3 The company's submission
3.1 The company presented evidence on a narrower population than covered by the marketing authorisation based on feedback from UK clinicians on the appropriate population for omalizumab in England. It positioned omalizumab in adults and young people aged 12 years and over with chronic spontaneous urticaria, previously treated with H1‑antihistamines (which have been used at up to 4 times the licensed dose), leukotriene receptor antagonists (LTRAs) and H2‑antihistamines (also referred to as H2‑receptor antagonists), whose disease is responding inadequately to whichever combination of therapies they are currently having.
3.2 The company carried out a systematic review that identified 6 trials evaluating omalizumab compared with placebo in patients with refractory chronic spontaneous urticaria. These included 3 phase III studies (GLACIAL, ASTERIA I and ASTERIA II), 2 phase II studies (MYSTIQUE and X‑CUISITE) and 1 small (n=10) study by Gober et al. (2008). To estimate clinical effectiveness, the company considered only the GLACIAL trial. The company included the methods and results of the ASTERIA I and ASTERIA II trials as an appendix to its submission. The companydid not include X‑CUISITE or the Gober et al. study, noting that the dosage of omalizumab used in these studies was different from the licensed dose (300 mg). The company considered the MYSTIQUE trial 'not important', even though the trial evaluated 300 mg omalizumab, noting that the data from the 3 large phase III trials were sufficient for this appraisal.
3.3 The primary objective of the GLACIAL trial was to evaluate the safety of the licensed dose of omalizumab (300 mg) over the 24‑week treatment period; another objective was efficacy. GLACIAL was a multicentre, international, randomised, double‑blind, placebo‑controlled, parallel‑group trial. Sixty‑five centres in 7 countries (including 4 centres in the UK) participated. The trial included patients aged 12–75 years with chronic spontaneous urticaria for more than 6 months, which was refractory to:
H1‑antihistamines (up to 4 times the approved dose) and either H2‑antihistamines or LTRAs, or
all 3 drugs in combination.
3.4 Patients were randomised in a 3:1 ratio to omalizumab (n=252) or placebo (n=84). The demographics and clinical characteristics of patients at baseline were similar between the omalizumab and placebo groups. The mean age of patients was 43.1 years, 71.9% were women, the mean BMI was 29.8 kg/m2, 89.0% were white and the median time since diagnosis was 3.6 years (range 6 months to 54.1 years). The mean number of previous medications for chronic spontaneous urticaria was 5.9 (standard deviation [SD] 2.5) in the omalizumab group and 6.4 (SD 2.9) in the placebo group.
3.5 Outcome measures of itch: The daily itch severity score is the average score from measuring twice daily (morning and evening) on a scale of 0 (none) to 3 (severe). The weekly itch severity score is the sum of the daily itch severity scores over 7 days and ranges from 0 to 21. A higher itch severity score indicates more severe itching. In the trials, the baseline weekly itch severity score was the sum of the daily itch severity scores over the 7 days before the first treatment. In the GLACIAL trial, the mean values for weekly itch severity score at baseline were 14.0 (SD 3.6) in the omalizumab group and 13.8 (SD 3.6) in the placebo group. In the trials, a 'minimum important difference' was defined as a decrease of at least 5 points in the weekly itch severity score.
3.6 Outcome measures of urticarial activity: The urticaria activity score (UAS) is a composite of scores on a scale of 0 (none) to 3 (intense/severe) for the number of wheals (hives) and the intensity of the itch, measured twice daily (morning and evening). The daily UAS is the average of the morning and evening scores (ranging from 0–6) and the UAS7 is the sum of the daily UAS over 7 days (ranging from 0–42). A higher UAS indicates more urticaria activity. Baseline UAS7 was calculated using data from the 7 days before the first treatment date. The mean values for UAS7 at baseline were 31.2 (SD 6.6) for the omalizumab group and 30.2 (SD 6.7) for the placebo group. In the trials, a 'minimum important difference' was defined as a decrease of at least 11 points in the UAS7.
3.7 At baseline, 54.4% (137/252) of those in the omalizumab group and 49.4% (41/83) of those in the placebo group had angioedema. Patients were tested for the presence of anti‑omalizumab antibodies and all but 1 patient tested negative at baseline.
3.8 The duration of the trial was 24 weeks, during which patients had omalizumab, with a follow‑on 16‑week observational period. However, the primary efficacy outcome was the change in the mean weekly itch severity score from baseline to 12 weeks. Secondary outcomes included changes from baseline to week 12 in:
the weekly number of hives score
the weekly size of largest hive score and
the proportions of patients whose disease showed a 'minimum important difference' in these outcomes.
The results showed that omalizumab improved weekly itch severity score compared with placebo (−8.6, 95% confidence interval [CI] −9.3 to −7.8 for omalizumab compared with −4.0, 95% CI −5.3 to −2.7 for placebo; p<0.001). Omalizumab improved all the other reported clinical efficacy outcomes, including change in UAS7 (−19.0, 95% CI −20.6 to −17.4 for omalizumab compared with −8.5, 95% CI −11.1 to −5.9 for placebo; p<0.001).
3.9 Omalizumab provided more rapid relief in symptoms than placebo, as measured by the median time to a minimum important difference in weekly itch severity score (2 weeks compared with 5 weeks, p<0.001). The mean change from baseline in weekly itch severity score was lower in patients randomised to omalizumab than in patients randomised to placebo from as early as week 1, and remained lower than placebo up to week 24. During the post‑treatment follow‑up (week 24 to week 40), the mean weekly itch severity score in the omalizumab arm gradually increased to values similar to the placebo group, with no differences between the omalizumab and placebo groups at week 40.
3.10 The company submitted a subgroup analysis of the GLACIAL trial, which it defined post hoc and which investigated the efficacy of omalizumab in patients who took H1‑antihistamines, H2‑antihistamines and LTRAs (instead of just taking 2 drugs: H1‑antihistamines and either H2‑antihistamines or LTRAs). The company analysed individual patient data to estimate the change in UAS7 and Dermatology Life Quality Index score from baseline to 12 and 24 weeks of treatment. The results of the subgroup analysis are academic in confidence and, although considered by the Committee, cannot be presented here.
3.11 ASTERIA I (n=319) and ASTERIA II (n=322) were international, phase III, multicentre, randomised, double‑blind, placebo‑controlled, parallel‑group trials. The primary end point of these trials was the change in weekly itch severity score from baseline to week 12. The trials differed from each other only in the duration of treatment: 24 weeks (6 doses) in ASTERIA I and 12 weeks (3 doses) in ASTERIA II. The trials enrolled patients aged 12 years to 75 years who had chronic spontaneous urticaria for more than 6 months, which was refractory to licensed doses of H1‑antihistamines for at least 8 consecutive weeks. Patients were randomised to omalizumab 75 mg, 150 mg or 300 mg or to placebo in a 1:1:1:1 ratio. The company considered that the demographics and clinical characteristics of the patients at baseline were well balanced across study groups in both trials.
3.12 The ASTERIA I and ASTERIA II trials showed that omalizumab 300 mg improved most outcomes at week 12 compared with placebo.
3.13 The company identified 1 prospective and 9 retrospective non‑randomised studies evaluating omalizumab in patients with chronic spontaneous urticaria. The company's submission summarised the methodology and results of these studies. In the company's view, the non‑randomised studies suggested further benefits of omalizumab, such as reducing the need for concomitant medications including corticosteroids, and showing that re‑treatment with omalizumab is effective. However, because these were observational studies, the results may be biased by confounding.
3.14 For evidence relating to the comparators listed in the scope, the company identified 3 randomised controlled trials (RCTs) and 5 non‑randomised studies that included treatment with 1 or more of the comparators. The company identified 2 RCTs and 2 non‑randomised studies for ciclosporin; 1 RCT and 1 non‑randomised study for methotrexate; and 1 non‑randomised study for mycophenolate mofetil. The company did not identify any head‑to‑head trials of omalizumab with these comparators. The company stated that it did not compare omalizumab with any of the potential comparators indirectly because the evidence base for the comparator technologies was characterised by: different outcomes; small sample sizes; differences in treatment duration and disease severity at baseline; and different concomitant therapies used.
3.15 The company presented data from the GLACIAL trial on adverse events during the 24‑week treatment period and the subsequent 16‑week follow‑up. At 24 weeks, the incidence of adverse events was similar in the omalizumab and placebo groups (65.1% compared with 63.9% respectively). During the treatment plus follow‑up period of 40 weeks, the company saw comparable rates in:
1 or more adverse events (83.7% with omalizumab compared with 78.3% for placebo)
1 or more adverse events suspected to be caused by the drug (11.1% with omalizumab compared with 13.3% for placebo)
1 or more serious adverse events (7.1% with omalizumab compared with 6.0% for placebo) and
adverse events leading to withdrawal (1.2% in both groups).
In both groups, the most frequent treatment‑related adverse events were infections and infestations (36.9% with omalizumab compared with 30.1% for placebo), gastrointestinal disorders (15.9% compared with 14.5%), and skin and subcutaneous disorders (16.7% compared with 14.5%). Headache (8.7% compared with 3.6%) and upper respiratory tract infections (7.1% compared with 2.4%) were more common in the omalizumab group, whereas sinus congestion (1.2% compared with 4.8%), migraine (1.6% compared with 3.6%) and idiopathic urticaria (2.8% compared with 7.2%) were more common in the placebo group.
3.16 The summary of product characteristics for omalizumab notes that type 1 local or systemic allergic reactions, including anaphylaxis and anaphylactic shock, may occur with omalizumab, even after a long treatment duration. The company noted that anaphylaxis occurs rarely (in 0.09% of patients) when using omalizumab to treat allergic asthma.
3.17 The most frequent treatment‑related adverse events in both the omalizumab and placebo groups of the ASTERIA II trial were infections and infestations (35.4% compared with 38.0% respectively), gastrointestinal disorders (11.4% compared with 15.2%) and skin and subcutaneous disorders (17.7% compared with 8.9%). The company labelled the adverse events data from the ASTERIA I trial as academic in confidence, so these cannot be presented here.
3.18 The ERG commented that the company identified the relevant studies for this appraisal. The ERG noted that the population of the GLACIAL trial differed from that of the NICE scope (the scope specified people aged 12 years and over with chronic spontaneous urticaria that had an inadequate response to H1‑antihistamines), nor was it in line with the company's decision problem because only some of the people in the trial were unsuccessfully treated with H1‑antihistamines (up to 4 times the licensed dose), LTRAs and H2‑antihistamines in combination. The ERG did not agree with the company that the ASTERIA I and ASTERIA II trials are not relevant for this appraisal. Specifically, the ERG noted that the ASTERIA trial populations are in line with the scope and the marketing authorisation for omalizumab and, as with the GLACIAL trial, some patients in the ASTERIA trials (although a smaller proportion than in GLACIAL) matched the population specified in the company's decision problem.
3.19 The ERG was unable to assess the quality of the included trials completely because: the company provided few details; published abstracts were not sufficiently detailed; and the ERG received the clinical study reports too late to include them in its critique of the company's submission. The ERG agreed that, taking them at face value, the trials appeared well conducted and of reasonably good quality.
3.20 The ERG commented that the effectiveness of omalizumab appeared greater in ASTERIA I and ASTERIA II than in the GLACIAL trial. The ERG noted that, in all 3 trials, patients in the treatment and placebo groups had lower weekly itch severity scores, and commented that the company did not address this apparent placebo effect. The ERG noted that the trials did not provide data on reducing or stopping corticosteroids, as specified in the scope. The ERG also noted that the definitions used by the company to define the minimum important difference in itch severity score and UAS7 were based on a small study (n=73) by Mathias et al. (2012), and are not widely accepted. The ERG also noted that the company did not present EQ‑5D results from the individual trials despite presenting pooled data from 3 trials to inform the health economic model.
3.21 The ERG commented that the Committee should interpret the results of the subgroup analysis with caution. The ERG would have preferred the company to compare the subgroup with the other patients not in the subgroup, as opposed to comparing the subgroup with all patients in the trial.
3.22 The ERG performed study‑level meta‑analyses of the GLACIAL, ASTERIA I and ASTERIA II trials, which the company had not done. This included the differences at week 12 in the mean change from baseline in the weekly itch severity score and the UAS7, calculated by pooling the results from GLACIAL, ASTERIA I and ASTERIA II trials, but not including MYSTIQUE. Using a fixed‑effect model, the summary effect measure estimated a mean difference of −5.00 (95% CI −5.94 to −4.06) in the weekly itch severity score and of −11.39 (95% CI −13.38 to −9.41) in UAS7. The pooled results for both outcomes remained unchanged for both the fixed‑effect and random‑effects models. For the trials evaluating the comparators listed in the scope, the ERG largely agreed with the company that the trials were too different for the results to be compared.
3.23 The ERG agreed that the incidences of adverse events and serious adverse events were similar in the omalizumab 300 mg groups and the placebo groups in the 3 trials included in the company's submission, but noted that the company did not test the observed differences statistically.
3.24 The company submitted a de novo Markov model. The company assumed that omalizumab improves qualify of life, but does not extend life. The model evaluated the cost–utility of omalizumab for patients with an inadequate response despite combining H1‑antihistamines (up to 4 times the licensed dose) with either H2‑antihistamines or LTRAs, or having combined all 3 drugs together, compared with 'no further pharmacological treatment'. The model adopted a 10‑year time horizon, with a cycle length of 4 weeks. The model's perspective was that of the NHS and personal social services. All future costs and benefits were discounted at a rate of 3.5%.
3.25 The model comprised 5 discrete health states based on the severity of the symptoms, as measured by 'urticaria activity score over 7 days' (UAS7). These states, and the corresponding scores, were:
severe urticaria (28–42)
moderate urticaria (16–27)
mild urticaria (7–15)
well‑controlled urticaria (1–6)
In addition, the model included health states for relapse and death. All modelled patients were in either the moderate or severe urticaria health state at baseline and were treated either with omalizumab 300 mg plus background medications, or only background medications. Patients could move from the baseline states to any of the 5 health states.
3.26 Patients in the omalizumab arm continued to get omalizumab for 4 cycles and were then assessed at 16 weeks to be classified as 'responders' (that is, patients whose disease had responded to treatment defined by the health states 'urticaria‑free' or 'well‑controlled urticaria', or defined by a UAS7 of 6 or less) or 'non‑responders' (that is, patients whose disease had not responded to treatment). 'Responders' had a further 8 weeks of omalizumab treatment. During weeks 16 to 24, 'responders' could only move between 'urticaria‑free' and 'well‑controlled' urticaria health states. 'Non‑responders' (patients in mild, moderate or severe urticaria states) stopped omalizumab after 16 weeks but remained on background medication and could move to any of the 5 states. The company explored a different definition of response in a scenario analysis, considering the mild urticaria health state as a response (UAS7 of less than 15). Patients in the comparator arm had background medication throughout the model. After 24 weeks (6 cycles) 'responders' could relapse, and all modelled patients could go into spontaneous remission or die.
3.27 The company modelled the effect of treatment with omalizumab expressed as the proportion of patients within each of the 5 health states in the omalizumab and comparator arms at a given time. The company used individual patient data from the GLACIAL trial to estimate the proportions, and the model included only patients who had moderate and severe urticaria at the start of the treatment. The model included data up to week 24 for 'responders' (determined at week 16), and included data only up to week 16 for 'non‑responders'. To replace missing data caused by loss to follow‑up, the company used the 'last observation carried forward' method in the base‑case analysis. In scenario analyses, the company used the 'baseline observation carried forward' method or used the observed data without substituting the missing data. The company provided the distribution of patients between health states at each time point for both omalizumab and comparator arms but, because the company labelled these results academic in confidence, they are not presented here.
3.28 In the original model, before consultation 'relapse' was defined as moderate or severe urticaria (UAS7 of 16 or more) after a previous response. Patients whose disease had relapsed remained in a 'relapse' health state for 1 cycle and then moved back to the baseline (moderate or severe urticaria) health states. The company assumed that all 'responders' (unless they had gone into spontaneous remission or died) relapsed by 16 cycles (64 weeks) in the base case. The company based this assumption on an observational study by Metz et al. (2014), a review of 51 patients with chronic urticaria treated with omalizumab at a single study centre in Germany, which included 20 patients with chronic spontaneous urticaria. According to Metz et al., in most patients the disease relapsed 4–8 weeks after stopping omalizumab and the longest observed period without reappearance of symptoms after omalizumab treatment was 16 months. The company also did a scenario analysis, which assumed that 'responders' could remain relapse‑free beyond 16 months.
3.29 Relapse rates in the model were based on data from the GLACIAL trial's 16‑week follow‑up period, which followed the 24‑week treatment period. The company estimated the proportion of patients who had a relapse after 24 weeks of omalizumab treatment at 28, 32, 36 and 40 weeks using patient‑level data stratified by health state (urticaria‑free, well‑controlled urticaria and mild urticaria). To estimate the probability of relapse after the treatment and follow‑up period (40 weeks), the company used a logarithmic curve fitted to the 4 data points (28, 32, 36 and 40 weeks). The company assumed that all patients relapsed by 64 weeks after the end of treatment (48 weeks beyond the end of the data provided by GLACIAL).
3.30 In the base case, the company assumed that all patients re‑treated with omalizumab would have a response (and therefore move to the urticaria‑free or well‑controlled health states) by the end of the 24‑week course. The company assumed that all patients being re‑treated had a response when first treated with omalizumab. In a scenario, the company assumed instead that some patients would not have a response when re‑treated with omalizumab, and that the proportion with no response when re‑treated would be the same as the proportion with no response when first treated.
3.31 'Spontaneous remission' meant that all the patient's symptoms resolved. Patients who had a spontaneous remission remained in the urticaria‑free health state (UAS7=0) for the remainder of the time horizon. The company applied a probability of spontaneous remission to all patients in both arms. The company stated that a patient could not experience (spontaneous) remission while being treated, but applied a cumulative remission probability (calculated from cycle 1 to the cycle in which treatment ends) at the end of treatment.
3.32 To model spontaneous remission in the base case, the company used data on remission rates from a prospective study of 5 years' duration in patients (n=228) with moderate to severe chronic spontaneous urticaria conducted in Italy (Nebiolo et al. 2009). The company used scenario analyses to explore the effect of using alternative remission rates from other studies (Beltrani et al. 2002, Toubi et al. 2004 and van der Valk et al. 2002). The company chose a log‑logistic distribution to fit the data from Nebiolo et al., as well as for data from Beltrani et al. For the Toubi et al. and van der Valk et al. studies, the company considered the log‑normal distribution to be the best fit.
3.33 The company used the term 'drop‑out' to refer to patients in the GLACIAL trial who had omalizumab but whose UAS7 data were missing at the end of treatment (week 24). In the company's model, drop‑outs did not mean patients were lost to follow‑up. To account for these missing observations in the modelled trial data, the company calculated 4‑week 'drop‑out' rates from the GLACIAL trial data for both arms, stratified according to the baseline health state of the model (moderate or severe urticaria). The company assumed that patients moved to a moderate urticaria health state if drop‑out occurred.
3.34 In the GLACIAL trial, patients could stop omalizumab for reasons other than it not improving symptoms; these other reasons included adverse events, disease progression, physician decision or patient choice. The company estimated the risk of stopping omalizumab from the proportion of patients who stopped the study drug (because of the above‑mentioned reasons) in the GLACIAL trial. The model allowed for different stopping rates during the first and later treatments; however, because there were no trial data on the probability of stopping associated with omalizumab re‑treatment, the company assumed the same probabilities for stopping for first and subsequent courses of omalizumab. After stopping omalizumab, patients remained on the background medications. Patients who stopped omalizumab because of adverse events, disease progression, physician decision or patient choice were not re‑treated with omalizumab in the model. The probabilities of them moving between health states were based on the placebo arm of the GLACIAL trial.
3.35 The adverse events included in the company's model were sinusitis, headache, arthralgia, injection site reactions and upper respiratory tract infection. The company stated that no meaningful differences in the rates of adverse events between omalizumab and placebo were reported in the trials.
3.36 The company did not assume in the model that chronic urticaria increases mortality or that omalizumab extends life. The company sourced all‑cause mortality data from the UK Office for National Statistics (2011) and calculated the mean mortality by age group and sex, assuming a 50:50 men to women ratio.
3.37 The company calculated pooled EQ‑5D scores from the GLACIAL, ASTERIA I and ASTERIA II trials to estimate the utility values in the model. For each of the 5 health states in the model, it used a mixed‑effect regression model to estimate the following utility values (rounded mean UAS7 in brackets): severe urticaria 0.712 (34.1); moderate urticaria 0.782 (21.9); mild urticaria 0.845 (11.4); well‑controlled urticaria 0.859 (3.1); and urticaria‑free 0.897 (0.0). Disutility values for the adverse events were sourced from published literature and were as follows: sinusitis (−0.0022); headache (−0.0297); arthralgia (−0.0402); upper respiratory tract infection (−0.0022); and injection site reaction (−0.0040).
3.38 The company incorporated 3 categories of resource use in the model that included treatment, health state and adverse event costs. The treatment costs for omalizumab included costs for: drug acquisition; administration (£14.21 per administration); and monitoring (£42.64 for the first 3 administrations and £21.32 for the fourth administration). Treatment costs also included the cost of background medications for both arms (H1‑antihistamines [£0.21 per day], LTRAs [£0.36 per day] and H2‑antihistamines [£0.33 per day]) based on unit costs of the medications from the British national formulary (BNF).
3.39 Health‑state costs comprised accident and emergency visits, outpatient attendance and laboratory tests. The costs for emergency and outpatient visits were from NHS reference costs 2012–13 (updated to 2014) and the laboratory tests from the National Institute for Health Research Industry Costing Template (2013). The number of accident and emergency visits, outpatient visits and laboratory tests were estimated from the ASSURE study, an unpublished, company‑sponsored, retrospective observational study designed to measure the burden of illness of chronic spontaneous urticaria. Costs associated with health states were reported as academic in confidence and therefore are not presented here.
3.40 The costs of treating adverse events were also incorporated in the model. The company took the unit cost of a GP appointment from the Personal Social Services Research Unit 2013 (updated to 2014) and the cost of an antibiotic (for sinusitis and upper respiratory tract infections) from the BNF price for a course of ampicillin. The company applied an additional cost of £97.80 for identifying a relapse, which is based on the mean cost of outpatient appointments across several specialities.
3.41 The company's deterministic base‑case result showed that, with the patient access scheme (implemented for NICE's technology appraisal guidance on omalizumab for treating severe persistent allergic asthma), omalizumab was associated with a total incremental cost of £7459 with an additional gain of 0.38 quality‑adjusted life years (QALYs). This resulted in an incremental cost‑effectiveness ratio (ICER) of £19,632 per QALY gained.
3.42 The ERG commented that the structure of the company's economic model was reasonable and consistent with the clinical pathway for urticaria. The ERG commented that the time horizon of 10 years was appropriate given that data from observational studies on the natural history of the disease suggests that, in most patients, the entire disease lasts less than 10 years. The ERG noted that the model structure did not permit comparison with other comparators such as ciclosporin.
3.43 The ERG noted that the company did not provide details on how it assured quality in the patient‑level data analysis. The ERG noted a minor difference in the proportions of patients with a UAS7 of 0 at week 12 in the omalizumab arm between the data used in the model and the published data. The ERG noted that correcting this would not substantially affect the results.
3.44 When estimating remission rates, the ERG acknowledged that the company had correctly extracted data from the text of the Nebiolo et al. (2009) study, but noted that the study reported different values between the text and the published Kaplan–Meier curves. The ERG commented that this meant the company's approach to extrapolating the log‑logistic function resulted in an extremely poor fit to the Kaplan–Meier curves in the Nebiolo et al. paper, overestimating remission up to around 24 months and underestimating remission over longer time periods. The ERG also calculated the median duration of chronic spontaneous urticaria from the company's base‑case log‑logistic function, noting that 20.8 years was implausibly high. The ERG commented that the company's extrapolated remission rates (22.73% at 1 year, 36.00% at 5 years and 42.65% at 10 years) did not represent the natural history of the disease. The clinical advice received by the ERG suggested a spontaneous remission of around 50% to 70% within 2 years and 70% to 90% within 10 years. The ERG extracted the data from the Kaplan–Meier curves published in Nebiolo et al. and, using exponential, Weibull and log‑logistic parametric functions for remission, estimated a median duration of disease as 6–7 years. The ERG conducted exploratory analyses using exponential and log‑logistic functions for spontaneous remission and noted that these increase the ICERs to £22,341 and £21,730 per QALY gained respectively.
3.45 For relapse, the ERG noted that the model could extrapolate the GLACIAL trial data using either a log‑normal distribution, as in the base case, or with a linear extrapolation. The ERG noted that using a linear extrapolation increased the company's base case from £19,632 per QALY gained to £23,065 per QALY gained.
3.46 The ERG was concerned with the company's approach to estimating probability of relapse specifically in patients whose disease had initially responded to omalizumab. Therefore, the ERG reconstructed the company's curve‑fitting exercise. The ERG considered that an exponential curve fitted the observed trial data better than a log‑normal extrapolation, and explored a scenario analysis using alternative probabilities of relapse on cost effectiveness. The ERG reported that using an exponential fit increased the ICER from £19,632 to £22,003 per QALY gained.
3.47 The ERG could not independently verify drop‑out rates used by the company in the model because the company provided only limited information in its submission. The ERG noted that, to model all‑cause mortality, the company assumed an equal proportion (50:50) of men and women in the modelled population, whereas in the GLACIAL trial population there were fewer men than women (30:70). The ERG did not anticipate that this had a substantial impact on the results. The ERG commented that the company collected utility estimates for the health states from a large sample of a directly‑relevant population, but noted that the utility decrements the company used for adverse events were sourced from populations not relevant for this appraisal. The ERG was satisfied with the resource use included in the model.
3.48 The ERG suggested that a more appropriate base case would include remission rates derived from an exponential fit to the Kaplan–Meier curve of Nebiolo et al. (2009) and relapse probabilities calculated from survival analyses using the exponential fit to relapse found in the GLACIAL trial. This scenario produces an ICER of £24,989 per QALY gained.
3.49 In response to consultation, the company revised its model and provided revised results from base‑case analyses. The new assumptions included:
A different definition of relapse for patients with severe urticaria at baseline, now defined as having a UAS7 of 16 or less instead of 6 or less, as in the original base case. The company did not change the definition of response in the patients with moderate urticaria at baseline, which remained a drop to a UAS7 of 6 or less. In practice, the new response criteria mean that patients' scores must drop at least 2 health states for their disease to be considered to have responded.
A new early stopping rule for the patients whose disease did not respond to omalizumab. Instead of 16 weeks (after 4 doses), as in the original base case, the company assumed that patients whose disease does not respond stop treatment at 8 weeks (after 2 doses).
Corrected data for spontaneous remission from Kaplan–Meier curves from the Nebiolo et al. (2009) paper.
Revised estimates of relapse rates from the follow‑up period of the GLACIAL trial accounting for some patients spontaneously remitting.
Linear extrapolation of relapse rates from the follow‑up period of the GLACIAL trial.
3.50 The company's revised deterministic base‑case result showed that, with the patient access scheme, omalizumab was associated with a total incremental cost of £7222 with an additional gain of 0.263 QALYs, which resulted in an ICER of £27,469 per QALY gained.
3.51 The company's revised probabilistic base‑case results, based on running the model with 1000 iterations, showed that the average incremental cost was £7191 and the average incremental QALY gain was 0.26, which resulted in an ICER of £27,707 (95% CI 27,548 to 27,886) per QALY gained. The probabilistic analysis indicated that there is a 0.2% and 80.7% probability of omalizumab being cost effective, at the maximum acceptable ICERs of £20,000 and £30,000 per QALY gained respectively.
3.52 The company presented deterministic sensitivity analyses varying various parameters of the model one at a time. To vary the modelled clinical effectiveness of omalizumab, the company presented 4 analyses that varied the proportion of patients in the urticaria‑free and well‑controlled urticaria health states separately, for each arm. For each analysis, the company calculated the percentage variation between the proportion of patients in the specific 'responder' health state (urticaria‑free or well‑controlled urticaria) at 24 weeks and the upper and lower limits of its 95% confidence interval. The company applied the same percentage variation at 4, 8, 12, 16, 20 and 24 weeks simultaneously in that health state and distributed the remaining proportions across the mild, moderate and severe health states, keeping the proportion in the other 'responder' state (urticaria‑free or well‑controlled urticaria) unchanged. Varying the proportion of patients in the urticaria‑free health state in the omalizumab arm had the most impact (compared with other health state or treatment combinations) and the ICER, with its upper and lower variation, ranged from £26,726 to £28,336 per QALY gained.
3.53 The company also provided deterministic sensitivity analyses by varying other parameters. The results showed that the revised base‑case ICER was most sensitive to change in the cumulative relapse rate in people free of urticaria, the acquisition cost of omalizumab, the cost of the severe urticaria health state, and the discount rates for outcomes and costs.
3.54 The company also provided a fully incremental analysis assessing the impact on cost effectiveness of different stopping rules for people whose disease does not respond after the first, second, third and fourth dose, and assuming no early stopping (that is, all patients had omalizumab for 6 doses irrespective of response). The company also provided data on the cumulative response seen in the GLACIAL trial (expressed as a proportion of all treated patients) by dose (see section 3.59). The results showed that more patients benefited from the first dose than from subsequent doses of omalizumab (0.225 QALYs gained for the first dose) and the incremental QALY benefits with subsequent doses were marginal and ranged between 0.038 (for the second dose) to −0.001 (for the sixth dose). The fourth dose dominated (more effective and less costly than) the sixth dose. The corresponding fully incremental ICERs ranged from £26,824 per QALY gained (for the first dose) to £32,493 per QALY gained (for the third dose).
3.55 The company responded to the Committee's request for a scenario analysis including waning of treatment effect during repeated courses of omalizumab by presenting scenarios in which fixed proportions of prior 'responders' did not respond on re‑treatment (varying from 1% to 10%). In these analyses, both incremental costs and QALYs reduced with increasing proportions of prior 'responders' not responding on re‑treatment. The proportionate reduction in QALYs was slightly greater than the proportionate reduction in costs, leading to a small increase in the ICER from £27,469 per QALY gained in the revised base case to £28,748 per QALY gained in the scenario in which 10% of prior 'responders' did not respond on re‑treatment.
3.56 As requested by the Committee, the company also provided separate results for patients with moderate and severe urticaria at baseline as opposed to patients with moderate and severe urticaria combined. The ICERs were £29,951 and £26,278 per QALY gained respectively.
3.57 In response to a request by the Committee for the company to provide a clear and quantified explanation for the difference in benefits seen in the GLACIAL trial and those estimated by the model, the company compared the trial results and the model predictions in terms of the proportion of patients with a UAS7 of 0 and a UAS7 of 6 or less at 40 weeks, corresponding to the end of the GLACIAL trial. To generate model outcomes comparable to the trial results, the company changed the following model settings:
24 weeks of treatment for all patients (no early stop for 'non‑responders')
imputing missing data using the baseline observation carried forward method (as done in the clinical trial analysis)
assuming no re‑treatment with omalizumab
assuming no death occurred.
3.58 For patients receiving omalizumab, the proportion of patients with a UAS7 of 0 and a UAS7 of 6 or less predicted by the revised model (12.3% and 19.5% respectively) was similar to that seen in the GLACIAL trial (12.3% and 19.8% respectively). However, for patients in the comparator arm, the model and trial results differed. The proportion of patients with a UAS7 of 0 and a UAS7 of 6 or less among patients having only background medication in the GLACIAL trial was 13.3% and 20.5% at 40 weeks. The corresponding values predicted by the model were 1.8% and 8.0%.
3.59 The company presented further analyses of the GLACIAL trial, exploring the response in patients with each subsequent dose of omalizumab. The company presented analyses separately for patients with moderate and severe urticaria at baseline. For moderate urticaria, the cumulative response after each of 6 doses was: 53.4%, 71.2%, 78.1%, 82.2%, 82.2% and 83.6%. For severe urticaria, using the revised definition of response (UAS7 of 16 or less), the cumulative responses were: 47.5%, 57.0%, 62.6%, 68.7%, 71.5%, and 72.6%.
3.60 The company also provided the number of courses (of 6 doses each) which a patient re‑treated with omalizumab would need over the entire time horizon of the model, assuming the patient's condition had previously responded to treatment. The revised model predicted that 'responders' would need 6.61 courses on average over a time horizon of 10 years. The company also provided the average number of treatment courses after adjusting for patients whose disease may spontaneously remit, who may stop omalizumab because of adverse effects, or who may die. The analysis showed that, on average, a 'responder' who continues to have omalizumab for the entire time horizon would receive 13.69 courses of omalizumab.
3.61 The ERG commented that the revised definition of response may still underestimate the proportion of patients having a clinically significant response. For example, to be considered a 'responder', a patient with severe disease and the lowest UAS7 (28) would need to improve by 12 points, whereas a patient with severe disease and the highest UAS7 (42) would need to improve by 26 points. The ERG also noted that, in the company's revised sensitivity analyses, the hazard ratios for remission, health state costs and utility values varied by arbitrary percentage points instead of varying within a clinically meaningful range as requested by the Committee.
3.62 The ERG commented that comparing the results from the GLACIAL trial with the model estimates (see section 3.58) indicated that the model did not perform well in predicting 40‑week outcomes. It noted that the clinical trial results suggested little difference in the proportion of patients with a UAS7 of 6 or less between omalizumab and placebo arms, whereas the model predicted a substantial benefit with omalizumab. The ERG noted that the company did not adequately explain the difference between the results from the trial and the model, further noting that it may reflect a serious flaw in the model.
3.63 Full details of all the evidence are available.