Summary of the Manufacturer’s Pharmacoeconomic Resubmission

The manufacturer submitted a new cost-utility analysis as part of its resubmission to CADTH, which compared onabotulinumtoxinA (Ona A) injections with best supportive care (BSC; assumed by the manufacturer as an unspecified group of acute-pain medications) for prophylaxis of headaches in adults with chronic migraine (CM). CM is defined as at least 15 days per month with headache lasting at least four hours a day or longer.⁶ A scenario analysis comparing Ona A with topiramate was also presented by the manufacturer to account for a change in clinical practice since the original submission. Both analyses assumed concurrent use of acute-headache medications, including triptans, based on headache frequency. The cost-utility analysis was conducted as a Markov cohort-state transition model from the perspective of a Canadian publicly funded health care payer and utilized a three-year time horizon with 12-week cycles.

The manufacturer modelled six headache frequency health states (0 to 3, 4 to 9, 10 to 14, 15 to 19, 20 to 23, and 24 to 28 headache days per month [HDPM]) for each treatment status (i.e., on treatment or discontinued treatment due to treatment failure) and a death state (13 total health states). Six of the health states reflected HDPM observed in CM (i.e., 15 to 19, 20 to 23, and 24 to 28 health states) and six others reflected HDPM observed in episodic migraine (EM) (i.e., 0 to 3, 4 to 9, and 10 to 14 health states). The baseline characteristics of the modelled population, treatment efficacy, and discontinuation differed by treatment and were based on pooled data from two trials within the manufacturer’s PREEMPT trial program, which compared Ona A injections with placebo injections. The placebo injection arm was used to inform BSC in the model. Patients enter the model in one of the six health states (predominantly in one of the three CM health states) and could transition to other headache frequency health states every 12 weeks. The transition probabilities were informed by individual patient data (IPD) from the PREEMPT trials populations and subgroups based on treatment response (using the “stopping rules” defined later) and prior treatment failure.⁶ The model was structured so that patients with the same headache frequency could be assigned two different health-state utilities depending on whether they received Ona A or BSC or discontinued treatment. These utility values were mapped from Migraine-Specific Quality of Life Questionnaire (MSQ) results from the PREEMPT trials.

Patients discontinuing Ona A or BSC were assumed to be treated with BSC (patients treated with BSC effectively did not discontinue). A “stopping rule” based on treatment response was applied such that those who do not experience a reduction of at least 50% in headache frequency after two cycles of treatment (24 weeks) discontinued treatment (red arrows in Figure 1), reflecting current clinical guideline⁷ and reimbursement criteria.⁵ The manufacturer also incorporated scenarios with other “stopping rules”: less than 30% reduction in headache frequency within the first 24 weeks, and patients who achieved zero to three HDPM. In the latter scenario, patients who achieved zero to three HDPM were also assumed to discontinue treatment independently of the < 50% HDPM reduction treatment-discontinuation rule. These patients were assumed to restart treatment when headache frequency increased to at least 15 HDPM again for more than 12 weeks (i.e., one cycle).

Patients could also transition to the death state based on the mortality risk of the general Canadian population.

Health care resource utilization (i.e., physician consultation, neurologist consultation, emergency department [ED] visit, hospitalization, and triptan treatment) was assumed to differ by whether patients had EM (fewer than 14 headaches per month) or CM (at least 15 headaches per month), based on the rates observed in the International Burden of Migraine Study (IBMS).⁸ Treatment-emergent adverse events that occurred in more than 2% of patients in the PREEMPT trials were modelled (i.e., neck pain, muscular weakness, musculoskeletal stiffness, myalgia, and eyelid ptosis). These adverse events were not assumed to affect health utilities in the base-case analysis, but half of the adverse events were assumed to result in a physician visit. Direct medical costs were included in the model. Drug acquisition, administration, and monitoring costs for the comparator treatments were included. Generally, unit costs from Ontario Case Costing Initiative, Ontario schedule of benefits, and Canadian Institute for Health Information were incorporated. The cost of Ona A administration was based on the Alberta Health Care Insurance Plan, and the cost of triptan treatment was based on IQVIA PharmaStat private claims data.⁶

The manufacturer’s resubmission incorporated revisions to account for several limitations identified in the original submission of Ona A for CM. Additional “discontinue treatment” health states in the new pharmacoeconomic model allowed the resubmission to be more transparent about patients who discontinue treatment. The manufacturer’s scenario analyses of longer time horizons, < 30% headache frequency reduction stopping rule, and discontinuation of some patients with improved headache symptoms were aimed at addressing concerns regarding these aspects identified in CADTH’s appraisal of the previous submission.

Manufacturer’s Base Case

The base case results are presented in Table 2. Compared with BSC, Ona A accrued 0.10 incremental QALYs at an additional cost of $3,430, with an incremental cost-utility ratio (ICUR) $34,407 per additional QALY. The manufacturer reported that Ona A had a 64% probability of being cost-effective at a willingness-to-pay (WTP) threshold of $50,000 per QALY, and a 77% probability of being cost-effective at a WTP threshold of $100,000 per QALY.

Table 2

Summary of Results of the Manufacturer’s Base Case.

Summary of Manufacturer’s Key Scenario Analysis

As part of the resubmission, the manufacturer presented new evidence, in the form of a scenario analysis in which Ona A was compared with topiramate. This analysis was based on inputs derived from the FORWARD trial, a 36-week randomized open-label trial that compared Ona A (155 Allergan units) and topiramate (25 mg daily, titrated up to 100 mg daily). Transition probabilities for the Ona A and topiramate arm of the scenario analysis were derived from IPD from the corresponding arms of the FORWARD trial. As the trial did not have a BSC arm, patients who discontinued treatment in the scenario analysis were assumed to follow the same transition probabilities of the patients who discontinued treatment in the base-case analysis, which were informed by IPD from the placebo injection arm of the PREEMPT trials. The health utilities in the model were treatment-independent and derived from the IBMS.⁶ Treatment-emergent adverse events that occurred in more than 2% of patients in the FORWARD trial were modelled. As in the manufacturer’s base case, these events were not assumed to affect health utilities and half of the events were assumed to result in a physician visit.⁶ The manufacturer reported that compared with topiramate, Ona A was associated with an ICUR of $13,283 per QALY. As no information on the total costs for each treatment was provided, CADTH replicated the manufacturer’s analysis to derive the results presented in Table 3.

Table 3

Summary of Results of the Manufacturer’s Base Case.

Summary of Manufacturer’s Additional Analyses

The manufacturer also conducted a number of additional scenario analyses (Table 14). The pharmacoeconomic model was most sensitive to a change in perspective (to societal), time horizon, different target population, treatment stopping rule, consideration of drug wastage, and the use of treatment-independent health utilities. The alternate assumptions using a societal perspective, longer time horizon, target population with prior oral prophylaxis treatment failures, and treatment discontinuation response criteria of less than 30% reduction in headache frequency all resulted in a reduced ICUR (range: $24,000 per QALY to $31,000 per QALY). Having no response-based rules, including drug wastage, and using treatment-independent health-state utility values resulted in an increased ICUR (range: $40,000 per QALY to $45,000 per QALY).

The results of the manufacturer’s topiramate comparison scenario analysis were relatively stable in two additional analyses of different discontinuation rates (Table 14).

Limitations of Manufacturer’s Submission

CADTH identified the following key limitations in the resubmission:

• It is uncertain whether the model structure sufficiently captures a spectrum of health states that are meaningful to patients with chronic migraine: Results from the IBMS, which reported EuroQol 5-Dimensions questionnaire (EQ-5D) utility values for the same health states (defined by headache frequency) used by the manufacturer, are not informative as to whether the health states defined by the manufacturer are clinically distinct.⁹ The HDPM ranges assigned to each health state were not based on clinical evidence that showed meaningful quality-of-life differences, but were instead inappropriately based in part on baseline HDPM distribution from PREEMPT trials.⁶ The same study showed that there is a statistically significant difference in utility values between patients with EM (< 15 HDPM) and patients with CM (≥ 15 HDPM),⁹ indicating that reducing health states to these two conditions may better reflect the meaningful difference in quality of life. As noted in the CADTH Clinical Review report, there is uncertainty around the absolute numerical difference between groups of approximately one to two headache and migraine days reported in PREEMPT-1 and PREEMPT-2 in the original CADTH Common Drug Review, and whether they are clinically meaningful.
  As the manufacturer acknowledged in its resubmission, a model structure based on headache frequency does not capture other clinically meaningful aspects of CM, such as headache severity.⁶ The clinical expert consulted by CADTH indicated that the relationship between headache frequency and severity was not linear (i.e., patients who have fewer headaches could have fewer or more severe headaches). Thus, the manufacturer’s model may overestimate the impact of a change in HDPM in patient quality of life. The manufacturer assigned treatment-dependent health-state utility values to patients experiencing the same headache frequency in an attempt to incorporate headache severity and duration into the model. The use of treatment-specific utilities is not recommended by CADTH,¹⁰ and headache severity and duration should have been incorporated in the model in a treatment-independent manner.
  Overall, the model structure based on treatment-specific headache frequency health states has limited face validity. The direction and the magnitude of impact on the model’s cost-effectiveness results are unclear.
• Clinical management of patients with episodic migraine is uncertain: The manufacturer assumes that patients who initially have CM (i.e., ≥ 15 HDPM) and improve to EM (i.e., < 15 HDPM) continue to receive treatment in the model, which does not reflect existing Ona A public drug insurance coverage in Ontario.⁵ Feedback from the clinical expert consulted by CADTH indicated that a potential treatment-and-relapse cycle may arise when patients are forced to discontinue and are allowed to restart treatment only after relapsing back into CM. The health-state transition probabilities used in the manufacturer’s base case assigned higher probabilities of transitioning to EM states to Ona A-treated patients than to BSC-treated patients, which resulted in patients on Ona A accruing more QALY benefit than patients on BSC. However, it is uncertain whether such stopping criteria would be enacted across all Canadian jurisdictions, which may affect the overall cost-effectiveness estimates.
• Key comparators were not comprehensively incorporated in the base case: The manufacturer’s base-case analysis presented a comparison between Ona A and BSC. According to the clinical expert consulted by CADTH, topiramate and erenumab are considered to be key comparators for prophylactic therapy in patients with CM in Canada. Although the manufacturer conducted a scenario analysis that compared Ona A with topiramate, an analysis of all comparators simultaneously allowing for a sequential analysis would have been the preferred approach for the base case.
• Uncertain comparative clinical evidence between Ona A and key comparators: According to the clinical expert consulted by CADTH, it is uncertain what combination of therapies would constitute BSC in Canada. A number of migraine prophylaxis treatments are available (Table 7 and Table 8). The data used to inform BSC in the model was from the placebo injection arm from the PREEMPT trials. It is unclear whether the efficacy and safety profile of placebo injection is similar to BSC as assumed by the manufacturer. Furthermore, the manufacturer incorporated transition probabilities based on subgroups of patients with different numbers of prior prophylactic treatment failures or different response-based stopping rules (i.e., discontinue if < 30% or < 50% headache frequency reduction in 24 weeks). CADTH clinical reviewers could not identify evidence regarding statistically significant differences in efficacy based on these subgroups, and it is unclear whether clinically meaningful differences exist between these subgroups.
  The FORWARD trial was used to inform the manufacturer’s scenario analysis comparison of Ona A and topiramate. CADTH clinical reviewers determined the comparative evidence from this trial was uncertain due to a number of limitations, including the open-label trial design, and high discontinuation rate in the topiramate arm compared with the Ona A arm. Data from a published indirect treatment comparison indicated that, based on a Bayesian network meta-analysis, Ona A was not favoured over topiramate or calcitonin gene-related peptide inhibitors in terms of change from baseline in monthly migraine days, change from baseline in monthly headache days, and all-cause discontinuation. CADTH clinical reviewers considered these results to be limited by heterogeneity that was not systematically evaluated and generalizability to the patient population of interest.
• Extrapolation of short-term data is uncertain: The transition probabilities for week 24 to year 3 (approximately 132 weeks) of the model were informed by short-term data for Ona A (32-week period from PREEMPT trials) and BSC patients (12-week period from PREEMPT trials) data. These transition probabilities allow perpetual improvement and worsening of headache frequency, which may not reflect clinical observations. The clinical expert consulted by CADTH provided feedback that patients who continue to receive treatment for CM would experience a plateau in their reduction in HDPM and maintenance of headache frequency.
• Inclusion of adverse-event data in the model is associated with uncertainty: The manufacturer modelled adverse events based on the percentages reported between week 24 and 36 of the PREEMPT trial. Given that evidence for the first 24 weeks is available, CADTH considered that incorporating the totality of evidence would have been preferred. As this period was the start of the open-label phase of the trial, the adverse events do not reflect a difference between Ona A and placebo safety profiles that is controlled by double-blinding. Additionally, the manufacturer also did not use an appropriate methodology¹¹ to transform probabilities based on the length of the trial (24 weeks for PREEMPT, 36 weeks for FORWARD) to the model’s cycle length of 12 weeks.
  The manufacturer also incorporated an arbitrary health disutility of 0.05 for each of the adverse events in a number of scenario analyses. It is uncertain how this arbitrary value relates to the true utility decrement associated with the adverse events. Furthermore, the disutility calculation in the model was not tied to proportion of treated or live patients, reducing the validity of these scenario analyses.
• Patient characteristics were assumed to differ between treatment groups at baseline: Patients entered the model health states in different proportions based on whether the patients received Ona A or BSC. Patients who received BSC were assumed to have lower overall headache frequency than those who received Ona A at baseline. While this may better approximate the baseline characteristics of PREEMPT, it is unlikely to be reflective of clinical practice, and any differences in model results between treatment groups should be solely due to different efficacy and harm profiles of the treatments that are explicitly reflected in the model structure. ▬▬▬▬▬
• Inappropriate cost assumptions and inputs: The manufacturer did not consider drug wastage and used the 2009 unit cost for ED visits. Incorporating drug wastage and using more recent ED visit costs would better reflect the current costs to the health care system, although the proportions of patients requiring an ED visit for CM remain uncertain.
  Furthermore, CADTH reviewers identified that the model does not consider adverse-event costs in patients who discontinue initial treatment. It is uncertain how accounting for these adverse events would further change cost differences between the compared treatments.

CADTH Common Drug Review Reanalyses

CADTH could not address model limitations associated with the lack of comprehensive consideration of key comparators, uncertain comparative evidence, and the clinical management of patients with EM.

Other limitations were addressed as possible:

1. Alternate treatment-independent utility values directly based on, instead of derived from, IBMS:

   (a)

   by HDPM health states, based on IBMS survey respondents who completed EQ-5D and HIT-6.

   (b)

   by HDPM health states, based on IBMS survey respondents who completed EQ-5D and MSQ.

   (c)

   by CM and EM health states, based on IBMS survey respondents.

2. Alternate adverse-event probabilities based on the 24-week double-blind period of PREEMPT trials, converted to 12-week cycle probabilities using Fleurence and Hollenbeak’s methodology.¹¹
3. Same age, sex, and headache frequency baseline characteristics assumed for all treatment groups based on both arms of PREEMPT trials, no patients below 15 HDPM at baseline.
4. Alternate transition probabilities:

   (a)

   transition probabilities of patients who discontinued treatment based on intention-to-treat (ITT) population.

   (b)

   long-term treatment transition probabilities (beyond 24 weeks) based on ITT population.

   (c)

   transition probabilities based on 4a and 4b.

5. Long-term plateauing and maintenance of headache frequency reduction efficacy.
6. Updated ED cost and drug-wastage consideration.

In the CADTH base case, consisting of: 1c, 2, 3, 4c, 5, and 6, Ona A produced an additional 0.03 QALYs at an incremental cost of $4,168 compared with BSC, resulting in an ICUR of $134,601 per QALY (Table 4). At a willingness to pay of $50,000 per QALY, Ona A was associated with a 9% probability of being the optimal intervention compared with BSC.

Table 4

CADTH Reanalysis (OnabotulinumtoxinA vs. Best Supportive Care).

CADTH conducted additional scenario analyses exploring the use alternate utilities, treatment stopping rule, time horizon, costs, and long-term transition probabilities (Table 17). The CADTH base case was found to be relatively robust in the majority of the scenarios explored. A longer time horizon of 10 years reduced the ICUR to $73,092 per QALY.

As topiramate was considered a relevant comparator, a similar set of reanalyses was also applied to the manufacturer’s scenario analysis comparison of Ona A and topiramate (Table 5), altered to accommodate two different adverse-event reanalyses:

1. Alternate treatment-independent utility values based on IBMS:

   (a)

   by HDPM health states, based on IBMS survey respondents who completed EQ-5D and HIT-6

   (b)

   by HDPM health states, based on IBMS survey respondents who completed EQ-5D and MSQ

   (c)

   by CM and EM health states, based on IBMS survey respondents

2. Alternate adverse-event data:

   (a)

   alternate adverse-event probabilities based on the 36-week FORWARD trial period, converted to 12-week cycle probabilities using Fleurence and Hollenbeak’s methodology¹¹

   (b)

   corrected adverse-event disutility calculation, adverse-event rates are based on proportion of on-treatment patients rather than on proportion of patients that enter the model

3. Same age, sex, and headache frequency baseline characteristics assumed for all treatment groups based on both arms of PREEMPT trials, with no patients below 15 HDPM at baseline
4. Alternate transition probabilities:

   (a)

   transition probabilities of patients who discontinued treatment based on ITT population

   (b)

   long-term treatment transition probabilities (beyond 24 weeks) based on ITT population

   (c)

   transition probabilities based on 4a and 4b

5. Long-term plateauing and maintenance of headache frequency reduction efficacy
6. Updated ED cost and drug-wastage consideration.

Table 5

CADTH Reanalysis (OnabotulinumtoxinA vs. Topiramate).

CADTH scenario analysis of the Ona A and topiramate comparison consisted of reanalyses 1c, 2a, 3, 4c, 5, and 6, and did not incorporate the arbitrary 0.05 disutility that the manufacturer used in the manufacturer’s scenario analysis. Ona A produced an additional 0.13 QALY at an incremental cost of $3,648 compared with topiramate, resulting in an ICUR of $28,968 per QALY (Table 5). At a WTP threshold of $50,000 per QALY, Ona A was associated with an 85.5% probability of being the optimal intervention compared with topiramate. The same set of additional scenario analyses were conducted for the CADTH scenario analysis of comparison of Ona A and topiramate (Table 19). CADTH could not address the substantial limitations with regards to the comparative effectiveness of Ona A and topiramate identified by the clinical reviewers, and as such, the results should be viewed with caution.

Table 6

CADTH Price Reduction Scenarios.

Issues for Consideration

• Feedback from the clinical expert consulted by CADTH suggested that in clinical practice, a 30% reduction in headache frequency may be considered an appropriate response. This differs from the current clinical criteria in Ontario⁵ and the current treatment guidelines (which focus on EM),⁷ which recommend a reduction of at least 50% in headache frequency after 24 weeks. Furthermore, based on the criteria observed in Ontario, patients are required to stop receiving treatment three months after achieving < 15 HDPM (EM).⁵ However, patients often relapse after discontinuing treatment due to achieving EM parameters and will have to start treatment again when headache frequency increases to CM levels (≥ 15 HDPM).
• Based on the feedback from the clinical expert consulted by CADTH, the use of Ona A for prophylaxis of CM may extend outside of the Health Canada indication to continuing treatment for patients once they achieve EM.
• According to a clinical expert consulted by CADTH, Ona A administration cost may be borne by patients as an out-of-pocket cost because the procedure may not be covered by a public health insurance plan, as is the case in Ontario. CADTH reanalyses have explored this scenario.
• The burden of CM to the patients may extend beyond health implications to social considerations, including ability to work. The manufacturer has conducted a scenario analysis from a societal perspective that accounts for potential productivity loss.

Patient Input

Input was received from Migraine Canada, which partnered with Migraine Québec to undertake two online patient surveys that included CM patients. The patients reported that migraine is an important cause of visits to the ED, confirming the importance of capturing the impact of CM prophylaxis on this health care resource use as already done by the manufacturer. Patient inputs also considered Ona A to be a long-term therapy for CM, suggesting the importance of considering a longer time horizon. The majority of patients surveyed reported CM affected their professional life, although only 15% of participants reported being able to go back to work in some capacity after using Ona A. It is uncertain how this finding relates to the manufacturer’s societal perspective scenario analysis, which captured productivity loss by headache frequency based on an American study.⁶

The patients reported that CM is associated with anxiety and depression, with 51% of patients reporting a moderate-to-severe effect of migraine on their mood. The impact of prophylactic treatments on these comorbid conditions is not explored in this review. Existing prophylactic therapies are also reported to be poorly tolerated, indicating a need to explore the impact of adverse events in the economic evaluation. The current review is not able to explore this consideration in detail. The patients also reported that, although Ona A may be used as a monotherapy for prevention, it may be combined with another preventive medication or multiple preventive medications. These reports point to a possible research gap that could be filled by additional pharmacoeconomic analyses involving comparisons of combination treatments and treatment sequences, pending available evidence in this treatment space.

Conclusions

CADTH identified several key limitations with the manufacturer’s model. Based on a series of reanalyses, CADTH estimated that the ICUR of Ona A compared with BSC was $134,601 per QALY. At a WTP threshold of $50,000 per QALY, Ona A was associated with a 9% probability of being the optimal intervention compared with BSC. A price reduction of more than 75% is required for Ona A to achieve an ICUR of less than $50,000 per QALY compared with BSC.

However, given the limitations with the model structure and comparative effectiveness data that could not be adequately addressed in CADTH reanalyses, the cost-effectiveness of Ona A for the prophylaxis of CM remains uncertain.