Of the 1156 patients randomized, 218 patients (88 in the placebo group and 130 in the selexipag group) prematurely discontinued treatment; defined as discontinuing treatment without experiencing a primary end-point event. The median treatment duration was 24.0 weeks for the placebo-treated patients and 16.7 weeks for the selexipag-treated patients. The amount of missing observation time relative to the maximum theoretical
observation time for the primary end point was 16.3% in the placebo group and 18.3% in the selexipag group.
According to the protocol and the statistical analysis plan, all randomized patients were included in, and contributed to, the primary end-point analysis, which is presented in the manuscript and in Table S1. In this analysis, the 218 patients who prematurely discontinued study treatment were censored at the time of treatment discontinuation. Importantly, these patients provided information during their time on study treatment for the Kaplan–Meier estimates and the log-rank tests.
Sensitivity analyses were performed on the primary end point to assess the impact of premature treatment discontinuations on the estimate of the treatment effect, as follows:
(a) The first sensitivity analysis (Table S1), prespecified in the statistical analysis plan, consisted of imputing a morbidity event in patients with signs of disease progression at the time of premature treatment discontinuation. Patients with signs of disease progression were identified if they met any of the following prospectively-defined criteria:
1) Last available WHO functional class is greater than baseline functional class 2) Last available 6-minute walk distance decreased by at least 15% of the baseline
value with no evidence that this decrease is unrelated to pulmonary arterial hypertension worsening
3) Pulmonary arterial hypertension medication started within 4 weeks prior to treatment discontinuation
4) An adverse event suggestive of ongoing pulmonary arterial hypertension worsening at treatment discontinuation
This analysis identified 67 additional patients with events (placebo, n=32; selexipag, n=35).
This reduced the proportion of patients censored at the time of treatment discontinuation from 19% (218 of 1156) to 13% (151 of 1156).
b) The second sensitivity analysis was conducted post hoc and consisted of imputing a primary end-point event at the time of treatment discontinuation for all 218 patients who prematurely discontinued treatment (Table S1). The assumption for this analysis was that treatment discontinuation was a signal for treatment failure.
(c) The third sensitivity analysis, prespecified in the statistical analysis plan, was conducted as an exploratory end point, whereby the primary outcome measure was analyzed up to the end of the study. Death and morbidity events occurring after treatment discontinuation were collected for the 80 patients who consented to be followed during the posttreatment
observation period. Deaths alone were collected for the 138 patients who did not consent to be followed during the posttreatment observation period (Table S1). This analysis contained the 397 events (placebo, n= 242; selexipag, n=155) collected in the primary analysis and 46
additional events collected during the posttreatment observation period (placebo n=16;
selexipag, n=30).
2. All-cause Mortality Up to End of Study
At the end of study, vital status was recorded for 1101 patients (placebo, n=552; selexipag, n=549) and was unknown for 55 patients (placebo, n=30; selexipag, n=25). These 55 patients included 45 patients who prematurely discontinued treatment and 10 patients who discontinued treatment subsequent to a primary end point event. Overall, for all-cause mortality up to the end of the study, the amount of missing observation time relative to the maximum theoretical observation time was 3.7% in the placebo group and 2.4% in the selexipag group.
According to the protocol and statistical analysis plan, all randomized patients were included in the analysis set for survival regardless of premature treatment or study discontinuation.
The 55 patients (5%) with unknown vital status at the end of study were censored at the date of last contact. These patients provided information on survival during their time in the study for the Kaplan–Meier estimates and the log-rank tests; these data are presented in the manuscript and Table S4.
(a) A sensitivity analysis, conducted post hoc, consisted of imputing a death at the date of last contact for the 55 patients with missing vital status at the end of the study (Table S4).
The assumption for this analysis was that lack of vital status information at the end of the study could indicate that the patient had died.
9. IMPUTATION RULES FOR MISSING 6-MINUTE WALK DISTANCE
Missing 6-minute walk distance values at week 26 were imputed based on the following prespecified rules:
1) For patients unable to walk at week 26, 0 meters was imputed
2) If the above rule 1) did not apply, 10 meters was imputed. The 10-meter value was the second lowest observed 6-minute walking distance value at week 26, irrespective of study treatment group.
10. FIGURE S1. DOSE ADJUSTMENT (TITRATION) SCHEME
Figure S1. Dose Adjustment.
Within 28 days after screening, patients were randomly assigned, in a 1:1 ratio stratified by center, to placebo or selexipag. During the 12-week dose-adjustment phase, study drug was initiated at 200 µg twice daily and titrated weekly in 200 µg twice-daily increments until prostacyclin-associated adverse effects such as headache and jaw pain were unmanageable. The dose was then decreased by 200 µg in both daily doses, giving the highest tolerated dose. The maximum dose allowed was 1600 µg twice daily. After 12 weeks, patients entered the maintenance phase of the study. From week 26, doses could be increased at scheduled visits; dose reductions were allowed at any time.
11. FIGURE S2. GRIPHON STUDY DESIGN
Figure S2. Study Design.
Within 28 days after screening, patients were randomly assigned, in a 1:1 ratio stratified by center, to placebo or selexipag. Study drug was initiated and the dose adjusted as outlined in section 10 above.
The end of treatment was either at the end of the study, after the occurrence of a primary end-point event or occurred prematurely for various reasons, such as an adverse event. The end of treatment differed for each patient and was defined as the last intake of double-blind treatment (placebo or selexipag) plus 7 days. The end of the study was declared when the prespecified number of primary end-point events was reached. Change in 6-minute walk distance, absence of worsening in WHO functional class and change in NT-proBNP were evaluated at week 26. Vital status was recorded at the end of the study.
Patients who discontinued double-blind treatment and provided written informed consent for further follow-up were followed until the end of the study, in the blinded posttreatment observation period (see section 7 above). Patients who had a nonfatal primary end-point event discontinued double-blind treatment and were eligible to receive open-label selexipag or commercially available drugs, as were patients receiving double-blind treatment at the end of the study.
11. FIGURE S3. EFFECT OF SELEXIPAG ON THE PRIMARY COMPOSITE END POINT EXCLUDING EVENTS THAT OCCURRED PRIOR TO INCREASING THE SAMPLE SIZE
Figure S3. Primary Composite End Point Excluding the 45 Events That Occurred Prior to Increasing the Sample Size (Kaplan–Meier Estimates).
Shown are Kaplan–Meier curves for the primary composite end point of death (from any cause) or a complication related to pulmonary arterial hypertension (disease progression, or worsening of pulmonary arterial hypertension that resulted in hospitalization, initiation of parenteral prostanoid therapy or long-term oxygen therapy, or the need for lung transplantation or balloon atrial septostomy) up to the end of the treatment period in the selexipag and placebo groups, excluding the 45 events that occurred prior to increasing the sample size show a significant treatment effect in favor of selexipag versus placebo (hazard ratio 0.61, 99% CI, 0.46 to 0.81, P<0.001 with the use of a one-sided log-rank test). The analysis took into account all available data, whereas the Kaplan–Meier curve is truncated at 36 months.
12. FIGURE S4. EFFECT OF SELEXIPAG ON THE PRIMARY COMPOSITE END POINT IN PATIENTS GROUPED BY PRESPECIFIED SELEXIPAG INDIVIDUAL MAINTENANCE DOSE VERSUS PLACEBO
Figure S4. Primary Composite End Point in Patients Grouped by Prespecified Selexipag Individual Maintenance Dose vs. Placebo (Kaplan–Meier Estimates).
Shown are Kaplan–Meier curves for the primary composite end point of death (from any cause) or a complication related to pulmonary arterial hypertension (disease progression, or worsening of pulmonary arterial hypertension that resulted in hospitalization, initiation of parenteral prostanoid therapy or long-term oxygen therapy, or the need for lung transplantation or balloon atrial septostomy) up to the end of the treatment period in the selexipag and placebo groups, according to prespecified selexipag individual maintenance dose strata: low (200, 400 µg twice daily), medium (600, 800, 1000 µg twice daily), and high (1200, 1400, 1600 µg twice daily). The individualized maintenance dose was defined as the dose that a patient received for the longest duration. The effect of selexipag was consistent in the low (hazard ratio 0.60, 95% CI, 0.41 to 0.88), medium (hazard ratio 0.53, 95% CI, 0.38 to 0.72), and high (hazard ratio 0.64, 95% CI, 0.49 to 0.82) dose strata. The analysis took into
account all available data, whereas the Kaplan–Meier curve is truncated at 30 months. In the selexipag group, 14 patients discontinued the 200 µg twice-daily dose during the dose-adjustment phase and 1 patient received doses different from the per protocol dosing, all 15 were assigned to 0 µg and are not reported here. The medium dose stratum (n=180) includes one patient treated with 900 µg twice daily.
13. FIGURE S5. EFFECT OF SELEXIPAG ON THE PRIMARY COMPOSITE END POINT BY PRESPECIFIED SUBGROUP
Figure S5. Effect of Selexipag on the Primary Composite End Point by Prespecified Subgroup.
Shown is the forest plot for the primary composite end point of death (from any cause) or a complication related to pulmonary arterial hypertension (disease progression, or worsening of pulmonary arterial hypertension that resulted in hospitalization, initiation of parenteral prostanoid therapy or long-term oxygen therapy, or the need for lung transplantation or balloon atrial septostomy) up to the end of the treatment period in the selexipag and placebo groups, according to prespecified subgroups. A consistent treatment effect was observed across prespecified subgroups: pulmonary arterial hypertension therapy at baseline, WHO functional class at baseline, sex, age at screening, pulmonary arterial hypertension etiology, and geographical region. The vertical solid line references the overall treatment effect.
14. TABLE S1. SENSITIVITY ANALYSES (A, B, AND C) OF THE PRIMARY ANALYSIS
Table S1. Sensitivity Analyses (a, b, and c) of the Primary Analysis.
Selexipag vs. Placebo
Hazard Ratio (CI) P Value*
Primary analysis 0.60 (99%, 0.46–0.78) <0.001
Sensitivity analysis (a) 0.65 (95%, 0.54–0.78) <0.001
Sensitivity analysis (b) 0.82 (95%, 0.70–0.96) 0.007
Sensitivity analysis (c)† 0.65 (95%, 0.54–0.79) <0.001
CI denotes confidence interval.
* One-sided unstratified log-rank.
Sensitivity analysis (a): The occurrence of a complication related to pulmonary arterial hypertension (a morbidity event) was imputed for patients with signs of disease progression, at the time of premature treatment discontinuation.
Sensitivity analysis (b): a primary end-point event was imputed for patients who prematurely discontinued treatment, at the time of treatment discontinuation. Three patients who withdrew consent and did not receive study drug were included as having an event on day 1.
† Exploratory end point of time to first primary end-point event up to the end of study. Sensitivity analysis (c) contained the 397 first events collected in the primary end-point analysis and 46 additional events collected during the posttreatment observation period.
15. TABLE S2. INDIVIDUAL MAINTENANCE DOSE
Table S2. Individual Maintenance Dose.*
Prespecified Dose Strata Twice-daily Dose
Placebo
Low-dose stratum 200 µg 15 (2.6) 68 (11.8)
400 µg 18 (3.1) 65 (11.3)
Medium-dose stratum 600 µg 20 (3.4) 62 (10.8)
800 µg 21 (3.6) 82 (14.3)
1000 µg 27 (4.6) 35 (6.1)
High-dose stratum 1200 µg 20 (3.4) 42 (7.3)
1400 µg 55 (9.5) 41 (7.1)
1600 µg 393 (67.5) 163 (28.4)
Other than per-protocol dosing 4 (0.7) 2 (0.3)
* The individual maintenance dose was defined as the twice-daily dose that a patient received for the longest duration in the maintenance period, or, for patients who did not enter the maintenance period, as the highest tolerated twice-daily dose that a patient received during the dose-adjustment period.
Patients (placebo, n=9; selexipag, n=14) in the 200 µg twice-daily dose strata who prematurely discontinued the study were assigned to 0 µg and are not reported here.
16. TABLE S3. ABSENCE OF WORSENING IN FUNCTIONAL CLASS FROM BASELINE TO WEEK 26
Table S3. Patients With and With No Worsening in WHO Functional Class (FC) from Baseline to Week 26. All patients.
Baseline
FC denotes WHO functional class.
* Missing data at week 26 were included in the analysis as worsening.
17. TABLE S4. SENSITIVITY ANALYSIS (A) OF THE SECONDARY END POINT ALL-CAUSE MORTALITY UP TO END OF STUDY ANALYSIS
Table S4. Sensitivity Analysis (a) of the Secondary End Point All-cause Mortality up to End of Study Analysis.
Selexipag vs. Placebo
Hazard Ratio (95% CI) P Value*
All-cause mortality, protocol-defined analysis 0.97 (0.74–1.28) 0.42
Sensitivity analysis (a) 0.95 (0.74–1.21) 0.33
CI denotes confidence interval.
* P values were calculated with the use of a one-sided unstratified log-rank test.
Sensitivity analysis (a): death was imputed at the date of last contact for patients who prematurely discontinued treatment without a primary end-point event or discontinued treatment with a nonfatal primary end-point event and had missing survival status at the end of the study.
18. TABLE S5. CHANGE IN N-TERMINAL PRO-BRAIN NATRIURETIC PEPTIDE (NT-proBNP) FROM BASELINE TO WEEK 26 Table S5. Change in N-terminal pro-brain natriuretic peptide (NT-proBNP)* from Baseline to Week 26.
Placebo
* NT-proBNP was quantified using Elecsys® 2010 (Roche Diagnostics, Indianapolis, IN).
† Absolute median change was calculated on paired data.
Only patients with a non-missing value at both the baseline and week 26 are included.
19. TABLE S6. PROSTACYCLIN-ASSOCIATED ADVERSE EVENTS REPORTED IN THE DOSE-ADJUSTMENT AND MAINTENANCE PHASES
Table S6. Prostacyclin-associated Adverse Events Reported in the Dose-Adjustment and Maintenance Phases.*
Dose Adjustment Maintenance Patients with ≥1 prostacyclin-associated
adverse event – no. (%) 303 (52.5) 499 (86.8) <0.001 238 (46.9) 367 (72.1) <0.001
Adverse event – no. (%)
Headache 163 (28.2) 363 (63.1) <0.001 99 (19.5) 203 (39.9) <0.001
Diarrhea 67 (11.6) 205 (35.7) <0.001 67 (13.2) 151 (29.7) <0.001
Nausea 79 (13.7) 163 (28.3) <0.001 51 (10.0) 100 (19.6) <0.001
Pain in jaw 22 (3.8) 140 (24.3) <0.001 20 (3.9) 105 (20.6) <0.001
Myalgia 26 (4.5) 86 (15.0) <0.001 16 (3.1) 48 (9.4) <0.001
Vomiting 25 (4.3) 86 (15.0) <0.001 28 (5.5) 39 (7.7) 0.21
Pain in extremity 29 (5.0) 84 (14.6) <0.001 31 (6.1) 66 (13.0) <0.001
Flushing 22 (3.8) 60 (10.4) <0.001 16 (3.1) 52 (10.2) <0.001
Dizziness 51 (8.8) 57 (9.9) 0.55 58 (11.4) 53 (10.4) 0.62
Arthralgia 30 (5.2) 41 (7.1) 0.18 27 (5.3) 47 (9.2) 0.02
Musculoskeletal pain 5 (0.9) 17 (3.0) 0.01 10 (2.0) 11 (2.2) 1.00
Temporomandibular joint syndrome 2 (0.3) 3 (0.5) 0.69 1 (0.2) 3 (0.6) 0.62
* Of the patients randomized to placebo, four did not receive study drug and were excluded from the safety analysis, and one received a single dose of eight tablets of selexipag and was assigned to the selexipag group for the safety analysis.