Abstract
Background and Aims
Effective therapies that target three main signalling pathways are approved to treat pulmonary arterial hypertension (PAH). However, there are few large patient-level studies that compare the effectiveness of these pathways. The aim of this analysis was to compare the effectiveness of the treatment pathways in PAH and to assess treatment heterogeneity.
Methods
A network meta-analysis was performed using individual participant data of 6811 PAH patients from 20 Phase III randomized clinical trials of therapy for PAH that were submitted to the US Food and Drug Administration. Individual drugs were grouped by the following treatment pathways: endothelin, nitric oxide, and prostacyclin pathways.
Results
The mean (±standard deviation) age of the sample was 49.2 (±15.4) years; 78.4% were female, 59.7% had idiopathic PAH, and 36.5% were on background PAH therapy. After covariate adjustment, targeting the endothelin + nitric oxide pathway {β: 43.7 m [95% confidence interval (CI): 32.9, 54.4]}, nitric oxide pathway [β: 29.4 m (95% CI: 22.6, 36.3)], endothelin pathway [β: 25.3 m (95% CI: 19.8, 30.8)], and prostacyclin pathway [oral/inhaled β: 19.1 m (95% CI: 14.2, 24.0), intravenous/subcutaneous β: 24.4 m (95% CI: 15.1, 33.7)] significantly increased 6 min walk distance at 12 or 16 weeks compared with placebo. Treatments also significantly reduced the likelihood of having clinical worsening events. There was significant heterogeneity of treatment effects by age, body mass index, hypertension, diabetes, and coronary artery disease.
Conclusions
Drugs targeting the three traditional treatment pathways significantly improve outcomes in PAH, with significant treatment heterogeneity in patients with some comorbidities. Randomized clinical trials are warranted to identify the most effective treatment strategies in a personalized approach.
Keywords: Pulmonary arterial hypertension, Treatment pathway, Comparative effectiveness, Treatment heterogeneity, Network meta-analysis
Structured Graphical Abstract
Structured Graphical Abstract.
Comparison of the effectiveness and heterogeneity of treatment effects in pulmonary arterial hypertension. 6MWD, 6 min walk distance; CI, confidence interval; IV/Sc, intravenous/subcutaneous; PO/Inh, oral/inhaled.
See the editorial comment for this article ‘Treatments for pulmonary arterial hypertension: navigating through a network of choices', by T. Pitre et al., https://doi.org/10.1093/eurheartj/ehae106.
Introduction
Pulmonary arterial hypertension (PAH) is a chronic progressive disease that is characterized by obstructive proliferative remodelling of the small muscular pulmonary arteries.1,2 This pathological remodelling results in an elevated pulmonary vascular resistance, leading to right ventricular dysfunction, dyspnoea, limitation of physical activity, and reduced quality of life and survival. The incidence and prevalence of PAH in the Western world has been estimated at 5–10 cases per million per year and 48–55 cases per million adults, respectively.3,4 Five-year survival in PAH patients is ∼60%.4,5
Over the past two decades, effective therapies for PAH have been approved by regulatory boards. Five classes of drugs (encompassing 15 different compounds or routes of administration) that target 3 main signalling pathways are currently approved for PAH: endothelin receptor antagonists, phosphodiesterase type 5 inhibitors, soluble guanylate cyclase activators, prostacyclin analogues, and prostacyclin receptor agonists.2–4,6 Endothelin receptor antagonists block endothelin-1, which is a potent pulmonary vasoconstrictor. Phosphodiesterase type 5 inhibitors and soluble guanylate cyclase activators target the nitric oxide pathway and increase cyclic guanosine monophosphate, which is vasodilatory and anti-proliferative. Prostacyclin analogues and prostacyclin receptor agonists increase cyclic adenosine monophosphate, which is anti-proliferative, vasodilatory, and anti-thrombotic.
The goal of treatment in PAH is to improve how patients feel, function, and/or survive. However, due to the challenges of performing large randomized clinical trials (RCTs) in this rare disease and the evolution of available treatments, there are few studies of the comparative effectiveness of drugs targeting different pathophysiologic pathways. Previous meta-analyses have compared the effectiveness of PAH drugs;7,8 however, these used study-level aggregate data rather than patient-level data, did not assess treatment effects on predicted risk status, and did not assess treatment heterogeneity across patient subgroups (e.g. age, sex, and comorbidities). Randomized clinical trials reporting treatment heterogeneity in PAH have mostly involved only a single active intervention and were mostly underpowered to detect statistically significant interactions.9–14
The use of individual participant data (IPD) from multiple trials enables adjustment for confounders, a consistent definition of outcomes, and provides more power to detect treatment–covariate interactions.15,16 Understanding the heterogeneity of treatment pathway effects could improve clinical decision-making, support a precision medicine approach, enable trial enrichment, and provide data to justify head-to-head trials of different treatment approaches in patients.
We, therefore, sought to compare the effectiveness of targeting different treatment pathways in patients with PAH and to assess the heterogeneity of treatment effects using IPD.
Methods
Eligibility criteria and study selection
This IPD meta-analysis was registered with the Research Registry (reviewregistry1558). We included RCTs of PAH therapy that were submitted to the US Food and Drug Administration (FDA), which provided us with IPD from these RCTs (see Supplementary data online, Table S1A). The major eligibility criteria for these trials are given in Supplementary data online, Table S2. Our study sample consisted of PAH patients in Phase III RCTs of the five classes of approved PAH drugs. There were no restrictions on the comparison arm or on a minimum follow-up time. Eligibility criteria for inclusion in this analysis are given in Supplementary data online, Table S1B, and were applied at an individual level.
Data collection and data items
We harmonized the IPD across all trials as previously described.17–19 Briefly, we used the Study Data Tabulation Model (Version 1.4) to organize the IPD into domain datasets across the trials. For example, we used consistent units for variables and a consistent definition of clinical worsening events across the trials. The University of Pennsylvania Institutional Review Board considered the harmonization and secondary use of these data as exempt from approval.
We grouped the individual drugs into treatment pathways based on their mechanism of action.2–4,6 Ambrisentan, bosentan, macitentan, and sitaxentan were considered ‘endothelin pathway’; sildenafil, tadalafil, and riociguat were considered ‘nitric oxide pathway’; and iloprost, selexipag, treprostinil, and epoprostenol were considered ‘prostacyclin pathway’. The prostacyclin pathway drugs were stratified into oral/inhaled (PO/Inh) and intravenous/subcutaneous (IV/Sc) routes of administration.
Primary outcomes were change in the 6 min walk distance (6MWD) from baseline to 12 or 16 weeks (based on the length of follow-up in the individual trial) and time to first clinical worsening (up to the end of the randomized phase of each trial). Clinical worsening was defined as any one of the following: all-cause death, hospitalization for worsening PAH, lung transplantation, atrial septostomy, discontinuation of study treatment (or study withdrawal) for worsening PAH, initiation of parenteral prostanoid therapy, or a decrease of at least 15% in 6MWD from baseline combined with either a worsening of World Health Organization functional class (WHO FC) from baseline or the addition of an approved PAH treatment.9,11,13 The IV/Sc prostacyclin pathway group was not assessed for the clinical worsening outcome because the epoprostenol trials did not capture most clinical worsening events, and the initiation of parenteral prostanoid therapy, which is a clinical worsening event, was the intervention in this group.
Secondary outcomes were time to death (up to the end of the randomized phase of each trial); attaining the minimal clinically important improvement in the 6MWD at 12 or 16 weeks (defined as a change in 6MWD from baseline ≥33 m);20 and achieving or maintaining a 6MWD of >440 m at 12 or 16 weeks.3 Other outcomes included change from baseline to Week 12 or 16 in REVEAL Lite 2 risk score, WHO FC, N-terminal pro-B-type natriuretic peptide (NT-proBNP), Medical Outcome Study Short Form-12 or -36 (SF-12/SF-36), and haemodynamic parameters [mean right atrial pressure (RAP), mean pulmonary artery pressure (PAP), cardiac index, pulmonary vascular resistance (PVR), and pulmonary artery wedge pressure (PAWP)].
Covariates considered for treatment heterogeneity included age, sex, race and ethnicity, geographical region, body mass index (BMI), PAH aetiology, use of background PAH therapy, estimated glomerular filtration rate,21 and comorbidities including systemic hypertension, diabetes, arrhythmias, and others.
Data analysis
We used mean ± standard deviation, median (and inter-quartile range), and counts (and proportions) to summarize data as appropriate and standardized differences22 to assess the balance of baseline variables across PAH treatment pathways.
We used an IPD network meta-analysis to assess the effectiveness of PAH treatment pathways. Network meta-analysis involves the simultaneous comparison of multiple treatments in a single analysis by combining direct (within-trial comparisons) and indirect evidence (across-trial comparisons) within a network of RCTs.23 Key assumptions for a valid network meta-analysis include transitivity (no systematic differences between the comparisons other than the treatments being compared) and consistency (agreement between direct and indirect evidence). We evaluated transitivity using the knowledge of each trial and by assessing the distribution of baseline variables across treatment arms. We evaluated the consistency of the direct and indirect evidence using both node-splitting (a visual inspection of the forest plots of direct and indirect evidence and contrasting direct and indirect evidence using a z-test) and design-by-treatment interaction models.24,25
The advantages of the IPD network meta-analysis over network meta-analysis using aggregate data extracted from study publications include harmonization of covariates and outcomes, adjustment for unbalanced covariates to improve transitivity and consistency, and more power to detect treatment-by-covariate interactions.16 IPD meta-analysis could use a one-step or two-step approach.26 In the one-step approach, a single multi-level model that accounts for clustering of participants within individual studies is specified. In the two-step approach, the data are first analysed within each study, and the effect estimates are synthesized across studies using traditional meta-analysis techniques.
For the one-step network meta-analysis, we used arm-based two-level hierarchical models (with a random intercept term for trial) to account for the clustering of participants within trials.26,27 We used restricted maximum likelihood random effects linear regression for the change in 6MWD, REVEAL Lite 2, NT-proBNP, SF-12/SF-36, RAP, PAP, cardiac index, PVR, and PAWP outcomes; Cox regression (with a frailty term) for the clinical worsening and mortality outcomes; random effects logistic regression for the clinically important improvement in the 6MWD and achieving or maintaining 6MWD >440 m outcomes; and the proportional odds mixed model for WHO FC. The proportional hazards assumption was assessed using statistical tests of zero slopes in the Schoenfeld residuals, and the proportional hazards assumption was not violated. All models were adjusted for age, sex, race, BMI, PAH aetiology, background PAH treatment, and baseline 6MWD. These variables were selected from Table 1 and were adjusted for to improve the homogeneity of the comparisons.
Table 1.
Baseline characteristics of study participants
| Characteristic | N | Overall (N = 6811) | Endothelin + nitric oxide pathways (N = 306) | Nitric oxide pathway (N = 999) | Endothelin pathway (N = 1414) | Prostacyclin pathway (PO/Inh) (N = 1321) | Prostacyclin pathway (IV/Sc) (N = 330) | Placebo (N = 2441) | Std diffa |
|---|---|---|---|---|---|---|---|---|---|
| Age (years) | 6808 | 49.2 ± 15.4 | 56.5 ± 13.8 | 52.0 ± 15.7 | 48.6 ± 15.6 | 48.3 ± 14.9 | 45.8 ± 15.1 | 48.4 ± 15.3 | 0.23 |
| Female sex | 6811 | 5342 (78.4) | 226 (73.9) | 778 (77.9) | 1113 (78.7) | 1034 (78.3) | 279 (84.5) | 1912 (78.3) | 0.07 |
| Race | 6811 | 0.49 | |||||||
| White | 4722 (69.3) | 284 (92.8) | 757 (75.8) | 978 (69.2) | 800 (60.6) | 280 (84.8) | 1623 (66.5) | ||
| Asian | 955 (14.0) | 5 (1.6) | 152 (15.2) | 158 (11.2) | 281 (21.3) | 6 (1.8) | 353 (14.5) | ||
| Black | 279 (4.1) | 12 (3.9) | 50 (5.0) | 70 (5.0) | 38 (2.9) | 20 (6.1) | 89 (3.6) | ||
| Multiple/other | 125 (1.8) | 5 (1.6) | 12 (1.2) | 15 (1.1) | 35 (2.6) | 10 (3.0) | 48 (1.9) | ||
| Unknown | 730 (10.7) | 0 (0.0) | 28 (2.8) | 193 (13.6) | 167 (12.6) | 14 (4.2) | 328 (13.4) | ||
| Ethnicity | 6811 | 0.50 | |||||||
| Not Hispanic/Latino | 5688 (83.5) | 274 (89.5) | 927 (92.8) | 1206 (85.3) | 1091 (82.6) | 218 (66.1) | 1972 (80.8) | ||
| Hispanic/Latino | 693 (10.2) | 31 (10.1) | 70 (7.0) | 208 (14.7) | 115 (8.7) | 15 (4.5) | 254 (10.4) | ||
| Unknown | 430 (6.3) | 1 (0.3) | 2 (0.2) | 0 (0.0) | 115 (8.7) | 97 (29.4) | 215 (8.8) | ||
| Body mass index (kg/m2) | 6811 | 27.1 ± 5.7 | 29.0 ± 6.5 | 27.3 ± 5.7 | 26.7 ± 5.6 | 27.0 ± 5.7 | 26.5 ± 5.9 | 27.1 ± 5.6 | 0.13 |
| Region | 6811 | 0.50 | |||||||
| Europe, Australia | 2910 (42.7) | 167 (54.6) | 478 (47.8) | 566 (40.0) | 580 (43.9) | 89 (27.0) | 1030 (42.2) | ||
| USA, Canada | 2551 (37.5) | 137 (44.8) | 346 (34.6) | 515 (36.4) | 419 (31.7) | 235 (71.2) | 899 (36.8) | ||
| Asia | 815 (12.0) | 2 (0.7) | 138 (13.8) | 131 (9.3) | 241 (18.2) | 0 (0.0) | 303 (12.4) | ||
| Latin America, South Africa | 535 (7.9) | 0 (0.0) | 37 (3.7) | 202 (14.3) | 81 (6.1) | 6 (1.8) | 209 (8.6) | ||
| PAH aetiology | 6801 | 0.27 | |||||||
| Idiopathic | 4061 (59.7) | 158 (51.6) | 602 (60.3) | 844 (59.8) | 813 (61.7) | 175 (53.0) | 1469 (60.3) | ||
| Associated with connective tissue disease | 1954 (28.7) | 118 (38.6) | 274 (27.4) | 421 (29.8) | 362 (27.5) | 97 (29.4) | 682 (28.0) | ||
| Associated with congenital heart disease | 518 (7.6) | 5 (1.6) | 75 (7.5) | 91 (6.4) | 87 (6.6) | 58 (17.6) | 202 (8.3) | ||
| Drug- and toxin-induced | 120 (1.8) | 9 (2.9) | 23 (2.3) | 25 (1.8) | 26 (2.0) | 0 (0.0) | 37 (1.5) | ||
| Heritable/familial | 70 (1.0) | 10 (3.3) | 12 (1.2) | 13 (0.9) | 15 (1.1) | 0 (0.0) | 20 (0.8) | ||
| Associated with HIV infection | 65 (1.0) | 6 (2.0) | 2 (0.2) | 17 (1.2) | 15 (1.1) | 0 (0.0) | 25 (1.0) | ||
| Portopulmonary hypertension | 13 (0.2) | 0 (0.0) | 11 (1.1) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 2 (0.1) | ||
| Time since PAH diagnosis (months) | 6727 | 11 (3, 36) | 1 (0, 2) | 9 (2, 34) | 8 (2, 29) | 13 (4, 36) | 15 (5, 51) | 14 (4, 42) | |
| Comorbidities | |||||||||
| Hypertension | 6811 | 2183 (32.1) | 155 (50.7) | 368 (36.8) | 494 (34.9) | 393 (29.8) | 66 (20.0) | 707 (29.0) | 0.21 |
| Dyslipidaemia | 6811 | 1075 (15.8) | 78 (25.5) | 197 (19.7) | 223 (15.8) | 200 (15.1) | 12 (3.6) | 365 (15.0) | 0.19 |
| Heart failure | 6811 | 811 (11.9) | 21 (6.9) | 112 (11.2) | 237 (16.8) | 144 (10.9) | 7 (2.1) | 290 (11.9) | 0.18 |
| Arrhythmia | 6811 | 837 (12.3) | 47 (15.4) | 145 (14.5) | 174 (12.3) | 175 (13.2) | 8 (2.4) | 288 (11.8) | 0.14 |
| Coronary artery disease | 6811 | 611 (9.0) | 51 (16.7) | 88 (8.8) | 179 (12.7) | 100 (7.6) | 2 (0.6) | 191 (7.8) | 0.20 |
| Diabetes mellitus | 6811 | 626 (9.2) | 51 (16.7) | 112 (11.2) | 126 (8.9) | 126 (9.5) | 16 (4.8) | 195 (8.0) | 0.13 |
| Hepatic disorderb | 6811 | 317 (4.7) | 14 (4.6) | 63 (6.3) | 55 (3.9) | 59 (4.5) | 14 (4.2) | 112 (4.6) | 0.04 |
| Chronic renal failure | 6811 | 294 (4.3) | 37 (12.1) | 47 (4.7) | 58 (4.1) | 43 (3.3) | 8 (2.4) | 101 (4.1) | 0.11 |
| Estimated glomerular filtration rate (mL/min/1.73 m2) | 6534 | 87.9 ± 25.7 | 87.5 ± 22.4 | 83.2 ± 25.5 | 90.2 ± 24.7 | 87.8 ± 25.9 | 93.6 ± 27.0 | 88.0 ± 26.1 | 0.14 |
| Background PAH therapy | 6811 | ||||||||
| None | 4327 (63.5) | 306 (100.0) | 672 (67.3) | 1096 (77.5) | 533 (40.3) | 330 (100.0) | 1390 (56.9) | ||
| PDE5i | 1030 (15.1) | 0 (0.0) | 0 (0.0) | 284 (20.1) | 304 (23.0) | 0 (0.0) | 442 (18.1) | ||
| ERA | 880 (12.9) | 0 (0.0) | 304 (30.4) | 0 (0.0) | 245 (18.5) | 0 (0.0) | 331 (13.6) | ||
| ERA and PDE5i | 504 (7.4) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 239 (18.1) | 0 (0.0) | 265 (10.9) | ||
| PCA | 44 (0.6) | 0 (0.0) | 21 (2.1) | 13 (0.9) | 0 (0.0) | 0 (0.0) | 10 (0.4) | ||
| PCA and PDE5i | 24 (0.4) | 0 (0.0) | 0 (0.0) | 21 (1.5) | 0 (0.0) | 0 (0.0) | 3 (0.1) | ||
| PCA and ERA | 2 (0.0) | 0 (0.0) | 2 (0.2) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | ||
| WHO FC | 6807 | 0.34 | |||||||
| I | 48 (0.7) | 0 (0.0) | 13 (1.3) | 5 (0.4) | 13 (1.0) | 0 (0.0) | 17 (0.7) | ||
| II | 2230 (32.8) | 95 (31.0) | 361 (36.1) | 555 (39.3) | 440 (33.3) | 26 (7.9) | 753 (30.9) | ||
| III | 4280 (62.9) | 211 (69.0) | 613 (61.4) | 809 (57.2) | 825 (62.5) | 263 (79.7) | 1559 (63.9) | ||
| IV | 249 (3.7) | 0 (0.0) | 12 (1.2) | 45 (3.2) | 42 (3.2) | 41 (12.4) | 109 (4.5) | ||
| Six-minute walk distance (m) | 6808 | 345.1 ± 85.6 | 346.8 ± 91.3 | 348.8 ± 77.5 | 354.0 ± 90.1 | 346.6 ± 75.0 | 315.4 ± 95.3 | 341.5 ± 88.3 | 0.14 |
| REVEAL Lite 2 | 6806 | 0.31 | |||||||
| Low risk | 1802 (26.5) | 110 (35.9) | 292 (29.2) | 423 (29.9) | 355 (26.9) | 26 (7.9) | 596 (24.4) | ||
| Intermediate risk | 3038 (44.6) | 75 (24.5) | 441 (44.1) | 676 (47.8) | 570 (43.2) | 203 (61.5) | 1074 (44.0) | ||
| High risk | 1966 (28.9) | 121 (39.5) | 266 (26.6) | 315 (22.3) | 395 (29.9) | 101 (30.6) | 769 (31.5) | ||
| Mean right atrial pressure (mmHg) | 1475 | 8.0 (5.0, 12.0) | 7.0 (4.0, 10.0) | 7.0 (4.0, 10.0) | 10.0 (6.0, 14.3) | 8.0 (5.0, 12.0) | 0.17 | ||
| Mean pulmonary artery pressure (mmHg) | 1618 | 53 (43, 63) | 50 (39, 60) | 54 (45, 64) | 52 (44, 59) | 57 (49, 66) | 54 (44, 64) | 0.27 | |
| Cardiac output (L/min) | 1599 | 4.00 (3.20, 4.90) | 4.20 (3.40, 5.20) | 4.09 (3.47, 5.00) | 3.60 (3.12, 4.57) | 3.68 (2.80, 4.70) | 3.95 (3.10, 4.86) | 0.19 | |
| Cardiac index (L/min/m2) | 1589 | 2.23 (1.80, 2.78) | 2.38 (1.92, 2.92) | 2.30 (1.90, 2.80) | 2.01 (1.79, 2.44) | 2.12 (1.63, 2.60) | 2.20 (1.80, 2.69) | 0.14 | |
| Pulmonary vascular resistance (Wood units) | 1507 | 10.8 (7.3, 15.3) | 9.2 (6.2, 13.7) | 10.3 (7.2, 14.1) | 12.1 (9.6, 15.0) | 12.7 (9.1, 17.4) | 11.2 (7.7, 16.0) | 0.28 | |
| Pulmonary artery wedge pressure (mmHg) | 1533 | 9.0 (6.0, 11.0) | 9.0 (6.0, 11.0) | 9.0 (7.0, 11.0) | 7.0 (5.0, 9.8) | 10.0 (7.0, 12.0) | 9.0 (6.0, 11.5) | 0.07 |
Values are n (%), mean ± SD, or median (inter-quartile range).
ERA, endothelin receptor antagonist; PDE5i, phosphodiesterase type 5 inhibitor; PCA, prostacyclin analogue.
aThe mean of standardized differences of each treatment pair. Standardized differences >0.20 may suggest imbalance.
bViral or autoimmune hepatitis, cirrhosis, and/or hepatocellular carcinoma.
Sensitivity analyses included a two-step network meta-analysis (the effect estimate was calculated within each trial and included in a frequentist contrast-based random effects network meta-analysis using the weighted least squares regression approach implemented in the netmeta package in R),26–28 exclusion of patients on background PAH therapy, exclusion of patients on riociguat from the nitric oxide pathway group, and adding time since PAH diagnosis and the start year of the trial to the primary multivariable models. The treatment pathways were ranked using the surface under the cumulative ranking curve (SUCRA) method.29
Treatment pathway-by-covariate interactions were assessed by including interaction terms in the one-step models. Within-trial and across-trial information was separated by centring the covariates around their study-specific mean and adding a variable for the mean to the model.30
Missing data were not imputed. The risk of bias in individual trials was not done as this has been extensively reported in previous meta-analyses.8,31 α was set at 0.05. All data analyses were done using R version 4.1.0 (18 May 2021). Multi-level models were run using the lme4 and ordinal packages in R.
Results
Study participants
Of the 23 available RCTs, 20 Phase III RCTs were included in this analysis (Figure 1, see Supplementary data online, Table S2). Of the 6825 PAH patients in the included trials, we excluded 14 patients who did not receive the study intervention, leaving 6811 patients (Figure 1). The network map is shown in Figure 2. There were 306 patients in the endothelin + nitric oxide pathway combination group, 999 in the nitric oxide pathway group, 1414 in the endothelin pathway group, 1321 in the PO/Inh prostacyclin pathway group, and 330 in the IV/Sc prostacyclin pathway group. Apart from the endothelin + nitric oxide pathway combination group (indirect comparison), all single treatment pathways had direct comparison paths with placebo.
Figure 1.
Participant flowchart
Figure 2.
Network diagram for primary outcomes
The baseline characteristics of the study sample are provided in Table 1. The mean ± standard deviation age of the sample was 49.2 years (±15.4); 78.4% were female, 69.3% were White, and 59.7% had idiopathic PAH. There was a higher proportion of Whites, a lower proportion of patients with idiopathic PAH, and a higher proportion of patients with PAH-connective tissue disease in the endothelin + nitric oxide pathway combination group compared with the other groups. A total of 2484 (36.5%) patients were on background therapy at baseline. The mean ± standard deviation 6MWD at baseline was 345.1 ± 85.6 m and was lowest among patients in the IV/Sc prostacyclin pathway group (315.4 ± 95.3 m).
Primary outcomes
At 12- or 16-week follow-up, 6MWD increased by 4.7 m [standard error (SE): 2.5] in the placebo group, 46.6 m (SE: 5.6) in the endothelin + nitric oxide pathway combination group, 33.7 m (SE: 3.7) in the nitric oxide pathway group, 29.4 m (SE: 3.1) in the endothelin pathway group, 23.1 m (SE: 3.2) in the PO/Inh prostacyclin pathway group, and 31.1 m (SE: 5.1) in the IV/Sc prostacyclin pathway group. After covariate adjustment, treatments targeting the various pathways significantly increased 6MWD at 12 or 16 weeks compared to placebo (Table 2).
Table 2.
Treatment effect on primary outcomes (one-step approach)
| Outcome | Treatment pathway | Change from baseline or incidencea | Est (95% CI)b | P-value | aEst (95% CI)c | P-value |
|---|---|---|---|---|---|---|
| Six-minute walk distance (m) | Placebo | 4.7 ± 2.5 | Ref. | Ref. | ||
| Endothelin + nitric oxide | 46.6 ± 5.6 | 41.9 (30.9, 52.8) | <.001 | 43.7 (32.9, 54.4) | <.001 | |
| Nitric oxide | 33.7 ± 3.7 | 29.0 (22.0, 36.0) | <.001 | 29.4 (22.6, 36.3) | <.001 | |
| Endothelin | 29.4 ± 3.1 | 24.7 (19.1, 30.3) | <.001 | 25.3 (19.8, 30.8) | <.001 | |
| Prostacyclin (PO/Inh) | 23.1 ± 3.2 | 18.4 (13.5, 23.4) | <.001 | 19.1 (14.2, 24.0) | <.001 | |
| Prostacyclin (IV/Sc) | 31.1 ± 5.1 | 26.4 (17.0, 35.9) | <.001 | 24.4 (15.1, 33.7) | <.001 | |
| First clinical worsening eventd | Placebo | 528/1484.8 (0.36) | Ref. | Ref. | ||
| Endothelin + nitric oxide | 54/457.7 (0.12) | 0.26 (0.17, 0.41) | <.001 | 0.25 (0.16, 0.38) | <.001 | |
| Nitric oxide | 88/414.3 (0.21) | 0.42 (0.30, 0.59) | <.001 | 0.47 (0.34, 0.65) | <.001 | |
| Endothelin | 288/1337.2 (0.22) | 0.55 (0.44, 0.69) | <.001 | 0.61 (0.49, 0.76) | <.001 | |
| Prostacyclin (PO/Inh) | 244/1026.9 (0.24) | 0.72 (0.60, 0.86) | <.001 | 0.70 (0.58, 0.84) | <.001 | |
| Prostacyclin (IV/Sc) | ||||||
| First clinical worsening (only death or hospitalization due to PAH) evente | Placebo | 326/1548.9 (0.21) | Ref. | Ref. | ||
| Endothelin + nitric oxide | 30/478.8 (0.06) | 0.22 (0.12, 0.40) | <.001 | 0.17 (0.08, 0.38) | <.001 | |
| Nitric oxide | 47/425.7 (0.11) | 0.41 (0.26, 0.65) | <.001 | 0.39 (0.20, 0.76) | .005 | |
| Endothelin | 182/1362.2 (0.13) | 0.55 (0.41, 0.75) | <.001 | 0.60 (0.38, 0.94) | .027 | |
| Prostacyclin (PO/Inh) | 186/1060.9 (0.18) | 0.88 (0.70, 1.10) | .26 | 0.87 (0.65, 1.16) | .33 | |
| Prostacyclin (IV/Sc) |
aMean ± SE change from baseline at 12 or 16 weeks for 6MWD; number of events/person-years (incidence rate) for clinical worsening.
bEst, estimate. Placebo-adjusted change in 6MWD from baseline or hazard ratio of clinical worsening.
cAdjusted for age, sex, race, BMI, PAH aetiology, background PAH treatment, and baseline 6MWD.
dAll-cause death (n = 183), lung transplantation (n = 3), atrial septostomy (n = 0), hospitalization for worsening PAH (n = 489), discontinuation of study treatment (or study withdrawal) for worsening PAH (n = 72), initiation of parenteral prostanoid therapy (n = 53), or a decrease of at least 15% in 6MWD from baseline combined with either a worsening of WHO FC from baseline or the addition of an approved PAH treatment (n = 402).
eAll-cause death (n = 204) and hospitalization for worsening PAH (n = 567).
A Kaplan–Meier plot of time to clinical worsening is shown in Figure 3. A total of 528 (356 per 1000 person-years) patients had at least one clinical worsening event in the placebo group, 54 (120 per 1000 person-years) in the endothelin + nitric oxide pathway combination group, 88 (212 per 1000 person-years) in the nitric oxide pathway group, 288 (215 per 1000 person-years) in the endothelin pathway group, and 244 (238 per 1000 person-years) in the PO/Inh prostacyclin pathway group, respectively. A breakdown of the clinical worsening events is given in Supplementary data online, Table S3 and Figure S1. After covariate adjustment, all treatment pathways significantly reduced the likelihood of having a clinical worsening event compared with placebo (Table 2). Except for the PO/Inh prostacyclin pathway group, the hazard ratios for death or hospitalization due to PAH were similar to those obtained when using the broader definition of clinical worsening.
Figure 3.
Kaplan–Meier plot of time to clinical worsening by treatment pathway
Head-to-head comparisons of all treatment pathways are shown in Figures 4 and 5. The endothelin + nitric oxide pathway combination therapy significantly improved the primary outcomes compared with the single treatment pathways (except the IV/Sc prostacyclin pathway for the 6MWD outcome). The treatment effects of the endothelin, the nitric oxide, and the PO/Inh and IV/Sc prostacyclin pathways on the change in 6MWD did not significantly differ from each other. Nevertheless, treatment targeting the nitric oxide pathway significantly reduced the likelihood of clinical worsening compared with the endothelin and with the PO/Inh prostacyclin pathways, and the endothelin pathway significantly reduced the likelihood of clinical worsening compared with the PO/Inh prostacyclin pathway.
Figure 4.
Head-to-head comparisons of the treatment effect on 6 min walk distance (two-step approach)
Figure 5.
Head-to-head comparisons of the treatment effect on clinical worsening (two-step approach)
In sensitivity analyses, treatment effect estimates when patients on background PAH therapy were excluded (see Supplementary data online, Table S4) and when the two-step approach was used (Figures 4 and 5) were similar to effect estimates obtained from the primary analysis. The exclusion of patients receiving riociguat from the nitric oxide pathway group did not markedly affect the results. Similarly, adding time since PAH diagnosis and the start year of the trial to the primary multi-variable models did not change the results.
Secondary outcomes
A Kaplan–Meier plot of overall survival is shown in Supplementary data online, Figure S2. A total of 276 patients died (52 per 1000 person-years) during the randomized phase of the trials: 112 (65 per 1000 person-years) in the placebo group, 9 (18 per 1000 person-years) in the endothelin + nitric oxide pathway combination group, 16 (36 per 1000 person-years) in the nitric oxide pathway group, 50 (35 per 1000 person-years) in the endothelin pathway group, 76 (68 per 1000 person-years) in the PO/Inh prostacyclin pathway group, and 13 (184 per 1000 person-years) in the IV/Sc prostacyclin pathway group, respectively. These estimates did not include deaths which occurred after patients terminated the study or during subsequent open-label studies. Treatment targeting both the endothelin and nitric oxide pathways and the endothelin pathway alone significantly reduced the likelihood of mortality compared with placebo (Table 3).
Table 3.
Treatment effect on secondary outcomes (one-step approach)
| Outcome | Treatment | Change from baseline or incidencea | Est (95% CI)b | P-value | aEst (95% CI)c | P-value |
|---|---|---|---|---|---|---|
| Mortality | Placebo | 112/1722.8 (0.07) | Ref. | Ref. | ||
| Endothelin + nitric oxide | 9/492.9 (0.02) | 0.27 (0.08, 0.86) | .026 | 0.23 (0.09, 0.61) | .003 | |
| Nitric oxide | 16/443.0 (0.04) | 0.56 (0.22, 1.41) | .22 | 0.53 (0.26, 1.08) | .08 | |
| Endothelin | 50/1416.9 (0.04) | 0.57 (0.31, 1.05) | .07 | 0.54 (0.33, 0.88) | .013 | |
| Prostacyclin (PO/Inh) | 76/1110.7 (0.07) | 1.16 (0.77, 1.75) | .47 | 1.24 (0.87, 1.79) | .23 | |
| Prostacyclin (IV/Sc) | 13/70.6 (0.18) | 0.72 (0.36, 1.45) | .35 | 0.80 (0.41, 1.57) | .52 | |
| Change in 6MWD ≥33 m at 12 or 16 weeks | Placebo | 608/2202 (27.6) | Ref. | Ref. | ||
| Endothelin + nitric oxide | 146/281 (52.0) | 3.28 (2.30, 4.67) | <.001 | 3.65 (2.55, 5.23) | <.001 | |
| Nitric oxide | 455/938 (48.5) | 2.14 (1.70, 2.68) | <.001 | 2.25 (1.78, 2.85) | <.001 | |
| Endothelin | 538/1278 (42.1) | 1.90 (1.57, 2.30) | <.001 | 2.01 (1.65, 2.43) | <.001 | |
| Prostacyclin (PO/Inh) | 458/1184 (38.7) | 1.67 (1.41, 1.97) | <.001 | 1.75 (1.47, 2.09) | <.001 | |
| Prostacyclin (IV/Sc) | 116/293 (39.6) | 1.87 (1.36, 2.57) | <.001 | 1.71 (1.23, 2.38) | .001 | |
| 6MWD > 440 m at 12 or 16 weeks | Placebo | 374/2202 (17.0) | Ref. | Ref. | ||
| Endothelin + nitric oxide | 85/281 (30.2) | 1.87 (1.26, 2.79) | .002 | 2.69 (1.63, 4.42) | <.001 | |
| Nitric oxide | 269/938 (28.7) | 1.53 (1.17, 2.01) | .002 | 2.50 (1.85, 3.38) | <.001 | |
| Endothelin | 358/1278 (28.0) | 1.56 (1.25, 1.95) | <.001 | 1.87 (1.45, 2.42) | <.001 | |
| Prostacyclin (PO/Inh) | 252/1184 (21.3) | 1.50 (1.22, 1.84) | <.001 | 1.86 (1.45, 2.42) | <.001 | |
| Prostacyclin (IV/Sc) | 40/293 (13.7) | 1.07 (0.69, 1.68) | .75 | 1.20 (0.71, 2.05) | .49 | |
| REVEAL Lite 2 risk score | Placebo | −0.05 ± 0.06 | Ref. | Ref. | ||
| Endothelin + nitric oxide | −1.26 ± 0.13 | −1.20 (−1.46, −0.95) | <.001 | −1.23 (−1.48, −0.97) | <.001 | |
| Nitric oxide | −0.69 ± 0.09 | −0.64 (−0.80, −0.47) | <.001 | −0.64 (−0.80, −0.48) | <.001 | |
| Endothelin | −0.57 ± 0.07 | −0.51 (−0.64, −0.39) | <.001 | −0.51 (−0.64, −0.38) | <.001 | |
| Prostacyclin (PO/Inh) | −0.35 ± 0.08 | −0.29 (−0.41, −0.18) | <.001 | −0.30 (−0.41, −0.19) | <.001 | |
| Prostacyclin (IV/Sc) | −0.68 ± 0.19 | −0.63 (−1.0, −0.24) | .001 | −0.60 (−0.99, −0.22) | .002 | |
| Lower WHO FC at 12 or 16 weeks | Placebo | Ref. | Ref. | |||
| Endothelin + nitric oxide | 2.42 (1.62, 3.61) | <.001 | 2.64 (1.77, 3.97) | <.001 | ||
| Nitric oxide | 2.41 (1.84, 3.16) | <.001 | 2.63 (2.01, 3.46) | <.001 | ||
| Endothelin | 1.98 (1.60, 2.44) | <.001 | 2.02 (1.63, 2.51) | <.001 | ||
| Prostacyclin (PO/Inh) | 1.38 (1.15, 1.65) | <.001 | 1.38 (1.15, 1.66) | <.001 | ||
| Prostacyclin (IV/Sc) | 6.08 (3.08, 11.99) | <.001 | 5.69 (2.92, 11.08) | <.001 | ||
| NT-proBNP (pg/mL) | Placebo | 155.7 ± 65.1 | Ref. | Ref. | ||
| Endothelin + nitric oxide | −968.5 ± 102.7 | −1124.2 (−1382.1, −866.4) | <.001 | −1144.7 (−1393.5, −895.8) | <.001 | |
| Nitric oxide | −242.4 ± 78.5 | −398.0 (−590.5, −205.6) | .001 | −400.8 (−596.8, −204.7) | .002 | |
| Endothelin | −620.0 ± 127.0 | −775.7 (−1051.4, −499.9) | <.001 | −771.4 (−1047.9, −494.9) | <.001 | |
| Prostacyclin (PO/Inh) | 54.6 ± 75.0 | −101.0 (−233.8, 31.7) | .14 | −106.1 (−238.7, 26.6) | .12 | |
| Prostacyclin (IV/Sc) | ||||||
| SF-12/SF-36 PCS score | Placebo | 0.57 ± 0.41 | Ref. | Ref. | ||
| Endothelin + nitric oxide | 3.92 ± 0.69 | 3.35 (1.92, 4.78) | <.001 | 3.52 (2.08, 4.96) | <.001 | |
| Nitric oxide | 3.61 ± 0.48 | 3.04 (1.99, 4.09) | <.001 | 3.12 (2.06, 4.18) | <.001 | |
| Endothelin | 2.87 ± 0.39 | 2.30 (1.45, 3.15) | <.001 | 2.26 (1.41, 3.11) | <.001 | |
| Prostacyclin (PO/Inh) | 1.57 ± 1.24 | 1.00 (−1.44, 3.47) | .42 | 0.99 (−1.52, 3.50) | .44 | |
| Prostacyclin (IV/Sc) | ||||||
| SF-12/SF-36 MCS score | Placebo | −0.02 ± 0.70 | Ref. | Ref. | ||
| Endothelin + nitric oxide | 4.14 ± 1.09 | 4.16 (2.00, 6.32) | <.001 | 4.26 (2.09, 6.43) | <.001 | |
| Nitric oxide | 2.53 ± 0.82 | 2.55 (0.95, 4.15) | .002 | 2.50 (0.89, 4.10) | .003 | |
| Endothelin | 2.78 ± 0.69 | 2.79 (1.54, 4.05) | <.001 | 2.75 (1.50, 4.00) | <.001 | |
| Prostacyclin (PO/Inh) | 4.23 ± 1.98 | 4.25 (0.42, 8.09) | .030 | 3.94 (0.08, 7.80) | .045 | |
| Prostacyclin (IV/Sc) |
aMean ± SE change from baseline at 12 or 16 weeks for REVEAL Lite 2, NT-proBNP, SF-12/SF-36 physical component summary (PCS) score, and SF-12/SF-36 mental component summary (MCS) score; number of events/person-years (incidence rate) for all-cause death; number of events/number of patients (proportion) for change in 6MWD ≥ 33 m and 6MWD > 440 m at 12 or 16 weeks.
bEst, estimate. Hazard ratio of all-cause death; odds ratio of change in 6MWD ≥ 33 m and 6MWD > 440 m at 12 or 16 weeks; placebo-adjusted change of REVEAL Lite 2, NT-proBNP, SF-12/SF-36 PCS score, and SF-12/SF-36 MCS score from baseline; odds ratio of having a lower WHO FC at 12 or 16 weeks after adjusting for baseline WHO FC.
cAdjusted for age, sex, race, BMI, PAH aetiology, background PAH treatment, and baseline 6MWD.
A total of 608 (28%), 146 (52%), 455 (49%), 538 (42%), 458 (39%), and 116 (40%) patients had a change in 6MWD ≥ 33 m at 12 or 16 weeks in the placebo, endothelin + nitric oxide combination, nitric oxide, endothelin, PO/Inh prostacyclin, and IV/Sc prostacyclin pathway groups, respectively. All treatment pathways significantly improved the odds of a change in 6MWD ≥ 33 m at 12 or 16 weeks compared with placebo (Table 3).
A total of 374 (17%), 85 (30%), 269 (29%), 358 (28%), 252 (21%), and 40 (14%) patients achieved or maintained a 6MWD of >440 m at 12 or 16 weeks in the placebo, endothelin + nitric oxide combination, nitric oxide, endothelin, PO/Inh prostacyclin, and IV/Sc prostacyclin pathway groups, respectively. All treatment pathways except the IV/Sc prostacyclin pathway significantly improved the odds of achieving or maintaining a 6MWD of >440 m at 12 or 16 weeks compared with placebo (Table 3).
Table 3 provides the impact of the treatment pathway on REVEAL Lite 2 risk score, WHO FC, NT-proBNP, SF-12/SF-36 physical component summary (PCS) score, and SF-12/SF-36 mental component summary (MCS) score at 12 or 16 weeks compared with placebo.
In studies where haemodynamics were assessed, treatment targeting the nitric oxide, endothelin, and IV/Sc prostacyclin pathways, respectively, significantly lowered RAP, PAP, and PVR and increased cardiac index at 12 or 16 weeks compared with placebo (Table 4).
Table 4.
Treatment effect on haemodynamics (one-step approach)
| Outcome | Treatment | Change from baselinea | β (95% CI) | P-value | aβ (95% CI)b | P-value |
|---|---|---|---|---|---|---|
| Mean right atrial pressure (mmHg) | Placebo | 0.82 ± 0.28 | Ref. | Ref. | ||
| Endothelin + nitric oxide | ||||||
| Nitric oxide | −0.56 ± 0.32 | −1.38 (−2.06, −0.70) | <.001 | −1.33 (−2.03, −0.63) | <.001 | |
| Endothelin | −0.47 ± 0.61 | −1.29 (−2.49, −0.08) | .037 | −1.49 (−2.70, −0.27) | .018 | |
| Prostacyclin (PO/Inh) | ||||||
| Prostacyclin (IV/Sc) | −1.07 ± 0.39 | −1.89 (−2.63, −1.14) | <.001 | −1.88 (−2.64, −1.13) | <.001 | |
| Mean pulmonary artery pressure (mmHg) | Placebo | −0.27 ± 0.48 | Ref. | Ref. | ||
| Endothelin + nitric oxide | ||||||
| Nitric oxide | −3.83 ± 0.60 | −3.55 (−4.70, −2.40) | <.001 | −3.50 (−4.67, −2.32) | <.001 | |
| Endothelin | −4.57 ± 1.10 | −4.30 (−6.41, −2.19) | <.001 | −4.43 (−6.56, −2.30) | <.001 | |
| Prostacyclin (PO/Inh) | −1.35 ± 1.28 | −1.08 (−3.50, 1.34) | .38 | −0.75 (−3.19, 1.68) | .54 | |
| Prostacyclin (IV/Sc) | −3.83 ± 0.68 | −3.56 (−4.77, −2.35) | <.001 | −3.62 (−4.84, −2.39) | <.001 | |
| Cardiac index (L/min/m2) | Placebo | −0.06 ± 0.04 | Ref. | Ref. | ||
| Endothelin + nitric oxide | ||||||
| Nitric oxide | 0.35 ± 0.05 | 0.41 (0.32, 0.50) | <.001 | 0.42 (0.32, 0.51) | <.001 | |
| Endothelin | 0.24 ± 0.09 | 0.31 (0.14, 0.47) | <.001 | 0.31 (0.14, 0.48) | <.001 | |
| Prostacyclin (PO/Inh) | 0.08 ± 0.10 | 0.14 (−0.05, 0.33) | .14 | 0.13 (−0.07, 0.32) | .20 | |
| Prostacyclin (IV/Sc) | 0.25 ± 0.05 | 0.31 (0.21, 0.41) | <.001 | 0.30 (0.20, 0.40) | <.001 | |
| Pulmonary vascular resistance (Wood units) | Placebo | 0.44 ± 0.23 | Ref. | Ref. | ||
| Endothelin + nitric oxide | ||||||
| Nitric oxide | −2.31 ± 0.28 | −2.75 (−3.32, −2.18) | <.001 | −2.71 (−3.30, −2.11) | <.001 | |
| Endothelin | −2.61 ± 0.53 | −3.05 (−4.08, −2.02) | <.001 | −3.07 (−4.14, −2.00) | <.001 | |
| Prostacyclin (PO/Inh) | −1.08 ± 0.65 | −1.52 (−2.76, −0.28) | .017 | −1.45 (−2.72, −0.18) | .025 | |
| Prostacyclin (IV/Sc) | −2.51 ± 0.34 | −2.94 (−3.59, −2.30) | <.001 | −2.83 (−3.48, −2.17) | <.001 | |
| Pulmonary artery wedge pressure (mmHg) | Placebo | 0.36 ± 0.26 | Ref. | Ref. | ||
| Endothelin + nitric oxide | ||||||
| Nitric oxide | 0.60 ± 0.32 | 0.25 (−0.39, 0.88) | .44 | 0.24 (−0.41, 0.89) | .47 | |
| Endothelin | 1.21 ± 0.60 | 0.85 (−0.31, 2.01) | .15 | 0.75 (−0.41, 1.92) | .20 | |
| Prostacyclin (PO/Inh) | 1.09 ± 0.72 | 0.74 (−0.65, 2.12) | .29 | 0.88 (−0.51, 2.27) | .21 | |
| Prostacyclin (IV/Sc) | −0.13 ± 0.38 | −0.48 (−1.19, 0.23) | .18 | −0.53 (−1.25, 0.18) | .33 |
aMean ± SE change from baseline.
bAdjusted for age, sex, race, BMI, PAH aetiology, background PAH treatment, and baseline 6MWD.
Consistency of the evidence and ranking
There was no evidence of inconsistency for any of the primary and secondary endpoints [overlapping confidence intervals for direct and indirect evidence (Figures 4 and 5); P-values from node-splitting and design-by-treatment interaction models were >.05 suggesting no significant disagreement between the direct and indirect evidence].
The ranking of the treatment pathways based on SUCRA is shown in Supplementary data online, Figure S3. The endothelin + nitric oxide pathway combination group was often more likely to be in the top rank, followed by the IV/Sc prostacyclin pathway, nitric oxide pathway, endothelin pathway, and then the PO/Inh prostacyclin pathway. The SUCRA ranks were consistent with the treatment effect sizes. Adjusting for the start year of the trial to account for possible chronological differences did not change these rankings.
Treatment heterogeneity
Interactions of treatment pathways with potential effect modifiers (i.e. heterogeneity of treatment effects) are shown in Figure 6 and Supplementary data online, Table S5. There were significant interactions between treatment pathway and age (P = .004) and diabetes (P = .040) for the change in 6MWD outcome and between treatment pathway and age (P = .001), BMI (P = .002), hypertension (P < .001), and coronary artery disease (P = .048) for the clinical worsening outcome (see Supplementary data online, Table S5). As shown in the interaction plots (Figure 6), treatment effects decreased with increasing age for all treatment pathways (except the IV/Sc prostacyclin pathway), with the nitric oxide pathway having the greatest decline. For the clinical worsening outcome, the treatment effect of the endothelin pathway did not vary by BMI; however, the treatment effect of the nitric oxide pathway decreased with increasing BMI, while the treatment effect of the PO/Inh prostacyclin pathway was greater with increasing BMI. These differences in treatment effectiveness across BMI values were not seen for the 6MWD outcome. The effect of therapies was decreased in the presence of hypertension, diabetes, or coronary artery disease, with the nitric oxide pathway being most affected.
Figure 6.
Heterogeneity of treatment effects on 6 min walk distance and clinical worsening
Discussion
In this IPD network meta-analysis including 20 clinical trials and 6811 PAH patients, we showed that the three traditional treatment pathways in PAH had different effects on 6MWD, clinical worsening events, mortality, WHO FC, NT-proBNP, and haemodynamic parameters. Based on the magnitude of the effect sizes, head-to-head comparisons, and treatment rankings (SUCRA), we confirmed that randomization to combination therapy targeting the endothelin and nitric oxide pathways was associated with a greater treatment effect than any of the single pathway treatment that was assessed. Our findings were robust in sensitivity analysis that excluded patients who were on background therapy and were consistent in secondary outcomes. We also found significant heterogeneity of treatment effect by age, BMI, hypertension, diabetes, and coronary artery disease (Structured Graphical Abstract).
The pathophysiology of PAH is complex and involves many mechanisms.3,4,32,33 Guidelines recommend single therapy or dual or triple combination of therapies based on predicted risk.2,3,6,34 While no trial in this analysis included upfront triple combination therapy, upfront dual combination therapy targeting the endothelin and nitric oxide pathways was more likely ranked higher than the single pathway therapies for most outcomes. We were not able to consider PAH background therapy as an intervention in our comparative effective analysis since PAH background therapy was not randomized and confounding by severity of illness would be likely. Nevertheless, we showed that adjustment for background therapy in our multi-variable models and exclusion of patients on background therapy did not markedly change our effect estimates. We did not find significant interactions between the treatment pathway and the use of background therapy.
All targeted treatment pathways significantly improved 6MWD at 12 or 16 weeks compared with placebo. With the exception of the PO/Inh prostacyclin pathway {β: 19.1 m [95% confidence interval (CI): 14.2, 24.0]}, the placebo-adjusted change in the treatment pathways exceeded the suggested minimally important difference for group mean comparisons (24 m).20 Also, all targeted treatment pathways increased the likelihood of an individual patient attaining a clinically important improvement in the 6MWD (≥33 m)20 and, with the exception of the IV/Sc prostacyclin pathway, the likelihood of achieving or maintaining the recommended absolute low-risk threshold of >440 m.3 Patients in the IV/Sc prostacyclin pathway group generally had more severe disease (lowest 6MWD) at baseline, likely attributable to the trials of epoprostenol that were older. Our findings were similar to those of previous network meta-analyses that used aggregate data and a wider follow-up range (8–26 weeks).8,31 To our knowledge, this is the first meta-analysis to report the treatment effect in individual patients in terms of a minimal clinically meaningful response in the 6MWD or attaining/maintaining an absolute threshold of >440 m.
All targeted treatment pathways significantly reduced the likelihood of having a clinical worsening event compared with placebo. Previous network meta-analyses that used aggregate data and heterogeneous definitions of clinical worsening reported similar findings.8,31 We harmonized the definition of clinical worsening across trials and incorporated components that are of importance to patients. In our study, the treatment effects for the PO/Inh prostacyclin pathway on hospitalization due to PAH or death were not statistically significant. Treatments for the nitric oxide pathway and the prostacyclin pathway (PO/Inh and IV/Sc) were not associated with the risk of mortality. Low event rates could have affected the power to detect these relationships. Most of the trials in this analysis excluded patients with severe PAH [only 249 (3.7%) patients with WHO FC IV were included]; hence, these findings may not be generalizable to patients with severe disease.
In head-to-head comparisons, combination therapy significantly improved the primary outcomes, consistent with the treatment ranking using SUCRA. Compared with placebo, almost all targeted treatment pathways significantly improved the secondary outcomes at follow-up including WHO FC, NT-proBNP, and RAP, PAP, cardiac index, and PVR in studies that performed right heart catheterization at baseline and follow-up.
We identified significant heterogeneity of treatment effects, which was made possible (unlike in prior studies)9–14,35 by analysing IPD. Notably, the treatment effect of all pathways (except the IV/Sc prostacyclin pathway) decreased with older age with some differences among the treatment pathways. Also, the treatment effect of drugs targeting the nitric oxide pathway significantly weakened with increasing BMI and in the setting of systemic hypertension, diabetes, or coronary artery disease, which may be explained by the inclusion of patients with subclinical left ventricular or systemic vascular disease. It is possible that patients with some comorbidities including older age, obesity, systemic hypertension, diabetes, and coronary artery disease may have more systemic vascular remodelling, and the presence of such comorbidities may facilitate disease progression.36 While Rosenkranz et al.37 found no effect modification of selexipag by comorbidities, post hoc analyses of the AMBITION trial38 and a real-world cohort of PAH patients39 suggested that treatment effectiveness of combination therapy was attenuated in the presence of comorbidities. We have reported variation of specific treatment pathways across specific comorbidities. For example, while the nitric oxide pathway may be less effective with increasing age, the IV/Sc prostacyclin pathway may be more effective in older patients. These analyses are hypothesis generating but could provide support for designing precision medicine approaches to treatment and clinical trial enrichment. We previously published a study that suggested that endothelin pathway therapies were less effective in men and possibly in African American patients, which was not seen in this broader study that included all treatment classes.40
Our study had strengths and weaknesses. To our knowledge, this is the first study to report the comparative effectiveness of PAH treatment pathways using IPD from multiple trials of PAH therapy. We used harmonized definitions across trials, adjusted for relevant covariates, assessed treatment-by-covariate interactions on an individual level, parameterized outcomes such as 6MWD and clinical worsening in multiple ways, and used relevant sensitivity analyses to assess the robustness of our findings. In pooling individual treatments and doses at a pathway level, we assumed that drugs within a pathway with various doses have the same effectiveness, which may not be the case. Also, while the 20 trials included in this analysis have previously been reported as having a low risk of bias,8,31 our analysis is likely to suffer from selection bias because (by default) positive clinical trials are submitted to the FDA in support of drug approval rather than null clinical trials. Randomized clinical trials performed after drug approval or recently were also not included. Nevertheless, this is a ‘living IPD meta-analysis’, and we aim to update the analysis as patient-level data from clinical trials in PAH become available to us. This includes data from trials of novel PAH therapy and trials that were not submitted to the FDA. Further, the placebo group received heterogeneous interventions such as oral placebo, subcutaneous placebo, or conventional therapy. The historical epoprostenol trials were open label and did not have the comprehensive data that the other studies had available. This IPD meta-analysis is observational, and differences in disease phenotypes and severity of illness may confound or bias some of the results, which should not be used to guide treatment without additional studies. Also, the results should be interpreted with caution, as some of the included trials assessed highly controlled interventions in specific populations, which may differ from the current real-life use. For example, IV/Sc prostanoids are currently used at higher doses, in faster up-titration, and in combination with other therapies.41,42
Conclusions
Drugs targeting the three traditional treatment pathways significantly improve outcomes in PAH. Combination therapy targeting the endothelin and nitric oxide pathways was associated with the greatest benefit in terms of 6MWD, time to clinical worsening, mortality, and NT-proBNP compared with single therapy targeting the nitric oxide, endothelin, or prostacyclin pathways. Pulmonary arterial hypertension drugs may be less effective in patients with some comorbidities, including older age, obesity, systemic hypertension, diabetes, and coronary artery disease, particularly those drugs targeting the nitric oxide pathway. Randomized clinical trials are warranted to identify the most effective treatment strategies in a personalized approach.
Supplementary data
Supplementary data are available at European Heart Journal online.
Supplementary Material
Contributor Information
Jude Moutchia, Department of Biostatistics, Epidemiology, and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
Robyn L McClelland, Department of Biostatistics, School of Public Health, University of Washington, Seattle, WA, USA.
Nadine Al-Naamani, Department of Medicine, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
Dina H Appleby, Department of Biostatistics, Epidemiology, and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
John H Holmes, Department of Biostatistics, Epidemiology, and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
Jasleen Minhas, Department of Medicine, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
Jeremy A Mazurek, Department of Medicine, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
Harold I Palevsky, Department of Medicine, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
Corey E Ventetuolo, Department of Medicine and Health Services, Policy and Practice, Brown University, Providence, RI, USA.
Steven M Kawut, Department of Medicine, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
Declarations
Disclosure of Interest
R.L.M. received personal fees for serving on a data safety monitoring board for the RURAL study. J.H.H. received personal fees for serving on a data safety monitoring board for Pan-European Response to the ImpactS of COVID-19 and future pandemics and epidemics. J.A.M. received institutional funding as grants from Corvia, Tenax, and Merck; personal fees for consulting from Janssen PH; and personal fees for continuing medical education from Janssen PH, United Therapeutics, and Merck. H.I.P. received personal fees for serving on a data safety monitoring board from United Therapeutics. C.E.V. received institutional funding as grants from Altavant Sciences; personal fees for continuing medical education from Gather-ED; personal fees for serving on a data safety monitoring board for U01 SATURN; personal fees for serving on an advisory board from Altavant Sciences and Acceleron/Merck; and personal fees for editorial work and for travel from the American Thoracic Society. S.M.K. received personal fees for consulting from Janssen, Regeneron, PureTech Health, and Morphic; personal fees from Janssen and institutional funding from Accredo, Actelion, Aerovate, Bayer, Inari Medical, Merck, United Therapeutics, Janssen, Liquidia, and Pfizer for continuing medical education; travel support from Aerovate; personal fees for serving on a data safety monitoring board from United Therapeutics, Keros, Acceleron, Vivus, Aerovate, and Proteo Biotech; editorial fees from the European Respiratory Journal; has stock or stock options in Verve Therapeutics; and has received in-kind remote monitoring equipment from PhysIQ. All other authors declare no disclosure of interest for this contribution.
Data Availability
The data used for this paper were provided by the FDA for secondary analyses of the included clinical trials. Due to agreements with the data provider, these data are unavailable to be shared.
Funding
This work was completed with support from the Cardiovascular Medical Research and Education Fund (S.M.K.), the National Institutes of Health (K24 HL103844, S.M.K.; K23 HL141584, N.A.-N.; T32 HL007891, J.M.), and the ATS-PHA Rino Aldrighetti grant (N.A.-N.).
Ethical Approval
The University of Pennsylvania Institutional Review Board considered the harmonization and secondary use of these data as exempt from approval.
Pre-registered Clinical Trial Number
Not applicable.
References
- 1. Humbert M, Guignabert C, Bonnet S, Dorfmüller P, Klinger JR, Nicolls MR, et al. Pathology and pathobiology of pulmonary hypertension: state of the art and research perspectives. Eur Respir J 2019;53:1801887. 10.1183/13993003.01887-2018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Hassoun PM. Pulmonary arterial hypertension. N Engl J Med 2021;385:2361–76. 10.1056/NEJMra2000348 [DOI] [PubMed] [Google Scholar]
- 3. Humbert M, Kovacs G, Hoeper MM, Badagliacca R, Berger RMF, Brida M, et al. 2022 ESC/ERS guidelines for the diagnosis and treatment of pulmonary hypertension: developed by the task force for the diagnosis and treatment of pulmonary hypertension of the European Society of Cardiology (ESC) and the European Respiratory Society (ERS). Endorsed by the International Society for Heart and Lung Transplantation (ISHLT) and the European Reference Network on rare respiratory diseases (ERN-LUNG). Eur Heart J 2022;43:3618–731. 10.1093/eurheartj/ehac237 [DOI] [PubMed] [Google Scholar]
- 4. Lau EMT, Giannoulatou E, Celermajer DS, Humbert M. Epidemiology and treatment of pulmonary arterial hypertension. Nat Rev Cardiol 2017;14:603–14. 10.1038/nrcardio.2017.84 [DOI] [PubMed] [Google Scholar]
- 5. Farber HW, Miller DP, Poms AD, Badesch DB, Frost AE, Muros-Le Rouzic E, et al. Five-year outcomes of patients enrolled in the REVEAL registry. Chest 2015;148:1043–54. 10.1378/chest.15-0300 [DOI] [PubMed] [Google Scholar]
- 6. Galiè N, Channick RN, Frantz RP, Grünig E, Jing ZC, Moiseeva O, et al. Risk stratification and medical therapy of pulmonary arterial hypertension. Eur Respir J 2019;53:1801889. 10.1183/13993003.01889-2018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Schlueter M, Beaudet A, Davies E, Gurung B, Karabis A. Evidence synthesis in pulmonary arterial hypertension: a systematic review and critical appraisal. BMC Pulm Med 2020;20:202. 10.1186/s12890-020-01241-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Pitre T, Su J, Cui S, Scanlan R, Chiang C, Husnudinov R, et al. Medications for the treatment of pulmonary arterial hypertension: a systematic review and network meta-analysis. Eur Respir Rev 2022;31:220036. 10.1183/16000617.0036-2022 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Galiè N, Barberà JA, Frost AE, Ghofrani H-A, Hoeper MM, McLaughlin VV, et al. Initial use of ambrisentan plus tadalafil in pulmonary arterial hypertension. N Engl J Med 2015;373:834–44. 10.1056/NEJMoa1413687 [DOI] [PubMed] [Google Scholar]
- 10. McLaughlin V, Channick RN, Ghofrani H-A, Lemarié J-C, Naeije R, Packer M, et al. Bosentan added to sildenafil therapy in patients with pulmonary arterial hypertension. Eur Respir J 2015;46:405–13. 10.1183/13993003.02044-2014 [DOI] [PubMed] [Google Scholar]
- 11. Pulido T, Adzerikho I, Channick RN, Delcroix M, Galiè N, Ghofrani H-A, et al. Macitentan and morbidity and mortality in pulmonary arterial hypertension. N Engl J Med 2013;369:809–18. 10.1056/NEJMoa1213917 [DOI] [PubMed] [Google Scholar]
- 12. Galiè N, Brundage BH, Ghofrani HA, Oudiz RJ, Simonneau G, Safdar Z, et al. Tadalafil therapy for pulmonary arterial hypertension. Circulation 2009;119:2894–903. 10.1161/circulationaha.108.839274 [DOI] [PubMed] [Google Scholar]
- 13. Sitbon O, Channick R, Chin KM, Frey A, Gaine S, Galiè N, et al. Selexipag for the treatment of pulmonary arterial hypertension. N Engl J Med 2015;373:2522–33. 10.1056/NEJMoa1503184 [DOI] [PubMed] [Google Scholar]
- 14. Ghofrani H-A, Galiè N, Grimminger F, Grünig E, Humbert M, Jing Z-C, et al. Riociguat for the treatment of pulmonary arterial hypertension. N Engl J Med 2013;369:330–40. 10.1056/NEJMoa1209655 [DOI] [PubMed] [Google Scholar]
- 15. Freeman SC. Individual patient data meta-analysis and network meta-analysis (NMA). In: Evangelou E, Veroniki AA (eds.), Meta-Research: Methods and Protocols. New York: Springer, 2022, 279–98. [DOI] [PubMed] [Google Scholar]
- 16. Richard DR, Sofia D, Sarah D, Jayne FT, Lesley AS, Orestis E, et al. Using individual participant data to improve network meta-analysis projects. BMJ Evid Based Med 2023;28:197–203. 10.1136/bmjebm-2022-111931 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Min J, Appleby DH, McClelland RL, Minhas J, Holmes JH, Urbanowicz RJ, et al. Secular and regional trends among pulmonary arterial hypertension clinical trial participants. Ann Am Thorac Soc 2022;19:952–61. 10.1513/AnnalsATS.202110-1139OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Urbanowicz RJ, Holmes JH, Appleby D, Narasimhan V, Durborow S, Al-Naamani N, et al. A semi-automated term harmonization pipeline applied to pulmonary arterial hypertension clinical trials. Methods Inf Med 2022;61:3–10. 10.1055/s-0041-1739361 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. McCarthy BE, McClelland RL, Appleby DH, Moutchia JS, Minhas JK, Min J, et al. Body mass index and treatment response in patients with pulmonary arterial hypertension: a meta-analysis. Chest 2022;162:436–47. 10.1016/j.chest.2022.02.041 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Moutchia J, McClelland RL, Al-Naamani N, Appleby DH, Blank K, Grinnan D, et al. Minimal clinically important difference in the six-minute walk distance for patients with pulmonary arterial hypertension. Am J Respir Crit Care Med 2023;207:1070–9. 10.1164/rccm.202208-1547OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF III, Feldman HI, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med 2009;150:604–12. 10.7326/0003-4819-150-9-200905050-00006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Yang D, Dalton JE. A unified approach to measuring the effect size between two groups using SAS. In: SAS Global Forum. Abstract Paper 335-2012. Orlando: SAS Institute Inc., 2012.
- 23. Rouse B, Chaimani A, Li T. Network meta-analysis: an introduction for clinicians. Intern Emerg Med 2017;12:103–11. 10.1007/s11739-016-1583-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. van Valkenhoef G, Dias S, Ades AE, Welton NJ. Automated generation of node-splitting models for assessment of inconsistency in network meta-analysis. Res Synth Methods 2016;7:80–93. 10.1002/jrsm.1167 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Jackson D, Barrett JK, Rice S, White IR, Higgins JP. A design-by-treatment interaction model for network meta-analysis with random inconsistency effects. Stat Med 2014;33:3639–54. 10.1002/sim.6188 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Burke DL, Ensor J, Riley RD. Meta-analysis using individual participant data: one-stage and two-stage approaches, and why they may differ. Stat Med 2017;36:855–75. 10.1002/sim.7141 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Karahalios A, McKenzie JE, White IR. Contrast-based and arm-based models for network meta-analysis. Methods Mol Biol 2022;2345:203–21. 10.1007/978-1-0716-1566-9_13 [DOI] [PubMed] [Google Scholar]
- 28. Schwarzer G, Carpenter JR, Rücker G. Network meta-analysis. In: Schwarzer G, Carpenter JR, Rücker G (eds.), Meta-Analysis with R. Cham: Springer, 2015, 187–216. [Google Scholar]
- 29. Mbuagbaw L, Rochwerg B, Jaeschke R, Heels-Andsell D, Alhazzani W, Thabane L, et al. Approaches to interpreting and choosing the best treatments in network meta-analyses. Syst Rev 2017;6:79. 10.1186/s13643-017-0473-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Riley RD, Debray TPA, Fisher D, Hattle M, Marlin N, Hoogland J, et al. Individual participant data meta-analysis to examine interactions between treatment effect and participant-level covariates: statistical recommendations for conduct and planning. Stat Med 2020;39:2115–37. 10.1002/sim.8516 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Jain S, Khera R, Girotra S, Badesch D, Wang Z, Murad MH, et al. Comparative effectiveness of pharmacologic interventions for pulmonary arterial hypertension: a systematic review and network meta-analysis. Chest 2017;151:90–105. 10.1016/j.chest.2016.08.1461 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Thenappan T, Ormiston ML, Ryan JJ, Archer SL. Pulmonary arterial hypertension: pathogenesis and clinical management. BMJ 2018;360:j5492. 10.1136/bmj.j5492 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Lan NSH, Massam BD, Kulkarni SS, Lang CC. Pulmonary arterial hypertension: pathophysiology and treatment. Diseases 2018;6:38. 10.3390/diseases6020038 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Beshay S, Sahay S, Humbert M. Evaluation and management of pulmonary arterial hypertension. Respir Med 2020;171:106099. 10.1016/j.rmed.2020.106099 [DOI] [PubMed] [Google Scholar]
- 35. Sitbon O, Chin KM, Channick RN, Benza RL, Di Scala L, Gaine S, et al. Risk assessment in pulmonary arterial hypertension: insights from the GRIPHON study. J Heart Lung Transplant 2020;39:300–9. 10.1016/j.healun.2019.12.013 [DOI] [PubMed] [Google Scholar]
- 36. Lang IM, Palazzini M. The burden of comorbidities in pulmonary arterial hypertension. Eur Heart J Suppl 2019;21:K21–8. 10.1093/eurheartj/suz205 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Rosenkranz S, Channick R, Chin KM, Jenner B, Gaine S, Galiè N, et al. The impact of comorbidities on selexipag treatment effect in patients with pulmonary arterial hypertension: insights from the GRIPHON study. Eur J Heart Fail 2022;24:205–14. 10.1002/ejhf.2369 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. McLaughlin VV, Vachiery J-L, Oudiz RJ, Rosenkranz S, Galiè N, Barberà JA, et al. Patients with pulmonary arterial hypertension with and without cardiovascular risk factors: results from the AMBITION trial. J Heart Lung Transplant 2019;38:1286–95. 10.1016/j.healun.2019.09.010 [DOI] [PubMed] [Google Scholar]
- 39. Badagliacca R, D’Alto M, Ghio S, Argiento P, Brunetti ND, Casu G, et al. Relevance of comorbidities on initial combination therapy in pulmonary arterial hypertension. ERJ Open Res 2022;8:00298–2022. 10.1183/23120541.00298-2022 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Gabler NB, French B, Strom BL, Liu Z, Palevsky HI, Taichman DB, et al. Race and sex differences in response to endothelin receptor antagonists for pulmonary arterial hypertension. Chest 2012;141:20–6. 10.1378/chest.11-0404 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Bergot E, Sitbon O, Cottin V, Prévot G, Canuet M, Bourdin A, et al. Current epoprostenol use in patients with severe idiopathic, heritable or anorexigen-associated pulmonary arterial hypertension: data from the French pulmonary hypertension registry. Int J Cardiol 2014;172:561–7. 10.1016/j.ijcard.2013.12.313 [DOI] [PubMed] [Google Scholar]
- 42. Vizza CD, Lang IM, Badagliacca R, Benza RL, Rosenkranz S, White RJ, et al. Aggressive afterload lowering to improve the right ventricle: a new target for medical therapy in pulmonary arterial hypertension? Am J Respir Crit Care Med 2021;205:751–60. 10.1164/rccm.202109-2079PP [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data used for this paper were provided by the FDA for secondary analyses of the included clinical trials. Due to agreements with the data provider, these data are unavailable to be shared.