Mirabegron for treating symptoms of overactive bladder

Comparisons with placebo

3.10 Urinary frequency was measured as change from baseline in mean number of micturitions in 24 hours (measured by micturition diary). The results from the full set analyses of SCORPIO, ARIES and CAPRICORN all indicated statistically significant improvements in change from baseline to week 12 in the mean number of micturitions in 24 hours in the mirabegron groups compared with placebo (SCORPIO mean difference −0.60, 95% confidence interval [CI] −0.90 to −0.29; ARIES mean difference −0.61, 95% CI −0.98 to −0.24; CAPRICORN mean difference −0.42, 95% CI −0.76 to −0.08). In the DRAGON and 178-CL-048 trials there was no statistically significant difference between mirabegron and tolterodine tartrate. In the 178-CL-045 trial significant improvements were seen with mirabegron compared with placebo (mean difference −0.74, 95% CI −1.12 to −0.36). The results of the manufacturer's pooled analysis of 3 trials (ARIES, CAPRICORN and SCORPIO) showed a statistically significant difference between mirabegron 50 mg and placebo (adjusted mean difference −0.55, 95% CI −0.75 to −0.36, p<0.001).

3.11 Frequency of incontinence was measured as the mean number of incontinence episodes per 24 hours and assessed in the manufacturer's submission by the full analysis set – incontinence population. The results from the SCORPIO, ARIES and CAPRICORN trials all indicated statistically significant improvements in change from baseline to week 12 in mean number of incontinence episodes per 24 hours versus placebo (SCORPIO mean difference −0.41, 95% CI −0.72 to −0.09; ARIES mean difference −0.34, 95% CI −0.66 to −0.03; CAPRICORN mean difference −0.42, 95% CI −0.76 to −0.08). In the DRAGON and 178-CL-048 trials there was no statistically significant difference between mirabegron and tolterodine tartrate. In the 178-CL-045 trial significant improvements were seen with mirabegron compared with placebo (mean difference −0.40, 95% CI −0.67 to −0.13). The results of the manufacturer's pooled analysis for the 3 trials showed a statistically significant difference between mirabegron 50 mg and placebo (mean difference −0.40, 95% CI −0.58 to −0.21 p<0.001).

3.12 The manufacturer presented results for the secondary outcomes reported in the trials. These were mean volume voided per micturition, frequency of urge urinary incontinence, mean level of urgency, the number of urgency episodes (grade 3 or 4) per 24 hours, and nocturia episodes per 24 hours.

Subgroup analyses

3.13 The manufacturer performed pooled subgroup analyses (from SCORPIO, ARIES and CAPRICORN) for men versus women, for the 2 primary outcomes. Mirabegron 50 mg was numerically more effective in the female subgroup than the male subgroup for the pooled trial data for the change in the number of incontinence episodes and micturitions per 24 hours between baseline and final visit. The adjusted mean difference versus placebo for incontinence episodes per 24 hours (incontinence full analysis set) in the male mirabegron 50 mg treatment group was −0.07 (95% CI −0.50 to 0.36), which was not statistically significant. The change in the adjusted mean difference versus placebo for incontinence episodes per 24 hours (incontinence full analysis set) in the female mirabegron 50 mg treatment group was −0.47 (95% CI −0.67 to −0.26), which was statistically significant. The adjusted mean difference versus placebo in the mean number of micturitions per 24 hours was −0.37 (95% CI −0.74 to −0.01) in the male mirabegron 50 mg treatment group, which was statistically significant. The adjusted mean difference versus placebo in the mean number of micturitions per 24 hours in the female mirabegron 50 mg treatment group was −0.62 (95% CI −0.85 to −0.39), which was statistically significant. Tests for treatment by sex interactions were non-significant for both outcomes (p value=0.22 for the change in the mean number of incontinence episodes per 24 hours; p=0.16 for change in the mean number of micturitions per 24 hours) indicating that there was no differential treatment effect between men and women.

3.14 The manufacturer also performed a pooled subgroup analysis using data from SCORPIO, CAPRICORN and ARIES comparing previously treated and treatment-naive groups. Mirabegron 50 mg was effective in both populations for the 2 primary outcomes. The change in the mean number of incontinence episodes per 24 hours between baseline and final visit (incontinence full analysis set) in the previously treated mirabegron 50 mg treatment group was −1.49 (95% CI −1.66 to −1.32) and the adjusted mean difference versus placebo was −0.57 (95% CI −0.81 to −0.33). The change in the mean number of incontinence episodes per 24 hours between baseline and final visit (incontinence full analysis set) in the treatment-naive mirabegron 50 mg treatment group was −1.50 (95% CI −1.71 to −1.29) and the adjusted mean difference versus placebo was −0.15 (95% CI −0.44 to 0.14). The change in the mean number of micturitions per 24 hours between baseline and final visit (full analysis set) in the previously treated mirabegron 50 mg treatment group was −1.67 (95% CI −1.86 to −1.48) and the adjusted mean difference versus placebo was −0.74 (95% CI −1.01 to −0.47). The change in the mean number of micturitions per 24 hours between baseline and final visit (full analysis set) in the treatment-naive mirabegron 50 mg treatment group was −1.84 (95% CI −2.04 to −1.64) and the adjusted mean difference versus placebo was −0.33 (95% CI −0.62 to −0.05). Mirabegron 50 mg was effective in reducing the mean number of incontinence episodes and micturitions per 24 hours from baseline to final visit for both previously treated and treatment-naive patients. Tests for interaction indicated no treatment effect for either outcome (p=0.095 for the change in the mean number of incontinence episodes and p=0.10 for the mean number of micturitions).

Safety outcomes

3.15 Data on adverse events presented in the manufacturer's submission were largely derived from the 52-week study TAURUS (mirabegron versus tolterodine tartrate). Additionally, information was available from the SCORPIO, ARIES and CAPRICORN trials. Treatment-emergent adverse events were adverse events that occurred after treatment, while treatment-related adverse events were directly linked to treatment, although the criteria used by the manufacturer for establishing a direct link with treatment were unclear. The incidence of treatment-related adverse events was similar in the 2 groups, with 26.2% in the mirabegron 50 mg group and 27.6% in the tolterodine tartrate group in TAURUS. In TAURUS the overall incidence of treatment-emergent adverse events was similar across the mirabegron 50 mg (59.7%) and tolterodine tartrate (62.6%) groups. Most treatment-emergent adverse events were mild or moderate in all treatment groups. The overall incidence of treatment-emergent severe adverse events was 5.2% in the mirabegron 50 mg group and 5.4% in the tolterodine tartrate group in TAURUS (safety analysis set). For treatment-emergent adverse events leading to permanent discontinuation of the study drug the rates were 5.9% in the mirabegron 50 mg group and 5.7% in the tolterodine tartrate group in TAURUS (safety analysis set). In SCORPIO, ARIES and CAPRICORN the incidence of treatment-emergent adverse events was similar across all the treatment groups (safety analysis set population).

3.16 The adverse events that were designated of interest because of their association with antimuscarinic drugs were dry mouth and constipation. The rates of constipation by treatment group for TAURUS were 2.8% and 2.7% for mirabegron and tolterodine tartrate respectively. For dry mouth, the rates were 2.8% and 8.6% for mirabegron and tolterodine tartrate respectively. In SCORPIO, the rates of dry mouth by treatment group were 1.8% and 9.5% for mirabegron and tolterodine tartrate respectively. The rates were not given for SCORPIO for constipation, or for either adverse event in ARIES and CAPRICORN.

Mixed treatment comparison

3.17 Based on the 40 studies identified in the manufacturer's literature review, the manufacturer conducted a Bayesian mixed treatment comparison (MTC) to estimate the relative efficacy and safety of mirabegron compared with oxybutynin hydrochloride (5 mg and 10 mg), fesoterodine fumarate (4 mg and 8 mg), trospium chloride (60 mg), tolterodine tartrate (4 mg), solifenacin succinate (5 mg and 10 mg) and placebo. This was done through a network analysis, using direct comparisons when available, or indirect comparison via placebo when available, or through another comparator if necessary. Oxybutynin hydrochloride and tolterodine tartrate have both MR and immediate-release formulations available, which were examined together for efficacy, but separately for safety, by the manufacturer. Analyses were conducted for micturition, urgency, urinary incontinence, dry mouth, constipation, blurred vision and frequency of incontinence for the general OAB population, but the manufacturer deemed that there were insufficient data available to conduct MTC analyses for the subgroups (including sex) identified in the final NICE scope. For each population, fixed and random effects models were used with a non-informative prior distribution. Quality of fit was assessed through the Bayesian deviance information criterion, with the model with the lowest deviance information criterion selected.

3.18 The results of the manufacturer's MTC for the outcome of number of micturitions per 24 hours indicated that the effect of mirabegron 50 mg did not differ significantly from any other treatments, except for solifenacin succinate 10 mg, which was more effective than mirabegron (odds ratio [OR] −0.583 to 95% credible interval [CrI] −0.8324 to −0.3326) and tolterodine tartrate 4 mg, which was less effective than mirabegron (OR 0.157, 95% CrI −0.0002 to 0.3154). For the outcome of number of incontinence episodes per 24 hours, the results indicate that there was no statistically significant difference between mirabegron and any of the comparators.

3.19 The manufacturer's MTC results for the adverse effects of dry mouth, constipation and blurred vision indicated that the mirabegron group had probabilities similar to the placebo group for all effects. All antimuscarinic drugs had a significantly higher risk of dry mouth compared with mirabegron 50 mg. The odds ratios for constipation for antimuscarinic drugs compared with mirabegron 50 mg were not statistically significantly different, except for solifenacin succinate 5 mg (OR 2.501, 95% CrI 1.41 to 4.127) and 10 mg (OR 4.369, 95% CrI 2.54 to 7.071), fesoterodine fumarate 8 mg (OR 1.926, 95% CrI 1.142 to 3.059) and trospium chloride 60 mg (OR 7.604, 95% CrI 2.08 to 22.59). These were associated with higher risks of constipation compared with mirabegron 50 mg. No differences in risk between mirabegron 50 mg and other treatments were found, with wide credible intervals around the odds ratios for blurred vision due to the rarity of this adverse event.

Quality of life

3.20 Health-related quality of life and treatment satisfaction were assessed using generic scales (EQ-5D, EQ-5D VAS, TS-VAS and WPAI:SHP) and disease specific scales (OABq and PPBC) in SCORPIO, ARIES and CAPRICORN. Data from SCORPIO, ARIES and CAPRICORN were pooled for a post-hoc analysis of EQ-5D results in the modified intention-to-treat set and no data from individual trials were presented in the manufacturer's submission. Adjusting for baseline confounders (the manufacturer's submission did not provide details of which confounders), mirabegron 50 mg had greater mean change from baseline to 12-week utility scores than tolterodine tartrate MR 4 mg in the pooled analysis of ARIES, CAPRICORN and SCORPIO (0.045 and 0.026 respectively, p≤0.05). The change between baseline and 12-week utility scores was not statistically significantly different between mirabegron 50 mg and placebo (placebo utility change 0.038, p=0.30 for difference).

3.21 SCORPIO, ARIES and CAPRICORN all showed that there was a greater improvement in quality of life in the mirabegron group than in the placebo group as measured by the OAB-q, which reached statistical significance in SCORPIO and ARIES (SCORPIO estimated mean difference 2.3, 95% CI 0.2 to 4.5; ARIES mean difference 4.1, 95% CI 1.6 to 6.6; CAPRICORN mean difference 1.2, 95% CI −1.0 to 3.4). The mirabegron, placebo and tolterodine tartrate groups achieved a clinically meaningful improvement of 10 points above baseline.

ERG critique of clinical-effectiveness evidence

3.22 The ERG noted key strengths and weaknesses in the evidence submitted by the manufacturer. The ERG considered ARIES, CAPRICORN and SCORPIO to be well-designed trials, and that the results for the effectiveness of mirabegron were consistent across the trials. The ERG's clinical expert noted that it was recommended in NHS clinical practice that pharmaceutical treatments for OAB are assessed after 3 months. The ERG therefore considered the duration of treatment and follow-up of the trials included in the manufacturer's submission to be sufficient to assess the efficacy and safety of treatment with mirabegron.

3.23 The ERG questioned the omission of the data from DRAGON, 178-CL-045 and 178-CL-048, as well as TAURUS, from the primary analyses. The ERG also noted that no direct comparison between mirabegron and tolterodine tartrate was drawn, even though SCORPIO, DRAGON, TAURUS and 178-CL-048 used tolterodine tartrate as an active control. The ERG acknowledged that these trials were not powered to evaluate the superiority or non-inferiority of mirabegron versus tolterodine tartrate, but that exclusion of data from the tolterodine tartrate group in SCORPIO limited the evidence available that was relevant for the decision problem. The ERG requested more detailed information on these trials during the clarification phase, and included it in its report. Although one of the manufacturer's exclusion criteria for 178-CL-045 and 178-CL-048 was that they were exclusively conducted in Japan, the ERG's clinical expert stated that ethnicity is unlikely to influence the development of symptoms of OAB and therefore trials from any country and any population involving patients with OAB were likely to be representative of patients with OAB in England and Wales.

3.24 The ERG considered the use of the full analysis set population appropriate, as was using the last-observation-carried-forward methodology for missing data (in SCORPIO, ARIES, CAPRICORN and TAURUS). The ERG noted that the intention-to-treat population was not reported across all trials for comparators. The ERG considered the multiplicity adjustments used by the manufacturer in SCORPIO, ARIES and CAPRICORN to be reasonable, in order to account for the multiple outcomes and the resulting increased probability of type I errors. The ERG noted that no statistical assessment of heterogeneity was performed on the pooled analysis.

3.25 For the comparison of mirabegron with tolterodine tartrate, the ERG performed an additional meta-analysis on the data from the 3 randomised controlled trials that had an active control of tolterodine tartrate (SCORPIO, DRAGON and 178-CL-048) for the outcomes of relevance to the cost-effectiveness analysis, but did not include TAURUS in this meta-analysis because the patients were mainly recruited from SCORPIO and ARIES, and this could have led to an 'enriched' dataset that was biased. The results showed that treatment with mirabegron 50 mg led to statistically significantly fewer micturitions per 24 hours compared with treatment with tolterodine tartrate MR 4 mg (mean difference −0.27, 95% CI −0.48 to −0.06, p=0.01). Conversely, data from TAURUS favoured tolterodine tartrate MR 4 mg, although the difference was not statistically significant (mean difference 0.12, 95% CI −0.11 to 0.35, p value not given). For incontinence episodes per 24 hours, the meta-analysis of the 3 trials showed that treatment with mirabegron 50 mg was statistically significantly more effective than treatment with tolterodine tartrate MR 4 mg (mean difference −0.21, 95% CI −0.41 to −0.01, p=0.04). However, the data from TAURUS showed that mirabegron was associated with statistically significantly more episodes per 24 hours (mean difference 0.25, 95% CI 0.01 to 0.49, p=0.04).

3.26 The ERG noted concerns with the inclusion and exclusion criteria for the manufacturer's MTC, and considered that the results should be interpreted cautiously. The ERG was concerned about potential clinical and methodological heterogeneity in the included studies, the inconsistency identified in 1 or more treatment comparisons for multiple outcomes, and the number of iterations used for sampling the posterior distributions (which may be an indicator of poor mixing of data within the model). The ERG noted that, for the random effects models used, there were no estimates of the between pairwise comparisons of heterogeneity given by the manufacturer.

3.27 The ERG re-ran the MTC with different inclusion and exclusion criteria on the same 40 studies identified by the manufacturer to perform an analysis on a more homogeneous dataset. The ERG excluded trials that included patients other than those with OAB; that were carried out in a single sex population; that reported on outcomes available at a time point other than 12 weeks; or that were deemed to be of poor methodological quality based on the manufacturer's summary (less than 4 'yes' responses in the first 4 categories assessed). The ERG included only outcomes, treatment formulations and doses used in the economic model supplied by the manufacturer. This decreased the number of studies to 22 and led to a greater degree of concordance and consequently more reliable results.

3.28 For the outcome of micturition episodes per 24 hours, the ERG found no significant difference between mirabegron 50 mg and any of the other active treatments assessed in their MTC. The manufacturer's analyses indicated that the only significant difference was that solifenacin succinate 10 mg was statistically significantly more effective than mirabegron at reducing the number of micturition episodes (mean difference −0.583, 95% CrI −0.832 to −0.333).

3.29 For the outcome of incontinence episodes per 24 hours, the results of the ERG's MTC concurred with the manufacturer's MTC results, with the exception that mirabegron 50 mg was statistically significantly less effective in reducing the frequency of incontinence episodes than solifenacin succinate 5 and 10 mg (solifenacin succinate 5 mg mean difference −0.386, 95% CrI −0.717 to −0.055, solifenacin succinate 10 mg mean difference −0.380, 95% CrI −0.694 to −0.067). The manufacturer's analyses also indicated that solifenacin succinate was numerically more effective than mirabegron, but this result was not statistically significant.

3.30 The ERG also analysed adverse events in its revised MTC analysis. The ERG found that mirabegron was statistically significantly less likely to be associated with constipation than fesoterodine fumarate 8 mg (OR 2.12, 95% CI 1.13 to 3.64), solifenacin succinate 5 mg (OR 2.11, 95% CI 1.16 to 3.59) and 10 mg (OR 4.52, 95% CI 2.60 to 7.47), and trospium chloride MR 60 mg (OR 7.63, 95% CI 2.12 to 22.95). Mirabegron was also found to be associated with a statistically significantly lower risk of dry mouth compared with all other antimuscarinic drugs assessed. Only oxybutynin hydrochloride 15 mg had a statistically significantly higher rate of discontinuation than mirabegron.

Model overview

3.32 In the manufacturer's model, simulated patients are either allocated mirabegron or treatment A (a single scope comparator). At the end of each monthly cycle patients can remain on the same medication, switch medication or stop all medication. The next line of therapy is considered to have cost, efficacy and safety equivalent to solifenacin succinate 5 mg. Once 2 drugs have failed, or 1 drug has failed followed by a cycle off any drug, botulinum toxin is available as a treatment option (see section 3.36).

3.33 The manufacturer's model simultaneously simulated 2 key symptoms: frequency of micturition and incontinence. Each symptom was categorised into 5 severity levels, resulting in 25 possible combinations of micturition and incontinence. At the end of each month, a person's symptoms could stay the same, improve or deteriorate. The transitions between symptom severity states were determined by multinomial logistic regression using patient-level data from the SCORPIO trial and defined as a function of treatment, symptom severity in previous month, age and sex. For each symptom (micturition and incontinence), 3 transmission probability matrices were produced: i) transition between baseline and month 1, ii) transition between month 1 and month 2, and iii) transition between month 2 and month 3. For patients remaining on treatment beyond 3 months, the third matrix was applied until discontinuation. To develop transition matrices for antimuscarinic drugs not studied in the mirabegron clinical study programme, a calibration approach was adopted to determine the beta coefficients for use in the multinomial logistic regression model.

3.34 The only adverse events incorporated into the model were dry mouth and constipation. The manufacturer stated that expert opinion suggested that these 2 were the most likely to occur with antimuscarinic drugs. Monthly probabilities of adverse events were obtained from SCORPIO for mirabegron and tolterodine tartrate MR 4 mg and from the MTC for the other antimuscarinic drugs. It was assumed that people who were on no treatment experienced no adverse events.

3.35 Discontinuation of treatment was incorporated into the model as a combination of background persistence with OAB medication and the occurrence of adverse events. The background persistence rate for the base case was taken from a published study (Wagg 2012) and a sensitivity analysis was performed on the estimate. The discontinuation rate used in the base-case model was 72%, and was based on that observed with tolterodine tartrate MR 4 mg. The manufacturer estimated that 54.7% of patients without an adverse event would discontinue treatment by 12 months. Discontinuation due to adverse events was based on expert opinion and set at 90%. The manufacturer did not identify any literature on treatment re-initiation rates after treatment discontinuation. The manufacturer assumed that 50% of patients who had stopped treatment with mirabegron or tolterodine tartrate (in the base-case model) would restart treatment annually (5.6% per month) without immediately switching to another drug. Of these, a third would go back to their previous drug, a third would receive next line A, and a third would receive next line B.

3.36 No data were available to the manufacturer on the probability of moving to botulinum toxin. The model assumed that 1% of people who had discontinued 2 therapies or discontinued 1 and gone to no treatment would receive botulinum toxin. The probability of success of botulinum toxin was taken from a previously published cost-effectiveness analysis. The model assumed that people in whom botulinum toxin was successful moved to the lowest level of symptoms for micturition and incontinence. For those in whom botulinum toxin failed, pre-botulinum toxin symptom severity levels were assumed.

3.37 Utility values assigned to the different symptom severities used in the base case were derived from EQ-5D index scores, based on health-related quality-of-life data, which were collected in SCORPIO. A linear regression model was used to estimate utilities for each of the 25 combinations of symptoms, with adjustment for age, sex and country (as random effect). No interaction between micturition frequency and incontinence was assumed, despite borderline statistical significance (p=0.0566). The regression model was based on all treatment arms of SCORPIO, and predicted utility values ranging from 0.85 (for people with micturition frequency and incontinence severity levels of 1) to 0.73 (for people with micturition frequency and incontinence severity levels of 5). Patients experiencing an adverse event had an associated disutility of 0.0357 (if they remained on treatment). Additionally, a sensitivity analysis was performed estimating utilities based on OAB-q and EQ-5D collected in the SCORPIO, ARIES and CAPRICORN trials.

3.38 Costs included in the model were the acquisition prices of the drugs, cost of treatment with botulinum toxin, GP visits, specialist visits and cost of incontinence pads. It was assumed that GP consultations would be 1 visit at the start and then at every treatment switch, and that specialist consultations would occur at every switch and on average 1.5 specialist consultations at the start of treatment. Incontinence pad costs were included in the model and the number of pads was determined by the level of incontinence obtained from SCORPIO. There were no costs associated with managing adverse events except specialist referral in case of a switch in treatment.

3.39 The manufacturer performed deterministic and probabilistic sensitivity analyses on assumptions and parameter estimates in the model. It also performed subgroup analyses of the base case for men and women, and treatment-naive and previous treatment groups.

Model results

3.40 The manufacturer's base-case result comparing mirabegron with tolterodine tartrate MR 4 mg based on the SCORPIO trial gave an incremental cost-effectiveness ratio (ICER) of £4386 per QALY gained. The manufacturer's probabilistic ICER was £4886 per QALY gained. The results of the secondary analysis using the effectiveness results from the MTC gave ICERs of £340 per QALY gained for solifenacin succinate 10 mg versus mirabegron, £3607 per QALY gained for fesoterodine fumarate versus mirabegron, £3715 per QALY gained for tolterodine tartrate MR 4 mg versus mirabegron, £3878 per QALY gained for oxybutynin hydrochloride MR 10 mg versus mirabegron, £8881 per QALY gained for trospium chloride MR 60 mg versus mirabegron, £12,493 per QALY gained for solifenacin succinate 5 mg versus mirabegron, and £14,234 per QALY gained for oxybutynin hydrochloride 10 mg versus mirabegron.

3.41 The manufacturer performed a fully incremental analysis assuming mirabegron persistence is equivalent to solifenacin succinate. Treatments are dominated when an alternative treatment is less expensive and has greater QALY gains, and a treatment is extendedly dominated if its ICER is higher than the next, more effective, alternative. When the dominated and extendedly dominated options are excluded, the incremental analysis shows that solifenacin succinate 5 mg has an ICER of £10,814 per QALY gained compared with oxybutynin hydrochloride 10 mg, and mirabegron has an ICER of £12,493 per QALY gained when compared with solifenacin succinate 5 mg.

3.42 Results from the manufacturer's base-case 1-way deterministic sensitivity analyses indicated that the primary base-case results were relatively insensitive to variation in parameter estimates, the transition probabilities between symptom severities having the highest impact. Uncertainty was explored by sensitivity analyses on the model time horizon, impact of OAB-related comorbidities (depression, fractures, urinary tract infections and skin infections) and the use of disease-specific health-related quality-of-life measures. The results of the sensitivity analyses in the primary and secondary base cases were similar. The effects of 1-way sensitivity analyses on the primary base-case parameters that the model results were most sensitive to were: the transition probabilities between symptom levels for incontinence and micturition for tolterodine tartrate; incontinence severity distribution across levels at baseline; the monthly probability of having botulinum toxin injections; and the transition probability between symptom levels for micturition for mirabegron.

3.43 The manufacturer submitted primary base-case ICERs (mirabegron versus tolterodine tartrate MR 4 mg) for subgroups by treatment status and by sex. The primary base-case ICERs were £3836 per QALY gained for the previously treated patient subgroup and £5315 per QALY gained for the treatment-naive subgroup. The primary base-case ICERs for the subgroups by sex were £38,708 per QALY gained for the male subgroup and £3091 per QALY gained for the female subgroup. The ICER fell to £2266 per QALY gained in the female subgroup but rose to £65,968 per QALY gained in the male subgroup if utilities derived from the OAB-q were used, rather than those from the EQ-5D.

ERG critique of the manufacturer's model

3.44 The ERG commented that it thought the manufacturer's model was well constructed, transparent and accurate. The ERG noted that the manufacturer's primary base-case cost-effectiveness analysis was generally robust. The ERG considered that the use of deterministic rather than probabilistic results was appropriate, given the high level of consistency between the deterministic and probabilistic results. The ERG noted that the manufacturer's 1-way sensitivity analyses were thorough, and that the primary base-case result was relatively robust and insensitive to individual parameter estimate changes.

3.45 The ERG identified several areas of inaccuracy or uncertainty. These included:

3.46 The ERG noted that the manufacturer considered the Pearson's correlation coefficient to assess any potential relationship between the frequency of micturition and incontinence. There was a small positive correlation (r=0.19094, p<0.0001) detected. Within the model, the manufacturer assumed that the frequency of micturition was independent of the frequency of incontinence, which the ERG considers may have compromised the accuracy of the model in the respect of the distribution of patients across different symptom levels. The ERG also noted that the correlation between these outcomes is unlikely to be affected by treatment and therefore may not result in model bias either towards or against mirabegron. The ERG accepted that dry mouth and constipation were likely to be the main drivers of adverse event-related discontinuation, and therefore considered that it was unlikely that exclusion of other adverse events would bias the model either towards or against mirabegron. The ERG noted that most of the parameters were based on clinical opinion that was estimated through open discussion, rather than generated through the use of elicitation techniques, which would lead to greater parameter uncertainty.

3.47 The ERG's clinical expert indicated that a 90% discontinuation rate with adverse events would be likely to be too high, but acknowledged that the manufacturer included a sensitivity analysis that assumed a 50% adverse event-related discontinuation rate, which had a limited effect on the ICER (£4585 per QALY gained compared with the base-case ICER of £4386 per QALY gained). The ERG noted that there were issues with the disaggregated discontinuation rate (rates for adverse event-related discontinuation, and for discontinuation due to other causes). The probability of other-cause discontinuation was assumed to be treatment specific and, in the manufacturer's base cases (primary and secondary), was derived from the published literature to exclude adverse event-related discontinuation. The ERG considered the application of the adverse event-related discontinuation rate inappropriate to other-cause discontinuation rates.

3.48 The ERG agreed with the manufacturer that the use of regression analysis to estimate transition probabilities in the model was appropriate, so that the potentially confounding factors of age and sex could be taken account of, and could minimise the risk of overestimating the utility benefit of mirabegron. The ERG also noted that the manufacturer stated that the interaction between the numbers of micturition and incontinence episodes was tested (Wald test: p=0.0566) and found not to be significant. The ERG commented that covariate selection was neither systematic nor rigorous and that expert clinical advice was not sought in the formulation of the linear regression models, but that in comparison with published literature the ERG considered the manufacturer's utility values generated by the regression model reasonable. The ERG also considered the selection of the covariates in the manufacturer's repeated regression model reasonable. The ERG noted that utility values from SCORPIO were comparable to those in the published literature, and considered the use of trial-based data to be appropriate. The ERG thought that the SCORPIO utility data would be likely to be biased against the more effective treatment, as would the use of EQ-5D rather than OAB-q health-related quality-of-life data.

3.49 The ERG commented that the subgroup analyses indicated that the manufacturer's primary base-case ICER was robust with respect to the subgroups considered, except for the male subgroup. The ERG noted that the proportion of men recruited for the trial was lower, therefore reducing statistical power to detect differences in efficacy. The manufacturer and the ERG also noted that male patients displayed lower baseline severity levels of OAB, and experienced a higher placebo response.

3.50 The ERG considered that the manufacturer's assumption of long-term use, based on the TAURUS study, was reasonable. The ERG noted that the relative difference between mirabegron and tolterodine tartrate in adverse event rates was higher in SCORPIO than in TAURUS, which was longer term. The primary base-case ICER decreased by £72 (from £4386 to £4314 per QALY gained) when using adverse event data from TAURUS rather than SCORPIO.

3.51 The ERG's additional sensitivity analyses (described in 3.47 to 3.50) cumulatively increased the primary base-case ICER from £4386 to £5272 per QALY gained. However, the impact on the secondary base case had greater effects. The secondary fully incremental analysis of the ERG's cumulative sensitivity analyses included the assumptions that the persistence rate with mirabegron was 28% and the probability of re-initiating original therapy was set to 0, as well as using the adverse event rates from TAURUS and NHS reference costs for botulinum toxin injections and outpatient specialist visits. The results of the analyses were largely consistent with the manufacturer's analyses (the ICER for trospium chloride MR 60 mg compared with oxybutynin hydrochloride increased by £586, and the ICER for solifenacin succinate 5 mg compared with trospium chloride MR 60 mg increased by £899). The largest change was for the ICER of mirabegron 50 mg compared with solifenacin succinate 5 mg, which changed from £12,493 to £32,712 per QALY gained, an increase of £20,219. The impact of the sensitivity analyses on the ICERs for mirabegron versus solifenacin succinate 10 mg ranged from £573 to mirabegron being dominated by solifenacin succinate 10 mg.

3.52 The ERG was unable to quantify the impact of using alternative assumptions or parameters for all the uncertainties it identified, including the difference between the manufacturer's MTC (with no statistically significant difference between mirabegron and solifenacin succinate in reducing incontinence episodes) and the ERG's MTC (in which solifenacin succinate 5 mg was statistically significantly more effective at reducing incontinence episodes than mirabegron). The ERG considered that its ICER was likely to be conservative, and that using the ERG's MTC data was likely to result in a higher ICER than £32,712 per QALY gained for mirabegron versus solifenacin succinate 5 mg.

3.53 Full details of all the evidence are in the manufacturer's submission and the ERG report.