| Baseline characteristics | NA; sponsor did not use patient baseline characteristic inputs | NA |
| Efficacy | IN glucagon was assumed to be equally efficacious as IM glucagon. | See |
| Bystanders to an SH event were assumed to be twice as likely to attempt to use glucagon when IN glucagon is available compared to when IM glucagon is available. | See |
| The proportion of bystanders to an SH event attempting glucagon treatment was based on the sponsor’s CRASH web survey study.25 | Acceptable. However, there is some uncertainty regarding this parameter estimate as the survey participants ▬. It is also unclear how the study’s exclusion of patients ▬▬▬▬▬ would impact the parameter estimate. |
| The difference in the probability of a successful glucagon administration between the IN and IM formulations of glucagon was informed by the sponsor’s Yale et al. (2017) study.3 | Inappropriate. Yale et al. (2017) defined only a full dose injection as a successful treatment with IM glucagon.3 However, according to the clinical expert consulted by CADTH, less than a full dose injection of glucagon may also be effective, especially for pediatric patients who only require half of a dose based on the IM glucagon product monograph.16,17 Therefore, a more appropriate definition of a successful glucagon injection would also include successful partial dose injections. The use of the more restrictive definition of successful treatment favoured IN glucagon. CADTH clinical reviewers also noted uncertain generalizability due to the limitations in the ability of simulated scenarios to mimic real-world conditions, and the fact that study participants who were caregivers or patients with diabetes did not own a glucagon device or never had one, in contrast to the HC-indicated population who are likely to own one. |
| Natural history | The probability that patients who do not successfully receive glucagon treatment by a bystander would be privately transported to an ED was based on a 2015 Canadian ED chart review study.11 | Uncertain. The input data are based on the proportion of patients with an SH event who were reported to have independently travelled to an ED in the Rowe et al. (2015) study.2,11 This proportion is based on a denominator that represents a different population (i.e., patients with SH who visited ED) compared to those who are reflected in the probability (i.e., patients with SH with impaired consciousness that preclude oral carbohydrate treatment, who did not receive a successful glucagon treatment from a bystander). Additionally, the sponsor’s approach did not replicate the proportion of patients observed in the Rowe et al. (2015) study as was seemingly intended. As the sponsor’s decision tree also modelled additional paths for patients to be transported to the ED via EMS, the overall proportion of modelled patients who visit the ED in the model were lower than observed in the Rowe et al. (2015) study. Although this may favour IN glucagon because the model assumes that more patients could be diverted from the ED, the direction of bias is uncertain as it is unknown how the population in the Rowe et al. (2015) study is related to the HC-indicated population for IN glucagon. |
| The probability of an EMS visit leading to an ED visit, and the probability of an ED visit leading to a hospital admission were based on a 2018 Canadian review of paramedic and ED records.10 | Acceptable. |
| Utilities | Mean disutilities associated with SH event resolution without health care resource use, SH event involving EMS, SH event involving ED visit, and SH event involving inpatient admission were approximated by pediatric epilepsy-related disutilities reported in the Lee et al. (2013) cost-effectiveness analysis.18 | Inappropriate. The sponsor inappropriately applied the disutility value associated with an admission to intensive care unit to model the reduction in QALYs associated with an inpatient admission, and also inappropriately applied the disutility value associated with an inpatient admission to model the reduction in QALYs associated with an ED visit. Although the Lee et al. (2013) study applied the disutility values to a 30-day period,18 the sponsor effectively applied the disutilities over the one-year time horizon. Furthermore, the methodological validity of the disutility values is unknown as the full methodology behind the disutilities, including the study country and the method of elicitation, is proprietary information that was not reported in the Lee et al. (2013) study.18 The disutilities also reflect a pediatric population and their generalizability to the HC-indicated population is unknown. CADTH explored the uncertainty associated with health care resource use-dependent disutility values in a sensitivity analysis. |
| Adverse events | Adverse events were not modelled | NA |
| Mortality | Mortality was not modelled | NA |
| Resource use and costs |
|---|
| Drug | Cost for IN glucagon was based on sponsor’s submission.25 | Appropriate. |
| Cost of IM glucagon kit was based on costs of Glucagen17 and glucagon16 from the Ontario Drug Benefit formulary7 weighted by market share data from IQVIA reported by the sponsor.2 | Appropriate. |
| Events | Resource use associated with follow-up care with a health care professional was based on reported referrals to endocrinologist, general internist, and primary care physician from the Rowe et al. study.11 | Inappropriate. The sponsor selected only some of the reported referrals in the Rowe et al. (2015) study and excluded other referrals such as diabetes education, a relevant follow-up cost.11 However, this selective costing does not impact the results of the pharmacoeconomic submission as the application of the follow-up costs did not differ between treatment arms by design. |
| Physician unit costs were based on Ontario Schedule of Benefits,8 and were inflated to 2019 Canadian dollars using the Canadian General Consumer Price Index.27 | Appropriate. |
| The cost of EMS was derived from 2017 Toronto Paramedic Services Annual Report.19 | Acceptable. The sponsor’s approach was based on CADTH costing guidelines.21 However, this estimate is uncertain as the cost estimate approach divided the gross operating budget by number of patients transported. As not every EMS response results in patient transportation, this approach likely overestimates the cost of the average EMS response. In order to explore this uncertainty, CADTH conducted a scenario analysis with a lower cost estimate based on the annual number of EMS responses. |
| The costs of an ED visit and an inpatient admission were based on the Ontario Case Costing Initiative database.6 | Appropriate. |
