3.1. Manufacturer’s Base Case
The manufacturer’s primary analysis in the submission was an analysis of onabotulinumtoxinA versus BSC in the subpopulation of CM patients who had failed three or more prior oral prophylactic medications. In this subpopulation, treatment of patients with onabotulinumtoxinA resulted in a total cost of $8,151 per patient over the approximately three-year horizon, compared with $5,339 per patient with BSC. Patients receiving onabotulinumtoxinA had significantly fewer headache attack days and fewer days per year with headaches lasting for at least four hours compared with BSC. These clinical improvements translate to 1.75 QALYs gained with onabotulinumtoxinA versus 1.64 QALYs gained with BSC, leading to an incremental gain of 0.11 QALYs achieved with onabotulinumtoxinA therapy. This resulted in an incremental cost per QALY of $25,470 ().
Summary of Results of the Manufacturer’s Base Case.
3.1.1. Scenario analysis
The manufacturer also undertook a scenario analysis of the full population. In this analysis, onabotulinumtoxinA resulted in an incremental cost per QALY of $28,940 versus BSC ().
Summary of Results of the Manufacturer’s Scenario Analysis of the Full Population.
The manufacturer stated that a stopping rule was included in the model; patients who did not achieve a 30% reduction in headache days were discontinued from study treatment and received BSC for the remainder of the time horizon. These patients were indicated to have been included as “discontinued” patients in the model; however, patients who did not achieve the designated improvement were grouped with patients that discontinued for other reasons within the trial. It would have been more transparent had these patients been separated within the model.
Other scenario analyses that the manufacturer should have undertaken include:
An analysis using data from PREEMPT-1 only (Study 191622-079)
An analysis using data from PREEMPT-2 only (Study 191622-080)
An analysis removing any patient that improved to the point of having fewer than 15 HDPM by week 12 and by week 24.
3.2. Summary of the Manufacturer’s Sensitivity Analyses
3.2.1. One-way sensitivity analyses
The manufacturer conducted several one-way sensitivity analyses, based on either standard errors or assumptions of a 10% change from the base-case value for the input parameters, as well as alterations to the model assumptions. The manufacturer reported that the results were robust to changes in model assumptions, with the ICURs of most scenarios falling under the $50,000 per QALY willingness-to-pay threshold. The results ranged from $5,984 (societal perspective including work productivity) to $66,774 (reducing the time horizon to one year).
The frequency of migraine-related hospitalizations in the CM health states and utility values in the EM health states were found to be the most influential parameters.
The values used for the sensitivity analyses were generally appropriate.
3.2.2. Probabilistic sensitivity analysis
The manufacturer undertook a probabilistic sensitivity analysis (PSA) using 10,000 simulations. The PSA indicated that approximately 88% of iterations fell below a willingness-to-pay threshold of $50,000 per QALY.
The values used for the sensitivity analyses were generally appropriate.
3.3. CADTH Common Drug Review Analyses
The main areas of uncertainty identified by CDRs critique apply to the modelling of CM in both the full population and the subpopulation. The manufacturer made several assumptions in the structure of the model, which prevented CDR from conducting reanalyses of interest that would better inform reimbursement recommendations and decisions. The model appears to include patients who have improved from CM (≥ 15 HDPM) to EM (< 15 HDPM) and who continue to be treated with onabotulinumtoxinA. As onabotulinumtoxinA is indicated for CM only,1 the manufacturer should have factored this stopping rule into its model. It appears as though discontinuation was modelled as an absorbing state; however, this is not transparent in the model, and these patients may have cycled back into the model at a later time point.
The trial data indicate that just 2% of onabotulinumtoxinA patients (14 out of 688 patients) who were in CM at week 12 remained in the group of patients that were in CM at week 24. The rest of the patients had either improved from CM (≥ 15 HDPM) to episode migraine (< 15 HDPM), or had discontinued (although six patients who had improved at week 12 to EM regressed back to CM at week 24). The open-label extension data indicate that fewer than 5% of patients had CM at week 36, at week 48, and at the end of the study. Between week 24 and the end of the study, 41 patients who had improved to EM (< 15 HDPM) transitioned back to CM. The study data indicate that from week 12 to the end of study, between 48% and 59% of study patients were no longer in the CM health state, and therefore would not be eligible to receive treatment with onabotulinumtoxinA. Thus, the inclusion of these patients for the duration of the model is not appropriate.
A model with either a stopping rule for patients who are no longer chronic migraineurs or a stopping rule that cycles these patients out of the model and back in with a recurrence of CM would have been more appropriate. CDR acknowledges that this would have been difficult to undertake given the available clinical data, and that any estimate would be conservative, as the proportion of patients transitioning back to CM is likely to be higher in patients who are not being treated with onabotulinumtoxinA. However, given the high placebo effect in the trials, the true rate of relapse is not known.
CDR did note other assumptions that could be considered through reanalyses: the cost of physician visits for onabotulinumtoxinA, the cost of administration of onabotulinumtoxinA, and the cost of drug acquisition.
3.3.1. CADTH Common Drug Review reanalysis 1
CDR clinical guidance has indicated that onabotulinumtoxinA will take an extra 15 to 30 minutes to administer; thus, one of the following physician visit codes would have been more appropriate (A385 Limited consultation/A186 Repeat consultation)17 for onabotulinumtoxinA, while the manufacturer specified code (A188) would remain appropriate for BSC ():
CDR Analyses — Revised OnabotulinumtoxinA Consultation Price.
3.3.2. CADTH Common Drug Review reanalysis 2
Should onabotulinumtoxinA be listed on drug plan formularies, the cost of the injection is likely to be listed by the plan as well given this is a medical procedure, thus increasing the cost of treatment. Ontario has a cost of injection for onabotulinumtoxinA of $120.00 (for other indications),17 which may be at the upper limit of the cost as the costs of administration in other provinces have been reported to be between $57 and $75. Incorporating this upper range into the manufacturer’s submission, the results are seen in .
CDR Analyses — Include Administration Cost.
3.3.3. CADTH Common Drug Review reanalysis 3
Although the manufacturer based its analysis on the average number of units used in the trial (164 U), onabotulinumtoxinA is supplied as 50 U, 100 U, and 200 U vials. Vials cannot be reused, and thus, there will be associated wastage, which was not taken into account in the drug acquisition cost. The updated drug acquisition costs are shown in .
CDR Analyses — Drug Acquisition Cost.
3.3.4. CADTH Common Drug Review reanalysis 4
The proportion of patients who were hospitalized or who visited an ER may be lower than the proportion indicated in the manufacturer’s submission. shows the results of altering these proportions to 0% for all health states to denote a lower range. also presents a range, based on information from the IBM Study (Sanderson et al.),14 which uses values for the CM and EM health states of 0.52 and 0.35 respectively for ER admissions, and values of 0.22 and 0.16 respectively for hospitalizations. Although these values were for the full population, they have been used in the subpopulation analysis as well.
CDR Analyses — Event Treatment Costs: Hospitalization and Emergency Room Visits.
CDR Analyses – Event Treatment Costs: Hospitalization and Emergency Room Visits.
3.3.5. CADTH Common Drug Review reanalysis 5
Including the revised assumptions for CDR reanalyses 1, 2, and 3 within the economic evaluation simultaneously result in the ICURs identified in .
CDR Reanalysis — Incorporating Reanalyses 1, 2, and 3.
3.3.6. Pricing analysis
Given the identified issues and uncertainties with the submitted structure for modelling CM, a price analysis was also undertaken to determine the price reduction required to achieve certain lower ICURs ().
CDR Analyses – Price Analyses for OnabotulinumtoxinA.
To achieve an ICUR of $25,000, a 42% to 50% price reduction of onabotulinumtoxinA would be required.
