The manufacturers have provided a semi-Markov model to consider the cost-effectiveness of glycerol phenylbutyrate (GPB) in four distinct subgroups:

• Subgroup 1: Patients with no prior treatment with sodium phenylbutyrate (NaPBA) or currently on treatment with sodium phenylbutyrate, with disease onset after two years old. Average starting age in the model is four years old.
• Subgroup 2: Patients previously treated with NaPBA but who discontinued treatment due to uncontrolled ammonia levels or were unable to tolerate NaPBA, with disease onset after two years old. Average starting age in the model is 20 years old.
• Subgroup 3: Patients with no prior treatment with NaPBA or currently on treatment with sodium phenylbutyrate, with disease onset between birth and two years old. All patients entered the model (as provided) at birth.
• Subgroup 4: Patients previously treated with NaPBA but who discontinued treatment due to uncontrolled ammonia levels or were unable to tolerate NaPBA, with disease onset between birth and two years old. Average starting age in the model is eight years old.

In Subgroup 1 and Subgroup 3, GPB was compared with NaPBA, while dietary control alone was used as a comparator for Subgroup 2 and Subgroup 4. The perspective of the analysis was from the Canadian health care system (payer perspective), and health benefits were reported in quality-adjusted life-years (QALYs).

The schematic for this model is provided as Figure 1 of APPENDIX 3. In the base-case, this model considered six health states, and the main event modelled was hyperammonemic crisis (HAC). As well as requiring costly treatment, these events were modelled to include a possibility of death, and those who had experienced one or more crises were modelled as separate groups from those who had never experienced a crisis.

All patients in Subgroup 2 and Subgroup 4 were assumed to have had a HAC previously and began in the “Post-HAC” health state. From this state, patients could remain in this state, transit to the “HAC” state, die (to “Other death”), or receive a transplant (“Liver transplant”). After a HAC, the patients could return to the “Post-HAC” state, die (“HAC-related death”), or receive a transplant. Both death states (“HAC-related death” and “Other death”) and the “Liver transplant” state were modelled as absorbing states.

For the other subgroups, the model is broadly similar. With the exception of starting age, the only difference in modelling was that 30% of patients in Subgroup 1 and 10% of patients in Subgroup 3 did not have a HAC before they become eligible for GPB. The model for Subgroup 1 and Subgroup 3 allows this portion of patients to start in a “No previous HAC” state, with the remainder starting in the “Post-HAC” state. Once in that state, they were assumed to remain there until a first HAC (or death, or transplant) occurred.

Within the economic model, the short-term success of treatments lies in reducing the probability of HACs and, in the longer term, success was based on reduction in mortality following HACs. While the model considers the possibility of chronic conditions following HACs (as a form of disease progression), this was presented only in a sensitivity analysis.

Short-term clinical effectiveness was based on an estimation of the impact of treatments on ammonia levels from short-term trials.² Evidence for the comparative efficacy of GPB versus NaPBA regarding ammonia levels was based on a pooled analysis of one double-blind crossover trial (HPN-100-006), and three open-label fixed sequence switchover trials (HPN-100-005, -012, and UP 1204-003). Ammonia-level data for patients on dietary control were from an observational study.³ In order to obtain an estimate of HAC rates, the manufacturer reanalyzed an existing estimated relationship using methods differing from the analysis that produced the estimated relationship. Raw data⁴ were available to base estimates on, but were not used by the manufacturer directly.

Drug monthly costs for GPB and Buphenyl (NaPBA) were provided by Horizon Pharma, with the cost of Pheburane (NaPBA) based on Quebec list prices.² In the base-case analysis, only Pheburane was compared with GPB, as equivalent efficacy for Buphenyl and Pheburane was assumed, with Pheburane being cheaper. The monthly costs of treatment for GPB (based on body surface area) and Pheburane (based on weight for those < 20 kg, and surface area for those > 20 kg) increases with age.

Health-state costs were based on the cost of maintenance in non-HAC states, costs of HACs, and costs of liver transplantation. A large number of assumptions were used to construct these estimates. The HAC cost was based on a microcosting of treatment (Ammonul, L-arginine) and hospital stays (general ward and intensive care units). In all cases, HACs were costly, with first HACs ranging from around $69,000 to $147,000 and subsequent HACs costing between $42,000 and $120,000 per event. Following HACs, monthly maintenance costs were assessed at $10 for adults, and around $74 to $79 for children. Liver-transplantation costs were based on a published figure from Alberta.²

For details on data sources and manufacturer’s assumptions, refer to APPENDIX 3.