Association between Initial Oral Therapy and Outcomes in Systemic Sclerosis-related Pulmonary Arterial Hypertension: Observations from PHAROS

Abstract

Objective

To compare time to clinical worsening (TTCW) based on initial oral PAH therapy in systemic sclerosis (SSc) patients with pulmonary arterial hypertension (PAH).

Methods

Using data from the PHAROS registry (a multicenter prospective observational study enrolling SSc patients with incident pulmonary hypertension), patients with group I PAH and 6 months of initial therapy with an endothelin-receptor antagonist (ERA), phosphodiesterase-5 inhibitor (PDE5i), or a combination of ERA/PDE5i were included. The main outcome was TTCW, defined as the first occurrence of death, PAH-related hospitalization, lung transplant, initiation of parenteral prostacyclin, or worsening symptoms.

Results

Ninety-eight patients (initial ERA=24, initial PDE5i=59, initial ERA/PDE5i=15) were included; no significant differences in baseline variables existed. TTCW was significantly worse in patients initially started on ERA compared to PDE5i or ERA/PDE5i (p=0.0001). Ten patients (41.6%) in the ERA group died over the 3 year observation period, compared to 4 (6.8%) in the PDE5i group and 1 (6.7%) in the combo ERA/PDE5i group (p=0.004). Baseline factors independently associated with shorter TTCW were initial ERA (HR 2.63, p=0.009), lower DLCO (HR 0.69 per 10% change, p=0.04), and higher PVR (HR 1.10 per Wood unit change, p=0.007).

Conclusions

Compared to PDE5i or combination ERA/PDE5i, initial therapy with an ERA in SSc-PAH patients was associated with a significantly worse TTCW, even after adjustment for commonly accepted prognostic factors. Further study into the optimal initial oral therapy in patients with SSc-PAH is needed.

Systemic sclerosis (SSc), commonly referred to as scleroderma, is a multi-organ heterogeneous disorder characterized by endothelial dysfunction and vasculopathy, inflammation, fibroblast dysregulation, and abnormal immune system functioning. Hemodynamically-confirmed pulmonary arterial hypertension (PAH) complicates SSc with an estimated prevalence of 8–12% [1] and is a leading cause of death in this patient population; PAH accounts for up to 30% of premature deaths in SSc patients [2]. Despite advances in treatment, patients with SSc-PAH still have a threefold higher risk of death compared to patients with idiopathic PAH [3]. Further, the response to therapy, as assessed by changes in functional capacity and survival, seems to be less robust in SSc-PAH compared to other forms of PAH [4, 5]. The optimal treatment strategy for patients with SSc-PAH remains to be defined.

Drug studies in PAH have enrolled heterogeneous patient populations, with a mix of idiopathic PAH, heritable PAH, SSc-PAH, and other associated causes of WHO Group 1 pulmonary hypertension (PH). Few subgroup analyses have focused on patients with systemic sclerosis [6–10]. Although more recent studies have included patients on background therapy, there is a relative lack of head-to-head comparisons between initial oral medications for PAH. There is also a need for descriptions of long-term clinical experience of medication use outside of carefully controlled randomized trials. Using data from the PHAROS registry of SSc-PAH patients, we performed a retrospective cohort study to examine whether initial oral medication choice is associated with differential outcomes. We hypothesized that there would be no difference in time to clinical worsening or survival in patients initially started on an endothelin-receptor antagonist (ERA) compared to those treated with a phosphodiesterase-5 inhibitor (PDE5i). Additionally, we hypothesized that, given the potential for synergist effects, clinical outcomes would be improved when patients were started on both an ERA and PDE5i compared to either agent given as initial monotherapy.

Patients and Methods

The Pulmonary Hypertension Assessment and Recognition of Outcomes (PHAROS) registry is a North American multi-center, prospective observational study established in 2006 enrolling SSc patients at high risk for developing or with incident PH [11]. This registry was set up to understand the natural history of PH development in SSc and observe the course of disease progression in those with incident PH. The current analysis is a retrospective cohort study utilizing data from the prospective PHAROS registry. Institutional review board approval was obtained for this analysis (Tulane IRB #685867).

Patient selection and definitions

Of the 178 patients judged by the investigators to have incident group I PAH in the PHAROS registry at the time of the data download (performed on May 14, 2014), 98 patients were included in this analysis (Figure 1). Patients were included if they had Group 1 PAH [12] based on right heart catheterization performed within 6 months prior to enrollment in the registry. This time point for the landmark analysis [13] was chosen a priori. Patients were excluded if they had a pulmonary artery wedge pressure (PAWP) >15 (to exclude those with PH due to left heart disease) or a forced vital capacity (FVC) and total lung capacity (TLC)<65% of predicted and moderate/severe fibrosis on CT (to exclude those with PH predominantly from interstitial lung disease). Included subjects also had to have 6 months of initial therapy exclusively with either an endothelin-receptor antagonist (ERA), phosphodiesterase-5 inhibitor (PDE5i), or a combination of ERA/PDE5i. Patients initially treated with prostanoids were excluded. These 4 exclusions account for the difference between the number of group I PAH patients in the registry (n=178) and the sample size of this analysis (n=98). To attempt to control for therapies added after this initial 6 month period, time on initial therapy (ERA, PDE5i, or combo ERA/PDE5i) was calculated. For example, if a patient was started on an ERA and then 12 months later a PDE5i was added, their “time on initial therapy” would be 1 year. To account for any delay in initiation of therapy, time from diagnostic RHC to start of medication was compared between groups. To avoid immortal time bias, the starting point for all analyses was the date when initial therapy was started. Duration of disease was calculated based on the time from onset of Raynaud’s phenomenon to enrollment in the registry. SSc subtype (limited vs. diffuse) was assigned based on standard definitions [14].

Figure 1.

Patient selection for this retrospective cohort study of systemic sclerosis-PAH patients from the PHAROS registry. Abbreviations: PAWP=pulmonary artery wedge pressure; FVC=forced vital capacity; TLC=total lung capacity; F/U=follow-up; ERA=endothelin receptor antagonist; PDE5i=phosphodiesterase-5 inhibitor; Rx=treatment; PGI₂=prostanoid

Outcomes

The main outcome was time to clinical worsening (TTCW) over a 3 year period. Time to clinical worsening was defined as the first occurrence of all-cause death, PAH-related hospitalization, lung transplant, initiation of parenteral prostanoid therapy, or worsened symptoms occurring over 3 years of follow-up [7]. Worsening of symptoms was defined as a decrease in 6-minute walk distance (6MWD) by >15% and a worsening of NYHA functional class and addition of a PH-specific medication [7].

Statistical Analysis

Continuous baseline variables, as well as time on initial therapy, were compared between the 3 initial treatment groups using one-way ANOVA with a Tukey’s post-test. Categorical variables were compared using Chi square. For TTCW and survival, Kaplan Meier survival curves were constructed and log-rank was performed to compare outcomes between the 3 groups. Cox regression analysis was used to assess baseline factors associated with TTCW. Examining the proportionality assumptions of the Cox proportional hazards model using both log-log curves and Schoenfeld residuals demonstrated that all assumptions were met. Univariate variables with a p<0.16 were included in the multivariate model [15], and stepwise regression with backward elimination was conducted, with a p threshold for elimination=0.20. The regression was also repeated using 1) forward selection and 2) a sensitivity analysis using a more stringent p value threshold (p<0.10) along with variables shown in prior SSc-PAH cohorts to predict outcome. Colinearity between hemodynamic variables was assessed using linear regression models to calculate variance inflation factors (VIF), with a VIF of >10 indicating that variables were collinear [16]. There was no significant colinearity between mean pulmonary artery pressure, cardiac output (CO) and pulmonary vascular resistance (PVR). Statistical analyses were conducted using Graph Pad Prism (version 5, La Jolla, CA) and Stata (version 13, College Station, TX), and a p value <0.05 was considered to be statistically significant. Some of these results were previously published as an abstract [17].

Results

Patient characteristics

There were no statistically significant differences in baseline characteristics between the 3 initial therapy groups (Table 1); of note, there were no differences in variables known to be associated with severity of disease or poor prognosis, such as 6MWD, CO, or NYHA functional class [18–20]. There were no significant differences in the amount of time on initial oral therapy between the ERA (2.21 ± 1.94 years), PDE5i (1.98 ± 1.51 years), or combo ERA/PDE5i (2.00 ± 1.17 years) groups (p=0.10). There was no significant difference in the time from diagnostic RHC to start of therapy (ERA: median 32 days [IQR −1, 66], PDE5i: 18 days [5, 57], combo ERA/PDE5i: 1 day [−2, 47]; p=0.38). For the ERA group, 16 were treated with bosentan and 8 with ambrisentan; for the initial PDE5i group, 43 were on sildenafil and 16 tadalafil. Patients in the initial combination group were treated with sildenafil + bosentan (n=5), tadalafil + ambrisentan (n=8), or sildenafil + ambrisentan (n=2). Please see the Supplemental Figure for a description of medications that were added after the first 6 months of initial therapy in each of the groups. Follow-up for vital status (alive or dead) was complete in 88% of patients.

Table 1.

Baseline characteristics of 98 Group I SSc-PAH patients from the PHAROS registry

		ERA (n=24)	PDE5i (n=59)	combo ERA/PDE5i (n=15)	p value
Age (years)		61.8 ± 10.3	60.0 ± 11.1	60.3 ± 12.0	0.78
Sex (% female)		79%	83%	80%	0.53
SSc duration (years)		10.9 ± 9.8	10.6 ± 9.5	12.4 ± 11.5	0.82
SSc subtype (% limited)		71%	69%	93%	0.22
FVC (% predicted)		78.3 ± 19.7	78.6 ± 15.9	86.9 ± 13.0	0.23
DLCO (% predicted)		35.8 ± 16.6	43.9 ± 16.9	48.0 ± 18.5	0.08
FVC/DLCO ratio		2.0 ± 0.6	2.1 ± 0.8	2.0 ± 0.7	0.90
6MWD (meters)		373 ± 80	324 ± 130	355 ± 127	0.30
Creatinine (g/dL)		1.05 ± 0.50	1.03 ± 0.81	0.97 ± 0.32	0.93
mean PAP (mmHg)		38.0 ± 9.8	35.6 ± 8.4	39.3 ± 10.7	0.42
PAWP (mmHg)		9.3 ± 3.4	10.1 ± 3.1	9.5 ± 2.6	0.50
CO (L/min)		5.2 ± 2.0	5.1 ± 1.6	5.1 ± 1.8	0.95
PVR (Wood units)		6.3 ± 3.6	5.8 ± 4.6	7.7 ± 5.5	0.34
Pericardial effusion (%)		25%	35%	20%	0.49
Oxygen use (%)		29%	32%	47%	0.47
NYHA functional class
1		13%	12%	13%	0.96
2		38%	39%	33%
3		50%	47%	47%
4		0%	2%	7%
Time on initial therapy (years)		2.21 ± 1.94	1.98 ± 1.51	2.00 ± 1.17	0.10
Time from RHC to start of therapy (median days)		32 [IQR −1, 66]	18 [IQR 5, 57]	1 [−2, 47]	0.65

Abbreviations: ERA=endothelin-receptor antagonist; PDE5i=phosphodiesterase-5 inhibitor; SSc=systemic sclerosis; FVC=forced vital capacity; DLCO=diffusion capacity for carbon monoxide; 6MWD=six-minute walk distance; PAP=pulmonary artery pressure; PAWP=pulmonary artery wedge pressure; CO=cardiac output; PVR=pulmonary vascular resistance; NYHA=New York Heart Association; RHC=right heart catheterization; IQR=interquartile range

Time to Clinical Worsening

For the TTCW analysis, there were a total of 187.6 patient-years at risk; the mean time at risk was 2.0 ± 1.0 years. Overall, the qualifying event for TTCW was death in 8 patients (8%), hospitalization in 14 (14%), and initiation of parenteral prostanoid in 8 (8%); 68 (69%) patients did not have a TTCW qualifying event over the 3 year observation period (Figure 2). Ten patients (41.6%) in the ERA group died over the 3 year observation period, compared to 4 (6.8%) in the PDE5i group and 1 (6.7%) in the combo ERA/PDE5i group (p=0.004). Patients initially treated with an ERA had a significantly shorter TTCW (p=0.0001, Figure 3), with a median TTCW of 2.4 years. The percentage of patients free of clinical worsening was lower in the ERA group (63.0% at 1 year, 52.3% at 2 years, and 34.8% at 3 years) compared to the initial PDE5i (85.9%, 83.7%, 80.8%) or combo ERA/PDE5i (85.7%, 77.9%, 68.2%) groups.

Figure 2.

Qualifying events for time to clinical worsening (all-cause death, PAH-related hospitalization, initiation of parenteral prostanoid, or no clinical worsening event for TTCW during the observation period). Data are expressed as percentage of patients having each qualifying event, stratified by initial oral therapy group. No patients experienced transplant or worsening symptoms as their TTCW event.

Figure 3.

Kaplan-Meier curves for time to clinical worsening (see Methods for definition). Data are expressed as percentage of patients in each group free from of a clinical worsening event.

Univariate factors associated with a shorter time to clinical worsening were initial ERA use (compared to initial PDE5i, hazard ratio 5.0, p<0.0001), lower DLCO (HR 0.73 per 10% of predicted change, p=0.04), lower PAWP (HR 0.88 per 1 mmHg change, p=0.02), lower CO (HR 0.77 per 1L/min change, p=0.03), higher PVR (HR 1.11 per Wood unit change, p=0.001), and a shorter duration on their initial therapy (HR 0.96 per 50 day change, p=0.05) (Table 2). In a multivariate analysis, the factors independently associated with a shorter TTCW were initial medication (ERA vs. PDE5i or combo, HR 2.63, p=0.009), lower baseline DLCO (HR 0.69 per 10% of predicted change, p=0.04), and higher PVR (1.10 per Wood unit change, p=0.007) (Table 3). When age, sex, and SSc disease duration were forced into the model, the results of the multivariate analysis were not appreciably changed. Additionally, when using forward selection or using a sensitivity analysis to limit the number of variables, there was no significant change in the results (data not shown).

Table 2.

Univariate factors associated with shorter time to clinical worsening

	HR	95%	CI	p
FVC, per 10% of predicted change	0.92	0.73	1.16	0.48
ERA vs. PDE5	5.00	2.27	11.11	<0.0001
ERA vs. combo	2.94	0.99	9.09	0.053
DLCO, per 10% of predicted change	0.73	0.54	0.99	0.04
6MWD, per 20 meter change	0.99	0.93	1.05	0.65
mPAP, per 10mmHg change	1.43	0.97	2.10	0.07
PCWP, per 1mmgHg change	0.88	0.78	0.98	0.02
CO, per 1L/min change	0.77	0.60	0.98	0.03
PVR, per Wood unit change	1.11	1.05	1.18	0.001
Age, per 10 year change	1.28	0.9	1.80	0.17
SSc duration, per year change	0.99	0.95	1.03	0.66
Sex, female vs. male	1.05	0.36	3.01	0.93
NYHA	1.26	0.73	2.20	0.41
SSc subtype, limited vs. diffuse	0.91	0.39	2.14	0.84
Initial med days, per 50 day change	0.96	0.92	0.99	0.046
Time to start initial med, per 50 day change	0.98	0.65	1.47	0.92

See Table 1 for abbreviations. Parameters in bold were included in the multivariate model

Table 3.

Stepwise backward regression for variables associated with shorter time to clinical worsening

	HR	95%	CI	p
Initial med, ERA vs. others	2.63	1.28	5.56	0.009
DLCO, per 10% of prediced change	0.69	0.49	0.99	0.04
PVR, per Wood unit change	1.10	1.03	1.18	0.007
PCWP, per 1mmHg change	0.89	0.77	1.02	0.10

See Table 1 for abbreviations. CO, mPAP, and initial med days were removed from the stepwise backwards regression because their p>0.2. Repeat analyses using 1) regression with forward selection or 2) a sensitivity analysis combining a more stringent p value threshold (p<0.10) and clinical previously shown to predict outcomes in SSc-PAH did not appreciably change these results.

Discussion

In this retrospective cohort study utilizing data from the prospective PHAROS registry of SSc patients with PAH, we found that initial therapy with an ERA alone was associated with worse outcomes compared to either a PDE5i alone or a combination of ERA and PDE5i. This unexpected finding was not explained by differences in measured baseline variables between the groups. After adjustment for other prognostic variables, initial ERA use was associated with more than a 2.5-fold increased risk for clinical worsening.

The optimal initial therapy for patients with SSc-PAH is undefined. While SSc patients have been included in most of the pivotal studies resulting in FDA approval for the currently available oral medications, there are limited analyses of the SSc subgroup. Furthermore, when analyzed they often did not show significant results, perhaps because the group sizes were relatively small. In the BREATHE-1 trial, a subgroup analysis of 66 patients with connective tissue disease (CTD)-related PAH (84% with SSc) randomized to bosentan or placebo for 12–16 weeks showed a non-significant 19 meter improvement in 6MWD, favoring bosentan [6]. In a subgroup analysis of the long-term SERAPHIN study of macitentan versus placebo, 185 patients with CTD-PAH had a non-statistically significant improvement in morbidity and mortality on the ERA [7]. In both the SUPER-1 study (sildenafil) and the PHIRST-1 study (tadalafil), the CTD-PAH subgroups had improvements in 6MWD after 12 weeks of the PDE5i [8, 9]. It is important to note that all of these studies compared the active agent to placebo (with or without background therapy), rather than head-to-head comparisons of ERA versus PDE5i. Further, with the exception of the SUPER-1 study (in which less than half of the CTD population had SSc), none of the other studies specified the proportion of patients with SSc-PAH compared to other CTD-PAH. Since the response to therapy within CTD-PAH subtypes varies and favors patients with non-SSc-PAH, the true response to therapy in SSc-PAH populations is not well described in clinical trials [21]. It does appear, however, that early recognition and treatment of PAH in SSc patients is associated with improved outcomes [22].

The recently published AMBITION study showed superiority for upfront combination therapy with ambrisentan and tadalafil compared to either medication alone for the PAH group as a whole [23]. Thirty-seven percent of included subjects in this study had CTD-PAH, although the number of patients with SSc was not reported and the effect on the CTD-PAH subgroup in this trial is not yet fully described. It is important to highlight that the SSc-PAH subgroup tends to respond less favorably to medical therapies that are beneficial in the PAH group as a whole [6, 7] and that other CTD-PAH subgroups, such as those with lupus [21], tend to do better than those with SSc. Although speculative, our findings may have also conflicted with those in AMBITION due to pharmacologic differences. The majority of ERA patients (67%) in this PHAROS analysis were treated with bosentan, which is a non-selective endothelin receptor antagonist. As noted above, bosentan (the predominant ERA prescribed in our analysis) did not show a significant benefit in the CTD-PAH subgroup in BREATHE-1 [6]. Long-term blockade of the ET-B receptor theoretically may be detrimental by promoting vasoconstriction and smooth muscle cell proliferation [24, 25]. Selective blockade of the ET-B receptor has been shown, in healthy individuals, to cause increased renal vascular resistance, impaired clearance of endothelin by the kidneys, and increased levels of circulating ET-1 [26]. This pharmacologic difference may limit extrapolation of our results to patients treated with the ET-A selective ERA ambrisentan (such as in AMBITION) or macitentan, which is a non-selective ERA but has much higher affinity for the ET-A receptor [27].

It is possible that our unexpected study results may be related to differences in patient characteristics between the groups; for example, if the patients started on ERAs were sicker at baseline, had a worse overall prognosis, or had a delay to amplification of therapy, this may explain their shorter TTCW and survival. However, there were no differences in any of the measured baseline characteristics or factors known to be associated with a poor prognosis, and the time to starting an additional agent was similar between the 3 groups. Additionally, ERA use was independently associated with worse outcomes even after adjusting for other prognostic factors. We also utilized a landmark analysis with a 6-month landmark to minimize the biases in a time-to-event outcome inherent in this observational study. This method allows for an unbiased estimate of conditional time-to-event probabilities, but is limited by the impact of the actual time landmark chosen; namely, an early landmark may lead to misclassification of events at a later time while a late landmark may lead to an omission of a large number of events and thus result in a significant loss of power [13]. Our selection of a 6-month landmark was based upon data from the SERAPHIN study of macitentan in PAH [7]. In this study, approximately 10% of treatment-naïve subjects receiving study drug and nearly 30% of those on placebo experienced clinical worsening by 6 months, highlighting the clinical relevance of this time-point. Selection of this time-point as the landmark minimizes the bias introduced by inclusion of patients in the study who have poorer prognosis and are more likely to fail early regardless of intervention.

Our analysis was not able to answer any mechanistic questions, but there are pharmacologic factors that could explain the worse outcomes seen with ERAs in our study. Fluid retention is a common side effect of ERAs, occurring in 6–28% of patients in clinical studies [28, 29], even being severe enough to cause drug discontinuation [30]. This potential for fluid retention may have driven TTCW events such as PAH-related hospitalization and prostacyclin initiation. Another potential mechanistic difference is the influence of PDE5-inhibition on right ventricular (RV) structure [31, 32], as changes in the RV are a strong determinant of outcomes in PAH [33–35].

Our study has several strengths. As opposed to most prior drug treatment studies which only assessed effects over a short period (12–20 weeks), this analysis had long-term follow-up, with a mean period of 2.4 years for the survival analysis. Medications that have benefit over a short time period may have detrimental effects that do not become evident until there is longer-term exposure. Additionally, our patients were carefully characterized at baseline, and follow-up for survival was 88%.

Our study also has limitations. Although the registry data were collected prospectively, this analysis was retrospective and therefore has inherent limitations. While we adjusted for many prognostic variables, there may be unmeasured factors that were unbalanced between groups and led to the differential outcomes. In particular, right atrial pressure, a well-described prognostic factor [18, 19, 36], was not collected in this dataset, although other important baseline hemodynamic variables were not statistically different between the 3 groups. Although our number of patients was larger than most prior SSc-PAH subgroup analyses, it is still a small sample size and there is a risk for Type II error, in that we may have not found differences in confounding baseline characteristics (e.g. DLCO) that may have driven the results. Our results are not definitive and require further study. By excluding those who did not have 6 months of follow-up, we could have introduced bias against sicker patients who may have died sooner. As discussed above, we chose this landmark analysis approach [13] to estimate time-to-event probabilities in an unbiased fashion. Of the 17 patients who were excluded for less than 6 months of follow-up, 9 of them simply had only one study visit. Of the 8 patients who died prior to 6 months, six of them were on prostanoid therapy, and thus would have been excluded from this analysis even if they had 6 months of follow-up. There were a limited number of deaths during the observation period, and therefore we were unable to perform a separate Cox regression survival analysis; however, it did appear that the rate of death was higher in the initial ERA group. The limited number of events may be explained by the fact that PHAROS was conducted at SSc centers focused on early detection and treatment of PAH. The effect of clinical center or individual practitioner preference of initial oral therapy may have influenced the results; however, the large number of centers prevents controlling for these factors. Since some patients progressed to a second oral agent, some progressed to inhaled prostanoid, and some had parenteral therapy initiated, a time-varying analysis was not performed. Lastly, outcomes such as hospitalization and initiation of parenteral prostacylins were not protocolized; however we feel that this could be a strength, as this may more accurately represent clinical care.

Conclusions

Compared to treatment with a PDE5i or combination ERA/PDE5i, initial therapy with an ERA was associated with significantly worse 3 year TTCW and survival in patients with SSc-PAH, even after controlling for other prognostic variables. This study highlights the need for randomized controlled trials performing head-to-head comparisons between PAH-specific medications specifically in SSc as well as registry studies with long-term clinical follow-up to properly define “real world” outcomes.

Supplementary Material

Supp FigureS1. Supplemental Figure.

Medications added after the 6 month period of initial therapy. “No change” means that the patient remained on their initial therapy for the duration of the observation period without another PH-specific medication added.