Table 1 Results of included studies

ERT/SRT

LSDs

<1: 50 000

soc

Wyatt [22]

Laronidase (Aldurazyme)

MPS1

1: 100 000

soc

Connock [21]

Agalsidase alfa and beta (Fabrazyme and Replagal)

Fabry disease

0.22: 100 000

soc

Markov-state-transition model

€3 282 252

Eleven Markov states were defined, including no symptoms, one or multiple complications and death, with a 1-year cycle period and a time horizon of 70 years. TPs, utilities and costs (both direct and indirect) were derived from retrospective and prospective follow-up data of the Dutch cohort. Utilities and costs were assumed to be equal in both arms as long as patients are in the same disease state, except for ERT costs.

18

Rombach [25]

Agalsidase alfa and beta (Fabrazyme and Replagal)

Fabry disease

0.22: 100 000

soc

Not specified

€351 622^b

Limited information reported in the literature precluded the development of a comprehensive model. The decision model considered a birth cohort of male patients. Survival and utilities for untreated patients were estimated from literature. ERT was assumed to restore patients to full health and no disease-specific mortality. Only direct costs were considered.

13

Connock [21]

Agalsidase alfa and beta (Fabrazyme and Replagal)

Fabry disease

0.22: 100 000

soc

Not specified

80 % probability that ERT has a positive net benefit at a willingness to pay per QALY of €266 520^b

Utilities were estimated on the basis of 2 studies. Costs were roughly estimated without using references. Bootstrapping was used to obtain ICER distributions. Results were plotted in a cost-effectiveness acceptability-curve. Primary objective of the study was not to calculate the ICER but to draw upper and lower limits of the current pricing structure.

8

Moore [16]

Velaglucerase alfa and imiglucerase (Vpriv and Cerezyme)

Gaucher disease

1: 100 000

soc

Markov-state-transition model

€432 540

Eight Markov states were defined, including no symptoms, one or multiple complications and death, with a 1-year cycle period and a time horizon of 85 years. TPs, utilities and costs (both direct and indirect) were derived from retrospective and prospective follow-up data of the Dutch cohort. Utilities and costs were assumed to be equal in both arms as long as patients are in the same disease state, except for ERT costs.

19

Van Dussen [13]

Alglucerase (Ceredase)

Gaucher disease

1: 100 000

soc

Not specified

€43 532^b (2.3U/kg three times weekly)

Estimate of ICERs for 3 drug dosing regimens was based on the assumption that alglucerase results in survival with perfect quality of life vs immediate death without treatment.

7

Beutler [14]

€66 631^b (30U/kg/2wk)

€130 595^b (60U/kg/2wk)

Alglucosidase alfa (Myozyme)

Pompe disease (infantile)

1: 14 000 to

soc

Patient-level simulation model

€1 043 868

Survival in both treatment arms was estimated on the basis of international studies, literature and their own patient cohort. Utilities and costs were derived from 6 and 12 patients resp., and were assumed to be equal in both arms except for ERT and infusion-related costs.

17

Kanters [24]

1: 300 000

Alglucosidase alfa (Myozyme)

Pompe disease (infantile)

1: 14 000 to

soc

Markov-state- transition model

England: €326 791^b

Two Markov states were defined: alive-symptomatic and dead, with a 1-year cycle and a time horizon of 20 years. Survival rate (75 vs 8 %) and costs were derived from literature and historic databases. It was assumed that ERT would result in higher utilities. Only direct costs were considered. Costs and QALYs were generated for 2 settings: England and Colombia.

16

Castro-Jaramillo [19]

1: 300 000

Colombia: €153 405^b

Eculizumab (Soliris)

Paroxysmal Nocturnal Haemoglobinuria

1: 500 000

soc

Markov-state-transition model

€1 620 256^b

Six different consequences from PNH or treatment were modelled. Patients may have more than 1 complication at the same time, resulting in a total of 47 disease states. Analysis was conducted for 6 patient strata to account for differences in thrombosis risk and transfusion requirements. TPs for both arms were estimated on the basis of cohort studies and clinical trials. Utilities were derived from literature. Only direct costs were considered.

17

Coyle [28]

Eculizumab (Soliris)

Paroxysmal Nocturnal Haemoglobinuria

1: 500 000

soc

Not specified

1) €358 655 (haemoglobin) €474 998 (LDH)

Due to the lack of reliable quantitative information about costs and benefit of eculilzumab, a fully informed estimation of cost-effectiveness was not feasible. Instead, 3 preliminary analyses were conducted to estimate the costs per: 1) patient with stabilised of haemoglobin and normalised LDH, 2) LYG under the assumption that eculizumab returns survival to normal, and 3) LYG based on reduced mortality rates from thrombosis.

9

Connock [28]

2) €697 500 - €1 953 000/LYG

3) €3 906 000 - €4 464 000/LYG

Mercaptamine (Procysbi)

Cystinosis

<1: 50 000

soc

Decision tree

Dominant; cysteamine is more effective and less expensive

Drug studied is cysteamine (Cystagon), which is the ‘reference drug’ for mercaptamine, i.e., both drugs are equal. Time to first renal failure was estimated to be 10 years for untreated and 15 years for treated patients based on literature. After first renal failure all patients were assumed to follow the same disease course. Only direct costs were considered. Costs were discounted, outcomes were not. Hence life expectancy increases by 5 years while costs for dialysis and transplants are delayed resulting in cost savings.

15

Soohoo [17]

Sildenafil (Revatio)

PAH

5: 1 000 000

one of: bosentan, treprostinil, epoprostenol, ilioprost, sitaxentan

Markov-state-transition model

Dominant; sildenafil is more effective and less expensive

Five states were defined (WHO functional classes and death), with a 3-month cycle period and a time horizon of 1 year. The model was based on 100 hypothetical patients with a gender and disease severity distribution based on previous studies. TPs were based on 3-months transitions reported for bosentan and estimated for the other drugs after adjustment for improvement in 6MWT. Reported utility values for each functional class were used, adjusted for route of administration. Only direct costs were considered.

15 (2)^a

Garin [15]

Iloprost (Ventavis)

PAH

5: 1 000 000

treprostinil

Markov-state-transition model

€1 228 055^b

See description above for Garin [15]

15 (2)^a

Garin [15]

epoprostenol

€1 016 649^b

Iloprost (Ventavis)

PAH

5: 1 000 000

treprostinil

Markov-state-transition model

Dominant; iloprost is more effective and less expensive

Five Markov states were defined representing the WHO functional classes and death, with a 3-month cycle period and a time horizon of 3 year. Treatment changes were allowed if patient status worsened. The model was based on 100 hypothetical patients with functional class III. TPs were based on 6MWT results obtained from clinical trials, utilities from the literature. Only direct costs were considered. All parameters were validated by an expert opinion.

15 (1)^a

Roman [26]

epoprostenol vs iloprost

€6 847 284

Bosentan (Tracleer)

PAH

5: 1 000 000

soc

Patient-level simulation model

Dominant; bosentan is more effective and less expensive

Time to progression functional class III to IV and utilities were estimated on the basis of international studies and literature. Life expectancy was assumed to be independent of the initial treatment. Time spent in the most severe disease state (with higher costs due to continuous treatment with prostaglandins in this state, and lower utility) is thus the only varying outcome parameter between the 2 treatment arms. Only direct costs were considered. 10,000 patients were simulated and replicated. For each patient a discrete event simulation approach was used in which the simulated time moved directly to the time of the next event (progression or death).

14 (1)^a

Stevenson [20]

Bosentan (Tracleer)

PAH

5: 1 000 000

one of: epoprostenol, treprostinil

Markov-state-transition model

Dominant; bosentan is more effective and less expensive

Five Markov states were defined representing the WHO functional classes and death, with a 3-month cycle period and a time horizon of 1 year. The model was based on 100 hypothetical patients. TPs were based on 3-months transitions reported for bosentan and estimated for the other drugs after adjustment for improvement in 6MWT. Utilities were estimated by a group of clinical experts. Only direct costs were considered.

14 (2)^a

Highland [18]

Bosentan (Tracleer)

PAH

5: 1 000 000

soc

Patient-level simulation model

€40 137 / LYG^b

Mortality rate in both treatment arms was estimated on the basis of international studies and literature. The key assumption was that bosentan increases life expectancy. Only direct costs were considered. 5,000 patients were simulated and replicated. At 6-months intervals the patient progressed through one or more health states (bosentan, conventional therapy or death) with a 15-year time horizon.

14 (1)^a

Wlodarczyk [27]

Bosentan (Tracleer)

PAH

5: 1 000 000

one of: epoprostenol, iloprost

Markov-state-transition model

Dominant; bosentan is more effective and less expensive

See description above for Garin [15]

15 (2)^a

Garin [15]

treprostinil

€65 282^b

sitaxentan

€2 621^b

Ambrisentan (Volibris)

PAH

5: 1 000 000

one of: iloprost, epoprostenol

Markov-state-transition model

Dominant; ambrisentan is more effective and less expensive

See description above for Garin [15]

15 (2)^a

Garin [15]

bosentan

Equal; no difference in effectiveness and costs

treprostinil

€65 282^b

sildenafil

Dominated; ambrisentan is equally effective but more expensive

sitaxentan

€2 621^b