synthesis.

Quantity and quality of evidence: randomised controlled trials

Of the 11 RCTs included in the review of effectiveness, 10 were relevant to the adult licence (age ≥ 12 years),19,28^–36 one was relevant to the children's licence, (age < 12 years)20 and one was relevant to both licences (age 6–20 years). 28 The criteria for the licence and their relationship to the inclusion criteria of included trials and their specified subgroups are shown in Table 2. Full details of the inclusion criteria and population characteristics of these trials are given in Appendices 3 and 4. The MS provided additional information on four RCTs and two observational studies, all of which were identified in the systematic review process. These data were used to supplement those obtained from the published papers. In particular the manufacturer provided data on the omalizumab responder population and on subgroups within the INNOVATE, Evaluate Xolair for Asthma as Leading Treatment (EXALT) and IA-05-EUP (European Union population) trials. Where data are derived from the MS this is explicitly stated. In all other instances the source was a publicly available (published or otherwise) document identified during the systematic review process.

In the case of studies in which the inclusion criteria did not determine that the trial population or a defined subgroup would correspond with the licence criteria, the reasons for concluding that a substantial, although undifferentiated, proportion of patients met these criteria are documented in Table 3.

Validity assessment and risk of bias of randomised controlled trials

The results of the validity assessment and the Cochrane risk of bias assessment for the RCTs are shown in Table 4, with full details in Appendix 7. Overall risk of bias was calculated as outlined in Methods for reviewing the efficacy of omalizumab (including long-term outcomes and steroid-sparing), Quality assessment. In some cases trials conducted by the manufacturer had unclear reporting of randomisation, allocation concealment and blinded outcome assessment, but it was indicated that the procedures had in fact been conducted using the manufacturer's standard approach. In these cases, the risk of bias from these measures was assumed to be low as the standard process documented in other manufacturer's trial reports represented a low risk of bias. This process was used for the following trials: SOLAR36 and the trials by Hoshino et al.,34 Ohta et al. 35 and Bardelas et al.,32 all of which were included as supportive trials. This materially affected the overall risk of bias judgement only for SOLAR and the trial by Ohta; in both cases the effect was to make the judgement ‘low’ rather than ‘unclear’.

Baseline medication

Exacerbation and treatment histories

The baseline exacerbation rate reported in the main adult trials was 2.5/14 months for INNOVATE and 2.1/year for EXALT. 19,29,31 It was not reported for EU IA-04, but over 99% of this subgroup had received ≥ 1 OCS course and the mean number of OCS courses in the past year was 4.1. 51 As OCSs are usually prescribed for clinically significant exacerbations, this is an indicator of the probable baseline exacerbation rate. The exacerbation rate in the IA-05 EUP subgroup (which meets the paediatric licence) was 2.8 per year. 20

Only INNOVATE reported the baseline severe exacerbation rate.

In line with the high baseline exacerbation rate, IA-04 EUP subgroup patients had higher rates of both hospitalisation (47%) and ER visits (92%) in the past year than patients in INNOVATE (39% and 56% respectively). 19,51 This also reflects the fact that the inclusion criteria required that one of the two qualifying exacerbations in the past year have resulted in hospitalisation or ER attendance. EXALT patients, by contrast, had substantially lower rates of both hospitalisation (22%) and ER visits (30%) compared with INNOVATE. 31 This is likely to be reflective of the less strict inclusion criteria with respect to exacerbation history.

Baseline exacerbation rates in the supportive adult trials were only reported in Hanania (1.95/year)33 and SOLAR (2.1/year);36 hospitalisation rates were reported only by the Ohta trial (9.8%). 35

The IA-05 EU subgroup had a hospitalisation rate of 12%;20 that in the supportive Busse et al. (2011) trial was substantially higher at 25%, reflecting the fact that this trial included a group of children and adolescents who were not receiving appropriate maintenance treatment. 28

Other parameters

Baseline FEV₁ was comparable between studies in the licensed adult population, ranging from 61% to 65% expected, although IA-04 did not use FEV₁ as an inclusion criterion. In supportive trials, FEV₁ ranged from 65% to 78% expected value. Mean age was also comparable between the adult trials, ranging from 39 to 47 years in the licensed populations and 38 to 55 years in supportive trials. FEV₁ was substantially higher in the children's trials at 82% for the IA-05 EUP group and 92% in the Busse trial.

Summary

There were four RCTs that met the adult licence criteria. Two of these were double blind and placebo controlled and were judged to be at low risk of bias; one of these was very small. The remaining two were well conducted but had a high risk of bias because of their open-label design; in one of these only a subgroup of patients met licence criteria. These open-label trials were not placebo-controlled but had a comparator of standard care without omalizumab. There were some differences in the baseline characteristics of these trial and subgroup populations, which are discussed above, but these were not sufficient to make comparisons between trials unreasonable.

One trial with a subgroup in the licensed paediatric population was identified; this was a good-quality double-blind placebo-controlled RCT with a low risk of bias.

Five RCTs were identified as providing supportive evidence for adults and one for the paediatric licence. Of these, only one large trial in adults was considered to be at low overall risk of bias. There were higher levels of clinical heterogeneity between the populations in these trials.

Quantity and quality of evidence: observational studies

The 13 observational studies included in the main review as supporting evidence of the effect of omalizumab in ‘real-world’ clinical situations are summarised in Table 6; full details of the inclusion criteria and population characteristics are in Appendices 5 and 6. Eleven of these studies related to the adult licence and two assessed efficacy in children. It had been anticipated that the observational studies would provide data on the longer-term efficacy of omalizumab but, in the event, this was relatively limited. One study (PERSIST) reported very limited data at 120 weeks follow-up, and only for about a third of the original patients. 39,60 Two relevant studies not identified by our systematic selection process were identified by the manufacturer: those by Britton et al. 61 and Tzortzaki et al. 62 (Two additional studies identified by the manufacturer as potentially relevant did not meet the inclusion criteria for population and were excluded. Chung et al. reported an extension study related to trial 011, in which patients were well controlled at baseline. 63 Storms et al. reported a study in which it was unclear how many patients were using a LABA at baseline and in which high attrition rates made it difficult to determine if outcome data were reported for any patients who met the license criteria. 64)

Tzortzaki et al. provided data on a number of relevant outcomes including clinically relevant exacerbations at 4 years,62 whereas Britton et al. reported mean follow-up of 982 days. These studies provided useful additional data on the longer-term efficacy of omalizumab but were small: Tzortzaki reported on 60 patients and Britton on 52.

Several studies reported data on only a small number of outcomes. Five additional studies were included only for the outcomes of persistence of response, OCS-sparing and safety of omalizumab; these studies are discussed in sections Results of review of clinical effectiveness: efficacy of omalizumab, Evidence of long-term-efficacy and persistence of response to Results of assessment of safety of omalizumab.

The results of the quality assessment for all observational studies, including those included in the assessments of safety and steroid-sparing are shown in Appendix 7. The quality of the observational studies was generally poor. Only one of the observational studies included for the main review, the PAX-LASER cohort, had a control group and in this study it was unclear whether outcome assessment was blinded. 47 The majority of studies reported eligibility criteria, but in most there was poor reporting of losses to follow-up or these losses were substantial (20% or greater). Few studies reported using reliable outcome measures. Six of the studies had sufficient follow-up for long-term assessment but in only three cases was this over 52 weeks (see above). Data derived from these observational studies are included as supportive data; these have relevance to real-world clinical practice but issues of uncertainty or low quality should be borne in mind throughout.

Response to treatment: global evaluation of treatment effectiveness

The global evaluation of treatment effectiveness (GETE) measures response to asthma treatment on a five-point scale: excellent (complete control of asthma), good (marked improvement of asthma), moderate (discernible, but limited improvement in asthma), poor (no appreciable change in asthma) to worsening of asthma.

GETE ratings were reported by four RCTs in adults (INNOVATE, EXALT, SOLAR and the trial by Bardelas et al. ) and by IA-05-EUP in children. The proportion of omalizumab and standard care patients with physician-rated GETE scores of good or excellent are shown in Table 7.

Response to treatment: Asthma Quality of Life Questionnaire change ≥ 0.5 points

IA-04 EUP and SOLAR reported the proportion of patients with a change from baseline in total AQLQ score ≥ 0.5 points, which represents the minimally important difference and is sometimes used as an alternative measure of response (Table 8). Data on this outcome were also reported for INNOVATE, EXALT and IA-05EUP, but these were not used to assess response to treatment by the study authors. There were no data from observational studies on response rate assessed using this criterion. The AQLQ criterion, representing as it does a minimally important difference, may result in an overestimation of the percentage of responders compared with evaluation using GETE. 20 This is supported by comparison of the two measures of response using data from the SOLAR trial.

Response rates from observational studies

Response rates were reported in six observational studies (Table 9): five measured by the GETE and one study in children by a combination of prednisolone dose, lung function and overall clinical status. 54 At 16 weeks the GETE response rate ranged from 69.6% to 86.4% and at 1 year it was reported as 72.3% or 77%.

Response rates: summary of omalizumab treatment effect

The omalizumab GETE response rate in adults reported in the double-blind RCT was around 55% with a RR of around 1.2. The open-label EXALT trial had a substantially higher proportion of omalizumab responders than INNOVATE (70% compared with 56.5%) and a much larger RR (2.24 vs. 1.38). This appears highly likely to be the result of the open-label design of the trial, as the proportion of patients classified as responders is likely to be elevated by the patients' and assessors' knowledge of their treatment allocation. The impact of these differential response rates is discussed in relation to the treatment effects observed in the responder analysis in sections Treatment effects of omalizumab: exacerbations and Hospitalisation and other unscheduled medical care requirements. It is worth noting that the response rate derived from EXALT is likely to be more representative of that seen in clinical practice, partly because of patients' knowledge of their treatment and partly because the assessment was conducted at 16 weeks, as it is in clinical practice. The EXALT omalizumab GETE response rate approaches, but is still lower than, those reported in observational studies.

In children the proportion of omalizumab responders in the IA-05 EUP subgroup was high at 74%, but a high proportion of the placebo group were also classified as responders (RR 1.15; 95% CI 0.95 to 1.39).

Total clinically significant exacerbations

All the included RCTs reported data on the primary outcome of clinically significant (CS) exacerbations, with the exceptions of Bardelas et al. 32 and Hoshino et al. 34 There was some degree of heterogeneity in the definition of clinically significant exacerbations within trials (see Appendix 8); however, this was not considered sufficient to preclude comparability. CS exacerbations are sometimes reported as total exacerbations, as exacerbations which are not clinically significant are not considered. A number of trials reported data on (or enabled the calculation of) the number of patients experiencing no CS exacerbations.

Results from trials providing data on total exacerbations are presented in Table 10. As can be seen, there is a consistent finding of benefit with omalizumab for both the incidence rate and proportion of patients with no exacerbations in the follow-up period, with the exception of the small trial of Chanez et al. 29 These benefits were statistically significant in all studies except SOLAR, in which a relatively low proportion of patients were taking a LABA. 36 A full report of the data reported for each trial is presented in Appendix 8.

There was some heterogeneity in the estimates of efficacy for rates of CS exacerbation for the individual trials. This appears to be primarily because of the trial design, with the open-label trials EXALT and IA-04 EUP showing larger estimates of effect than those that were double blind. It is notable that the overall estimate of effect for the Hanania trial was comparable to that for INNOVATE, with the M2 subgroup mirroring this effect. The lack of a treatment benefit in the M3 subgroup is notable but the subgroup was relatively small and underpowered. This cannot, therefore, be considered as evidence for a lower treatment effect in patients using OCSs as maintenance or repeat therapy.

There was considerable clinical heterogeneity across the main and supportive trials in terms of trial design and population. In the main trials, IA-05 was undertaken in children and could not therefore be combined with adult data. IA-04 EUP was a post hoc subgroup of a larger trial. EXALT was an open-label trial with a comparator of optimised standard care which provided consistently higher estimates of the omalizumab treatment effect compared with the double-blind placebo-controlled INNOVATE. Therefore, combining results in a meta-analysis was not considered appropriate.

Apart from the trial by Hanania, supportive trials did not report rates of exacerbations, but SOLAR and the trial by Ohta did report numbers of patients with zero exacerbations in adults. This was statistically significant in the Ohta trial but not the SOLAR trial. 35,36

In children the IA-05-EUP data for the 24-week constant treatment phase (the primary outcome) showed a statistically significant treatment benefit for omalizumab in terms of a reduced CS exacerbations rate. This benefit was seen for the total 52-week period of the trial; despite a steroid-sparing phase between weeks of the trial, it appeared that substantial benefit was accrued in the omalizumab group. The Busse trial also reported a statistically significant benefit of omalizumab in the number of patients with zero exacerbations in children and adolescents. 28

Clinically significant severe exacerbations and clinically significant non-severe exacerbations

Only three of the included trials reported the incidence of CSS exacerbations (Table 11); separate data on CSNS exacerbations in these trials were provided by the manufacturer. All of these were trials in which the inclusion criteria closely approximated the terms of the licence(s): INNOVATE19 and EXALT31 trials in adults and the IA-05 trial (EUP)20 in children.

Both INNOVATE and IA-05 EUP defined a CSS exacerbation as a clinically significant exacerbation with an FEV₁ (or PEF in the case of INNOVATE) of < 60% of personal best; EXALT used a slightly broader definition, having multiple additional options for meeting the criterion (see Appendix 9). 19,31 This may have resulted in more exacerbations being classified as severe in EXALT compared with INNOVATE or IA-05 EUP. Whether this might have favoured omalizumab in EXALT more than in the other main trials is unclear.

The two adult trials both found a statistically significant treatment benefit of omalizumab. As a result of clinical heterogeneity across the main and supportive trials in terms of trial design and population, combining results in a meta-analysis was not considered appropriate.

IA-05 EUP also reported relative risks in favour of omalizumab in children at 24 weeks, and at 52 weeks despite the steroid-sparing phase from week 28, but they were not statistically significant.

The result from INNOVATE suggests that omalizumab has no treatment benefit in terms of reducing CSNS: its benefit comes through reducing the rate of CSS exacerbations. In contrast, the results from both EXALT and IA-05 EUP suggest that omalizumab has an equally beneficial effect on CSS and CSNS exacerbations. A difference between INNOVATE and EXALT could be explained by the different definitions of CSS exacerbation adopted in these two trials or because one trial is double blind whereas the other is open label, but this cannot hold for the difference between INNOVATE and IA-05 EUP.

Exacerbations: subgroup analyses

Results for post hoc subgroups for INNOVATE and EXALT and IA-05-EUP were provided by the manufacturer (Novartis Pharmaceuticals) (in their submission to NICE or additionally to the assessment group). The subgroups were:

• patients with a history of hospitalisation

• patients on OCSs at baseline (not for IA-05-EUP because of the very small number of patients).

Data on patients not on OCSs at baseline and on those with an exacerbation history of ≤ 2 and ≥ 3 exacerbations per year at baseline were also provided to the assessment group and were used in regression analyses.

Data on total exacerbations, CSS exacerbations and CSNS exacerbations were reported. These are shown in Tables 12 and 13 for patients on OCSs or with a history of hospitalisation. Rate ratios were calculated by the AG using an approximation of the standard error (SE) to derive confidence intervals. 25 These data are presented with the caveat that they are derived from small post hoc subgroup analyses and confidence intervals are wide, representing the high uncertainty around the estimate. This is particularly the case with the IA-05 trial, in which we are considering post hoc subgroups of an a priori subgroup.

The rate ratios of the subgroup data indicated that there may be an increased treatment effect in patients on OCS maintenance therapy in the INNOVATE trial; however, this was not confirmed by a test for interaction for total exacerbations (OCS interaction with treatment: RR 0.85; 95% CI 0.44 to 1.66). Results of tests for interaction in the hospitalisation group of INNOVATE and in both subgroups in the EXALT trial data also did not show evidence of a significant interactions. Data on subgroups of patients with ≤ 2 exacerbations/year and ≥ 3 exacerbations/year at baseline did not suggest evidence of differential effectiveness in these groups. These data are not presented because they are commercial-in-confidence (CiC). For all outcomes there was considerable overlap of confidence intervals between all subgroups and the ITT population. For all of the subgroups it appeared that the treatment effect on total exacerbations was driven by the impact on CSS exacerbations to a greater degree in INNOVATE than in EXALT; this mirrors the pattern of observed effects in the ITT population.

Exacerbations: responder analyses

The manufacturer also supplied analyses based on patients who responded to omalizumab. For three trials (INNOVATE, EXALT and IA-05 EU subgroup): the responder subgroups were defined using a GETE rating of good or excellent (see section Treatment effects of omalizumab: response to treatment) at 28, 16 and 52 weeks respectively. 19,30,31 For the IA-04 EUP trial, responder status was defined using the criterion of an improvement in mini-AQLQ score of ≥ 0.5 points at 27.

In the responder analyses both INNOVATE and EXALT demonstrated statistically significant benefits of omalizumab for CS exacerbations and CSS exacerbations (Table 14); the differential estimate of benefit between these trials seen in the ITT analyses was not present. The responder analysis in the EU population subgroup of the double-blind children's trial IA-05 also showed a statistically significant benefit for CSS and CS exacerbations.

Although the open-label design of EXALT is likely to have inflated the response rate compared with that seen in the double-blind INNOVATE, this bias will have been moderated in the responder analysis as the responder population in EXALT probably contained a proportion of people who may be classed as false-positives in terms of response status.

Further responder analyses by subgroup were provided by the manufacturer. These are given in Appendix 10. The concerns about post hoc subgroups discussed in the case of the ITT analysis apply to an even greater extent in the responder analyses where the overall numbers are even smaller. Briefly, although the rate ratios appeared to show larger estimates of treatment effect in patients on maintenance OCS or with at least one hospitalisation in the previous year, confidence intervals were wide, overlapping those for the whole population of omalizumab responders. The subgroup data also appeared to show that response rates in EXALT were close to those of INNOVATE for the subgroups of patients with a history of hospitalisation, patients on maintenance OCS and patients with ≥ 3 exacerbations at baseline (data not presented because they are CiC). This suggests that the much higher response rate in the ITT population of EXALT may be driven by patients outside these groups who had less severe disease at baseline.

Exacerbation rates: data from observational studies

Data on total exacerbations were reported by nine observational studies (Table 15). As can be seen, the data indicate substantial reductions in the exacerbation rate from baseline, and where a treatment effect was reported this showed statistical significance. Studies which are of particular relevance to the appraisal are the Asthma Patient Experience on Xolair (APEX) study, which has high clinical relevance to UK clinical practice, and the Tzortzaki study, which showed sustained benefit over a 4-year period. However, the methodological problems with these studies, including their lack of control groups, should be borne in mind.

Data on severe exacerbations were reported by five observational studies (Table 16). Where reported the data indicated substantial reductions in incidence of severe exacerbations relative to baseline; in Korn et al. this was reported as being statistically significant. 44 However, the comparative PAX-LASER cohort showed statistically significant reductions in severe exacerbation rates in both the omalizumab and control arms, although the reduction was larger in the omalizumab arm (between-group comparisons were not reported). 47 The data from Deschildre et al. indicated substantial reductions in children (mean age 11.8 years). 42

Exacerbation rates: summary of omalizumab treatment effect

There was clear evidence of efficacy of omalizumab in RCTs and RCT subgroups in the adult licensed populations, with statistically significant benefits for the outcomes of total (CS) exacerbations; CSS exacerbations and CSNS exacerbations (where reported). 19,30,31 There was evidence of a larger treatment effect in the open-label trials than in the double-blind placebo-controlled trials.

There was also evidence of treatment benefit in wider populations in trials included as supportive evidence; in particular in the large (n = 850) trial of Hanania et al., which showed statistically significant benefits in the whole trial population and in the M2 subgroup of patients taking medication additional to ICS plus LABA. 33 Although the M3 subgroup of patients on maintenance OCS or with four or more exacerbations in the previous year did not show such a benefit, this group was underpowered. All adult trials except SOLAR (which has a low proportion of patients who potentially meet the licence criteria) and the small study of Chanez et al. showed a statistically significant benefit of omalizumab, with the SOLAR result close to statistical significance.

As the only two trials in adults that reported on CSS, both INNOVATE and EXALT found statistically significant reductions in CSS exacerbations with similar effect sizes, but only EXALT showed a statistically significant benefit for CSNS exacerbations. This may indicate that in the double-blind placebo-controlled trial much of the benefit in total exacerbation reduction was driven by reductions in severe exacerbations.

Responder analyses from INNOVATE and EXALT comparing omalizumab responders with all comparator patients showed a similar pattern to the ITT analyses with a statistically significant benefit for all licensed populations for total exacerbations and CSS and CSNS exacerbations. In contrast to the ITT analyses, there was little difference between the estimates of effect from the trials in total; as discussed above, this may be a consequence of the impact of trial design on the proportion of responders. However, it remained the case that INNOVATE showed a larger treatment benefit for CSS exacerbations than for CSNS exacerbations, whereas EXALT showed a similar effect size for both types of exacerbation.

There was limited evidence from observational studies but the available data reflected that from the RCTs, indicating substantial reductions from baseline and, where statistical tests were reported, these indicated that significant benefit was obtained for both total exacerbations and CSS exacerbations.

In children there was a statistically significant benefit of omalizumab on total exacerbation rate in the IA-05 EUP subgroup but not on the CSS or CSNS exacerbation rates. It is probable that this was a consequence of lack of power in the subgroup. The responder population showed a statistically significant benefit for both total and CSS exacerbations. The Busse et al. trial, which was included as supportive evidence, also showed a statistically significant benefit for the number of children and adolescents with zero exacerbations. 28 The limited evidence from a single uncontrolled observational study indicated a substantial reduction in severe exacerbations in children. 42 The value of this supportive evidence is limited by quality and issues of generalisability to UK clinical practice.

Hospitalisation

Hospitalisation data for adult populations were reported by the IA-04-EUP, EXALT and INNOVATE trials19,30,31 and the small study by Chanez et al. (which reported zero events)29 (Table 17). The relative treatment effect for hospitalisation rate favoured omalizumab in both INNOVATE and EXALT, but was statistically significant only in EXALT. No trial found a treatment effect of omalizumab in terms of the number of patients avoiding hospitalisation (zero hospitalisation), although EXALT did find a benefit when hospitalisation and ER visits were counted together (RR 1.24; 95% CI 1.08 to 1.41). In EXALT and IA-04 EUP the number of days in hospital was also reported but statistical comparisons were not reported.

In children, the IA-05 EUP showed no evidence of a difference in hospitalisation rates between the groups or in the number of patients with zero hospitalisations (see Table 21). The supporting Busse et al. trial,28 which did not report data separately for those aged < 12 years, reported a very small but statistically significant benefit of omalizumab. This may reflect the inclusion of patients not receiving appropriate maintenance therapy at baseline. 28

Data on hospitalisation rates from INNOVATE and EXALT were also reported in the MS for the subgroups of patients who had been hospitalised in the previous year and for patients on maintenance OCS (Table 18). The rate ratios are suggestive of a greater effect in these subgroups. However, CI calculated using an approximation of the standard error were wide and these are subgroup results, so great reliance should not be placed on them, particularly given the lack of a statistically significant interaction in the primary outcome of CS exacerbations.

Emergency care use: emergency department visits and unscheduled doctor visits

The results are presented in Table 19. A statistically significant reduction in total emergency visits was seen in all three of these adult trials. The small Chanez trial reported a non-significant difference in median change from baseline for emergency department and doctor visits combined. 29 Emergency department treatment and unscheduled doctor visits were reported separately by the INNOVATE, EXALT and IA-04 –EUP trials. 19,30,31 As with hospitalisation rates, the only study to show a statistically significant benefit of omalizumab for emergency department visits was EXALT; the INNOVATE and IA-04 EUP trials showed non-significant results favouring omalizumab. Unscheduled doctor visits showed a corresponding pattern, with only EXALT showing a statistically significant benefit of omalizumab in event rate. A statistically significant reduction in total emergency visits was seen in all three of these trials.

For children the IA-05 EUP subgroup showed no statistically significant difference between omalizumab and placebo at either 24 or 52 weeks for the incidence of emergency department attendance, unscheduled doctor visits and total emergency (Table 19), but the direction of effect favoured placebo, in contrast to the result in adults. Busse et al. 28 did not report data for these outcomes.

Data on emergency care use from INNOVATE, EXALT and IA-05 EUP were also reported in the MS for the subgroups of patients who had been hospitalised in the previous year or who were on maintenance OCS (Table 20), and the effects look favourable. However, CIs calculated using an approximation of the standard error were wide so great reliance should not be placed on them, particularly given the lack of a statistically significant interaction on the primary outcome of CS exacerbations.

Emergency care use: responder analysis

In the responder analyses of emergency care use, INNOVATE and EXALT showed a benefit of omalizumab although it was not always statistically significant, but in IA-04 EUP this was much less pronounced with very wide non-significant confidence intervals (Table 21). The differential estimate of benefit between INNOVATE and EXALT seen in the ITT analyses was not present. In children in the IA-05 EUP subgroup there was a statistically significant benefit in hospitalisation rates but non-significant benefits for other emergency care use measures.

Subgroup responder analyses for adult patients on OCS maintenance or hospitalisation in the previous year were presented by the manufacturer (see Appendix 10). These were suggestive of a greater magnitude of treatment effect on hospitalisation rates in both subgroups and on total unscheduled care in the OCS maintenance subgroup. As before, the small numbers, multiple subgroups and lack of variance estimates should be borne in mind in interpreting this evidence.

Emergency care use: data from observational studies

Hospital visits, ER attendance and unscheduled doctor visits were reported by nine observational studies; data are shown in Table 22. The APEX study reported statistically significant benefits of omalizumab for all three measures of unscheduled care, whereas Korn et al. reported such benefits for hospitalisation and a combined measure of emergency visits and PAX-LASER for combined hospitalisation and ER visits. 37,44,47 PAX-LASER also reported a statistically significant benefit over the comparator group for this outcome. Although other studies did not report statistical tests of difference from baseline, the data which were reported did support the pattern of a reduction in incidence of unscheduled care of all kinds associated with omalizumab treatment. In particular, Britton et al. reported numerical reductions in all forms of unscheduled care. The results of the APEX and Britton et al. studies are of particular relevance to UK clinical practice, representing data from UK severe asthma centres over periods of 12 months and 2.7 years respectively. 61,65 The controlled PAX-LASER study is also of importance in showing significant benefits over usual care. 47 The observational studies showed a consistent pattern of statistical significance which corresponds to that seen in the EXALT trial;31 benefits in the double-blind RCT did not reach statistical significance. This may reflect weaker methodology or may be indicative of the fact that in many of the studies (e.g. Britton et al. 61) it is unclear whether only data for responders are reported.

Hospitalisation and unscheduled care: summary of omalizumab treatment effect

There was limited evidence of benefit in the adult ITT populations: of the trials which reported data for these outcomes only EXALT showed statistically significant benefits. There was some indication of greater benefit in subgroups of patients taking maintenance OCS or with a history of hospitalisation in the previous year for the outcome of hospitalisation, but confidence intervals calculated using an approximation of the standard error were wide and a post hoc subgroup effect was not supported by tests for interaction on the outcomes of CS exacerbations.

Analyses comparing omalizumab responders with placebo/standard care patients showed evidence of statistically significant benefit for both INNOVATE and EXALT across the outcomes assessed with the exception of ER attendance in INNOVATE. This pattern of results is similar to that seen for exacerbations.

Only limited data on emergency care use were available from observational studies but they showed evidence of substantial reductions across all types of care; where statistical tests were reported these showed significant benefits of omalizumab treatment relative to baseline or standard care.

In children the IA-05 EUP group showed no significant differences between the groups for any emergency care use outcome in the ITT analysis. Supportive evidence from the trial by Busse et al. 28 indicated a statistically significant benefit of reduced hospitalisation, but this result may be driven by children/adolescents not on appropriate maintenance therapy.

Responder analysis results from IA-05-EUP indicated a statistically significant benefit in reduced hospitalisation rates in omalizumab responders, but only non-significant effects on other unscheduled care.

Symptom scores

A number of different scales were used to assess symptom control in the included trials (Table 23): the Wasserfallen asthma symptom score; ACT; ACQ; the Total Asthma Symptom Severity score; and an unspecified asthma symptom score. Full data on changes from baseline in total asthma clinical symptom scores were reported for INNOVATE, IA-04 EUP and IA-05 EUP in the MS, together with changes in the nocturnal symptom score, morning symptom score and daytime symptom score. These supplemented the more limited data reported in trial publications. Different measurement tools were used to assess change in asthma symptoms over time. The ACT scores symptoms on a scale of 1 (worst) to 5 (best), with a higher overall score denoting greater improvement. By contrast, a higher overall score using the ACQ, Total Asthma Symptom score and Wasserfallen symptom denotes a worsening in symptoms; a lower score represents better asthma control.

There were statistically significant benefits of omalizumab on change from baseline in the total symptom score in the INNOVATE trial. The IA-04 EU subgroup showed a statistically significant benefit of omalizumab on the Wasserfallen symptom score, whereas the EXALT trial showed a similar benefit on the ACQ.

In supportive trials in adults the SOLAR and Hanania et al. trials found statistically significant benefits on the Total Asthma Symptom Severity score and the Wasserfallen symptom score respectively. 33,36

In the IA-05 EU subgroup there were no statistically significant changes from baseline in total asthma clinical symptom score in children at either 24 or 52 weeks; similar results were found using the Wasserfallen symptom score. Busse et al. found a statistically significant benefit on ACT score in children aged ≤ 11 years but not in older children and adolescents. 28

Individual symptoms

INNOVATE and EXALT and the small Chanez et al. trial reported data on at least one individual asthma symptom for the licensed population in adults; in children data were reported for the IA-05 EU subgroup. Supportive trials reporting data were those of Bardelas and Ohta in adults and Busse et al. in children (Table 24). 28,32,35 Outcomes reported were night awakenings, days with/without symptoms and activity impairment. Individual components of the asthma symptom score reported above also addressed night-time, morning and daytime symptoms. The results were variable but there was some evidence of impact on disturbed sleep, with statistically significant results reported by EXALT and the Bardelas et al. trial, as well as the Busse et al. trial in children and adolescents; the Ohta trial and the small Chanez trial reported non-significant results.

Asthma symptoms data from observational studies

Six small observational studies reported on asthma symptoms (Table 25). Five of the studies reported ACT scores (four in adults and one in children) and one ACQ scores over 8 months in adults. Where adult studies reported difference from baseline these indicated statistical significance, in other studies it was unclear. The Brodlie et al. study54 found statistically significant increases in the ACT, providing some useful evidence of the impact of omalizumab on day-to-day asthma symptoms in UK children with severe OCS-dependent asthma.

Data on the percentage of patients experiencing daily asthma symptoms and night-time awakenings was also reported by Korn et al. who found statistically significant reductions (p < 0.001) in both measures at both 4 and 6 months. 44

Asthma symptoms: summary of omalizumab treatment effect

There was considerable heterogeneity in the assessment of asthma symptoms in the included studies:19,31–33,35,36,61,62,65 a wide range of scales and individual symptom measures were used to assess response to therapy. In RCTs there was evidence of a statistically significant benefit of omalizumab on symptom scales in the three licensed population groups in adults (INNOVATE, EXALT and IA-04-EUP) and also in the supportive SOLAR and Hanania et al. trials; the studies of Bardelas et al. and Ohta et al. showed non-significant benefits. The observational studies APEX and eXpeRience showed evidence of benefit on symptom scores but did not report statistical test results, although APEX reported that there were significant differences compared with baseline. Additional studies by Tzortzaki and Britton also showed evidence of benefit with Tzortzaki showing statistical significance at 4 years. Although these studies were small and uncontrolled, they showed evidence of sustained benefit of omalizumab treatment over a relatively long period of time in patients relevant to UK clinical practice.

There was mixed evidence of impact on individual symptom measures, with most evidence of a treatment benefit for outcomes related to disturbed sleep for which benefits were reported in EXALT and the trial by Bardelas et al. as well as the observational study of Korn et al.

There was limited evidence of efficacy in children. The IA-05-EUP showed a non-significant benefit of omalizumab on both the total asthma symptom score and disturbed sleep assessment. Supportive evidence from the Busse et al. trial indicated a significant benefit in ACT score in children aged > 12 years, but a non-significant effect in those aged < 12 years; individual symptom scores showed significant benefit for the whole trial population of children and adolescents. The small study by Brodlie et al. indicated highly statistically significant gains in asthma control in OCS-dependent children aged < 12 years as well as in older adolescents;54 these children correspond closely to those treated in severe asthma clinics in the UK.

Symptom control is one of the key aims of asthma therapy but evaluation of symptom outcomes is complicated by the wide range of measures used to assess day-to-day symptoms. A number of different scales were employed in the studies, including in the key trials, making comparisons, even within the licensed population, difficult. Different methods were used to appraise individual symptom occurrence, with few symptoms reported consistently across trials and observational studies.

Use of rescue medication: results from randomised controlled trials

A majority of trials reported some data on rescue medication use (Table 26). This was reported as either puffs required or number of days on which the medication was required. In the licensed population, INNOVATE, the IA-04 EU subgroup and the Chanez et al. trials reported data for adults and the IA-05 EU subgroup reported data for children. For INNOVATE and the IA-04 EU subgroup these data were drawn from the review by King et al., which was an appendix to the MS;66 data for the whole trial population were reported for slightly different outcomes (data not shown). Supportive trials reporting data in adults were SOLAR and the trials by Hanania, Bardelas and Ohta. With the exception of the IA-04 EUP and Hanania et al. trials, the differences between the groups favoured omalizumab but were not statistically significant; King et al. suggested that the IA-04 EUP result was anomalous with respect to repeated measures data throughout the trial. 66

Data on use of rescue mediation from observational studies

Only two observational studies41,59 reported on changes in rescue inhaler use. Both reported substantial reductions following omalizumab treatment (Table 27). Costello41 found a 56% reduction in the number of puffs for a group of omalizumab responders 6 months after treatment initiation. Another study59 showed that approximately 66% of its participants had either reduced or stopped using a rescue inhaler. Neither study specified which inhalers were used.

Summary of treatment effect on use of rescue medication

There was limited evidence of efficacy of omalizumab in reducing requirement for rescue medication. Of the trials in the adult licensed populations only the IA-04-EUP found a statistically significant benefit. Hanania et al. also found a statistically significant benefit. 33 There was extremely limited evidence from observational studies, with two studies reporting reduced use but no results of statistical tests.

In children, the IA-05-EUP initially showed a statistically significant benefit which lost significance following adjustment for multiple testing. There was no additional evidence from supporting RCTs or observational studies.

Forced expiratory volume in 1 second results from randomised controlled trials

All trials in adults except Hanania et al. 33 and Ohta et al. 35 reported change from baseline in percentage of predicted FEV₁ (Table 28): Ohta et al. reported changes in volume (ml). 35

All of the main adult RCTs showed a statistically significant impact of omalizumab on FEV₁% predicted, although the between-group differences in INNOVATE and EXALT were small in absolute terms at 2.8% and 4.4% respectively. The supportive SOLAR trial and Ohta trial reported a significant effect in the increase in FEV₁ in ml). 35,36 The small Chanez et al. trial, and the supportive trials by Bardelas et al. and Hoshino et al. did not find a statistically significant benefit. 29,32,34

FEV₁ was not reported for the paediatric IA-05 EUP subgroup, but the supportive Busse et al. trial found no difference between the treatment groups in children and adolescents. 28

Forced expiratory volume in 1 second from observational studies

Five observational studies37,39,41,58,62 reported changes in FEV₁% predicted following omalizumab treatment (Table 29). Of those, four37,39,41,62 showed statistically significant improvements in FEV₁% from baseline, and one58 reported no improvement in the longer term (mean 2.1 years).

The PERSIST39 study found a clinically and statistically significant increase of about 12 points in FEV₁% at 16 weeks (p < 0.001). This improvement was maintained after 1 year of treatment (p < 0.001). The APEX study, which retrospectively analysed patient data from 10 UK centres, reported a significant increase of about 8% at 16 weeks (p < 0.001, n = 111) and at up to 12 months treatment (p = 0.002, n = 32). Although it is unclear which proportion of patients included in this analysis strictly met the EU licence, these results are likely to reflect outcomes observed in UK practice. Ohta 201046 showed no significant change in a group of 133 moderate-to-severe asthma persistent patients after 48 weeks of treatment, but found a statistically significant improvement in a subgroup of 37 severe patients (from 1.76 to 1.89 l; p = 0.031). However, the subgroup in Ohta 201046 was classed as severe uncontrolled according to the Japanese label, which includes patients with less severe asthma than the EU licence. The Tzortzaki study showed evidence of sustained benefit over 4 years in patients who met licence criteria.

The study by Brodlie,54 which included children and adolescents treated in the UK on step 5 therapy, showed an increase in median FEV₁ from 2.10 to 2.25 l that was not statistically significant (p = 0.1) following 16 weeks of treatment. Similar results were reported for children aged ≤ 12 and adolescents between 12 and 16 years.

Three observational studies37,43,46,54 reported on changes in FEV₁ (l). The APEX study found an improvement that was significant at 12 months follow-up (from 1.99 to 2.22 l; n = 70, p < 0.001) but not at 16 weeks (from 1.99 to 2.10 l; n = 88, p = 0.22).

Forced expiratory volume in 1 second: summary of omalizumab treatment effect

Randomised controlled trial data indicated statistically significant benefits of omalizumab on FEV₁ as a percentage of the predicted value in the licensed population, although the absolute benefits were small. Supportive trials did not indicate a statistically significant benefit, but these were undertaken in populations with higher mean baseline FEV₁ values than the main trials. One supportive trial showed a benefit in FEV₁ measured in ml. Observational studies provided additional evidence that omalizumab leads to significant improvements in lung function in adults with uncontrolled severe asthma.

In children there was no evidence from the licensed population as IA-05-EUP did not assess FEV₁. The supportive Busse et al. trial found no evidence of an effect of treatment in children and adolescents. 28 Some improvements were reported in children and adolescents treated in severe asthma UK centres, although these results were not statistically significant and drawn from a single small observational study. 54

Quality-of-life results from randomised controlled trials

Some measure of asthma-related QoL was reported by six adult trials (INNOVATE, EXALT and IA-04 EU subgroup in the licensed population and the supportive trials SOLAR and the trials of Hanania and Hoshino) and by the IA-05 EUP subgroup in children. 19,20,31,33,34,36,51 The AQLQ or, in the case of IA-05 EUP, the paediatric AQLQ, was employed in all except the Ohta trial, which reported daily activity scores as a measure of QoL. EXALT reported EQ-5D scores in addition to AQLQ scores.

Data were reported on the mean difference in AQLQ score from baseline and/or on the proportion of patients who improved by ≥ 0.5 points, ≥ 1 point and ≥ 1.5 points from baseline; ≥ 0.5 points is defined as the minimally important difference (Table 30).

The treatment effect (Table 30) on both the main change from baseline and the number of patients reaching a minimally important difference in AQLQ was statistically significant in all cases in main and supportive trials (where reported), although the size of effect was substantially larger in the open-label EXALT and IA-04 (EU subgroup) trials than in INNOVATE. The supporting Hanania trial, which was double blind, also gave a more conservative estimate of treatment effect for mean change from baseline score. 33 The difference between the treatment groups in SOLAR just reached significance, because of a very high treatment response in the placebo arm. 36 The small open-label Hoshino trial did not calculate the between-group difference. 34 The trial by Ohta et al. found no significant difference from baseline in either group in daily activity scores. 35

In children, the IA-05 EU population subgroup showed no difference between groups in either measure of QoL, with very high response rate in the placebo arm.

EXALT showed no difference in change from baseline on the EQ-5D utility index score, but a statistically significant benefit of omalizumab on the EQ-5D health state assessment (p < 0.001).

Quality-of-life data from observational studies

In observational studies the reporting of changes in quality of life was variable; seven observational studies reported some measure of QoL (see Table 31). APEX reported a mean increase ≥ 2 points in the AQLQ at assessment at 16 weeks; a minority of participants were assessed after at least 12 months and reported comparable gains from baseline. 37 In eXpeRience 58.2% of patients reported the minimally important increase of ≥ 0.5 points, whereas in PERSIST this was higher at 84.4% although only 56.7% registered an improvement in utility on the EQ-5D scale. 39

Importantly, the Brodlie et al. study documented evidence of a statistically significant increase in mini-AQLQ scores in children aged both < 12 years and in those aged > 12 years. The numbers involved were very small but these data are valuable as there is very limited evidence on QoL impact associated with omalizumab treatment in children. As this study enrolled OCS-dependent children with very severe asthma attending UK clinics, it may be considered a useful indicator of the potential treatment effect of omalizumab on QoL in children. 54

Quality of life: summary of omalizumab treatment effect

Studies in the adult licensed population showed statistically significant evidence of benefit on the AQLQ. Supporting this, the Hanania et al. trial also showed a statistically significant benefit. 33 This benefit was not seen in the SOLAR trial where there was a substantial placebo response. 36 In children, the IA-05-EUP also demonstrated a substantial placebo response and showed no significant evidence of treatment benefit. 20 In the observational study by Brodlie et al. there was evidence of statistically and clinically significant increases in mini-AQLQ scores in OCS-dependent children in the UK, including those aged under 12 years. Although this population is small, it represents the only evidence on children with OCS-dependent asthma. 54

Withdrawals rates results from randomised controlled trials

Nine RCTs reported omalizumab discontinuation rates (see Table 32). Discontinuation rates varied across the trials, both in the omalizumab and comparator arms. The double-blind RCTs in adults reported withdrawal rates in the omalizumab arm of between 2.4% and 19.4%, compared with 7.7% and 22.2% on placebo. In the open-label trials the withdrawal rates were much higher in the comparator compared with the omalizumab arm. In the one trial in children (IA-05 EU subgroup) the rate of withdrawal was around 20% in both arms.

Three studies reported rates of discontinuation because of lack of treatment efficacy. 19,20,31 Rates were generally low and not dissimilar between treatment groups in two of these RCTs. 19,20 The open-label EXALT trial showed a marked difference between treatment groups, with a higher rate of withdrawals because of lack of treatment efficacy reported in comparator patients.

Withdrawal rates data from observational studies

The observational studies that reported data on withdrawals over a reported period of follow-up are listed in Table 33. The reporting of withdrawals was inconsistent, with a lack of clarity regarding the follow-up duration, timing of withdrawal and, in many cases, the reason for withdrawal. No withdrawals data were found for cohorts on OCS maintenance.

In clinical practice, response to omalizumab is checked at 16 weeks. Four observational studies reported withdrawal rates at this 16-week timepoint. 37,41,43,44,54 Rates ranged from 14.4% to 17.6%. Rates for withdrawal because of lack of efficacy at various time points were consistent at around 15–20% (Table 33). Withdrawal rates reported for longer periods of follow-up were variable: although four studies39,41,44,45 reported a rate around 30% at 6 months or 12 months, two others,40,54 including the largest study (postmarketing surveillance),53 reported lower rates of around 8.5%.

Withdrawal rates: summary

There were considerable variations in withdrawal rates between studies. The key INNOVATE study showed lower withdrawal rates than other trials, with a lower disparity between trial arms than the open-label trials EXALT and IA-04-EUP in which comparator arms showed a higher withdrawal rate than omalizumab arms. Withdrawal rates in observational studies did not appear markedly different to RCT data, although there was greater variation. The withdrawal rate in the APEX study, which might be considered most reflective of UK clinical practice was around 18% at 16 weeks. The IA-05 EUP trial in children had a withdrawal rate of 20%, which was at the upper end of the range for adult RCTs; there was no imbalance between the trial arms.

Studies not included in evaluation of long-term response

A small number of studies appeared to be long term but were not included in this section for the following reasons. Braunstahl53 reported a follow-up duration up to 2 years, but reported outcome data up to 8 months only. Gutierrez57 reported follow-up data at 18 months, but did not report data on the outcomes of interest. These two observational studies were therefore not included in the overview of long-term data. Domingo56 reported data at a mean follow-up of 17 months, but only reported data on OCS use. 56,67 To avoid duplication, this is discussed in the OCS-sparing review and is not reported in this section.

Specific serious adverse events

Concerns also exist suggesting that omalizumab may be associated with an increased risk of specific serious adverse events. The SPC18 highlights immune system disorders, including anaphylaxis, a numerical imbalance in malignancies arising in patients taking omalizumab and arterial thromboembolic events (such as stroke, myocardial infarction, and cardiovascular death). Clinical advisors to the assessment group highlighted anaphylaxis, malignancy and acute thrombotic events as important potential adverse effects to consider in this assessment. Mortality rates associated with treatment and withdrawals because of adverse events are also potentially important drivers for the economic model. The data on these adverse events from the existing reviews are summarised in Table 40. Reporting on adverse events of special interest was generally limited and where events were reported incidence was generally low.

Ongoing studies

A number of publications refer to an ongoing long-term safety study in patients with moderate-to severe asthma (EXCELS). Interim data (to November 2010)71 reports on malignancy rates in patients aged at least 12 years from US centres. The report comprises 18,860 person-years in the omalizumab cohort and 10,947 person-years in the non-omalizumab cohort. No statistically significant differences were shown in the incidence of study-emergent primary malignancy: risk difference (RD) −1.70 per 1000 person-years (95% CI −6.43 to 2.21 per 1000 person-years; see also Adverse events and serious adverse events of omalizumab from existing summaries and reviews). Twenty-four other ongoing studies were identified from the ClinicalTrials.gov website. There were insufficient data available to determine whether these studies met the criteria for inclusion in the review, and attempts to obtain further data or links to publications were unsuccessful.

Attempts were made to access data from a national audit of asthma deaths that is being led by Dr Nasser from Cambridge University Hospital. Unfortunately, data collection only commenced at the beginning of 2012 and data are therefore not yet available. However, Dr Nasser has been running a regional confidential enquiry into asthma deaths for many years and reported that the number of deaths is very small locally (approximately 20) (personal communication, 30 August 2011).

Any adverse event and serious adverse events

Adverse event data in adults and adolescents were reported in nine RCTs (see Appendix 14). Adverse event rates and serious adverse events were generally similar between treatment groups. Two RCTs in adults and adolescents31,36 showed statistically significant higher rates of adverse events in patients exposed to omalizumab (RR 1.25; 95% CI 1.05 to 1.50; and RR 1.14; 95% CI 1.01 to 1.28).

Adverse event data in children were reported in two RCTs (see Appendix 14). 20,28 The proportion of patients experiencing one or more adverse events in the two RCTs including children was similar between treatment groups. Serious adverse event rates, however, were statistically significantly higher in the placebo-treated groups in both RCTs (RR 0.45; 95% CI 0.24 to 0.85 and RR 0.49; 95% CI 0.26 to 0.94). A subgroup of 155 patients with severe asthma were assessed in the children-only trial and the rates of serious adverse events were no longer significantly different (six patients receiving omalizumab and eight receiving placebo). 21

Nine observational studies reported the number of patients experiencing one or more adverse events and eight reported on serious adverse events (see Appendix 15). The rates of adverse and serious adverse events ranged from 6.5% to 98.5%, and 0% to 24.4% respectively. The proportion of patients experiencing adverse events in the observational studies varied more than in patients receiving omalizumab in the RCT. Most RCTs reported more than 50% omalizumab patients experiencing adverse events compared with observational studies, which mainly reported figures < 50%. Serious adverse events generally occurred in < 20% of the population in both the RCTs and the observational studies. Follow-up durations for the majority of observational studies ranged from 4 months up to 2 years. In comparison, follow-up for the majority of RCTs was 48 weeks or less.

Specific serious adverse events

Serious adverse events of special interest (anaphylaxis, malignancy and thrombotic events) were rarely reported.

Withdrawals because of adverse events

Rates of withdrawals because of adverse events were similar between treatment groups in adults and adolescents. Three children withdrew from the children only trial because of adverse events; two (0.5%) in the omalizumab group (one of whom had severe asthma) and one in the placebo group, but the difference was not statistically significant. 20

Rates of withdrawals because adverse events were generally similar in the nine observational studies38–41,45,46,59,62,70 reporting this outcome compared with rates reported in the RCTs. Rates ranged from 1% to 12% in the observational and between 0% and 7.2% in the RCTs. 19,20,28–33,35,50

Hoes et al.82

This paper describes a systematic review of the adverse effects of low to medium low doses of OCSs (doses of ≤ 30 mg, with some flexibility). The literature searches included MEDLINE, EMBASE and CINAHL using appropriate search terms. The criteria for studies to be included in this review were: the study was of adults with inflammatory diseases treated with corticosteroids (glucocorticoids); dose of corticosteroids ≤ 30 mg (one study that used a higher dose for the first month was included); dichotomous adverse events data had to be reported; the study was reported in a full paper. Studies that included patients with previous long-term or recent experience of OCSs were excluded. We note that there is some ambiguity about whether or not the paper was purely about OCSs.

A potential limitation of this systematic review is that only papers that reported dichotomous data were included, with the risk that some potentially relevant data might be missing. In addition, the actual duration of each trial's follow-up is not reported, ignoring the possibility that event rates may change with time. For the purposes of the present appraisal, the results of this review are further limited by the fact that there were no included trials of asthma or chronic obstructive pulmonary disease (COPD). The results were reported overall and by diagnosis [rheumatoid arthritis (RA), polymyalgia rheumatica (PMR) and inflammatory bowel disease (IBD)] and rates were found to vary between indications, raising the question of the generalisability to asthma patients. Finally, the data from the IBD trials included in the review were different from those from other trials; however, this may reflect the fact that they are short-term trials: short-term trial results are more heterogeneous and with higher event rates than those of longer-term trials.

The Hoes et al. review calculated and reported rates of adverse events based on single-arm or uncontrolled data: all adverse events and also categories (by body systems) of adverse events. However, it did not report rates of individual adverse events, for example it reported ‘cardiovascular’ but not acute myocardial infarction (AMI). All results were reported as dichotomous data (events/patient-years) with calculated event rates (rate/100 patient-years) and 95% CI. The latter are given in Table 41. The paper stated that ‘comparison of low and medium dosages did not show dose dependency of any of the adverse events’. This could be a reflection of some flaw in the analysis, as it does not reflect other findings related to adverse events of OCSs: it could be that the dose range studied is too narrow, or the dichotomy between low and medium too crude, to reveal a dose-dependent effect.

In summary, although the Hoes et al. paper described a good-quality systematic review, unfortunately it included only single-arm or uncontrolled data and only presented data for the number of patients experiencing any adverse event or rates of classes of adverse events, for example the rate for ‘gastrointestinal’ events. This information was not useful for the purposes of the economic model.

Sarnes et al.83

This study was a semi-systematic review of the adverse events associated with oral and parenteral corticosteroids. It is, in part, an update of the review by Manson et al. 84 and so there is much overlap between these two publications. The study included appropriate literature searches of databases including MEDLINE, EMBASE and The Cochrane Library, with dates reported (up to 2009) and the inclusion criteria were reported. However, the synthesis was not systematic nor transparent. Some relevant data were presented together with costs in the US context.

The review included 47 studies, but four were excluded for being of too poor quality. Twenty-four of the studies were of OCSs: 19 were of parenteral corticosteroids or parenteral and oral mixed. Six studies were in paediatric patients, but results for adverse events in children were not presented separately. The results of this study were presented as risk ratios for specified adverse events associated with certain dose levels of corticosteroids, although some incidences are also reported. However, the sources of results are not reported consistently: sometimes the source is an individual primary study, or a narrative synthesis of primary studies, sometimes it is the results of another review article or articles. One additional problem was that the results for OCSs were not separated from parenteral corticosteroid use.

Unfortunately, this review did not use adequate systematic review methods and it is not possible to be certain that the results are reliable.

Manson et al.84

This was a semi-systematic review that involved literature searches of key databases including MEDLINE, EMBASE and The Cochrane Library, covering the period January 1990–March 2007. This aim of the review was to identify studies that considered adverse events because of OCS treatment. It specified criteria for inclusion of studies in the review: that studies reported on adverse effects/events of OCSs (prevalence of OCS adverse effects, relationship between OCS adverse effects and patient characteristics or duration of steroid use, dose–response relationship for OCSs and adverse effects, or threshold effect for OCS adverse events). Studies that investigated non- (or sub-)clinical adverse effects, e.g. effects on bone markers, were not included. Non-English-language papers were not included. However, the synthesis was poor, such that, although all studies were presented there was no clarity regarding the method of synthesis: the data synthesis was essentially narrative and not transparent.

The paper reported individual trial results and also reported relative risks for certain adverse effects, but only those where relative risks were reported in the primary publication. Unfortunately, the review did not report the variance. Importantly, the relative risks reported are just from individual studies with no explanation why synthesis was not attempted, or how studies or data were selected.

Allen et al.85

This study was a meta-analysis of data on the effects of corticosteroids on height attained by children. This meta-analysis was not based on a systematic review. Although the source data were reported as having been identified through an ‘exhaustive literature search of “leading medical journals” up to January 1993’, further details were not given. It was unclear what study designs were included and there was no attempt at quality assessment of the included studies.

It was a meta-analysis of studies comparing attained height with expected height. The analysis included 21 studies including 810 patients (395 of whom were on OCSs). It is unclear how representative the included studies are of all studies on the effects of OCSs on growth, given that only those that reported the precise numbers of children at or above their expected height were included. The use of meta-analysis appears appropriate. However, the results of the meta-analysis are presented only as a Z-value, p-value and mean correlation coefficient (r).

The results found that prednisone (and separately other OCSs) is associated with a statistically significant tendency to not attain expected height (Table 42). However, there was no information on how short of expected height these children were.

Although growth retardation is a known and concerning adverse effect of OCS use in children, an estimate of the size and clinical significance of this effect has not been identified from the literature.

Synthesis

The most useful and appropriate source of information on the adverse effects of OCSs was the semi-systematic review by Manson et al. 84 This was because it focused on OCSs, whereas the updated review by Sarnes et al. 83 included oral and i.v. administration. The Manson et al. 84 publication also presented relative risks or odds ratios which are required for the economic model. The analysis presented in the Novartis submission was also considered a good source of evidence, especially as the effect sizes reported in the Novartis submission were, in the most part, derived from the Manson et al. study. However, there were some inconsistencies between the information provided in the MS and the Manson et al. paper, and therefore the primary studies were checked and data used from those primary sources where necessary. The summarised estimates are listed in Table 43. These were taken, in most cases from Manson et al., and the primary study authors and citation number from Manson et al. are reported in the table. Where an estimate was taken directly from a primary source, that source is cited directly.

Adverse events of oral corticosteroids

The translation of any steroid-sparing effects of omalizumab into patient benefit is dependent on the avoidance of the adverse events associated with OCSs. Although OCS adverse events are widely recognised, there has been limited systematic appraisal of the level of risk associated with maintenance use of OCSs. All the evidence syntheses identified in our review were subject to limitations, and the reliability of the data were unclear. The most reliable source of evidence provided quantitative evidence for the known adverse events of fracture, diabetes, peptic ulcer, cardiovascular events including myocardial infarction and stroke, cataract and glaucoma, sleep and mood disturbance, and weight gain. 84 Increased fracture risk remains a long-term consequence even when OCS is discontinued as a consequence of irreversible osteoporosis. Weight gain has also been identified, both by consultee submissions to NICE for the appraisal of omalizumab and by the clinical advisors to this technology assessment, as being of key importance and as leading to a cycle of reduced asthma control, increased OCS requirement and further weight gain. There is some evidence of a relationship between childhood OCS treatment and failure to achieve expected adult height. 85

Manufacturer's submission for technology appraisal number 13388

The manufacturer approached the decision problem by looking at adults and adolescents with severe persistent allergic asthma in accordance with the EU/UK marketing authorisation. Omalizumab as an add-on therapy to standard care was compared with standard care alone. Standard care included high-dose ICS, long- and short-acting β₂-agonists, OCSs, leukotriene antagonists and, where appropriate, theophylline. The MS presented evidence on the clinical effectiveness of add-on therapy with omalizumab based on the results of the INNOVATE trial. The primary outcomes from this trial were the rate of CS asthma exacerbations, the rate of CSS exacerbations and the rate of emergency visits for asthma. The input parameters in the economic analysis were largely based on the INNOVATE study. 19

The Markov transition model had a lifetime of 40 years and consisted of five health states: day-to-day symptoms, CSNS exacerbation, CSS exacerbation, asthma-related death and death from other causes. In the model, it was assumed that patients on omalizumab were assessed for response to treatment at 16 weeks. The proportion of patients on omalizumab who were responders at 16 weeks was based on the proportion of responders observed in the INNOVATE study at 28 weeks. Non-responders were assumed to revert back to standard therapy and receive the same exacerbation rates and HRQoL as patients on standard care. Responders to omalizumab continued on omalizumab treatment for 5 years. During the period of treatment, responders to omalizumab were assumed to experience the exacerbation rates and HRQoL improvements observed in the omalizumab responders of the INNOVATE study. After treatment discontinuation (5 years), patients who were on omalizumab were assumed to experience the exacerbation rates and HRQoL of patients on standard care. HRQoL for day-to-day symptoms with omalizumab and standard care were estimated by mapping the AQLQ scores collected during INNOVATE for each treatment arm to EQ-5D utility scores using a published mapping function. 103 The loss of HRQoL associated with CSNS and CSS exacerbations were based on a published study by Lloyd et al. (2007). 104 Asthma-related mortality was assumed to occur only from a CSS exacerbation. As no deaths were observed in INNOVATE, an asthma-related mortality risk of 3.108% was obtained from a Swedish observational study on data collected between 1988 and 1990. 99 Costs were based on health-care resources consumed in INNOVATE with UK unit prices applied. The acquisition cost of omalizumab was based on the distribution of doses observed in INNOVATE, and assuming no vial wastage and re-use of unused vial portions. Appendix 16 presents the input parameters used in the MS for TA133.

The base-case analysis for the patient characteristics of the INNOVATE population produced an ICER of £30,647 per QALY gained. Two subgroup populations were also presented: (1) a high-risk hospitalisation subgroup, consisting of 39% of patients in INNOVATE, who had asthma exacerbations requiring hospital admission in the year prior to enrolling in the trial, and (2) a severe subgroup of patients from the IA-04 study who met the EU/UK marketing authorisation requirements for omalizumab. 30 The ICERs for the hospitalisation subgroup were £26,500 per QALY gained, whereas the ICERs for the IA-04 subgroup were £21,700 per QALY gained. Table 45 presents the results of the manufacturer's one-way sensitivity analysis. These suggested that the cost-effectiveness results were most sensitive to the asthma-related mortality risk, treatment duration and time horizon. Reducing the asthma-related mortality rate from 3.109% to 2.478% increased the ICERs from £30,647 to £33,468 per QALY gained, whereas a 0% mortality rate increased the ICER to £73,177.

The previous Evidence Review Group's critique102

The MS was considered to be of good quality and to meet the requirements of the NICE reference case. 105 The modelling approach, health states and structural assumptions were considered reasonable. However, the ERG identified a number of issues with the parameters used in the economic model and uncertainties relating to the cost-effectiveness analysis. Some data sources were not adequately justified, for example, the source used to inform asthma-related mortality. The one-way sensitivity analysis conducted by the manufacturer did not capture uncertainty adequately, as it was performed on a limited number of parameters and using inappropriate ranges of parameter values.

The ERG considered that the asthma-related mortality applied in the model may not be reflective of the mortality risk faced by patients in the UK. The asthma-related mortality used in the base-case analysis was obtained from a Swedish observational study that evaluated the impact of training ambulance crews on the management of acute asthma. Data on the number of calls concerning acute asthma and on the number of deaths following ambulance arrival were collected between 1988 and 1990. 99 It was unclear whether the results were generalisable to the UK setting or appropriate for the year of the appraisal (2006). In addition, the mortality rate observed in the Swedish study was for an average age of 62.3 years but the manufacturer applied the rate to a patient cohort starting in the model at 43 years of age. Furthermore, the definition of CSS exacerbations used in the model and INNOVATE, where the mortality is applied, may not correspond to the same definition of an acute asthma attack that prompted patients to call an ambulance as used in Lowhagen et al. 99

The ERG noted uncertainties surrounding the utility values assigned to CSNS and CSS exacerbations, and the cost of omalizumab on a per milligram basis. Therefore, the ERG performed an exploratory scenario analysis on alternative assumptions for these parameters. The ICERs for these scenarios ranged from £33,320 to £40,889 per additional QALY for the base-case population (base-case ICER = £30,647), between £29,849 and £34,303 per additional QALY for the hospitalisation subgroup (base-case ICER = £26,509) and between £24,698 and £30,715 for the IA-04 subgroup (base-case ICER = £21,660). The ERG also performed an amended probabilistic sensitivity analysis and estimated a mean probabilistic ICER of £38,900 per QALY gained, and a probability that omalizumab is cost-effective of 0.236 at the £30,000 per QALY threshold. The ERG concluded that, in addition to asthma-related mortality, the improvement in HRQoL from omalizumab and the assumptions used to calculate the cost of omalizumab were the key drivers of cost-effectiveness.

Manufacturer's submission for technology appraisal number 20189

The manufacturer approached the decision problem by looking at children aged 6–11 years with severe persistent allergic asthma in accordance with the EU/UK marketing authorisation. Omalizumab as an add-on therapy to standard care was compared with standard care alone from the UK NHS perspective over a lifetime horizon. Standard care included high-dose ICS, LABAs and, where appropriate, OCSs. The manufacturer undertook a systematic review of previously published economic evaluations relevant to the decision problem but no studies were found. Therefore, the manufacturer submitted a de novo economic model. The model had the same structure as that used for TA133. The exacerbation rates and resource use data were drawn largely from the preplanned subgroup IA-05 EUP of the IA-05 trial in children, corresponding to the EU/UK marketing authorisation.

Patients on omalizumab should be assessed for response at 16 weeks. The MS included a post hoc ‘responder’ subgroup of the EUP population. Responders were defined as children who were rated as excellent or good on the GETE scale at 52 weeks of treatment. The manufacturer used the response rate at 52 weeks as a proxy for the proportion of patients on omalizumab who were responders at 16 weeks. Non-responders were assumed to revert back to standard therapy and receive the same exacerbation rates, costs and HRQoL as patients on standard therapy alone. Responders to omalizumab and patients on standard therapy (or non-responders) were assumed to experience the exacerbation rates and resource use observed in the respective treatment arms of the IA-05 EUP study. No deaths were observed in the IA-05 EUP study; therefore, asthma-related mortality was obtained from an alternative published source (Watson et al. 106). Watson et al. 106 examined the rate of all-cause mortality following hospital admissions for asthma and acute severe asthma in the UK. Similar to the adult and adolescent's model, asthma-related mortality was assumed to occur only from a CSS exacerbation. The authors estimated an asthma-related mortality rate following hospital admission for acute severe asthma of 0.097% for children aged < 12 years, 0.319% for ages 12–16 years, 0.383% for ages 17–44 years and 2.48% for ages ≥ 45 years. No HRQoL values for children were available from the IA-05 EUP study. IA-05 EUP used the paediatric-AQLQ, but a non-significant difference was observed between treatment groups. Therefore, the base-case analysis assumed that there was no HRQoL improvement in day-to-day symptoms for omalizumab compared with standard therapy until patients reached the age of 12 years. After age 12 years, children were assumed to receive the HRQoL improvements observed in INNOVATE, based on the AQLQ improvement which was mapped onto EQ-5D (the same as the MS for TA133). The HRQoL values for CSNS and CSS exacerbations were based on values reported in Lloyd et al. (2007)104 (the same as for adults and adolescents in TA133). Costs were based on the resource use observed in IA-05 EUP with UK unit prices applied. For the acquisition costs of omalizumab, the same assumptions of no vial wastage and re-use of vials were employed as in the adults and adolescents model. More importantly, children were assumed to remain on the same baseline dose schedule throughout the entire treatment duration. Appendix 16 presents the input parameters used in the MS for TA201.

The base-case analysis corresponded to the patient characteristics observed in the IA-05 EUP population. The MS also presented a post hoc subgroup analysis for a high-risk population, the EUP hospitalisation subgroup, consisting of patients who experienced at least one hospitalisation for an asthma exacerbation in the year prior to study entry. The ICER for the base-case analysis was £91,188 per QALY gained, which was reduced to £91,169 followed a slight amendment to the model noted by the ERG. The ICER for the hospitalisation subgroup was £65,911 per QALY gained. The manufacturer conducted extensive one-way sensitivity analyses. Table 46 presents the results of the manufacturer's sensitivity analysis. Despite some scenarios having a substantial impact on the ICER, none reduced the ICER to below £68,029 per QALY gained (achieved by assuming that children aged < 12 years experience the same HRQoL improvement with omalizumab as adults). An ICER of £69,603 per QALY gained was achieved by doubling the asthma-related mortality rate for all ages. The ICER was most sensitive to the length of treatment duration, the HRQoL improvement assumed for omalizumab compared with standard therapy and the asthma-related mortality, suggesting that these parameters were the main drivers of cost-effectiveness. Probabilistic sensitivity analyses suggested that if maximum acceptable thresholds of £20,000 and £30,000, respectively, for an additional QALY gained were used, omalizumab had a 0% probability of being considered cost-effective.

The previous Evidence Review Group's critique for technology appraisal number 201107

As with TA133, the ERG considered the economic submission to be of good quality, meeting most of the requirements of the NICE reference case, and that the structure of the Markov model was appropriate for the decision problem. Many of the key uncertainties, such as asthma-related mortality and treatment duration were explored through one-way sensitivity analysis for the base-case population but not for the hospitalisation subgroup.

The ERG undertook exploratory analysis to identify the factors underlying the cost-effectiveness results in children aged 6–11 years using alternative parameter values which matched those used in TA133 for adults and adolescents. The exploratory analysis focused on the hospitalisation subgroup and the parameter values for exacerbation rates, proportion of responders, asthma-related mortality and HRQoL. The exploratory analysis showed that applying the efficacy rates for CSNS and CSS exacerbations from INNOVATE (as used in TA133) to patients aged ≥ 12 years in the hospitalisation subgroup resulted in an increase in the ICER from £65,911 to £73,779. Applying an improvement in HRQoL associated with omalizumab relative to standard care for day-to-day symptoms for children < 12 years decreased the ICER to £53,133. The exploratory analysis demonstrated that asthma-related mortality was the key driver of cost-effectiveness. The asthma-related mortality used in the children's submission was substantially lower than that applied in the submission for TA133. The model for adults and adolescents (aged ≥ 12 years) considered a cohort with an average age of 45 years and an asthma mortality risk of 3.109%. Applying the higher mortality of 3.109% from TA133 to the children's model (average age 9 years) once patients reach the age of 12 years, reduced the ICER to £31,737. The ERG expressed the view that the higher mortality rate may be appropriate for patients aged > 45 years but it was unlikely to be appropriate for younger populations.

The previous ERG identified a number of potential weaknesses and remaining uncertainties in the economic submission for TA201. These included: (1) the use of 52-week data as a proxy for 16-week assessment of response to treatment (the period specified in the marketing authorisation); (2) the assumption that exacerbation rates remain constant over time in children and adolescents, especially as adolescent growth can have an impact on asthma; (3) no systematic literature searches were undertaken to identify key parameters such as asthma-related mortality; (4) no uncertainty was considered in the cost estimates as part of the probabilistic sensitivity analysis; (5) the cost of an exacerbation was not differentiated according to severity; (6) a treatment duration of 10 years was assumed without providing justification; (7) HRQoL values for children were informed by studies in adults; and (8) other more severe subgroup populations were not considered in the economic analysis, for example, patients with more than three exacerbations per year.

Responders to omalizumab therapy

The proportion of patients responding to omalizumab treatment observed in the trials was used to inform the probability of being an omalizumab responder at 16 weeks. For the base case of adults and adolescents, the proportion of responders observed at 28 weeks in INNOVATE was used as a proxy for response at 16 weeks. For children, the proportion of responders observed at 52 weeks in IA-05 EUP was used as a proxy for response at 16 weeks. For the EXALT and APEX scenarios, the 16-week response rates reported in these studies were used. For the subgroup analyses, the response rate observed in each of the subgroups was used. Once patients were identified as responders, they were assumed to receive the exacerbation rates of responders over the entire duration of treatment.

Table 49 presents the proportion of responders to omalizumab therapy used in the economic model for the base-case populations, alternative scenarios and patient subgroups. The proportion of responders observed differed in the two double-blind RCTs: 56.5% in INNOVATE and 74.2% in IA-05 EUP. The proportion of responders in EXALT was greater than in INNOVATE at 69.9%. However, the assessment of response in EXALT may have been influenced by the open-label design of the trial. The proportion of responders in APEX was the highest at around 80%. This may reflect not only the selection of the most suitable patients for omalizumab in clinical practice but also the influence of knowing the treatment status of the patient when assessing for response. The proportion of responders in the patient subgroups was generally lower than in the overall population.

The approach assumes that the response to omalizumab treatment remains unchanged over time. However, evidence from EXALT suggests that this may not be the case; around 8.6% of responders at 16 weeks in EXALT were not considered responders at 32 weeks. Although these results may have been influenced by the open-label design of the trial, they indicate that response may not persist through time. Therefore, there may be patients who discontinue treatment after 16 weeks or patients who remain on treatment but experience a reduced treatment effect. The potential impact of this was not considered in the MS.

Exacerbation rates

The exacerbation rates observed during the trials were used to inform the probability of experiencing an exacerbation in the model. The exacerbation rates from the trial follow-up period were annualised and assumed constant throughout the model. Patients on standard care were assumed to experience the exacerbation rates observed in the standard care arm of the trials. During the first 16-week cycle, patients on omalizumab experience the exacerbation rates observed for all patients who were randomised to receive omalizumab in the trials, regardless of response rate. From 16 weeks onwards, omalizumab responders were identified and received the exacerbation rates observed by the responders in the trial. Non-responders were assumed to revert back to standard therapy and experience the exacerbation rates in the standard care arm of the trials. Similarly, once omalizumab treatment is discontinued omalizumab responders revert to standard care.

Table 50 summarises the values for the key parameters on treatment effectiveness used in the model.

The approach taken by the manufacturer seems appropriate in light of the available evidence. However, the exacerbation rates observed for patients in the placebo group may be lower than those experienced by patients on standard care in clinical practice, because of the increased contact with health-care professionals inherent to any RCT. If patients on standard care experience exacerbations more frequently than in INNOVATE, omalizumab may be more cost-effective than the base-case results suggest. In addition, some observational studies suggest that the likelihood of a future exacerbation is dependent on number of past exacerbations, that is exacerbation rates are not necessarily constant over time. 108^–110

All-cause mortality

All-cause mortality was based on interim life-tables for England and Wales for the years 2007–2009 from the Office for National Statistics. 115 However, asthma-related deaths were not removed from the life-tables and so there is some element of double counting of mortality in the model. However, because of the small number of asthma deaths in the general population, this is unlikely to be a significant issue.

Omalizumab therapy costs

Costs associated with omalizumab therapy include the costs of the drug itself and the costs of administration and monitoring. Omalizumab is administered as a subcutaneous injection every 2–4 weeks, and the exact dose depends on the patient's serum IgE and weight.

The dosing distribution of omalizumab used in the economic analysis refers to the ‘standard dose’ of treatment rather than the ‘expanded dose’. An expanded dose above 375 mg per administration and/or dosing for some lower weight patients with IgE of greater than 700–1500 international units (IU)/ml was included in the EU SPC in a January 2010 update. 26 However, the standard dose was applied in the earlier studies of INNOVATE, EXALT, APEX and IA-05 EUP.

Omalizumab is currently available as 75 mg and 150 mg prefilled syringes. 22 At the time of the previous STAs, omalizumab was only available as a 150-mg vial. Consequently, the assumptions regarding vial wastage and re-use in the previous appraisals are no longer relevant. For the base-case populations, the model assumes an average dose of omalizumab corresponding to the dose-distribution of the patient population in INNOVATE, EXALT, APEX and IA-05. Although children would be expected to increase in weight during the period of treatment duration, the model does not adjust for an increase in weight. However, the average cost per patient is similar across populations; therefore, the increase in weight is unlikely to change the results significantly.

Costs of administration and monitoring

The costs of administration were estimated by assuming 10 minutes of administration time and using the hourly cost of a specialist asthma nurse at £47/hour. 116 Monitoring costs for anaphylaxis were included up to and including the 16-week responder assessment. For the first three administrations, the monitoring was assumed to take 2 hours, while from the fourth administration up to the 16-week assessment, monitoring was assumed to take only 1 hour, with each hour costing 15 minutes of specialist asthma nurse time. The costs of administration and monitoring were considered appropriate by our clinical advisors.

Standard care costs: standard therapy and routine secondary care