Population Pharmacokinetic/Pharmacodynamic and Exposure‐Response Modeling Analyses of Sotatercept in Healthy Participants and Patients with Pulmonary Arterial Hypertension

Abstract

Sotatercept is a breakthrough, first‐in‐class biologic, recently approved by the Food and Drug Administration (FDA) for the treatment of pulmonary arterial hypertension (PAH). Exposure‐response (E‐R) analyses and pharmacokinetic/pharmacodynamic (PK/PD) modeling were performed for sotatercept after intravenous and subcutaneous (SC) administrations. Clinical endpoints included 6‐minute walk distance (6MWD), pulmonary vascular resistance (PVR), and probability of N‐terminal pro‐B natriuretic peptide (NT‐proBNP) concentrations < 300 pg/mL for efficacy, and hemoglobin (Hgb) for safety from two Phase 1 studies, two Phase 2 studies, and one Phase 3 study. E‐R models using nonlinear mixed effect modeling approach were developed for 6MWD and PVR, while Cox proportional hazards model and semi‐mechanistic PK/PD model were used for NT‐proBNP and Hgb. Covariate analyses were conducted to identify significant predictors of variability for each of these clinical endpoints. Modeling results showed that increasing sotatercept average concentration (C _avg) at week 24 is associated with increased predicted 6MWD, increased probability of NT‐proBNP concentration < 300 pg/mL, decreased predicted PVR, and increased Hgb which was clinically manageable. All these responses approached their corresponding plateaus at a C _avg range associated with the dose of 0.7 mg/kg Q3W SC. Statistically relevant covariates included age and iron supplementation which slightly increased Hgb‐mediated effect for 6MWD, PAH disease duration, and baseline therapy infusion with prostacyclin for PVR, and WHO functional class for NT‐proBNP. The magnitudes of the impact of these covariates are not clinically meaningful. Taken together, these results support an appropriate benefit–risk profile for the FDA‐approved target dose for sotatercept of 0.7 mg/kg Q3W SC.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

PAH is a rare disease causing a progressive increase in PVR resulting in right ventricular dysfunction, failure, and ultimately premature death. Sotatercept is the first activin signaling inhibitor therapy for PAH, representing a new class of therapy that works by improving the balance between pro‐ and anti‐proliferative signaling to regulate vascular cell proliferation underlying PAH.

• WHAT QUESTION DID THIS STUDY ADDRESS?

Data from two Phase 1 studies in healthy PMW, two Phase 2 studies, and 1 Phase 3 study in participants with PAH were used to develop a model for sotatercept after IV and SC administrations to evaluate its influence on hemoglobin, a safety endpoint and to characterize the relationships between sotatercept exposures with efficacy endpoints 6MWD, PVR, and NT‐proBNP.

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

The PK/PD model for hemoglobin as well as E‐R models for efficacy endpoints adequately characterized the observed data in all five clinical trials. The predicted 6 MWD and predicted probability of NT‐proBNP <300 pg/mL increased while the predicted PVR decreased with increasing sotatercept exposures approaching a plateau at the corresponding C _avg range of the clinical dose of 0.7 mg/kg SC Q3W, supporting the approved FDA‐label dose.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

This study has improved the scientific understanding of sotatercept and demonstrated that the dosing schedule of sotatercept is optimal for balancing efficacy and safety in participants with PAH.

Pulmonary arterial hypertension (PAH) applies to a group of diseases causing a progressive increase in pulmonary vascular resistance (PVR), resulting in right ventricular dysfunction and ultimately failure leading to premature death. ¹ The PAH pathophysiology involves pulmonary endothelial dysfunction, resulting in impaired production of vasodilators such as nitric oxide and prostacyclin, and overexpression of vasoconstrictors, abnormal proliferation of pulmonary vascular smooth muscle cells in pulmonary arterioles, which results in progressive pulmonary vascular remodeling, increased PVR, and eventually right‐sided heart failure. ² In the absence of treatment, the majority of patients succumb to heart failure within a few years of diagnosis. ³ Current background PAH therapies focus on symptom management by increasing blood flow through the pulmonary vasculature via pharmacologic manipulation of various pathways to relieve symptoms and slow clinical worsening of the disease. In addition to general supportive care agents (e.g., anticoagulants, diuretics, digoxin), current disease‐specific treatments for PAH include vasodilator‐type agents such as endothelin‐receptor antagonists, phosphodiesterase inhibitors, and prostanoids, in addition to general supportive care agents (e.g., anticoagulants, diuretics, digoxin).

Sotatercept, a recombinant fusion protein targeting the activin receptor type IIA (ActRIIA), has been recently approved by Food and drug Administration (FDA) for the treatment of PAH. Clinical trials in PAH patients have demonstrated the efficacy of sotatercept in reducing PVR and enhancing exercise capacity measured by the 6‐minute walk distance (6MWD). ⁴ , ⁵ Notably, a phase 2 trial ⁵ revealed a significant reduction in PVR at Week 24, with least‐squares mean difference (LSMD) in change from baseline at −145.8 dyn.sec/cm ⁵ (95% confidence interval [CI], −241.0 to −50.6; P = 0.003) between 0.3 mg/kg Q3W and placebo groups, and at −239.5 dyn.sec/cm ⁵ (95% CI, 329.3 to −149.7; P < 0.001) between 0.7 mg/kg Q3W and placebo groups. In this trial, both dose arms exhibited a substantial improvement of 6MWD, with LSMD in change from baseline at Week 24 at 29.4 m (95% CI, 3.8–55.0) for the 0.3 mg/kg Q3W arm and 21.4 m (95% CI, −2.8–45.7) for the 0.7 mg/kg Q3W arm. ⁵ In a subsequent phase 3 trial, sotatercept demonstrated a consistent reduction in PVR and an improvement of 15% in exercise capacity after 24 weeks of treatment. ⁴

A population PK model for sotatercept was previously published. ⁶ The objectives of the present work are to describe the relationships between sotatercept exposure and important efficacy and safety endpoints. More specifically, to describe the pharmacokinetic/pharmacodynamic (PK/PD) relationship between sotatercept and hemoglobin (Hgb, a safety endpoint) in adults including healthy participants and patients with PAH, and to characterize the relationships between sotatercept exposure with efficacy endpoints including 6MWD, PVR, and time to N‐terminal pro‐B natriuretic peptide (NT‐proBNP) to achieve serum concentrations < 300 pg/mL in patients with PAH.

METHODS

Clinical trials' designs and populations

Data used included two phase 1 trials in healthy post‐menopausal women (PMW) as a single ascending dose (SAD) study and a multiple ascending dose (MAD) study, two phase 2 trials (SPECTRA and PULSAR), and one phase 3 trial (STELLAR) in participants with PAH (Figure 1 ). All trials were conducted in accordance with the principles of the Declaration of helsinski, the ethical guidelines of the Council for International Organizations of medical Sciences, the Good Clinical Practice guidelines of the International Council of Harmonization, and all applicable laws and regulations. An institutional review board or independent ethics committee at each trial center approved the protocol and all participants provided written informed consent. In SAD, participants received a single subcutaneous (SC) injection of placebo or sotatercept at 0.03 or 0.1 mg/kg intravenously (IV) or a dose in the range of 0.01–3 mg/kg over 1‐hour infusion. In MAD, participants were planned to receive 1 SC injection of placebo or sotatercept in a dose range of 0.1–1 mg/kg on Days 1, 29, 57, and 85; not all doses were administered in this study. ⁷ In PULSAR, participants received 1 SC injection every 3 weeks (Q3W) for 24 weeks of either placebo or sotatercept at 0.3 or 0.7 mg/kg. Participants were monitored up to 30 months after Week 24. In SPECTRA and STELLAR, participants received 1 SC injection Q3W for 24 weeks at an initial dose of 0.3 followed by 0.7 mg/kg Q3W. In all trials PULSAR, SPECTRA, and STELLAR, dose modifications (delay, reduction to 0.3, 0.1, or 0.05 mg/kg, or discontinuation) were possible based on increased Hgb and decreased platelets. ⁸ For each efficacy and safety endpoint, all participants who contributed to a baseline measurement and at least 1 additional measurement after the first dose were included in the final analysis dataset. Further details on participants' dosing and sampling are reported in Supplementary Material S1 .

Figure 1.

General schematic of all clinical trials included in the population modeling analyses. *Includes samples from PULSAR participants who switched from placebo to treatment. PAH, pulmunary arterial hypertension; PMW, post‐menopausal women.

Data analyses and procedures

Data analyses were performed using SAS Version 9.4 ⁹ and KIWI Version 2022R1. ¹⁰ Population modeling was performed using NONMEM®, Version 7, Level 3.0. ¹¹ NT‐ProBNP modeling using Cox's proportional hazards modeling approach was conducted in R Software Version 4.1.3 ¹²

Population modeling of PK/PD and E‐R relationships

A population modeling approach using nonlinear mixed effect modeling (NONMEM) techniques was employed for developing exposure‐6MWD, and exposure‐PVR models. For exposure‐NT‐proBNP relationship, the endpoint was time to NT‐proBNP < 300 pg/mL (threshold associated with low risk according to the guidelines), ¹³ and a time‐to‐event modeling approach was used. Standard model development strategy was followed including exploratory data analysis, base structural model development, evaluation of covariate effects, model refinement, and model qualification, designed to properly assess the data and result in an unbiased and precise model, consistent with the quality and quantity of data available. ¹⁴ , ¹⁵

Assessing impact of covariates

Continuous and categorical covariates that were tested during model development are presented for the PK/6MWD, PK/PVR, and exposure‐NT‐proBNP (Table S1 ), and PK/Hgb (Table S2 ), and the hematological baselines for participants included in the PK/Hgb model (Table S3 ). The choice of covariates to be tested was based on scientific plausibility, prior knowledge, clinical relevance, and initial data exploration. The covariates tested on the different parameters from each PD variable are summarized in Table S4 .

Forward selection and backward elimination procedure. Forward selection of covariates was performed via stepwise analyses. Each covariate‐parameter relationship was tested for statistical significance. Based on α = 0.01, covariates decreasing the minimum value of objective function (VOF) of ≥ 6.64 (P < 0.01, 1 df) and resulting in a decrease in IIV of ≥ 5% in the parameter of interest were considered significant.

A stepwise backward elimination proceeded after any adjustments/refinements to the full multivariable model following the completion of the forward selection process. A covariate was considered significant at a decrease in VOF ≥ 10.83 (P < 0.001, 1df for χ² ‐distribution) when removed from the model. The backward elimination procedure was repeated until all remaining covariates were significant.

Model‐based simulations

To assess the clinical relevance of the efficacy and safety endpoints, stochastic simulations, including the contributions of covariates and parameter uncertainty, were performed using the final integrated E‐R models for 6MWD and PVR and PK/PD model for Hgb to predict efficacy and safety responses at Week 24 in a population of virtual patients with PAH. The virtual simulated population was defined by the observed covariate distributions in the STELLAR analysis population. The responses were predicted for standardized sotatercept exposures ranging from 0 (placebo) to 12,000 ng/mL, in 1,000 ng/mL increments, to include the expected range of average sotatercept concentrations within a dosing interval observed for a SC dose regimen of 0.7 mg/kg Q3W. For each sotatercept exposure level, simulations were replicated a total of 500 times to obtain and calculate the mean and the 90% CI of the mean PD responses.

RESULTS

Exposure‐response models for efficacy

6MWD

Observed 6MWD generally increased over time in participants treated with sotatercept in all 3 studies. In PULSAR, an increase in 6MWD was also observed for placebo patients, but the same was not true for the placebo patients enrolled in STELLAR. Six‐minute walk distance increased with increasing sotatercept exposures but appeared to reach a plateau. The final integrated Phase 2/3 E‐R model for 6MWD response was described by:

where 6MWD_BL is the baseline 6MWD in WHO Class III Patients (m); β ₁ is the slope of age effect centered at 48 years; Age is the patient's age in years; c is the median age among the Phase 2/3 population in years; Age_cut‐off is the cut‐off point for age in years; AgeI is the cut‐off point for age (1 = Age ≥ 48.7, 0 = Age < 48.7); β₂ is the proportional shift in WHO Class II Patients for 6MWD; $\mathrm{WHO}$ is the WHO functional class (0 Class III, 1 Class II); β₃ is the slope of hemoglobin Effect (m*dL/g); $\mathrm{Hgb}$ is the time‐varying value of hemoglobin; E _max is the maximum time‐dependent change in 6MWD (m); ET₅₀ is time to reach half E _max (month); $t$ is the 6MWD measurement time (in months); $C_{\mathrm{avg}}\left(t\right)$ is the average sotatercept concentration at time t; and TRTPOW is the power of C _avg effect.

Fixed and random effects parameters were estimated with reasonable precision (relative standard error expressed as a percent [%RSE] 35.6%) (Table 1 ). The final model adequately characterized the central tendency and variability of the 6MWD observed data (Figure S1 ).

Table 1.

Parameter estimates and standard errors for the final integrated phase 2/3 E‐R model for 6MWD and PVR

Parameter (unit)		Final parameter estimate		Magnitude of variability
		Population mean	%RSE	Final estimate	%RSE
6MWD
BL	Baseline 6MWD in WHO Class III Participants (m)	398	2.56	64.9 SD	7.95
	Proportional Shift in WHO Class II Participants (−)	0.139	13.3
	Cut‐off Age (years)	48.7	5.86
	Slope of Age Effect Centered at 48 Years	−3.17	21.0
E max	Maximum Time‐Dependent Change in 6MWD (m)	20.1	35.6	110 SD	20.4
T50	Time to Reach Half EMAX (month)	6.93	18.0	NE	NA
TRTPOW	Power of C avg Effect (−)	0.292	10.1	0.108 SD	32.8
HGB	Slope of Hemoglobin Effect (m'dL/g)	5.50	16.7	NE	NA
Residual variability 6MWD		903	6.92	30.1 SD	NA
PVR:
BL	Baseline PVR (dynes.s/cm5)	719.9	1.943	39.43 %CV	8.357
DRG	Power for average sotatercept concentration (ng/mL)	−0.05164	7.342	0.02384 SD	22.47
	Slope of (PAHDUR ‐ (7.1)) for DRG	0.001570	24.29
	Slope of (PVR2BL ‐ (664)) for DRG	−2.567E‐05	30.94
	Proportional shift in DRG for PROSINF = 1	−0.3623	19.58
Additive residual error PVR		18,140	12.33	134.7 SD	NA
Minimum Value of the Objective Function for 6MWD modeling = 24,557.333 Minimum Value of the Objective Function for PVR modeling = 11,643.851

Shrinkage estimates: For 6MWD: 5.3% for IIV in BL, 16.3% for IIV in EMAX, and 57.7% for IIV in TRTPOW. For PVR: 9.1% for IIV on BL and 40.7% for IIV on DRG.

%CV, coefficient of variation expressed as a percent; %RSE, relative standard error expressed as a percent; 6MWD, 6 minute walk distance; C _avg, average sotatercept concentration; CI, confidence interval; E‐R, exposure‐response; pulmonary arterial hypertension disease duration; IIV, interindividual variability; NA, not applicable; NE, not estimated; PROSINF, baseline infusion therapy; PVR, pulmonary vascular resistance; PVR2BL, baseline PVR; SD, standard deviation; WHO, World Health Organization.

Patients classified into WHO Class III were predicted to have an approximately 14% lower baseline 6MWD when compared to patients in WHO Class II. In addition, for each additional year of age beyond 48 years, there would be a decrease in the 6MWD of 3.17 m at baseline. Other covariates including bodyweight, sex, duration of PAH pathology, PAH etiology, prostacyclin infusion therapy at baseline, ethnicity, baseline standard of care (SOC) therapy, baseline PVR, and observed baseline 6MWD were also tested on both baseline and sotatercept effect parameters and were not found to be statistically significant predictors of 6MWD response. Stochastic simulations demonstrate that simulated 6MWD (Figure 2 a ) and its change from baseline (Figure 2 b ) increase with increasing the C _avg of sotatercept at Week 24 and approach a plateau at exposures corresponding to 0.7 mg/kg Q3W.

Figure 2.

Efficacy responses versus concentration average (C _avg) of sotatercept at week 24 in a virtual population of participants with pulmonary arterial hypertension for the final integrated phase 2/3 ER models for 6MWD, PVR, and NT‐proBNP.

PVR

Observed PVR did not visibly change in participants with PAH who received a placebo between baseline and Week 24. In contrast, a dose and concentration‐dependent decrease in PVR was observed in participants who received sotatercept. The final integrated exposure‐PVR model in patients with PAH was a decreasing power function of C _avg at week 24 described as:

$$\begin{array}{c}\mathrm{PVR}=\mathrm{PV}{\mathrm{R}}_{\mathrm{BL}}\times C_{\mathrm{avg}}^{{\theta }_i}\text{with}{\theta }_i=\mathrm{DRG}\times \left(1-{\beta }_1\times \text{PROSIN}{\mathrm{F}}_i\right)\hfill \\ +{\beta }_2\times \left(\text{PAHDU}{\mathrm{R}}_i-\text{PAHDU}{\mathrm{R}}_{\text{median}}\right)\hfill \\ -{\beta }_3\times \left(\text{PVRB}{\mathrm{L}}_{\mathrm{i}}-\text{PVRB}{\mathrm{L}}_{\text{median}}\right)\hfill \end{array}$$

where C _avg is the sotatercept C _avg; DRG is the power for average sotatercept concentration (ng/mL) effect in patients receiving baseline prostacyclin infusion therapy; $\text{PROSIN}{\mathrm{F}}_{\mathrm{i}}$ is the indicator of no baseline infusion therapy for subject i; β ₁ is the proportional shift in DRG for PROSINF_i, $\text{PAHDU}{\mathrm{R}}_{\mathrm{i}}$ is the PAH disease duration for subject i; PAHDUR_median is the median PAHDUR in the Phase 2/3 population; β ₂ is the slope for PAHDUR, PVRBL_i is the baseline PVR measurement for subject i; PVRBL_median is median PVRBL, and β ₃ is the slope of PVRBL for DRG. None of the other tested covariates on the baseline and sotatercept effect parameters were found to be statistically significant predictors of the PVR response. All fixed and random effects parameters were estimated with good precision (%RSE < 31%) (Table 1 ). The selected model adequately characterized the central tendency and variability of the observed data across the study phases (Figure S2 ).

Assuming C _avg value of 8,454 ng/mL (i.e., median of predicted sotatercept concentrations at Week 24), PVR was projected to be approximately 16% lower when patients are administered baseline therapy infusion with prostacyclin than without prostacyclin, about 7% higher for every 5 years increase in PAH disease duration, and approximately 2% lower for every 100 dynes.sec/cm⁵ increase in baseline PVR. Simulations demonstrate that PVR (Figure 2 c ) and its change from baseline (Figure 2 d ) decrease with increases in the C _avg and approach a plateau at exposures corresponding to 0.7 mg/kg Q3W.

NT‐proBNP

The baseline probability of achieving NT‐proBNP < 300 pg/mL was incorporated in the base Cox hazard model due to 41.5% of the participants demonstrating NT‐proBNP < 300 pg/mL at baseline. After incorporation of the baseline probability, C _avg was found to be a significant predictor of time to NT‐proBNP < 300 pg/mL. The final E‐R model for time to NT‐proBNP < 300 pg/mL model was a Cox proportional hazards model:

$$h\left(t\right)=h_0\left(t\right)\times \exp \left({\beta }_1\times \text{BSLFLA}{\mathrm{G}}_i+{\beta }_2\times C_{\mathrm{avg},\mathit{ij}}-{\beta }_3\times \mathrm{WH}{\mathrm{O}}_i\right)$$

where $h_0\left(t\right)$ is the baseline hazard; t is the time in weeks; ${\text{BSLFLAG}}_i$ is the baseline event flag for the ith patient; β ₁ is the slope for BSLFLAG_i; $C_{\mathrm{avg},\mathit{ij}}$ is the average concentration in the ith patient during the dosing interval at the jth time; β ₂ is the slope for C _avg,ij; ${\mathrm{WHO}}_i$ is the baseline WHO functional class (FC) for the ith patient; and β ₃ is the proportional shift in WHO FCII patients.

All fixed and random effects parameters were estimated with reasonable precision (%RSE = 38%) (Table 2 ). The selected model characterized the central tendency and variability of the observed data across all study phases but seemed to consistently underpredict the observed data (Figure S3 ).

Table 2.

Parameter estimates and standard errors from the exposure‐response base model for time to NT‐proBNP < 300 pg/mL using final data from studies PULSAR, SPECTRA, and STELLAR

Parameter (unit)	Coefficient	SE	%RSE	P value	Hazard ratio (95% CI)
BSLFLAG (−)	2.597676	0.164593	6.336168	<1E‐6	13.43248 (9.72869, 18.54634)
C avg (ng/mL)	0.000297	3.05E‐05	10.28303	<1E‐6	1.000297 (1.000237, 1.000357)
WHO Class III (−)	−0.30988	0.116874	37.71578	0.008016	0.7335335 (0.5833594, 0.9223667)

BSLFLAG, flag for participants whose event occurred at baseline; C _avg, average sotatercept concentration; CI, confidence interval; NT‐proBNP, N‐terminal prohormone of brain natriuretic peptide; P, probability; %RSE, relative standard error expressed as a percent; SE, standard error; WHO, World Health Organization.

A hazard ratio (HR) is a measure used in clinical trials to compare the risk of a particular event occurring at any given time between two groups. The HR for the effect of C _avg L was 1.000297 (95% CI = 1.000237, 1.000357), indicating that for each increase of 1 ng/mL of C _avg, the probability of NT‐proBNP < 300 pg/mL at any given time increased by 0.0297%. Similarly, the hazard ratio for the effect of baseline WHO FCIII on time to NT‐proBNP < 300 pg/mL was 0.734 (95% CI = 0.583, 0.922), indicating that for a patient with WHO FCIII at baseline compared WHO FCII at baseline, the predicted proability of NT‐proBNP < 300 pg/mL was 26.6% higher. Of all tested covariates on NT‐proBNP response (Table S4 ), WHO FC was identified as a statistically significant but not clinically relevant covariate. Simulations demonstrated that the probability of a NT‐proBNP < 300 pg/mL event by Week 24 increased with increasing C _avg and approached a plateau at exposures corresponding to 0.7 mg/kg Q3W dose (Figure 2 e ).

Population PK/PD model for hemoglobin

Figure S4 shows the PK/PD model for Hgb. ¹⁶ Table 3 lists the estimated parameters for the PK/Hgb model. Figure 3 depicts the Hgb response for sotatercept exposures for a dose regimen of 0.7 mg/kg SC Q3W. The VPCs in Figure S5 show that the final PK/PD model captured the central tendency and magnitude of variability of the observed Hgb concentrations generally well for patients with PAH, with a minor underprediction bias at later time points. In healthy participants, the median Hgb response (and variability) was overpredicted to various degrees across the study duration. Covariate analysis identified that iron supplementation was associated with approximately 5% lower RBC counts relative to patients without iron supplementation; iron supplementation also resulted in a relatively small (~2.5%) stimulation of the production and maturation rates of progenitor cells. None of the other covariates tested on the baseline and sotatercept effect parameters including age, BWT, eGFR, sex, or self‐identified‐racial classification were found to be significant descriptors of the variability in Hgb response.

Table 3.

Parameter estimates and standard errors for the final integrated phase 1/2/3 PK/PD model for hemoglobin

Parameter (unit)		Final parameter estimate			Magnitude of variability
		Population mean	%RSE	Bootstrap mean (95% CI)	Final estimate	%RSE	Bootstrap mean (95% CI)
SMAX	Maximum stimulation (−)	0.186	9.85	0.194 (0.155, 0.237)	85.8 %CV	17.0	88.0 %CV (68.9%, 113.6%)
SC50	Half‐stimulatory sotatercept concentration (ng/mL)	2,710	20.7	3,108 (1883,4,963)	NE	NA	NA
KP1	Rate of elimination of precursor 1 (1/day)	0.374 d	FIXED	NA	NE	NA	NA
KP2	Rate of elimination of precursor 2 (1/day)	2.05 d	FIXED	NA	NE	NA	NA
KRBC	Elimination rate of RBC (1/day)	0.0675	28.0	0.0699 (0.0277, 0.116)	NE	NA	NA
IC50	Half‐inhibitory change from hemoglobin baseline (g/dL)	36.3	19.7	38.8 (23.3, 53.3)	NE	NA	NA
GAMMA	Hill coefficient for sotatercept stimulation function (−)	1.00	FIXED	NA	NE	NA	NA
RET0	Baseline RET in healthy participants (109 cells/L)	50.3	4.77	50.3 (45.7, 54.6)	27.7 %CV	13.1	27.5 %CV (24.0%, 31.4%)
	Baseline RET in PAH participants (109 cells/L)	97.8	3.03	97.8 (91.9, 104.0)
RBC0	Baseline total RBC (1012 cells/L)	4.71	0.520	4.71 (4.66, 4.76)	10.9 %CV	7.00	10.9 %CV (10.2%, 11.7%)
	Additive shift for baseline iron supplementation (1012 cells/L)	−0.234	40.3	−0.234 (−0.405, −0.0583)
HGBRV	Scaling of hemoglobin RV in healthy participants (−)	3.04	0.781	3.04 (3.00, 3.09)	NE	NA	NA
	Scaling of hemoglobin RV in PAH patients (−)	2.80	0.764	2.80 (2.76, 2.84)
IRONST	Additive effect of iron supplementation on K in (−)	0.0247	60.2	0.0245 (−0.0014, 0.0582)	NE	NA	NA
cov (IIV in RET0, IIV in SMAX)		−0.0608a	30.9	−0.0606 (−0.0969, −0.0228)	NA	NA	NA
cov (IIV in RBC0, IIV in SMAX)		−0.0226b	27.1	−0.0232 (−0.0384, −0.0104)	NA	NA	NA
cov (IIV in RBC0, IIV in RET0)		0.00925c	37.2	0.0922 (0.0023, 0.0161)	NA	NA	NA
RETRV ‐ CCV RV in healthy participants		0.125	7.07	0.125 (0.109, 0.142)	35.3 %CV	NA	35.3 %CV (34.0%, 39.1%)
RBCRV ‐ Additive RV in healthy participants		0.0290	7.57	0.0292 (0.0252, 0.0342)	0.170 SD	NA	0.171 SD (0.159 SD, 0.185 SD)
RETRV ‐ CCV RV in PAH participants		0.0496	14.2	0.0494 (0.0380, 0.0668)	22.3 %CV	NA	22.2 %CV (19.7%, 26.3%)
RBCRV ‐ Additive RV in PAH participants		0.0685	5.89	0.0683 (0.0612, 0.0769)	0.262 SD	NA	0.261 SD (0.247 SD, 0.277 SD)
Minimum value of the objective function = 20,425.912

Shrinkage estimates: 30.3% for IIV in SMAX, 41.1% for IIV in RET0, and 1.9% for IIV in RBC0.

CCV, constant coefficient of variation; CI, confidence interval; cov, covariance; %CV, coefficient of variation expressed as a percent; IIV, interindividual variability; K _in, zero‐order input rate of early progenitors; NA, not applicable; NE, not estimated; PAH, pulmonary arterial hypertension; PK/PD, pharmacokinetic/pharmacodynamic; RBC, red blood cell; RET, reticulocyte; %RSE, relative standard error expressed as a percent; RV, residual variability; SD, standard deviation.

^a The calculated correlation coefficient (r) associated with cov (IIV in RET0, IIV in SMAX) was −0.301 with r ² = 0.0909.

^b The calculated correlation coefficient (r) associated with cov (IIV in RBC0, IIV in SMAX) was −0.279 with r ² = 0.0779.

^c The calculated correlation coefficient (r) associated with cov (IIV in RBC0, IIV in RET0) was 0.312 with r ² = 0.0972.

^d Parameters fixed from. ¹⁶

Figure 3.

Hemoglobin versus concentration average (C _avg ) of sotatercept at week 24 in a virtual population of participants with pulmonary arterial hypertension. (a) predicted week 24 hemoglobin vs. week 24 sotatercept C _avg; (b) predicted change from baseline in hemoglobin at week 24 vs. Week 24 Sotatercept C _avg.

DISCUSSION

The exposure‐safety and exposure‐efficacy relationships for sotatercept were analyzed in clinical trials including healthy PMW (two Phase 1), and PAH participants (two Phase 2 and one Phase 3). ⁴ , ⁵ , ¹⁷ , ¹⁸ The Phase 1 studies allowed characterization of the sotatercept PK and its tolerability in a wide range of single doses (0.01–3 mg/kg) as well as monthly doses (0.03–1 mg/kg), administered IV or SC. ⁶ The Phase 2/3 studies allowed the evaluation of the efficacy and safety of sotatercept in the target population at 0.3 or 0.7 mg/kg Q3W (Phase 2) and using a titration regimen with an initial dose of 0.3 mg/kg followed by the target dose of 0.7 mg/kg Q3W thereafter unless a dose reduction was warranted (Phase 3). ⁴

The efficacy (6MWD, PVR, and NT‐proBNP) and safety (Hgb) endpoints selected for sotatercept are important measures of PAH disease status for healthcare professionals and patients. The 6MWD represents the functional improvement (exercise capacity) in response to treatment in PAH patients. It is used as a key endpoint in many pivotal PAH clinical trials (including STELLAR) and is an accepted endpoint by regulatory agencies. ¹⁹ PVR reflects the functional status of pulmonary vascular endothelium/smooth muscle cell couple systems. This direct measure determines the impact of treatment on PVR and is routinely used to demonstrate proof‐of‐concept during the development of medicines that treat PAH. ²⁰ NT‐proBNP is a key biomarker for the function of the heart. It increases during heart failure and decreases with improved cardiac function. In PAH patients, NT‐proBNP levels are shown to correlate with 6MWD, PVR, right ventricular function, and other hemodynamic variables. ²¹ , ²² , ²³ Hemoglobin was evaluated as a safety measure to provide dosing guidance for PAH patients. Although the exact mechanism of sotatercept on erythropoiesis remains uncertain, it is believed to act via inhibition of Smad2/3 signaling. ²⁴ , ²⁵ In clinical studies, sotatercept has been shown to increase Hgb concentrations in some patients. Hence, dose modification guidance was included in the clinical studies including STELLAR to monitor Hgb prior to sotatercept dose administration.

The E‐R model for 6MWD characterized well the observed data from placebo and Phase 2/3 studies. The model suggests that an increase in Hgb level is associated with an increase in exercise capacity as measured by 6MWD. However, since sotatercept has effects on both Hgb and 6MWD, it is not possible to conclude whether this is a causal relationship or merely a correlation. Therefore, the effect of sotatercept and Hgb on 6MWD is confounded and the improvement in exercise capacity observed with the administration of sotatercept can only be partially attributed to increase in Hgb levels and suggests some direct effects of sotatercept on 6MWD. Participants classified into WHO Class II were predicted to have ~14% higher baseline 6MWD compared to participants in WHO Class III. This is consistent with the expected higher walking capacity in participants who experience milder symptoms of PAH. Furthermore, age had a significant effect on the 6MWD and baseline 6MWD appeared to decrease with increase in age with an estimated median age to be 48.7 years. This indicates that for each additional year of age beyond 48.7 years, there would be a decrease in the 6MWD of 3.17 m. This is consistent with the expectation of a slowing walking pace with increasing age. The pcVPC showed that the E‐R model for 6MWD adequately characterized the central tendency and the extent of variability of all observed data. Simulations demonstrated that 6MWD increased with increases in the Week 24 C _avg and approached a plateau at exposures corresponding to 0.7 mg/kg (Figure 2 a,b ), where the change from baseline was ~40 m, but in the absence of sotatercept, participants also exhibited a placebo effect of a 10 m improved 6MWD.

The relationship between exposure and PVR was best described by a saturable inhibitory function of model‐predicted C _avg at the time of PVR measurement. The analysis suggests that no significant change in PVR was observed in participants who received placebo plus background PAH therapy, while for participants receiving sotatercept, PVR decreased with increasing sotatercept exposure approaching a plateau at the C _avg range corresponding to the clinical dose of 0.7 mg/kg Q3W (Figure 2 c,d ). The mean PVR decrease from baseline at Week 24 corresponding to the median C _avg of 0.7 mg/kg was −217 dynes.sec/cm⁵, which is approximately a 30% improvement in participants with PAH. Further, PAH duration, baseline PVR, and prostacyclin infusion were identified as significant covariates impacting the exposure‐PVR relationship. Overall, simulated PVR response decreased with increasing sotatercept exposure and approached a plateau at the C _avg range corresponding to 0.7 mg/kg Q3W dose.

The threshold associated with low risk in PAH participants for NT‐proBNP at < 300 pg/mL is well documented. ¹³ , ²⁶ The baseline WHO FC was found to negatively impact the probability of NT‐proBNP < 300 pg/mL event for both placebo and sotatercept recipients. The estimate of the hazard ratio indicated that at baseline the predicted probability of NT‐proBNP < 300 pg/mL is 73.4% for a patient with WHO FCIII at baseline compared to a patient with WHO FCII. The model predicted that the probability of achieving NT‐proBNP < 300 pg/mL increased with increasing sotatercept exposures and approached a plateau at sotatercept exposures corresponding to 0.7 mg/kg (Figure 2 e ).

Hemoglobin levels increased with increasing sotatercept exposure in PAH participants enrolled in all sotatercept clinical studies. ⁴ , ⁵ , ¹⁷ , ²⁷ Literature reports that Smad2/3 protein signaling is overactive in erythroid progenitors and precursors of patients with myelodysplastic syndromes and β‐thalassemia with defective erythropoiesis. ²⁸ Sotatercept was also shown to improve anemia in patients with both myelodysplastic syndrome and β‐thalassemia. ²⁹ Sotatercept time‐course concentrations successfully drove the temporal changes of Hgb concentrations with Hgb concentrations expressed as a metric derived from predicted RET and mature‐RBC counts and the observed mean corpuscular Hgb. Overall, the developed semi‐mechanistic‐population PK/Hgb model ¹⁶ captured well all observed Hgb data, however, VPCs showed a slight underprediction bias at later time points and a slight overpredicted Hgb response in healthy participants, which should be interpreted with caution due to the small sample size. Interestingly, the model predicted an equivalent amplitude of maximal response at Week 24 irrespective of the administration of a lower starting dose. This observation is anticipated due to the lack of incorporation of the PD mechanism of tolerance in the model.

The maximum estimated drug effect on Hgb was modest in that the production and maturation rates of progenitor cells were predicted to increase by a maximum of approximately 18.6%. Participants with PAH were found to have approximately twofold higher baseline RET counts than healthy participants (97.8 vs. 50.3 (10⁹ cells/L)). This finding is consistent with the preclinical findings in a rat model of PAH with increased erythropoietin concentrations and a significantly higher total and immature RET counts compared to their wild‐type counterpart. ³⁰ The estimate of the concentration of sotatercept resulting in half the maximum stimulatory effect on Hgb dynamics was 2,710 ng/mL and was not influenced by disease status (i.e., healthy vs. PAH). Simulations demonstrated that Hgb and change in Hgb from baseline increased with increase in the Week 24 C _avg and approached a plateau (Hgb < 18 g/dL and ΔHgb < 2 g/dL) at exposures corresponding to 0.7 mg/kg (Figure 3 a,b ), however, these increases remain clinically manageable as demonstrated by the low number of participants who required dose reductions in the STELLAR trial (n = 21), where all but 2 participants were able to complete STELLAR. ⁴

Iron supplementation, which was more prevalent in participants with lower baseline RBC counts, was also found to be associated with slightly lower baseline RBC counts (4.5 vs. 4.7 (10¹² cells/L) in the presence and absence of iron supplementation) and to provide a small magnitude of stimulation of production and maturation of the progenitor cells (approx. 10‐fold lower than the estimated maximum sotatercept effect). As such, the median Week 24 change from baseline Hgb was slightly higher for participants who received iron supplementation (1.9 vs. 1.7 g/dL). All other covariates were not found to be statistically significant factors impacting Hgb dynamics or sotatercept effect parameters in the PK/Hgb model.

The FDA‐approved dosing regimen for sotatercept is a starting dose of 0.3 mg/kg followed by a target dose of 0.7 mg/kg Q3W by SC injection. This dose is well justified based on the totality of the observed safety and efficacy data from Phase 1, Phase 2, and Phase 3 studies. During the Phase 1 MAD trial, a 1 mg/kg Q4W dose was tested and led to an increase in hemoglobin levels and hypertensive adverse events, resulting in the study's interruption. Consequently, the highest dose of 0.7 mg/kg Q3W dose was selected for the Phase 2 dose‐ranging trial. This dose was found to be both effective and safe, increasing hemoglobin levels by < 2 g/dL from baseline. Hence, dosage for the pivotal Phase 3 STELLAR was selected based on safety and efficacy data from Phase 1 studies in PMW and Phase 2 studies in participants with PAH. Results from PULSAR showed that both the 0.3 and 0.7 mg/kg sotatercept dose levels resulted in significant improvements in PVR and 6MWD compared with placebo as well as comparable mean change in Hgb from baseline at Week 24 in the 0.3 and 0.7 mg/kg Q3W groups (1.2 and 1.5 g/dL). ¹⁷ , ¹⁸ The PULSAR study also showed that excursions in Hgb concentration above the upper limit of normal can be effectively managed by sotatercept dose modification rules. ⁸ Furthermore, clinical trial simulations suggested that the probability of having Hgb ≥ 18 g/dL and an increase in Hgb ≥ 2 g/dL is higher during the first 21 days after a dose of 0.7 than 0.3 mg/kg. Given numerically better PVR response in the 0.7 mg/kg group, and manageable Hgb profile in both 0.7 and 0.3 mg/kg Q3W, 0.7 mg/kg Q3W was selected as the target dose for STELLAR. Therefore, a starting dose of 0.3 followed by 0.7 mg/kg Q3W target dose was administered with appropriate dose modification rules was selected for the pivotal STELLAR trial. The Q3W frequency is consistent with the 21 days half‐life for sotatercept.

This selected dosing regimen was subsequently tested in STELLAR and showed superior efficacy as well as a favorable safety profile relative to placebo. The E‐R and PK/PD analyses predicted efficacy and safety responses approached plateaus at exposures corresponding to 0.7 mg/kg Q3W. Neither intrinsic nor extrinsic factors had a meaningful clinical impact nor were significant on the drug effect terms in any E‐R and PK/PD models. Therefore, the dosing regimen (0.3 mg/kg as the starting dose followed by 0.7 mg/kg as the maintenance dose) is considered appropriate for the entire target population. Hence, the final recommended dose regimen for sotatercept is a starting dose of 0.3 mg/kg followed by the target dose of 0.7 mg/kg Q3W by SC injection. Dosage adjustments may be required according to patient response and tolerability of treatment. ⁸

CONCLUSION

A consistent E‐R relationship was found for the efficacy endpoints (6MWD and PVR) and the efficacy biomarker (NT‐proBNP). Exposures achieved at 0.7 mg/kg SC Q3W fall near the top of the E‐R curves supporting the target 0.7 mg/kg dose from the labeled dose titration guidance as achieving near maximum efficacy. The safety endpoint (Hgb) increased with sotatercept exposures in a similar E‐R relationship to the efficacy measures. The elevations in Hgb are manageable, however, require monitoring and appropriate dose modification. Altogether, these effects support the dose regimen for sotatercept at a starting dose of 0.3 mg/kg followed by a target dose of 0.7 mg/kg Q3W by SC injection. This dose is well‐justified based on the totality of the observed safety and efficacy data from Phase 2 and Phase 3 studies.

FUNDING

These studies and analyses were funded by Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA.

CONFLICTS OF INTEREST

The authors declared no competing interests for this work. S.A.‐O., Z.H., S.H., J.O.P., L.W., and F.G. are current or former employees of Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA and may hold stock options in Merck & Co., Inc., NJ, USA. D.J., M.G., and H.B. are employees of Cognigen division of Simulations Plus Inc., Buffalo, NY and may hold stock options in Simulation Plus. S.B. is a former employee of Cognigen division of Simulations Plus Inc., Buffalo, NY and is currently an employee of Regeneron Pharmaceuticals, Inc and may hold stocks in Regeneron Pharmaceuticals, Inc. BB and BM were previous employees of Acceleron Pharma Inc., a subsidiary of Merck & Co., Inc., Rahway, NJ, USA. B.B. is currently an employee of Alector LLC, South San Francisco, CA, USA and may hold stock options in Alector LLC. B.M. is currently an employee of Insmed Incorporated, Bridgewater, NJ, USA and may hold stock options in Insmed.

AUTHOR CONTRIBITIONS

S.A.‐O. and F.G. wrote the manuscript, designed the research, performed the research, and analyzed the data. S.A.‐O., Z.H., L.W., and F.G. designed the research. S.A.‐O., D.J., Z.H., M.G., H.B., S. H., S.B., B.B., B.M., J.O.P., L.W., and F.G. performed the research and analyzed the data.

Supporting information

Data S1