Background
Pulmonary arterial hypertension (PAH) is a progressive and debilitating disease characterized by a pathological increase in the resistance of the pulmonary circulation [
1,
2]. The increased pulmonary vascular resistance (PVR) leads to right ventricular dysfunction, exertional impairment, and premature death [
3]. The United States national prospective registry for primary pulmonary hypertension reported the median survival for the idiopathic form of PAH to be only 2.8 years without treatment [
3].
There is currently no cure for PAH, however the past two decades have seen significant advances with the development and clinical implementation of a number of medications that specifically target the aberrant regulatory and structural changes in the pulmonary arterial bed [
4,
5]. Three classes of drugs have been developed and approved for the treatment of PAH: prostanoids, endothelin-1 receptor antagonists (ERAs), and phosphodiesterase type 5 (PDE5) inhibitors. All three classes of medication have been shown to favorably affect hemodynamic parameters as well as improve functional capacity and exercise tolerance [
4]. Although all three classes of drugs have been evaluated in well-designed clinical studies, only one early trial of intravenous epoprostenol was able to detect improvement in mortality in functional class III and IV patients [
6]. No other treatment has been demonstrated to have an impact on mortality. Futhermore, adequately powered trials could be considered ethically inappropriate considering the documented symptomatic and functional benefits of many treatments in PAH. This illustrates the role of a meta-analysis in determining the improvement in mortality with these other treatments.
Two meta-analyses have reviewed the treatments of PAH [
7,
8]. A meta-analysis by Macchia et al in 2007 included some patients with non-PAH pulmonary hypertension and the results of several trials have been reported since this publication [
7]. A meta-analysis by Galiè et al published in 2009 concluded that PAH treatment improved mortality, however this conclusion is limited by the pooling of all three classes of PAH treatment and the inclusion of multiple doses of medication, some of which are not approved for clinical use due to either increased adverse effects or lack of efficacy [
8]. The failure to include unpublished data in this meta-analysis may have also introduced a publication bias.
We sought to improve upon these previous meta-analyses by addressing these issues. By pooling the available literature, we sought to determine the effect of these classes of medication on total mortality and secondarily to assess their impact on other clinical endpoints, including dyspnea, exercise tolerance, hemodynamics, and adverse effects.
Methods
Literature search
We performed a literature search using the MEDLINE and EMBASE databases to identify randomized controlled trials that evaluate the effects of pharmacotherapy on outcomes in PAH. We used the following search terms: pulmonary hypertension, pulmonary arterial hypertension, pulmonary artery hypertension, pulmonary vascular disease, pulmonary heart disease, and pulmonary cardiac disease. The details of the search strategy are summarized in Additional file
1. We also searched the Cochrane Central Register of Controlled Trials and examined bibliographies of retrieved articles and other major review articles. Our search included articles and conference abstracts published from database inception to November 2009 (PUBMED from 1950 to November 2009, EMBASE from 1980 to November 2009). No language restriction was applied. Studies were included if they evaluated adults with PAH and had a follow-up of eight weeks or more. Studies were excluded if they were not double-blind randomized placebo-controlled trials. The only exceptions were studies that evaluated intravenous agents since the use of placebo may be considered unethical in some jurisdictions. We used the Jadad score and the Cochrane Collaboration's tool for assessing methodologic quality and risk of bias, and accepted only those trials with a score of three or greater (two or greater for trials of intravenous agents) using these scales [
9,
10]. We included studies published in abstract form if sufficient information was available to assess methodologic quality. The literature search, data abstraction, and methodologic grading were performed independently by two authors (CJR and SN) using a predefined standardized data abstraction form. All discrepancies were resolved by iteration and consensus.
Endpoints
The primary end point was total mortality from any cause. Secondary end points included 6 minute walk distance (6 MWD), Borg dyspnea scores, functional class (New York Heart Association (NYHA) or World Health Organization (WHO) scores), hemodynamic parameters, and adverse events. Hemodynamic parameters included mean pulmonary artery pressure (mPAP), mean right atrial pressure (mRAP), cardiac index, and pulmonary vascular resistance (PVR) obtained by right heart catheterization. For studies reporting PVR in Woods units, we multiplied this value by 80 to obtain the PVR in dyn-sec/cm5.
Statistical analysis
We pooled the data for each end point from individual studies to produce summary effect estimates. Where possible, the endpoints were analyzed based on intention-to-treat. We used the p value or CI when pooling data for studies reporting significance in multiple manners (e.g. p value, CI, standard error, standard deviation). For dichotomous outcomes we calculated a relative risk (RR) and 95% confidence interval (CI). We calculated weighted mean differences and 95% CI for continuous variables. In studies reporting only the placebo-corrected mean change, we used this value for the mean change in the intervention group and assigned a value of 0 for the placebo group. Heterogeneity was examined using a X2 test. For outcomes with significant heterogeneity (p ≤ 0.10) we used a random-effects model to pool the data; otherwise, a fixed-effects model was used. All analyses were conducted using Review Manager statistical software (version 5.0.17 Cochrane Collaboration; Oxford, England). A p-value of less than 0.05 was considered significant.
Discussion
The present study demonstrates that treatment with the prostacyclin and prostacyclin analogue class significantly improves mortality in patients with PAH. Prostanoids, ERAs, and PDE5 inhibitors provide symptomatic relief and improve the functional as well as hemodynamic status of patients with PAH.
The overall prognosis of PAH is very poor with an untreated median survival as short as 3.6 years [
3]. In the present study, the overall mortality in the untreated arm was 4.2% over 14.9 weeks. Thus, mortality reduction is a major goal of pharmacotherapy. However, owing to relatively small sample sizes and short duration of follow-up, the effect of various pharmacotherapies on mortality has been controversial. A recent meta-analysis by Galiè et al concluded that PAH treatment improved mortality, however this conclusion was limited by the combination of all three classes of PAH treatment [
8]. Additionally, this meta-analysis included multiple doses of medications, some of which are not approved for clinical use due to lack of efficacy or increased adverse effects. The relatively narrow search strategy used in this prior meta-analysis did not identify three additional studies that were included in the present study, including two abstracts and one full-text publication [
17,
18,
20]. We therefore build upon this meta-analysis by separately analyzing individual classes of PAH treatment, by including data only for approved medication doses, and by expanding the search to capture studies published in abstract form. In addition, we include a fourth study that has been recently published [
31].
The present study indicates that mortality reduction can be achieved using prostacyclin and prostacyclin analogues. Based on the pooled estimate, the number needed to treat to prevent one death would be thirty-two patients treated for sixteen weeks. This mortality signal is driven primarily by studies of intravenous prostanoids, particularly those studies that included a greater percentage of patients with functional class III or IV symptoms. With intravenous epoprostenol or treprostinil, only eight patients would require treatment for twelve weeks to prevent one death. The studies of intravenous therapy were among the first trials performed in PAH and typically included patients with the most severe disease. The placebo arm in these trials had a high mortality rate of 18%, improving the ability to detect mortality benefit from treatment.
Non-intravenous prostanoids, ERAs and PDE5 inhibitors were not associated with a change in mortality. However, these studies often excluded patients with the most severe disease or allowed concurrent therapy with other pulmonary vasodilators. These factors may have lowered the mortality rate in these studies and thus limited the ability to demonstrate improved mortality with these treatments. Treatment with ERAs resulted in a 42% non-significant improvement in mortality in the 947 subjects included in the pooled analysis. Considering the baseline mortality of only 2% in the placebo group of these studies, approximately 7,000 subjects would be required to detect this degree of mortality improvement with ERA treatment. This calculation illustrates that such a benefit of ERAs will not be shown with additional studies of similar design. In contrast, based on the relative mortality in the treatment and control groups, the pooled analysis of PDE5 inhibitors is only slightly underpowered to detect a significant difference in mortality.
The proven mortality benefit of intravenous prostanoids is consistent with the present guidelines, which recommend the use of intravenous epoprostenol as first-line therapy for patients with poor functional class (e.g. NYHA Class IV patients). Despite the lack of proven mortality benefit, non-intravenous prostanoids, ERAs, and PDE5 inhibitors provide improvements in functional class, exercise tolerance, and pulmonary hemodynamics. These drugs may therefore be reasonable therapies for patients with mild to moderate disease with significant functional limitations.
The benefits of these drugs must be carefully balanced against possible toxicities. Bosentan is associated with an increased risk of transaminitis, however our meta-analysis found no evidence of this risk with other ERAs such as sitaxsentan and ambrisentan. These data are consistent with previous studies. Two doses of sitaxsentan have been studied as an alternative in patients who failed bosentan therapy due to transaminitis, with only one of twelve patients developing a non-fatal and reversible transaminitis after thirteen weeks of sitaxsentan therapy [
35]. This effect appears to be dose-related with the standard dose of 100 mg daily being associated with fewer episodes of liver toxicity [
27]. In a second study, only one of thirty-six patients discontinuing bosentan or sitaxsentan developed a transient transaminitis upon starting ambrisentan [
36]. The data from the current study, while unable to conclude an absence of liver toxicity with sitaxsentan and ambrisentan, do provide further evidence that these ERAs have less liver toxicity than bosentan.
There are several limitations to this study. First, most of the included trials had a relatively small sample size and short follow-up. Thus, the effect of these drugs on long-term mortality and duration of survival improvement is uncertain, particularly for the ERA and PDE5 inhibitor classes. Second, pooling all trials within each class of medication can be criticized since trials were still heterogeneous, even within a single class. The relatively few trials for any single intervention also limited the ability to perform analyses on individual drugs within each class. Third, while several results of this meta-analysis are positive, it is not entirely clear that outcomes such as a small change in mPAP or small increases in 6 MWD (less than 30 meters) have a strong clinical impact. Fourth, we did not directly evaluate the impact of combined therapies, making it unclear whether an individual agent or the combination itself provides a more beneficial outcome. Only four randomized controlled studies have directly examined the potential benefits of combination versus single agent therapy [
16,
24,
30,
37]. The primary endpoint was not met in three of these studies [
16,
24,
37]. The single study showing benefit in the primary endpoint reported a small, but statistically significant placebo-corrected improvement of 28.8 m in 6 MWD [
30]. Finally, current guidelines recommend several other treatments for PAH [
4]. Supplemental oxygen and diuretics are recommended for symptomatic control while warfarin and calcium channel blockers are recommended in some forms of PAH. Our search did not identify any randomized placebo-controlled trials that evaluated oxygen, diuretics, warfarin, or calcium channel blockers, though several observational studies suggest their benefit in PAH [
38‐
40].
Conclusions
The present robust meta-analysis suggests that prostanoids, ERAs, and PDE5 inhibitors all confer a therapeutic benefit. Of these, only intravenous prostacyclins has a proven survival benefit, particularly in patients with severe disease. Non-intravenous prostanoids, ERAs, and PDE5 inhibitors have not been shown to improve mortality, however these agents have not been adequately studied in patients with the most severe disease. Additional studies will be required to determine the optimal dose and duration of these therapies in exacting the best possible outcomes at the lowest cost and risk of adverse events for patients.
Competing interests
The authors CJR, SN, DDS have no competing interests. JRS has received honoraria from Actelion Pharmaceuticals and Pfizer/Encysive for speaking engagements as well as participation in advisory boards for GSK, Pfizer, and Actelion Pharmaceuticals. Assistance for participation in educational activities has also been received from Actelion Pharmaceuticals and Pfizer/Encysive. JRS does not have ongoing contractual or financial relationships with any of these companies. There was no funding provided for this study.
Authors' contributions
CJR and DDS conceived the study design. CJR and SN performed the search and data abstraction. All authors participated in data analysis and interpretation. All authors participated in drafting the manuscript. All authors read and approved the final manuscript.