Clinical and cost-effectiveness of epoprostenol, iloprost, bosentan, sitaxentan and sildenafil for the treatment of pulmonary arterial hypertension within their licensed indications: a systematic review and economic evaluation

Clinical classification

PAH is one of five differing subtypes of pulmonary hypertension.

Pulmonary hypertension was traditionally classified into two categories, primary pulmonary hypertension (PPH) or secondary pulmonary hypertension, depending on the absence or presence of identifiable causes or risk factors. In 1998 the World Health Organization (WHO) co-sponsored a symposium on pulmonary hypertension, which took place in Evian, France. A new clinical classification of pulmonary hypertension based on pathophysiological mechanism, clinical presentation and therapeutic options was proposed in the symposium. This ‘Evian classification’ (or sometimes referred to as the WHO 1998 classification) includes five major categories, with PAH being one of the categories. The term ‘primary pulmonary hypertension’ was retained within this category and included subcategories of sporadic PAH and familial PAH (FPAH). It was agreed that the term ‘secondary pulmonary hypertension’ should be abandoned. In a subsequent symposium that took place in Venice, Italy, in 2003, the Evian classification was further modified. The term ‘primary pulmonary hypertension’ was removed and the subcategory of sporadic PAH was replaced by idiopathic PAH (IPAH). The details of the Venice 2003 clinical classification are listed in Table 1. 1

As PPH was widely used before the advent of the Venice 2003 classification a decision was made to retain this term in this report if it was used in the original publications/reports of individual studies. Where PPH is retained it is regarded as being interchangeable with IPAH.

Functional classification

Traditionally, patients with PAH are classified according to the classification of functional capacity developed by the New York Heart Association (NYHA) for patients with cardiac diseases based on clinical severity and prognosis. An adaptation of the NYHA functional classification specifically for patients with pulmonary hypertension was proposed in the aforementioned WHO symposium in Evian. The WHO classification and the NYHA classification are nearly identical and are sometimes referred to as the NYHA/WHO classification, which is listed in Table 2. 1

Drugs and toxins

Exposure to certain drugs and toxins might increase the risk of PAH. Evidence has been provided to associate the use of appetite suppressants structurally derived from amphetamine (aminorex, fenfluramine and dexfenfluramine) with a sixfold increase in the risk of developing PAH. Because of this adverse effect such suppressants have been removed from the market. 2

No significant difference has been reported between patients with PAH and the general population with regard to smoking habits. 4

Demographic and medical conditions

There is fairly clear evidence that, in adults, women tend to be more likely to develop PAH than men. Although the ratio of women to men varies from study to study it is of the order of 1.3:1 to 2.2:1. 4–6 In most trials women constitute the majority of patients.

No significant difference between PAH patients and the general population with regard to number of births per woman has been demonstrated. 4

Diseases

PAH is frequently associated with a number of other diseases. These associations are reflected in the subclassifications of PAH (see Table 1).

A relationship between HIV infection and PAH has been clearly demonstrated. 1 About 0.5% of patients infected with HIV will develop PAH. 2

Associated PAH (APAH) occurs in connective tissue diseases (CTD) and most commonly in scleroderma; around 12% of a hospital population of scleroderma patients suffer from PAH. 7 Despite similar haemodynamics survival in scleroderma APAH is worse than in IPAH, with a median survival of 1.2 years. 8

Congenital heart disease (CHD) with non-restrictive systemic to pulmonary shunts, such as ventricular septal defects, patent ductus arteriosus and large atrial septal defects, may lead to PAH. Eisenmenger syndrome develops when such patients develop severe PAH with reversal of flow across the shunt and cyanosis. Survival in untreated Eisenmenger syndrome is much longer than in IPAH although it is still markedly reduced compared with that in the normal population. 9

Portopulmonary hypertension, associated with liver disease and portal hypertension, is observed in 4–15% of patients who are evaluated for liver transplantation. 2

Hereditary

In 6–10% of patients PAH is suspected or proven to be of hereditary origin. In total, 50–90% of patients diagnosed with FPAH have mutations of the bone morphogenetic protein receptor, type II (BMPR2) gene. Patients with FPAH tend to suffer from more severe and quickly progressing disease. 2 In 2001 in the UK there were at least 20 families known to have FPAH. 10

Prognosis and prognostic factors

The prognosis for patients with PAH on supportive care (see Current service provision) is considered to be poor. In the 1980s median survival at the time of diagnosis for patients with IPAH (PPH) receiving supportive care was 2.8 years. 5 Percentages of patients surviving a specified period were estimated as 68% [95% confidence interval (CI) 61–75%] at 1 year, 48% (95% CI 41–55%) at 3 years and 34% (95% CI 24–44%) at 5 years. 5 One of the key factors influencing prognosis is functional class (FC). Patients with FCI or FCII PAH in the 1980s cohort had a median survival of 58.6 months, whereas those with FCIII had a median survival of 31.5 months. An extremely low median survival of 6 months was observed in patients with FCIV. 5 Given the greater awareness of PAH, the development of specialised PAH services and treatment algorithms and the potential for earlier diagnosis, median survival times from diagnosis may be longer today.

Haemodynamic variables related to decreased survival have been identified: increased mean (m)PAP, increased mean right arterial pressure (RAP)and decreased cardiac index. These variables also appeared in an equation predicting patient survival based on the results of a multivariate analysis of data from a registry established in the 1980s by the National Institutes of Health (USA). 5 The applicability of survival rates predicted by this equation, however, is questionable given the changes in medical practice as well as in other socioeconomic factors over the past few decades.

Exercise endurance, usually measured with the 6-minute walk test (6MWT), is also considered to be an important prognostic factor. One of the earliest drug trials for PAH demonstrated that the 6-minute walk distance (6MWD) was a predicator of survival, independent of treatment. 2,11

The progression of PAH symptoms in the context of change in clinical parameters is shown schematically in Figure 1.

FIGURE 1.

Schematic of progression of pulmonary arterial hypertension and change in clinical parameters. Adapted from Rich S. Primary pulmonary hypertension. Prog Cardiovasc Dis 1988;31:205–38. Copyright 1988, with permission from Elsevier.

Adapted from Rich S. Primary pulmonary hypertension. Prog Cardiovasc Dis 1988;31:205–38. Copyright 1988, with permission from Elsevier.

Measurement of disease

A number of measures are used clinically to monitor the severity, progression and response to treatment of PAH. Many of these can be related to exercise capacity, haemodynamics and/or cardiac performance. Clinically no single measure or composite measure is utilised to measure the disease. Severity, progression and response to treatment are assessed utilising a combination of measures. Some of the key measures are outlined in the following sections.

Anticoagulation

The aim of treatment with anticoagulants is to reduce the risk of venous thromboembolism, the risk of which is increased by PAH. 16 Usually, a target of an international normalised ratio (INR) ranging between 2.0 and 3.0 in Europe (and 1.5–2.5 in North America) is assumed. 1

The effectiveness of oral anticoagulants was originally demonstrated in retrospective single centre studies including only patients with IPAH and PAH related to anorexigens. Anticoagulation is also used in patients with other aetiologies of PAH, but all contraindications (such as a high risk of gastrointestinal bleeding in patients with CTD) have to be carefully considered. If there are no contraindications, patients treated with intravenous medication (e.g. epoprostenol; see Prostanoids, Epoprostenol) are also treated with anticoagulants, as they are at an increased risk of thrombosis associated with the use of a catheter. 1

In recent PAH randomised controlled trials (RCTs) the use of anticoagulants was reported in 51–86% of patients. 1 Warfarin is frequently the anticoagulant used in the treatment of PAH and, as with its use in other diseases, patients require frequent monitoring to reduce serious adverse effects such as haemorrhage.

Diuretics

Diuretics are used to prevent or reduce fluid retention. The aim of using diuretics in patients with PAH is to treat oedema or fluid retention connected with right heart failure, such as ankle swelling or ascites. 16 There are several classes of diuretics, including thiazides, loop diuretics and potassium-sparing diuretics. In recent PAH RCTs 49–70% of patients were treated with diuretics. Because of the lack of trials including specific classes of these drugs the choice of type and dose of medication are left to the decision of the physician. Monitoring of serum electrolytes and indices of renal function is advised in patients undergoing diuretic therapy. 1

Examples of diuretics used in the treatment of PAH include furosemide, amiloride and spironolactone.

Oxygen

Oxygen is used in patients with hypoxaemia, an abnormal deficiency of oxygen in arterial blood. 16 Hypoxaemia at rest is usually mild in patients with PAH. Some patients experience improvement in PAH with low-flow supplemental oxygen. Although the effect has not been proven in controlled studies it is considered important to maintain an oxygen saturation greater than 90% in patients with PAH. 1

The use of oxygen in patients suffering from PAH associated with some underlying conditions, such as cardiac shunts, can be controversial. A clinical trial1 assessing the efficacy of nocturnal use of oxygen in patients with PAH associated with Eisenmenger syndrome found no effect of oxygen therapy on haematological variables, quality of life or survival.

The need for oxygen often decreases in patients treated with epoprostenol. Patients without targeted treatment require more oxygen therapy.

Digoxin

The progression of right heart failure often results in depression of myocardial contractility. This condition can be treated with inotropic agents, for example agents that affect the force of muscle contraction.

Digoxin is used in patients with refractory right heart failure in sinus rhythm. 1 Digoxin is available in tablet form or as an injection. An increase in cardiac output, as well as a reduction in circulating noradrenaline levels, can be obtained by short-term intravenous use of digoxin. There is no evidence indicating the long-term efficiency of this drug. 1 Digoxin may be prescribed for improvement of cardiac output; however, now it is considered useful in rare cases of atrial fibrillation or atrial flutter to slow the ventricular rate. 16 It was used in 18–53% of patients taking part in recent PAH RCTs. 1

Calcium channel blockers

Calcium channel blockers are used in PAH patients with no right heart failure for reduction of PVR.

No more than 10% of IPAH patients respond acutely to vasodilator therapy. 17 Treatment of paediatric IPAH with calcium channel blockers has also shown some favourable results. It is less clear if therapy in patients suffering from PAH associated with other conditions is effective.

Only patients responding substantially in the short term to this therapy are considered for treatment with calcium channel blockers alone. 1 They are identified by means of an acute vasodilator challenge using short-acting agents, such as intravenous prostacyclin (see Description of technology under assessment, Prostanoids), adenosine or inhaled nitric oxide during right heart catheterisation. 18 As a result of a retrospective analysis of 557 patients tested with intravenous epoprostenol and inhaled nitric oxide, response criteria have been accepted of a fall in mPAP of 10 mmHg to an absolute mPAP ≤ 40 mmHg with an unchanged or increased cardiac output. 2

Nifedipine and diltiazem are the vasodilators most frequently used in clinical trials. There are also new-generation calcium channel blockers (e.g. amlodipine and felodopine). Limited reports on efficacy, tolerability and dosage are available.

The choice of a calcium channel blocker can be based on the patient’s heart rate, with relative bradycardia indicating nifedipine and relative tachycardia favouring diltiazem. The effective daily doses of these drugs tend to be high, ranging from 120 to 240 mg for nifedipine and from 240 to 720 mg for diltiazem. The advised procedure is to start with lower doses and gradually increase them to the highest tolerated ones. Usually, systemic hypotension and lower limb peripheral oedema limit the dose increase. The side effects can at times be decreased by the use of digoxin and/or diuretics (see earlier sections on these treatments). 1 This therapy requires close monitoring as the positive effect is not always maintained over time. 19

Epoprostenol

Epoprostenol is a synthetic sodium salt of prostacyclin. It is indicated for the intravenous treatment of PPH in NYHA FCIII and FCIV patients who do not respond adequately to conventional/background therapy. 20 For this indication epoprostenol is licensed in vial sizes of 1.5 mg. (1.5-mg vials along with 0.5-mg vials are also licensed for renal dialysis when the use of heparin is otherwise contraindicated or heparin use carries a high risk of causing or exacerbating bleeding. 20,25) Conventional/background therapy, although not explicitly defined, can be considered to be those treatments not classed as interventions in this assessment and as specified in the section on current service provision.

Epoprostenol is contraindicated in patients with known hypersensitivity to the drug, with congestive heart failure from severe left ventricular dysfunction and/or who develop pulmonary oedema during dose ranging. 20

Epoprostenol has a short half-life in the circulation (3–5 minutes) and therefore is administered continuously via pump into a central venous catheter (Hickman line). 1 Furthermore, once in solution epoprostenol is only stable for 8 hours at room temperature, requiring it to be kept cool before infusion with ice packs. Given the route of delivery, continuous administration and limited stability the treatment is not without complications. Not all patients are suitable for epoprostenol treatment as a great deal of self- or carer-ability and commitment is required to prepare and administer the drug under sterile conditions and to maintain sterility of the permanent central venous catheter. Ongoing patient/carer education and training are vital and these are delivered regularly by a specialist nurse.

Treatment must be initiated as an inpatient under specialist care because of the intensive training of patients and/or their carers and the close monitoring and emergency backup required. Initiation of treatment is by short-term dosing to determine the patient-specific infusion rate (this process can also be undertaken using a peripheral rather than a central line). Initially the infusion rate is 1–2 ng/kg/min and this is increased until maximum benefit on haemodynamic parameters is achieved and/or dose-limiting pharmacological effects occur.

Patients well enough to return home do so after this period, which usually lasts 1–2 weeks. Not all patients can safely manage epoprostenol treatment without help from carers.

Patients require two serviceable pumps at home in case one fails. These, along with a regular supply of sterile and other consumables and epoprostenol, are usually delivered by home care services to the patient. Patients have access to telephone support from the specialist centre, usually immediate access to outpatient and inpatient care and district nursing services.

Over time the infusion rate is gradually increased by 1–2 ng/kg/min steps to assess the clinical response and, overall, gradual dose increases are to be expected in most patients to arrest a deterioration in symptoms. 20 Typical doses might be in the range of 15–50 ng/kg/min (higher upper doses have been used in the USA) depending on the length of time on treatment, the resistance of the disease to adequate control and the severity of any adverse effects.

Patients who deteriorate appreciably while on treatment and/or who are not fit/able to return home after initiation of treatment usually require full-time hospitalisation.

Once initiated, withdrawal of epoprostenol treatment is problematic because of rebound pulmonary hypertension and rapid clinical deterioration, which may result in death. For this reason, once initiated, epoprostenol treatment is considered to be lifelong by many.

Because of the difficulties associated with epoprostenol treatment it is a very considered decision by both the patient/carer and clinical team whether and when to initiate treatment. For this reason the other interventions outlined in the following sections will be considered or utilised initially in preference. However, epoprostenol is considered to be the last defence against deterioration of the disease. It is therefore added to treatment regimes when other treatments begin to fail. Thus, many patients will be receiving epoprostenol, usually in combination with an oral treatment (see Endothelin receptor antagonists, phosphodiesterase-5 inhibitors and Current usage of technologies in the NHS). Patients presenting with aggressive disease and/or in FCIV will receive epoprostenol.

The price of epoprostenol is approximately £130–390 per day (15–45 ng/kg/min per 70-kg patient; one to three vials per day; net price). 26 This price only includes the epoprostenol powder and diluent and not the pumps, consumables, delivery or any other costs associated with administration (insertion of Hickman line), monitoring, inpatient time and training. The price for some of these items is difficult to ascertain and/or contained in confidential service agreements.

Iloprost (inhaled)

Iloprost is a stable prostacyclin analogue that has been developed for intravenous, oral and inhaled administration. Only administration by inhalation is part of this assessment.

Inhaled iloprost has EU marketing authorisation for the treatment of PPH patients in NYHA FCIII to improve exercise capacity and symptoms. 21 Two vial sizes, 1 ml and 2 ml, are licensed.

The administration of iloprost by inhalation is an attractive idea as potentially it is selectively delivered to the pulmonary circulation. To ensure distribution to the alveoli a delivery system is required to produce small enough aerosol particles. Three types of delivery systems (nebulisers) are available: compressed air, ultrasonic and vibrating mesh nebulisers. The recommended dose is 2.5 ∝g or 5.0 ∝g of iloprost (as delivered at the mouthpiece of the nebuliser) per inhalation session according to individual need and tolerability. One vial is sufficient for each inhalation session. Each inhalation session takes 3–10 minutes depending on the dose, the nebuliser and the patient breathing pattern. 21 The serum half-life of iloprost is about 20–25 minutes and this short duration requires six to nine inhalation sessions per day.

Treatment is usually initiated under specialist care with the patient admitted to hospital for about 3 days for training, education and monitoring of self-delivery. Patients can return home once stabilised and trained. Patients receive two nebulisers (one as backup) and consumables are delivered regularly to their home. Nebulisers are replaced approximately every 2 years. Support from the specialist centre is readily available.

Length of treatment is patient specific and unless discontinued for other reasons will continue until the patient’s condition deteriorates and epoprostenol treatment (see Epoprostenol) is accepted by the patient and initiated.

Inhaled iloprost is often seen as an additional treatment to the oral therapies in this assessment, bridging the gap for those patients in whom oral interventions do not adequately reduce progression of disease, but who are either not so severely affected that epoprostenol treatment is indicated or not suitable for epoprostenol treatment.

Iloprost is contraindicated in patients with known hypersensitivity to the drug, conditions in which activity on platelets might be undesirable (e.g. active peptic ulcers, intracranial bleeds, trauma), severe coronary disease events (e.g. severe artery disease, angina, recent myocardial infarction), recent cerebrovascular events (e.g. stroke), pulmonary hypertension due to veno-occlusive disease, valvular defects with clinically relevant myocardial function disorders unrelated to pulmonary hypertension, pregnancy and lactation. Furthermore, iloprost is not recommended in patients with unstable pulmonary hypertension, with advanced right heart failure. 21

The cost of iloprost nebuliser solution is approximately £85–127 per day (one vial six to nine times per day; net price same for each vial size). 26 This price is for the solution only and does not include the nebulisers, consumables, service, delivery, inpatient time and training. The price for some of these items is difficult to ascertain and/or contained in confidential service agreements.

Bosentan

Bosentan is an orally administered dual ET_A and ET_B receptor antagonist. It has UK marketing authorisation for PAH to improve exercise capacity and symptoms in patients in FCIII. 22 Two tablet sizes are available: 62.5 mg and 125 mg.

Treatment should be initiated and monitored under specialist care. Initially, dosing is 62.5 mg twice daily (morning and evening with or without food) for 4 weeks, increased thereafter to a maintenance dose of 125 mg twice daily. Doses for some patients may be increased to 250 mg twice daily but this is rare.

Patients are usually admitted to hospital as day cases under specialist care for the initiation of treatment. Some education is also given. Patients return home and drugs are usually delivered to them at regular intervals.

Length of treatment is patient specific. Limited or no response after 8–16 weeks of treatment or a deterioration of condition at any time requires re-evaluation of treatment. This usually entails either the addition of, or the replacement with, other treatments. Withdrawal of bosentan requires careful management.

Bosentan is metabolised by the liver and has been associated with a dose-dependant increase in the liver enzymes aspartate aminotransferase and alanine aminotransferase (more than eight times the upper limit of normal in some cases). Such elevation can be the marker of potentially serious liver injury. This is reflected in the recommended maintenance dose of 125 mg twice daily rather than 250 mg twice daily. This is not a unique feature of bosentan as it also occurs with sitaxentan. Regular monitoring of hepatic enzymes (usually monthly) is required for as long as the drug is taken.

Bosentan is not indicated in patients with a known hypersensitivity to the drug, those with hepatic impairment (including aminotransferase levels more than three times the upper limit of normal) and those taking ciclosporin A (amplifies the plasma concentration of bosentan by an unknown mechanism). Bosentan is contraindicated in pregnancy as it is assumed to be teratogenic and therefore women with childbearing potential should not receive bosentan unless using reliable contraception (bosentan may interact and lessen the effectiveness of hormonal contraception).

The cost of bosentan tablets is approximately £55 per day (2 · 62.5 or 2 · 125 mg as the net price is the same for each tablet size). 26 This price is for the drug only and does not include delivery, monitoring of liver function or inpatient time, etc.

Sitaxentan

Sitaxentan is an orally administered selective receptor antagonist for ET_A (but not ET_B). It has EU marketing authorisation for PAH to improve exercise capacity in FCIII. 23 One tablet size is available: 100mg.

Treatment should be initiated and monitored under specialist care. Dosing is 100 mg once a day with or without food. Patients are usually admitted to hospital as a day case under specialist care for the initiation of treatment. Some education is also given. Patients return home and drugs are usually delivered to them at regular intervals.

Length of treatment is patient specific. Limited response after 24 weeks of treatment or a deterioration of condition at any time requires re-evaluation of treatment. 23 This usually entails either the addition of, or the replacement with, other treatments. Withdrawal of treatment requires careful management.

As with bosentan, sitaxentan is associated with effects on liver enzymes and these require regular monitoring, with subsequent treatment adjustment if elevated more than three times the upper limit of normal. 23

Contraindications are similar to those of bosentan. There is significant drug interaction between sitaxentan and warfarin. Reducing the dose of warfarin upon starting sitaxentan and regular monitoring of the INR is required to reduce the risk of bleeding.

The cost of sitaxentan is approximately £55 per day (1 · 100 mg; net price). 26 This price is for the drug only and does not include delivery, monitoring of liver function or inpatient time, etc.

Sildenafil

Sildenafil is an orally administered specific inhibitor of phosphodiesterase-5. It has UK/EU marketing authorisation for PAH to improve exercise capacity in patients in FCIII. 24 It is available as 20-mg tablets.

Treatment should be initiated and monitored under specialist care. Dosing is 20 mg three times per day (6–8 hours apart) with or without food. 24

Patients are usually admitted to hospital as a day case under specialist care for the initiation of treatment. Some education is also given. Patients return home and drugs are usually delivered to them at regular intervals.

Length of treatment is patient specific. Deterioration of condition at any time requires a re-evaluation of treatment. 24 This usually entails either the addition of, or the replacement with, other treatments. Withdrawal of treatment requires careful management.

Contraindications for sildenafil include hypersensitivity to the drug; use with nitric oxide-producing treatment or nitrates is not recommended as sildenafil potentiates the hypotensive effects of these agents. It is also contraindicated in patients with severe hepatic impairment, recent history of stroke or myocardial infarction, and severe hypotension at initiation. Furthermore it is contraindicated in some specific eye conditions. 24

The cost of sildenafil is approximately £12.45 per day (20 mg · 3; net price). 26 This price is for the drug only and does not include delivery, inpatient time and training, etc.

Outcomes of interest

Selected outcomes of interest were specified in the review protocol, based upon the final scope issued by NICE for this technology appraisal. They were:

• survival

• time to clinical deterioration (including switch of drug therapy and lung transplantation)

• health-related quality of life

• exercise capacity (6MWT)

• symptomatic improvement

• frequency and duration of hospitalisation and outpatient/GP visits

• SAEs

• adverse events that are considered as being clinically relevant or having a potential impact on tolerability

• withdrawal for any reason

• withdrawal because of lack of efficacy

• withdrawal because of adverse events

• haemodynamic assessment, e.g. cardiac index, RAP, pulmonary arterial oxygen saturation, PAP and PVR.

Of these, sufficient data were available from the included trials and meta-analyses were carried out for the following outcomes:

• dichotomous outcomes: death, clinical worsening (as defined in individual trials), symptomatic improvement (change in functional class), SAEs and withdrawal for any reason

• continuous outcomes: exercise capacity (6MWT), haemodynamic assessment including mPAP, RAP, PVR and cardiac index.

When data were available, narrative summaries were also provided in this review for time-to-event analyses of survival and clinical deterioration, and for other outcomes related to symptomatic relief (such as dyspnoea or fatigue) and health-related quality of life.

Individual adverse events were not meta-analysed as adverse event profiles varied between the interventions being assessed, and data on the severity or seriousness of specific adverse events were usually not provided. Withdrawal because of lack of efficacy and withdrawal because of adverse events were not separately analysed as it became apparent during data extraction that lack of efficacy of treatment in PAH naturally leads to adverse events associated with disease worsening. It was therefore not possible to attribute withdrawal to either lack of efficacy or adverse events in many cases and withdrawal for any reason would be a more appropriate outcome covering both. None of the included RCTs reported the frequency and duration of hospitalisation and outpatient/GP visits and pulmonary arterial oxygen saturation.

Handling of data and presentation of results

For dichotomous outcomes, results are presented as RRs. For continuous outcomes, results are presented as weighted mean differences (WMD).

Relative risks for ‘FC improved or maintained’ were initially calculated to provide more stable estimates as the proportion of patients with FC either improved or deteriorated was expected to be small. However, it was felt that ‘FC improved’ alone was also clinically important and thus RRs for this outcome were also calculated and presented. In addition, when data specifically for FCIII patients were available from the RCTs, odd ratios were compiled for ‘FC improved’ and ‘FC worsened’ at 12 weeks to inform the independent economic assessment (see Chapter 4).

For outcomes with continuous data the values of mean change from baseline (i.e. mean value measured at the end of the trial minus the mean value measured at baseline) were used in meta-analysis. When possible the standard deviation (SD) was taken directly from the reported results or derived from the standard error of the mean (SEM) or CIs. For the 6MWD data of Barst et al. ,11 SDs for mean change from baseline were imputed using the SDs of baseline values and SDs of post-treatment values assuming an intercorrelation coefficient of 0.5. 32 For the 6MWD data of Badesch et al. ,33 for which only the SD for the post-treatment value was available, it was used as the SD for the mean change from baseline.

Approaches for meta-analysis

Meta-analyses were carried out using review manager version 4.2. Separate analyses were performed for each of the interventions being considered for the outcomes specified above. The primary analysis included data for licensed doses only (where appropriate) for patients with PAH (all subcategories in category 1 of the Venice 2003 clinical classification excluding the subcategory 1.5, persistent pulmonary hypertension of the newborn) in NYHA/WHO FCIII (and FCIV for epoprostenol) using the latest follow-up data available from the randomised controlled period of each trial. A random-effects model was used given the heterogeneous populations within PAH. Comparisons were made separately for:

• each of the interventions versus placebo/nothing with ongoing supportive treatments

• each of the interventions versus placebo/nothing with another ongoing intervention and ongoing supportive treatment (trials were available for iloprost versus placebo/nothing with ongoing bosentan and supportive treatment; and sildenafil versus placebo with ongoing epoprostenol and supportive treatment)

• comparison of the interventions against each other (trials were available for sitaxentan versus bosentan and sildenafil versus bosentan)

• comparison between different combinations of interventions (one trial was available for epoprostenol plus bosentan versus epoprostenol).

No indirect comparisons or mixed treatment comparisons were planned or performed.

Given the expected discrepancy between the scope of this technology appraisal (specific types of PAH, FC and dose for each of the drugs within their licensed indication) and the heterogeneous trial evidence actually available for each drug, several sensitivity analyses taking into account the population mix in terms of FC and pulmonary hypertension categories, intervention doses, trial design and data status, as well as subgroup analyses for IPAH and PAH associated with connective tissue disease (CTD-APAH), were planned. The primary analysis (analysis A) and other planned analyses (analyses B–H) are listed in Table 6. Whether each of the listed analyses was actually carried out depended on the availability of data, and this is stated in an analyses checklist within each of the specific comparisons. The aims of these analyses were to ensure that available evidence that was directly applicable to this technology appraisal (or the lack of such evidence) was highlighted, while allowing other potentially relevant evidence to also be considered.

Assessment of heterogeneity

Statistical heterogeneity between studies was assessed with the chi-squared test and I². The I² is a measure of inconsistency in studies’ results in meta-analysis. 34 It describes the percentage of the total variation across studies that is due to heterogeneity rather than to chance (sampling error), and lies between 0% (no observed heterogeneity) and 100% (significant heterogeneity). An I² of 25%, 50% and 75% would indicate low, moderate and high heterogeneity respectively. When there was evidence of statistical heterogeneity (p ≤ 0.10 for chi-squared test for heterogeneity or I² ≥ 50%), the values of I² were shown besides the pooled estimates within the results tables and the heterogeneity was discussed in the text. I² was reported for all of the pooled estimates quoted in the texts, irrespective of its value.

Assessment of publication bias

All manufacturers were requested to provide a list of all company-sponsored RCTs that were relevant to this appraisal. Requests were also made for reports of unpublished trials and data that are potentially available but not reported in published papers. Given that the lists of RCTs were provided by all of the companies and the number of trials for each of the technologies was small, publication bias was not formally assessed.

Quantity and quality of included studies

Three RCTs11,33,39,40 compared epoprostenol (added to supportive treatment) with supportive treatment. [An additional study (BREATHE-2; Bosetan Randomised Trial of Endothelin Antagonist Therapy)56 that compared the initiation of bosentan–epoprostenol combination with epoprostenol alone will be described in the bosentan section and a further study (PACES-1)36 that compared sildenafil with placebo in patients who were stable on epoprostenol treatment will be described in the sildenafil section.] In addition to the main publications associated with these trials, further data not included in these publications were available from the Cochrane review by Paramothayan and colleagues. 60 The clinical study report for one of the RCTs11 was made available to the assessment group by GlaxoSmithKline.

The characteristics of the three trials are summarised in Table 8. All were industry-sponsored multicentre studies conducted in the USA. The number of patients randomised ranged from 2339 to 11133 and study duration was between 839 and 1211,33 weeks. Rubin et al. 39 and Barst et al. 11 recruited exclusively patients with PPH, whereas Badesch et al. 33 recruited exclusively PAH patients with a scleroderma spectrum of disease. All three trials included patients with mixed FC, with 65–78% of patients in FCIII and 17–26% of patients in FCIV at baseline. The mean/median 6MWD at baseline was less than 300 metres in all three trials. The primary end point was change in 6MWD for Barst et al. 11 and Badesch et al. ,33 and was not stated for Rubin et al. 39

Quality assessment of these trials is summarised in Table 9. All of the trials were open-label studies as a double-blind, placebo-controlled design was not considered possible because of the known incidence of sepsis caused by central venous catheters and the unique or highly predictable symptoms during long-term epoprostenol treatment. 11 However, assessors for the 6MWT were blinded in Barst et al. 11 and Badesch et al. 33 With the exception of survival and 6MWD in Barst et al. 11 and 6MWD in Badesch et al. ,33 ITT analysis was not used. Treatment withdrawal/loss to follow-up was not clearly reported in Rubin et al. 39 and Badesch et al. 33

Epoprostenol (added to supportive treatment) versus supportive treatment

Planned meta-analyses for this comparison and those actually carried out are summarised in Table 10.

As outcome data stratified by FC were available neither from published papers nor from the clinical study report it was not possible to perform the planned primary analyses [analysis for PAH, by FC (FCIII and FCIV), treated with licensed doses]. Furthermore, some other planned analyses were also not possible or not required. The reasons for these are also given in Table 10.

All of the findings presented in this section are on analyses that could be performed and these are associated with patient populations of mixed FC (III and IV). The results of meta-analyses (or of individual trials when only one trial provided the data) are listed in Table 11. Results for individual outcomes are summarised in the following subsections.

Exercise capacity

The mean changes from baseline for the 6MWD for the three trials11,33,39 are shown in Figure 4. Improvements were seen in all three trials and the pooled result for the WMD was an increase of 81 metres (95% CI 45–117, I² = 25%) for epoprostenol groups compared with control groups.

FIGURE 4.

Forest plot: epoprostenol added to supportive treatment versus supportive treatment alone – change in 6-minute walk distance. Note: The standard deviations for Barst et al.11 and Badesch et al.33 were not reported and were imputed using the methods described in the section on handling of data and presentation of results. CI, confidence interval; CTD, connective tissue disease; IPAH, idiopathic pulmonary arterial hypertension; PAH, pulmonary arterial hypertension; SD, standard deviation; WMD, weighted mean difference.

Quantity and quality of included studies

Two RCTs [AIR (Aerosolized Iloprost Randomised study),41 AIR-236] compared iloprost (added to supportive treatment) with supportive treatment. The AIR-2 study36 was identified through industry submission and Schering Health Care provided the assessment group with an unpublished manuscript. The study had not been published at the completion of this report and data from the manuscript are considered academic-in-confidence. Two further RCTs [COMBI (Combination Therapy of Bosetan and Aerosolized Iloprost in Idiopathic Pulmonary Arterial Hypertension),58 STEP (Safety and Pilot Efficacy Trial in Combination with Bosetan for Evaluation in PAH)59] compared iloprost (added to ongoing bosentan therapy and supportive treatment) with ongoing bosentan therapy and supportive treatment.

The characteristics of these four trials are summarised in Table 12. All were industry-sponsored multicentre studies (COMBI58 was investigator-initiated, but was supported by the manufacturer58). The AIR study41 was a multinational study conducted in Europe and was the pivotal trial for iloprost. The AIR-236 and COMBI58 trials were conducted in Germany and the STEP study59 was conducted in the USA. The number of patients randomised ranged from 4058 to 20341 and the duration of study was 12 weeks for all four trials.

Both the AIR and the AIR-2 studies were carried out in mixed populations including IPAH and other PAH within category 1 of the Venice classification, as well as other pulmonary hypertension (mainly chronic thrombolic, Venice category 4). The COMBI study recruited exclusively IPAH patients and the STEP study included mixed PAH populations (all within Venice category 1). The trials were also different in the mix of patients in terms of baseline FC: the AIR study had mixed FCIII and FCIV patients, whereas the AIR-2 study also included patients in FCII. The COMBI study recruited exclusively patients in FCIII. The vast majority of patients in the STEP trial were also in FCIII at baseline. Mean 6MWD at baseline was lowest in the COMBI study (306 metres)58 and was highest [Academic-in-confidence information has been removed].

With regard to end points, the AIR study41 used a composite end point of ‘at least a 10% increase in 6MWD and improvement in FC without deterioration’ as the primary outcome. Change in 6MWD was the primary outcome for the COMBI study,58 whereas the AIR-236 and the STEP59 trials did not clearly state their primary end points.

Quality assessment of these trials is summarised in Table 13. The AIR and the STEP studies were double-blind, placebo-controlled trials, whereas the AIR-2 and the COMBI trials were open-label studies. Methods of randomisation and allocation concealment were adequate except in the COMBI study in which the method of randomisation was unclear. Neither of the open-label studies36,58 mentioned blinding of outcome assessors. Only the COMBI study used ITT analysis across all of the outcomes examined. The AIR study used ITT analysis for its primary composite end point, clinical worsening and 6MWD, but not for changes in FC and other measures. ITT analysis was not used in the AIR-2 and the STEP trials. As more patients from the iloprost arms than from the control arms were excluded from the analysis in these two studies there was a potential bias in favour of iloprost (if excluded patients had poorer outcomes) in these studies.

The results of the AIR and AIR-2 studies, and of the COMBI and STEP studies are described separately in the following sections given the different nature of the comparisons between the trials. Because of the relatively short half-life of iloprost (hence short acute effect) and the intermittent nature of drug inhalation (as opposed to continuous infusion in the case of epoprostenol), studies of iloprost frequently measure treatment effect both before and after iloprost inhalation, which corresponds to the expected trough and peak drug concentration/effect. The post-inhalation measures (which represent acute effects) are used in the analysis in this review, although relevant findings from preinhalation measures (which represent chronic effects) are also discussed.

Iloprost added to supportive treatment versus supportive treatment alone

This comparison was investigated in the AIR and AIR-2 studies. Planned meta-analyses for the comparison and those actually carried out are summarised in Table 14. Because both trials included non-PAH patients, and outcome data excluding these patients and stratified by FC were not available, it was not possible to perform the planned primary analysis (all PAH, FCIII). However, various sensitivity analyses and limited subgroup analysis were carried out to summarise the available evidence, which may help inform the technology appraisal. The results of meta-analyses (or of individual trials when only one trial provided the data) are listed in Table 15. Results for individual outcomes are summarised in the following subsections. As there was a paucity of results that were directly applicable to iloprost’s licensed indication (PPH, FCIII), findings presented here were mainly based on the overall results of the AIR and AIR-2 studies, which included patients with mixed pulmonary hypertension and FC.

Iloprost added to ongoing bosentan and supportive treatment versus ongoing bosentan and supportive treatment

This comparison was investigated in the COMBI58 and STEP59 studies. Planned meta-analyses for this comparison and those actually carried out are summarised in Table 16. Because all patients in the COMBI study and the vast majority of patients in the STEP study were in FCIII at baseline, both studies were included in the primary analysis (all PAH, FCIII) and no sensitivity analysis including mixed FC was performed. The results of meta-analyses (or of individual trials when only one trial provided the data) are listed in Table 17. Results for individual outcomes are summarised in the following subsections.

Exercise capacity

The mean changes from baseline for the 6MWD (post inhalation) for the two trials are shown in Figure 6. The mean 6WMD for the iloprost group compared with the placebo/control group increased by 26 metres in the STEP trial,59 but decreased by 10 metres in the COMBI trial. 58 Neither difference was statistically significant. The difference between treatment groups was smaller (18 metres) when 6MWD was measured preinhalation in the STEP trial. 59

FIGURE 6.

Iloprost added to ongoing bosentan and supportive treatment versus ongoing bosentan and supportive treatment – change in 6-minute walk distance. CI, confidence interval; FC, functional class; PAH, pulmonary arterial hypertension; SD, standard deviation; WMD, weighted mean difference.

Quantity and quality of included studies

Bosentan was investigated in six of the included RCTs. Four of these [Channick et al. ,15,43 BREATHE-1,45,46 BREATHE-547 and STRIDE (Sitaxsentan to Relieve Impaired Exercise study)-248] allowed the comparison between bosentan and placebo with ongoing background therapy. Another trial (BREATHE-256) compared the combination of epoprostenol plus bosentan with epoprostenol alone. The characteristics of these five studies are summarised in Table 18.

Bosentan was compared with sitaxentan in STRIDE-248 and with sildenafil in a further study by Wilkins and colleagues. 57 These direct comparisons will be described separately.

All of the five studies shown in Table 18 were industry-sponsored international studies (STRIDE-2 was sponsored by the manufacturer of sitaxentan). The number of patients randomised (excluding the sitaxentan arms in STRIDE-2) ranged from 3243 to 21345 and the duration of study was 12 weeks,43 18 weeks48 or 16 weeks. 45,47,56 The bosentan dose of 125 mg twice daily was used in all of the trials. In addition, BREATHE-1 also included the dose of 250 mg twice daily. In line with bosentan’s license, an initiation dose of 62.5 mg twice daily for the first 4 weeks was used before patients were uptitrated to the targeted doses in all trials.

The patient populations varied between trials in terms of PAH subcategories and FC. With the exception of BREATHE-5,47 which recruited exclusively patients with Eisenmenger syndrome, all of the trials included a mixed population of IPAH (59–84% within each trial) and CTD-APAH (16–30% within each trial). STRIDE-2 also included 11% of patients with CHD. The submission to NICE by Actelion indicates that patients with PAH associated with CHD (postsurgically corrected) were enrolled in BREATHE-1, although this was not stated in the published papers. 45,46 The Actelion submission also states that:

… as the post-surgical CHD patient numbers were small and since they are believed by clinicians to act like IPAH patients, the CHD patients were grouped with the IPAH patients for all analyses. Within this submission, the CHD group is never separated out from the IPAH sub-group, but is implicitly included.

Two43,47 of the five trials recruited exclusively patients in FCIII. Most of the patients in the other three studies were also in FCIII. STRIDE-248 and BREATHE-145 included a small proportion of patients in FCIV (4% and 8% respectively), whereas nearly one-quarter of patients in BREATHE-256 were in FCIV at baseline. STRIDE-248 also included a significant proportion of patients in FCII at baseline (37%). Baseline 6MWD was not reported in BREATHE-2 and was fairly similar for the other four trials (range 334–358 metres) despite the differences in FC mix. The primary outcome measure was change in 6MWD for Channick et al. ,43 BREATHE-145 and STRIDE-2;48 change in total pulmonary resistance (determined by right heart catheterisation) for BREATHE-2;56 and change in systemic pulse oximetry for BREATHE-5. 47

Quality assessment of the five trials is summarised in Table 19. The method of randomisation was adequate in Channick et al. ,43 BREATHE-547 and STRIDE-248 and was not clearly described in the published papers for BREATHE-145,46 and BREATHE-2. 56 Allocation concealment was also adequate in BREATHE-547 and STRIDE-248 and was not clearly described in the other three trials. All of the five trials were double-blind studies, except that the bosentan arm in the STRIDE-2 trial was open-label (the only open-label arm in the trial). The investigators stated that this was because ‘bosentan was only available commercially on a named-patient basis and blinded drug supplies were not available’. Nevertheless, the assessors for 6MWT, FC assessments and Borg dyspnoea scores were blinded. ITT analysis was used in all trials for most outcomes, but not for haemodynamic measures. The proportion of patients completing each study was slightly lower in the placebo group than in the bosentan group, except for the BREATHE-2 study in which a slightly lower proportion of patients treated with bosentan (plus epoprostenol) completed the trial compared with those treated with placebo (plus epoprostenol).

Given the different nature of the comparisons between the trials, the results of Channick et al. 43 and the BREATHE-1,45 BREATHE-547 and STRIDE-2 (bosentan versus placebo only)48 studies will be described separately from the results of the BREATHE-2 study.

Bosentan added to supportive treatment versus supportive treatment alone

This comparison was investigated in Channick et al. ,43 BREATHE-1,45 BREATHE-547 and STRIDE-2. 48 Planned meta-analyses for the comparison and those actually carried out are summarised in Table 20. As no data stratified by FC were available from BREATHE-1 (the largest among the bosentan trials) and the only stratified data available from STRIDE-2 were for change in FC, the planned primary analysis (all PAH, FCIII) included data only from Channick et al. 43 and BREATHE-547 for most outcomes. If the stratified data were available, 195 out of 213 patients in BREATHE-1 and 72 out of 122 patients in STRIDE-2 receiving either bosentan or placebo would have also been included in this analysis. However, sensitivity analyses including populations of mixed FC from these two trials were carried out. The results of meta-analyses (or of individual trials when only one trial provided the data) are listed in Table 21. Results for individual outcomes are summarised in the following subsections. Given the relatively small number of patients included in the primary analysis, results presented are mainly drawn from data of mixed FC. Findings specifically for FCIII are stated separately when appropriate.

When data from BREATHE-145 were included, the results from the two bosentan arms (125 mg twice daily and 250 mg twice daily) in the trial were combined unless otherwise specified. When STRIDE-248 is mentioned in this section it is only referred to with regard to data from the placebo and bosentan arms.

Exercise capacity

The mean changes from baseline for the 6MWD for the four trials43,45,47,48 (mixed FC) are shown in Figure 7. A significant increase in 6MWD for the bosentan groups compared with the placebo groups was observed in all trials, including those by Channick et al. 43 and BREATHE-5,47 which recruited only patients in FCIII. The pooled result from these two studies (analysis A, Table 21) was 59 metres in favour of bosentan (95% CI 20–99, I² = 0%).

FIGURE 7.

Forest plot: bosentan added to supportive treatment versus supportive treatment – change in 6-minute walk distance. CI, confidence interval; SD, standard deviation; WMD, weighted mean difference.

Bosentan plus epoprostenol versus epoprostenol

This comparison was investigated in the BREATHE-2 trial. 56 This trial was the only study included in this review that genuinely compared the initiation of a drug combination with a single drug (rather than comparing the addition of a drug to placebo with another ongoing drug). The characteristics and quality assessment of the study are shown in Tables 18 and 19 respectively. Methods of randomisation and allocation concealment were not clearly reported, and ITT analysis was not used for 6MWD (patients who were unable to or who did not perform the assessment were excluded from the analysis). Although the majority of patients included in this study were in FCIII at baseline, nearly one-quarter of the patients (8/33) were in FCIV. As data stratified by FC were only available for FC improvement, results described in this section were mainly based on patients with mixed FC. Given that the findings were from a single trial with a relatively small sample size (n = 33), only a narrative summary of the study findings will be provided.

Quantity and quality of included studies

Sitaxentan was investigated in three of the included RCTs. All three of the trials (STRIDE-1,49 STRIDE-2,48 STRIDE-437) compared sitaxentan with placebo in patients with ongoing supportive treatment. The STRIDE-2 trial48 also included an open-label bosentan arm. The bosentan–placebo comparison from this trial has already been described; the bosentan–sitaxentan comparison is described later in this chapter [see Direct (head-to-head) comparisons]. This section focuses on the comparison of sitaxentan added to supportive treatment versus supportive treatment alone. The characteristics of the three studies are summarised in Table 22. The STRIDE-2 trial48 has been listed in relevant tables in the section on bosentan, but is also listed in this section for the convenience of readers.

All three studies37,48,49 were industry-sponsored multicentre trials that randomised between 98 and 247 patients. The STRIDE-1 study was conducted in North America. The STRIDE-2 study was an international study and the STRIDE-4 trial was conducted mainly in South America, but also in Spain and Poland. The clinical study reports (commercial-in-confidence) for all three trials were made available to the assessment group by Encysive. The study duration was 12 weeks for STRIDE-149 and 18 weeks for STRIDE-248 and STRIDE-4. 37 The licensed dose (100 mg once daily) for sitaxentan was investigated in all three trials. In addition, STRIDE-149 included an above licensed dose of 300 mg once daily (included only in the relevant sensitivity analysis in this review) and STRIDE-248 and STRIDE-437 included a sublicensed dose of 50 mg once daily (not considered in this review).

All three trials recruited mixed PAH populations of IPAH (ranging from 53% in STRIDE-1 to 68% in STRIDE-4), CTD-APAH (15–30% within each trial) and CHD (11–24% within each trial). The majority of patients in STRIDE-1 and STRIDE-2 were in FCIII at baseline (66% and 59% respectively), whereas in STRIDE-4 only 38% were in FCIII at baseline (the majority being in FCII: 61%). The primary end point was per cent of predicted peak oxygen uptake (VO₂) in STRIDE-1 and change in 6MWD in STRIDE-2 and STRIDE-4.

Quality assessment of the three trials is summarised in Table 23 (only information relevant to the placebo and sitaxentan arms is listed). Methods of randomisation and allocation concealment were adequate in all three trials. 37,48,49 ITT analysis was used for most of the outcomes in STRIDE-149 and STRIDE-4. 37 STRIDE-248 excluded a small number of patients ([Academic-in-confidence information has been removed]) without a valid post-baseline 6MWT.

Sitaxentan (added to supportive treatment) versus supportive treatment

This comparison was investigated in all three trials. 37,48,49 Planned analyses and those actually carried out are summarised in Table 24. Results of the meta-analysis are listed in Table 25 according to planned comparisons. Results for individual outcomes are described in the following subsections. As all three trials included patients with mixed FC, and data stratified by FC were available only for the outcome of change in FC, the findings presented in this section are mainly based on meta-analysis results of mixed PAH populations. Findings specifically for FCIII are stated separately when appropriate.

Exercise capacity

The mean changes from baseline for 6MWD for the three trials37,48,49 are shown in Figure 8. Sitaxentan at the licensed dose significantly increased the 6MWD compared with placebo (32 metres, 95% CI 18–47, I² = 0%) in patients with mixed FC. Data specifically for FCIII patients were not available.

FIGURE 8.

Sitaxentan added to supportive treatment versus supportive treatment alone – change in 6-minute walk distance. CI, confidence interval; FC, functional class; PAH, pulmonary arterial hypertension; WMD, weighted mean difference.

Quantity and quality of included studies

Sildenafil was investigated in six of the included RCTs. Four of these [SUPER-1 (Sildenafil Use in Pulmonary Arterial Hypertension study),53 Bharani et al. ,35 Sastry et al. ,54 Singh et al. 55] compared sildenafil with placebo in patients with ongoing supportive treatment (patients in Bharani et al. 35 appeared to have stopped previous vasodilator therapy before entering the study). Another trial (PACES-138), identified through manufacturer submission, compared sildenafil with placebo in patients with ongoing epoprostenol and supportive treatment. The characteristics of these five studies are summarised in Table 26. Sildenafil was compared with bosentan in a further study by Wilkins and colleagues (SERAPH),57 which will be described later in this chapter [see Direct (head-to-head) comparisons].

The SUPER-153 and PACES-138 trials were industry-sponsored international studies that randomised 278 and 267 patients respectively. The clinical study reports (commercial-in-confidence) for both trials were made available to the assessment group by Pfizer. The study duration was 12 weeks for SUPER-153 and 16 weeks for PACES-1. The studies by Bharani et al. ,35 Sastry et al. 54 and Singh et al. 55 were small (n = 10, 22 and 20 respectively) single-centre, cross-over trials conducted in India. The study by Sastry and colleagues54 was sponsored by a not-for-profit organisation and the sponsorship for Bharain et al. 35 and Singh et al. 55 was not reported. The cross-over trials had a duration of 235–654,55 weeks for each treatment period. The doses investigated in these trials varied, but only the SUPER-1 study53 included a treatment arm using the licensed dose (20 mg three times daily). Above licensed doses of up to 80 mg three times daily were also investigated in the SUPER-1 trial53 and in all of the other trials (see Table 26).

Both the SUPER-153 and PACES-138 trials recruited mixed PAH populations of IPAH and CTD-APAH. The SUPER-1 study also included 6% of patients with CHD. The majority of patients in both trials were in FCIII at baseline; there were 39% and 26% of patients in FCII at baseline for the SUPER-1 and PACES-1 trials respectively. The primary end point was change in 6MWD for both trials. Bharani et al. 35 recruited patients in FCII–IV at baseline with various types of pulmonary hypertension including PPH, PAH associated with Eisenmenger syndrome and other forms of pulmonary hypertension. Sastry et al. recruited exclusively PPH patients, the majority of which (82%) were in FCII at baseline. 54 Singh et al. 55 recruited mixed populations of IPAH and Eisenmenger syndrome patients and the study included a significant proportion of children (as young as 3 years old). Given the large proportion of the study populations being outside sildenafil’s licensed indication in Bharani et al. ,35 Sastry et al. 54 and Singh et al. ,55 and their small sample sizes, the characteristics and study results of these three trials are only briefly listed/mentioned in the following sections and data from these studies were not meta-analysed.

Quality assessment of the five trials is summarised in Table 27. Both SUPER-153 and PACES-138 used adequate methods of randomisation and allocation concealment. ITT analysis was not used as the primary analysis, but was used as sensitivity analysis for a few outcomes. The proportion of patients who completed the trials was similar between treatment arms in SUPER-1, and was [Commercial-in-confidence information has been removed].

Because of the different nature of the comparisons between the trials, the results of SUPER-1 and PACES-1 will be described separately [see Direct (head-to-head) comparisons].

Sildenafil added to supportive treatment versus supportive treatment alone

This comparison was investigated in SUPER-1,53 Bharani et al. ,35 Sastry et al. 54 and Singh et al. 55 As previously stated, results from the last three trials will only be briefly mentioned and will not be combined with results from SUPER-153 because of the minimal relevance of their study populations to this technology appraisal. The findings presented in this section are therefore mainly based on a single trial (SUPER-153) rather than meta-analysis, but results will be presented in a format similar to previous sections. Planned comparisons and those actually available are summarised in Table 28. The results from SUPER-1 are listed in Table 29 according to planned comparisons. Results for individual outcomes are described in the following subsections. As data stratified by FC were only available for the outcome of change in FC, results described were mainly drawn from data of mixed FC. Findings specifically for FCIII are stated separately when appropriate.

Sildenafil added to ongoing epoprostenol versus ongoing epoprostenol

This comparison was investigated in the PACES-1 trial. 38 This study remained unpublished at the time of completion of this technology assessment. Data presented in this section were largely obtained from Pfizer’s submission to NICE for this technology appraisal. Additional data were sought from the clinical study report (commercial-in-confidence) of this study, which was made available to the assessment group by Pfizer. The characteristics and quality assessment of the PACES-1 study are shown in Tables 26 and 27 respectively.

As the results presented in this section were based on a single trial and the dose of sildenafil used (80 mg three times daily) was above its licensed dose, only a narrative summary of the study findings is provided. Most of the findings were based on the whole trial population, which included mixed types of PAH (IPAH and CTD-APAH) and mixed FC (I–IV).

Quantity and quality of included studies

Direct comparisons between the five technologies under assessment in this review were made in two of the included RCTs. Bosentan was compared with sitaxentan (both at the licensed dose) in the STRIDE-2 study. 48 The comparisons in this trial of both drugs with placebo have been included in the bosentan and sitaxentan sections respectively. Bosentan (licensed dose) was compared with sildenafil (above licensed dose) in a further study (SERAPH) by Wilkins and colleagues. 57 The characteristics of these two studies are summarised in Table 30 (STRIDE-248 is listed again for the convenience of readers; only relevant treatment arms are listed). Additional comparisons involving combinations of the technologies under assessment (iloprost added to ongoing bosentan versus ongoing bosentan; epoprostenol plus bosentan versus epoprostenol alone) have been described in the iloprost and bosentan sections respectively.

The study characteristics and quality assessment of the STRIDE-2 study48 have been described in previous sections. The SERAPH study57 was a 16-week, single-centre trial that randomised 26 patients. It was conducted in the Hammersmith Hospital, London and was funded by the British Heart Foundation. The dose of sildenafil used was 50 mg twice daily for the first 4 weeks, uptitrated to 50 mg three times daily thereafter (above licensed dose). The trial recruited exclusively patients in FCIII at baseline and the majority of patients had IPAH (23/26) with the rest having CTD-APAH. The mean 6MWD at baseline was 297 metres. The primary end point for SERAPH was change in right ventricular mass from baseline as measured by cardiovascular magnetic resonance. 57

The methods of randomisation and allocation concealment in the SERAPH study were adequate. Both ITT analysis and per-protocol analysis (including patients who completed the trial) were used. Except for one patient who died during the study, no patient was withdrawn from treatment. 57 The quality assessment for STRIDE-2 and SERAPH is summarised in Table 31.

Sitaxentan versus bosentan with ongoing supportive treatment

The results presented in this section are from STRIDE-248 rather than from meta-analysis, as this study was the only one that investigated this comparison. Planned comparisons and those actually performed are summarised in Table 32. Results are listed in Table 33 according to the planned comparisons, followed by a paragraph summarising the findings. As STRIDE-2 included patients with mixed FC, and data stratified by FC were available only for the outcome of change in FC, the findings presented are mainly based on the results of mixed PAH populations. Findings specifically for FCIII are stated separately when appropriate.

Sildenafil versus bosentan with ongoing supportive treatment

This comparison was investigated in the SERAPH study. 57 As the dose used in the study was above the licensed dose and the number of patients randomised was small (n = 26), only a narrative summary is provided below.

Survival

All of the RCTs included in this review were of a duration of 18 weeks or less. Death generally occurred more frequently in the placebo/control groups than in the treatment groups, but the numbers were very small within each trial. The epoprostenol trial by Barst and colleagues11 was an exception and was the only RCT that demonstrated a significant survival benefit within a trial. The pooled RRs for death were in favour of each of the technologies (except for sildenafil for which the result was based on two deaths in a single trial), but did not reach statistical significance as confidence intervals were wide.

A recent meta-analysis63 of treatments for pulmonary hypertension (i.e. a wider population than just PAH) reported a RR of death of 0.70 (95% CI 0.41–1.22) compared with control. The analysis pooled all disease-modifying technologies for pulmonary hypertension including those outside this technology assessment. The merits of this are debatable. However, the estimated non-significant 30% reduction in mortality led the authors to question the survival benefit offered by the technologies for pulmonary hypertension, and in particular whether the trials were adequately powered or of a long enough duration to adequately measure survival. The findings of this assessment report are in agreement with this conclusion.

Despite some methodological issues with this meta-analysis63 (e.g. the head-to-head trial SERAPH57 was included in the analysis with the bosentan arm being treated as a control), its finding is consistent with the finding of this technology assessment in that the overall direction of effect was in favour of active treatments and was consistent across different types of drugs. The key questions are therefore whether the magnitude of the effect varies between drugs and whether it changes over time. Unfortunately these questions are unlikely to be answerable with existing evidence because of the small numbers of deaths that occurred in the RCTs and their short duration. Increasing evidence from long-term observational studies (see Appendix 7) agrees with the potential survival benefit of these treatments observed during short-term trials, but unbiased comparison between drugs using observational data is difficult to achieve because of differences in patient populations, entry criteria, treatments offered and methods of follow-up.

In addition to the small numbers of patients/events and short duration of study, interpretation of results from RCTs in relation to mortality as well as other outcomes needs to take into account the following issues.

Clinical worsening

Clinical worsening events were not defined or reported in any of the epoprostenol trials. A significant reduction in clinical worsening events and/or an increase in time to clinical worsening was demonstrated in individual trials of bosentan,43,45 the pooled results of sitaxentan trials and the pooled results of bosentan trials (excluding STRIDE-248). Fewer clinical worsening events occurred in the active treatment arms than in the placebo arms in the pivotal trials for iloprost (AIR study41) and sildenafil (SUPER-153), but the results did not reach statistical significance.

Most points already discussed in relation to survival are also applicable to clinical worsening events. An additional issue for this outcome is that different definitions of clinical worsening have been used in different trials. It is likely that both the severity of disease in study participants at baseline and the definition of clinical worsening adopted in the trials had an impact (in addition to the treatment effect) on the event rates for clinical worsening. Standardisation of the definition in future trials would be helpful for comparison of results between trials.

Changes in functional class

All five technologies demonstrated a significant benefit compared with placebo/control with regard to having an improved FC on treatment. One issue that needs to be highlighted in relation to this outcome is the substantial variations between trials in the response rates in the placebo/control groups. The differences between trials may partly be attributable to the differences in the mix of FC at baseline. However, even when limiting the data to patients in FCIII at baseline, the proportion of patients having their FC improved still ranged from 29% (19/65) in BREATHE-145 (bosentan trial) to 6% (2/36) in the AIR study41 (iloprost trial). Again, the differences in trial populations and standard of (supportive) care need to be considered when interpreting the results. Within the context of RCTs, the nature of FC being a subjective outcome means that there was a possibility of misclassification of FC at baseline. Interpretation of baseline FC and outcomes related to FC changes should therefore be made in conjunction with objective outcome measures such as 6MWD.

The variations in the response rates in the placebo/control groups between trials mean that calculating the NNT, assuming a common ‘baseline risk’ across different technologies or even within a drug, may be problematic. The NNTs presented in Table 34 were calculated according to the pooled risk differences of the trials contributing to the data for each technology. Therefore they should not be compared against each other and should be interpreted with great caution, particularly when substantial heterogeneity in risk differences is shown (I² ≥ 50%).

In contrast to the significant effect on improving FC, most technologies, except sitaxentan, failed to demonstrate a statistically significant effect on having FC improved or maintained (i.e. not worsened). This was unexpected and was in part due to some of the trials failing to report changes in FC other than improvement (i.e. proportion of patients having FC maintained or worsened), and consequently the smaller number of patients included in the analysis of this outcome. Furthermore, it is likely that the exclusion of patients with unstable conditions from these trials made it more difficult for these studies to demonstrate the benefit of reducing FC deterioration within the short duration of the trials.

6-Minute walk distance

All five technologies demonstrated a significant effect compared with placebo/control on increasing exercise capacity as measured by 6MWT. The mean difference between the treatment and placebo/control groups appeared to be greatest in two of the epoprostenol trials33,39 (approximately 100 metres). The between-group difference varied from approximately 30 to 75 metres in the other trials, with wide confidence intervals because of the high variability between individual patients. Again, values from different trials are not directly comparable because of differences in patient populations in terms of types of PAH, baseline FC and exercise capacity. For example, a ceiling effect in 6MWD (those patients with milder disease/higher baseline 6MWD had less scope for improvement) was observed in a post hoc analysis of the STRIDE-1 study (sitaxentan trial). 50,51

Whether ITT analysis was used and the methods for imputing missing data can also have a substantial impact on the reported group means and differences for 6MWD, particularly in trials of small sample size. Excluding patients who had no post-baseline 6MWT would almost certainly bias the results. Different methods of imputing data (e.g. last observation carried forward; assuming no change compared with baseline; assuming a 6MWD of 0 for missing observations), however, could produce different results. Interpretation of trial results and comparison between studies therefore requires great caution.

Quality of life

The volume of evidence from RCTs with regard to the impact of treatment on health-related quality of life varied between technologies. No data were reported in the trials for bosentan. Other trials used different tools and the findings seemed inconclusive (see Table 34). Two studies measured the EQ-5D utility index (AIR study for iloprost;41 SUPER-1 study for sildenafil53); there was an improvement of approximately 0.1 (on a scale of 0–1) compared with placebo in both trials.

Haemodynamic measures

All five technologies demonstrated significant effects on haemodynamic measures, which are important indicators of PAH disease progression and/or survival. 5,65,66 In the context of clinical trials these measures were most susceptible to missing observations and were usually not analysed by ITT. Consequently, statistically significant findings in these outcomes may better be treated as ‘proof of concept’ for PAH treatment. The magnitude of the reported group means and differences, however, may not be clinically useful.

Serious adverse events and withdrawal for any reasons

The potential harm associated with any treatment for PAH needs to be weighed against the potential worsening of the disease without treatment. Worsening of PAH frequently incurs events that are classified as SAEs and may require withdrawal from RCTs. Effective treatments for PAH with an acceptable safety profile would therefore be expected to demonstrate a reduced risk of SAEs and withdrawal for any reason compared with placebo/control.

Data on the total number of patients experiencing at least one SAE were not available for the three epoprostenol trials11,33,39 and two of the bosentan trials. 43,45 The pooled data for the other two bosentan trials47,48 showed a significant reduction in the risk of experiencing at least one SAE for bosentan compared with placebo, but the pooled results (or result from the only trial) for other technologies failed to demonstrate a significant risk reduction. A significant reduction in the risk of withdrawal for any reason was observed in individual trials for epoprostenol11 and iloprost41 and in the pooled results for sitaxentan.

Poor reporting of the outcomes and the small number of patients with the events are possible reasons for the lack of consistent findings for these outcomes across technologies (contrary to some of the efficacy outcomes). Further evidence from comparative trials is needed as these drugs have different adverse effect profiles and differences between drugs in the maintenance of patients on treatment and the overall risk–benefit profile cannot be ruled out.

Economic evaluations

A total of four economic evaluations meeting the inclusion criteria were identified, none of which were a UK study. All four evaluations met at least eight of the 10 Drummond and Jefferson quality assessment criteria and 16 of the 19 CHEC-list criteria. Full details can be found in Appendix 8. The characteristics and the main results of the economic evaluations are summarised in Table 36. Einarson et al. 70 and Narine et al. 71 both compared treprostinil with epoprostenol, and several of the authors were involved in both papers. In essence, the same model was used; however, one study considered Canadian costs70 and the other considered US costs. 71 For simplicity these papers will be discussed in tandem as the model structure and data inputs are essentially the same. However, it should be noted that treprostinil is not a technology being evaluated as part of this assessment. The Highland et al. paper,72 a US-based study, compared bosentan with treprostinil or epoprostenol, and Wlodarczyk and colleagues73 from Australia also considered bosentan but in comparison to conventional therapy. Three out of the four studies had connections with industry; the exception was the study by Highland et al. 72 in which no reference was made to funding or conflicts of interests stated.

All four studies were model-based analyses. In the studies by Einarson et al. and Narine et al. a cost-minimisation analysis was conducted as they assumed that treprostinil and epoprostenol were clinically equivalent. 70,71 The Wlodarczyk et al. study, which describes the process of the Australian Pharmaceutical Benefits Scheme (PBS) listing for bosentan, conducted a cost-effectiveness analysis using survival as the outcome. 73 Only the model by Highland et al. 72 conducted a cost–utility analysis. This paper was not explicit about the population modelled in terms of FC, with the population described as a cohort of 100 PAH patients. The other papers all considered patients of FCIII and FCIV, although the Narine and Einarson model only considered patients who were non-responders to oral therapy.

The model presented in the Narine et al. and Einarson et al. papers was a decision-analytical spreadsheet model built in Microsoft excel. This model followed a cohort of patients over a 3-year period and was built to represent a logical sequence of clinical practice for PAH patients. Highland et al. presented a Markov cohort model that followed patients over a 1-year period with a cycle length of 3 months, with health states based on FC. The model in the Wlodarczyk et al. study was an individual patient level simulation, run over a time horizon of 15 years with a cycle length of 6 months. Within the model patients could improve, stabilise or not respond to a therapy.

The effectiveness data used in the model presented by both Narine et al. and Einarson et al. were obtained from preliminary data analysis, expert clinical opinion and also non-comparative studies. The two therapies in question were assumed to be clinically equivalent, based on results of a 3-year clinical trial showing equal survival. Highland et al. obtained transition probabilities for bosentan from Rubin et al. 45 Values for treprostinil and epoprostenol were based on bosentan probabilities and adjusted by the RR of improvement in the 6MWT for each therapy obtained from other trials. The model presented by Wlodarczyk and colleagues obtained effectiveness data from two clinical trials43,45 and a long-term open-label extension study provided by industry. Mortality data for conventional therapy were estimated using clinical data on haemodynamic parameters from a trial with long-term follow-up. 43,45 The data were entered into the NIH equation and mortality estimated using the survival model proposed by D’Alonzo et al. 5 Bosentan mortality data were obtained from the two clinical trials. In addition, data on withdrawal rates and the probability of hospitalisation were also estimated using trial data.

The analysis in Narine et al. was from a health-care perspective, with Einarson et al. widening the perspective that the analysis considered by including societal costs. Both models discounted costs only at a rate of 3%, as no measure of effectiveness was used. The models presented in the Wlodarczyk and Highland papers considered a health-care perspective only, with the former discounting costs and life-years at 5%. The Highland model did not require discounting as the time horizon was 1 year.

All four studies considered appropriate resource use items. These typically were the cost of the drugs, initiation of therapy, medical supplies, particularly those associated with the delivery of the drugs, primary and secondary care consultations, surgical and diagnostic procedures, including liver function tests for bosentan, and treatment of SAEs, in particular sepsis. Wlodarczyk et al. considered conventional therapy as the comparator, and this consisted of diuretics, oral anticoagulants, calcium channel blockers, oxygen therapy and digoxin. Unit costs were obtained from standard sources in all studies.

The model presented by both Einarson et al. and Narine et al. did not consider outcomes as the two therapies were considered to be of equal efficacy. The Wlodarczyk model considered outcomes in life-years and only the Highland model measured outcomes in quality-adjusted life-years (QALYs). The health state valuations were obtained from clinical experts. Using the EQ-5D questionnaire, a consensus was achieved on the extent of limitations in each of the five dimensions for each FC. Then the health state descriptions were adjusted for expected side effects associated with the treatments. An alternative set of values was also estimated by increasing FCI estimates by 0.04 and other estimates by this factor plus a further 0.02. A minimum of 0.1 was allowed for FCIV. The values produced are presented in Table 37.

The model used by the Einarson and Narine papers demonstrated savings when using treprostinil compared with epoprostenol. The analysis by Einarson et al. from a US perspective gave savings of US$37,433 over the 3-year time horizon, with an expected average cost saving per patient per year of US$12,478. The greatest savings were attributed to reducing hospitalisation for dose titration and treatment of adverse events, particularly sepsis. The savings reported from a Canadian perspective by Einarson et al. were C$2,610,642 overall, with an average annual saving of C$14,504 from a health-care perspective and C$15,452 from a societal perspective. Again, savings were attributed to reduced hospitalisations. Probabilistic sensitivity analysis (PSA) presented in both papers demonstrated almost a 100% probability of cost savings.

The results of the Highland et al. analysis, from a US perspective, showed bosentan to be dominant over epoprostenol. For a cohort of 100 patients, cost savings were US$3,631,900 with a QALY difference of 11. Sensitivity analyses included changing the RR of improvement and also using alternative utility values, but bosentan still remained cheaper and with greater QALYs. The Australian study by Wlodarczyk et al. demonstrated greater survival on bosentan than on conventional therapy, with 6.7 discounted life-years after 15 years for bosentan compared with 2.8 for conventional therapy. The discounted mean cost was A$234,618 for bosentan and A$18,287 for conventional therapy, giving an ICER of A$55,927 per QALY gained. After 5 years the ICER was a much higher A$84,231 per QALY gained. Sensitivity analyses considered issues such as continuation rules (addition of or switching to epoprostenol), and a series of one-way sensitivity analyses were conducted on many of the model parameters. Mortality was found to be a key variable that reduced the ICER, as was the inclusion of epoprostenol for a small proportion of patients.

Quality of life

A total of 16 quality of life studies (excluding the economic evaluation papers presenting quality of life) were identified; however, two of these74,75 were subsequently found to be general commentary papers that did not present empirical data. Five papers presented values using standard tools to elicit health state utilities. The remaining nine papers considered generic or disease-specific quality of life measures, the most common being the SF-36. One further published paper76 was identified from the manufacturer submissions, giving health state valuations for NYHA FCI–IV. This section will briefly describe the studies considering generic and disease-specific quality of life measures, and will concentrate on those studies containing utility values for health states.

Excluding those studies converting SF-36 values into utility values, quality of life was assessed using the SF-36 in a total of seven studies. 9,49,77–81 Chua et al. 81 also used the Minnesota Living with Heart Failure (MLHF) questionnaire and the Assessment of Quality of Life (AQoL) instruments. The MLHF tool was also used by Cenedese et al. 82 A paper published in 2006 by McKenna et al. 83 reported on the development of the disease-specific Cambridge Pulmonary Hypertension Outcome Review (CAMPHOR) questionnaire, the first PAH-specific tool. The EQ-5D was also administered in the validation stage; however, only correlations between the tools were presented.

Keogh et al. 84 considered the SF-36 and AQoL questionnaire in 177 patients receiving bosentan. Responses to certain items in the SF-36 were used to produce utility values for each FC. Two papers generated values using the EQ-5D. Olschewski et al. 41 conducted a study to evaluate the use of inhaled iloprost in patients with severe pulmonary hypertension compared with placebo. The questionnaire was administered at baseline and after 12 weeks, demonstrating an improvement in health state for patients on active therapy. However, utility values were not presented by FC, and at baseline there was a mix of FCIII and FCIV patients. Sitbon et al. 85 considered both the EQ-5D and SF-36 in 16 patients with HIV-associated PAH receiving bosentan for a total of 16 weeks. As in the previous study the questionnaire was administered at baseline and at the end of follow-up and demonstrated an improvement in quality of life on treatment.

Shafazand et al. 86 described quality of life in 53 patients with PAH, of whom 53% received epoprostenol and 75% were in FCIII or FCIV. The tools administered in the study were the Nottingham Health Profile, the Congestive Heart Failure Questionnaire and the Hospital Anxiety and Depression Scale. In addition, the authors used the VAS and standard gamble methods to elicit preferences for current health. Results were presented for all patients and those taking and not taking epoprostenol. The standard gamble results showed little difference between the epoprostenol and non-epoprostenol groups; however, the VAS score gave a slightly lower value with no epoprostenol, and overall the VAS values were lower. One drawback of the utilities gained in the study is that they were directly elicited from patients rather than from the general population.

The paper by Groen et al. 87 presented a model of lung transplantation for patients with end-stage pulmonary disease, which included pulmonary hypertension. Utility values were derived from the EQ-5D questionnaire administered to patients on the waiting list of a lung transplantation programme every 3 months and after transplantation; however, the values for different periods of time on the waiting list were not related to any health state, for example FC or specific condition. Kirsch76 considered the feasibility of defining a QALY from disease-specific data using the NYHA classifications using the time trade-off method (TTO) associated with the EQ-5D valuation method. The TTO valuations were conducted over a 2-year and a 10-year period for each of the health states (FC), and were elicited from a general population sample of 64 people via interview. Health state valuations by FC were also given in the paper by Highland et al. ,72 described earlier in this chapter, which presented a decision model comparing three treatments. For completeness the values used in this paper are presented in Table 37 alongside all of the utility values presented in the other papers for each FC. Table 38 presents the utility values for the other health states outlined in the papers.

Although four cost studies were identified, three88–90 were not UK studies. The remaining study91 concerned epoprostenol treatment in children alone and presented costs in US dollars. This population group is outside the remit of the appraisal and therefore this paper was also omitted.

Model description

A Markov model built in treeage pro^® was developed to determine the cost-effectiveness of each intervention therapy with supportive care for PAH compared with supportive care alone. The population considered was adults with PAH (category 1 of the Venice 2003 clinical classification1) in NYHA/WHO FCIII (and NYHA/WHO FCIV for epoprostenol) for whom calcium channel blockers were inappropriate or no longer effective. One reference case analysis was conducted using data on all category 1 PAH patients. A separate analysis for idiopathic PAH alone was proposed but a lack of data prevented this.

The five intervention therapies considered within their licensed indications for FCIII were epoprostenol (administered by continuous intravenous infusion), iloprost (administered by inhalation) and the oral therapies of bosentan, sitaxentan and sildenafil, with epoprostenol also considered for FCIV. Only the first use of the interventions was considered, and initiation of any of the interventions after failure of another listed intervention was not considered, with the exception of epoprostenol for patients in FCIV. Therefore, for all treatments the starting state was FCIII with a further analysis conducted with a starting state of FCIV for epoprostenol.

The time horizon of the model was the effective lifetime of patients (30 years), and a starting age of 50 years was used to represent the average age of patients with the disease. The general mortality data were weighted to take into account the ratio of women to men with the disease (1.5:1). A time cycle of 12 weeks was chosen as being sufficiently short enough to capture the effect of treatment, and this time period was in line with that used in the trials for measurement of treatment effect. Health states were based on FC, with a starting health state of FCIII for all therapies and FCIV when the model was run for intravenous epoprostenol for this patient population.

In the first cycle of treatment patients could improve from their starting state FC to the adjacent FC. In all cycles patients could also remain in the same health state or deteriorate and move to the next FC. In addition, patients were also at risk from PAH-related mortality or an age-related death due to other causes. In the intervention arm, once a patient deteriorated and moved to FCIV, the patient switched to intravenous epoprostenol alone, with the first-line therapy discontinued. Although in clinical practice the first-line therapy is unlikely to be stopped, this appraisal considers the treatments within their licensed indications only and therefore this was the only option considered. Data on the effectiveness of combination therapies was not available and therefore inclusion of combinations in FCIV would only have an impact on cost. In the supportive care arm, once deterioration to FCIV was reached, patients switched to intravenous epoprostenol. The only exception was for the model run concerning epoprostenol in FCIV patients and here the comparator was supportive care alone. For all active therapies patients also received supportive therapy.

As the model ran, costs and QALYs were accumulated dependent on the transitions between health states determining FC, therapy and survival. A half-cycle correction was applied. All costs and QALYS were discounted at the rate of 3.5% per year. The model is presented in Figure 9.

FIGURE 9.

Diagram of the decision model. PAH, pulmonary arterial hypertension.

Estimation of model parameters

Mortality

Mortality comprised two components: age-related general population mortality and PAH-related mortality. It was assumed that there was an additional mortality due to PAH, dependent on FC and treatment. This was assumed to be constant for each cycle. Mortality in FCII was assumed to be the general population mortality only. Details of the method used and uncertainty around it appear in Appendix 9. Transition probabilities for PAH-related mortality were entered in the model as beta distributions. The 12-week mortality rates for the intervention therapies are presented in Table 44. The corresponding mortality rates for supportive care are presented in Table 45.

Resource use and costs

The resource use was broadly concerned with the initiation and ongoing costs of each therapy, contacts with primary and secondary health care, adverse events and use of wider social services including palliative care. The perspective adopted for the reference case is that of the NHS/PSS, and a price year of 2006 was applied.

The cost of each of the therapies in question was calculated using British National Formulary (BNF) prices for March 2007,26 using the licensed dose (Table 46). For bosentan, in the first month of treatment it was assumed that the dose was 62.5 mg twice a day, with a dose of 125 mg twice a day for subsequent months. For inhaled iloprost and intravenous epoprostenol, for which the actual doses vary, estimates of average doses from clinical opinion were used. The amount of inhaled iloprost varies from patient to patient; however, as one vial is 10 ∝g and a patient opens a vial each time they nebulise, the cost of a 10-∝g vial (£14.15) was used for each inhalation. It was assumed that a patient nebulised seven times a day. The amount of epoprostenol required for infusion was approximately 15–20 ng/kg/min at the end of the first year, and an average of 17.5 ng/kg/min was used for the first year. Although it was assumed that the dose in the first months would be much lower (the iloprost manufacturer submission uses values of 2.2 ng/kg/min at baseline and 14.1 ng/kg/min at 3 months), comparison with this industry data demonstrated that using this mean over the whole year would not be inappropriate. An average dose of 40 ng/kg/min was used for the second and subsequent years as the range was between 30 and 50 ng/kg/min. A standard deviation around the point estimate was estimated by assuming that the difference between the mean and an upper (or lower) limit equalled two standard deviations. The cost per mg of the drug was £86.71 and an average patient weight of 70 kg was applied.

Further information was provided by the manufacturers with regards to the cost contract with the NHS. GlaxoSmithKline stated that epoprostenol was available at a discount [Commercial-in-confidence information has been removed] for this indication, [Commercial-in-confidence information has been removed]. Schering Health Care referred to a fixed fee system called VENTafee, in which iloprost was provided at a fixed cost of £7400 (excluding VAT) per quarter irrespective of the dose. For both drugs the BNF price was included in the reference case, with the price of the alternative arrangements included in a sensitivity analysis. This equates to £2269.13 for 4 weeks of iloprost and [Commercial-in-confidence information has been removed] per 4 weeks for the first and subsequent years of epoprostenol respectively.

The therapies included in the definition of supportive care were warfarin, furosemide, digoxin and oxygen and it was assumed that all patients would be on each therapy. This is likely to be an overestimate; however, the costs of supportive therapies are small in comparison with the intervention therapies. In addition, when supportive care was not used in conjunction with an active therapy, it was assumed that the patient was hospitalised until death in FCIV. Supportive therapy was assumed to be given to patients irrespective of being on active therapy, but the intensity of supportive therapy was dependent on FC, particularly oxygen therapy. The proportion of patients requiring oxygen in each FC was obtained from the iloprost manufacturer submission, with rates of 5%, 27% and 71% for FCII, FCIII and FCIV respectively. An assumption was made that all patients in FCIV taking supportive therapies would only be on oxygen. The intensity of oral supportive therapies may also increase with worsening FC; however, this level of detail was not available and therefore a standard dose for each drug was used. As the cost of these oral therapies was deemed very low, the impact of dose changes would be negligible. All units and costs are presented in Table 47.

The cost of warfarin therapy includes not only the drug but also regular monitoring to ensure that the patient lies within their therapeutic INR range, thus reducing the risk of thrombolic or haemorrhagic events. As there are different models of care for monitoring, an average cost per visit was used from a trial of 617 patients101 and applied to an assumed average frequency of a monitoring visit every 4 weeks.

For each active therapy an initiation cost was required. In the case of the three oral therapies the model assumed that the patient was on a day ward as a day case and any education by a nurse was assumed to be part of this day case cost. The unit cost used was that of a day case with cardiac catheterisation as this procedure would take place at this visit. An additional initiation cost for bosentan and sitaxentan was a liver function test. Patients commencing inhaled iloprost or intravenous epoprostenol therapy required a longer period of time in hospital and training in order to use the drug delivery system safely. For inhaled iloprost it was assumed that patients were admitted for 3 days with a specialist nurse spending 2 hours a day with the patient for training. Initiation costs for intravenous epoprostenol were higher as it was assumed that a patient would spend approximately 12 days in hospital. Much training is required to ensure that patients are familiar with mixing the drugs and keeping all equipment sterile to reduce the risk of infection. Therefore it was assumed that a specialist nurse would spend 2 hours a day, 5 days a week training the patient. In addition, the cost of the insertion of a central venous catheter for intravenous administration of epoprostenol was also included. The unit cost used here was an elective inpatient stay for catheterisation of 2 days. Therefore the cost of the additional 10 days was calculated using the daily inpatient rate. Other costs may be applicable to patients at initiation of therapy, particularly with regard to standard tests for PAH patients; however, these were not included as they were assumed to apply for all therapies.

Ongoing costs were attributed to each drug to take into account a service fee, which includes delivery of the drug and providing any equipment required for drug delivery. Costs presented here are strictly confidential. Because of the possibility of liver toxicity when taking bosentan or sitaxentan it was assumed that patients had a liver function test every 4 weeks. In addition, each therapy was associated with a number of adverse events, varying in severity and most common in the first period of taking therapy. However, it was decided that the model should only consider the most severe (and therefore costly) adverse events. Therefore only line infection and sepsis while on intravenous epoprostenol were considered, with 17% (line infection) and 4% (sepsis) of patients suffering these events over a 3-month period. 11,39

Primary and secondary care resource use was assumed to be related to FC. Social care and palliative care were also included in resource use, again related to FC. No published data were available on resource use; however, information was available in the iloprost industry submission, obtained from the research carried out by Schering Health Care. The overall costs per FC are presented in Table 48. NHS contacts included seeing hospital physicians and nurses, GP visits and A&E attendance. Personal and social services included residential, day and home care and hospice visits. Hospitalisations considered stays in general wards and intensive care and coronary care units and associated A&E attendance. Full details of resource use and unit costs used by this industry submission are presented in the Appendix 10. As the model assumes that patients on supportive care alone in FCIV will be hospitalised until death, the same resource use for FCIV was used except that the average hospitalisation costs were excluded and replaced by a cost of ongoing inpatient care for all patients at £188 a day.

Unit costs were obtained from a number of standard sources and are presented in Table 49. Drug costs were obtained from the most recent BNF (March 2007). 26 Staff costs and the cost of an inpatient stay were obtained from the Unit Costs of Health and Social Care. 102 Costs of procedures were obtained from NHS reference costs for 2005/6. 103 Warfarin monitoring costs were obtained from a trial dataset presented in Jowett et al. 101 and were inflated to 2006 costs. Other costs, for example liver function tests, were obtained from estimates used in the industry submissions.

Estimation of quality-adjusted life-years

As the model health states were based on FC, utility values also based on FC were sought from the literature and manufacturer submissions. Valuations based on FC were available from two quality of life studies,76,84 one economic evaluation72 and the iloprost and bosentan manufacturer submissions. 41,94 The data in the AIR study were analysed further to provide utility scores by FC, and values presented in the iloprost submission. The data presented here are from the simple pooling analysis. The data from Meads et al. 94 remains academic-in-confidence. The values used in the Highland et al. model72 were not utilised here as the values were gained by clinical consensus and a valuation of 0 was given for FCIV, i.e. the same as death, which was not deemed to be appropriate for this cost-effectiveness analysis. The values from Keogh et al. 84 were used in the base case as this study has the largest sample size and is not academic-in-confidence. However, it should be noted that, although the patient population comprised bosentan patients, the model assumes that these values are applicable for all therapies. The utility values were entered into the model as beta distributions (Table 50). Alternative values were used in the sensitivity analysis to investigate the impact on overall results (Table 51).

Model assumptions

Odds ratios were calculated for the deterioration from FCIII to FCIV for each therapy, to be applied to supportive care transition probabilities. In the absence of suitable mortality data for supportive care alone, it was assumed that the odds ratios for deterioration could also be used as odds ratios for mortality.

Although lung transplantation is an option available to PAH patients in FCIV, this was not included as an event in this model as very few PAH patients actually have a transplant

Bosentan is licensed at 125 mg twice daily and 250 mg twice daily, and consideration of the dose taken was required for the drug costs. Advice from clinical experts indicated that very few patients are on the 250-mg twice daily dose, as liver toxicity is greater and no significant improvement is seen on this higher dose. Accordingly, the model assumed that all patients were taking 125 mg twice daily.

Assessment of cost-effectiveness

The main results are presented as mean costs and QALYs from 10,000 simulations for the alternative policy options considered. Incremental costs and QALYs and, when appropriate, an estimate of the incremental cost per QALY are shown. Cost-effectiveness acceptability curves (CEACs) are included to give a measure of the uncertainty reflected in the model. Some exploration of the contribution of individual model parameters to this uncertainty is reported.

Non-reference case analyses

Additional model runs were undertaken to consider the three main issues. First, there was an investigation of the effect on the results of running the model for shorter time horizons of 10 and 20 years. Second, alternative therapy costs supplied by the manufacturers for inhaled iloprost and intravenous epoprostenol were incorporated. Finally, as there was more than one set of utility values to apply to the health states, those values not used in the reference case were explored.

Epoprostenol in addition to supportive care versus supportive care alone, functional class III

Reference case

Table 52 presents the results of the analysis for epoprostenol in FCIII. Compared with supportive care alone, epoprostenol alongside supportive care is more expensive but generates more QALYs, giving an ICER of £277,000 per QALY gained. The CEAC presented in Figure 10 shows that at willingness to pay thresholds of £20,000 and £30,000 per QALY gained, epoprostenol has a zero probability of being cost-effective.

TABLE 52 - Epoprostenol with supportive care versus supportive care alone, functional class III

ICER, incremental cost-effectiveness ratio; QALY(s), quality-adjusted life-year(s).

Strategy	Cost (£)	Cost difference (£)	QALYs	QALY difference	ICER (£/QALY)
Supportive care	479,000		2.056
Epoprostenol	697,000	218,000	2.843	0.787	277,000

FIGURE 10.

Cost-effectiveness acceptability curve for epoprostenol with supportive care versus supportive care alone, functional class III. Inset graph shows a larger x-axis scale. QALY, quality-adjusted life-year.

Analysis of the effect of single parameters in the reference case shows that in many cases the cost and QALY differences change significantly, but in the same direction, so that the difference in ICER is small. Full details are in Appendix 11.1.

Epoprostenol with supportive care versus supportive care alone, functional class IV

Reference case

In FCIV, epoprostenol has a much greater cost than supportive care alone and produces just over one extra QALY, resulting in an ICER of £343,000 per QALY gained (Table 54). At the £20,000 and £30,000 per QALY thresholds, the probability of epoprostenol being cost-effective is zero in both cases (Figure 11).

Analysis of the effect of single parameters in the reference case shows that in most cases the cost and QALY differences change noticeably, but in the same direction, so that the difference in ICER is small. Full details are in Appendix 11.2.

TABLE 54 - Epoprostenol with supportive care versus supportive care alone, functional class IV

ICER, incremental cost-effectiveness ratio; QALY(s), quality-adjusted life-year(s).

Strategy	Cost (£)	Cost difference (£)	QALYs	QALY difference	ICER (£/QALY)
Supportive care	128,000		0.829
Epoprostenol	531,000	403,000	2.003	1.174	343,000

FIGURE 11.

Cost-effectiveness acceptability curve for epoprostenol with supportive care versus supportive care alone, functional class IV. Inset graph shows larger x-axis scale. QALY, quality-adjusted life-year.