Exposure–Response Analyses of Ropeginterferon Alfa‐2b in Essential Thrombocythemia

Abstract

Ropeginterferon alfa‐2b is a new‐generation interferon therapy for the treatment of myeloproliferative neoplasms (MPNs), including essential thrombocythemia (ET). This investigation was designed to characterize its population pharmacokinetics–pharmacodynamics (PopPK‐PD) and delineate exposure–response (E‐R) relationships in ET. A population PK model was constructed utilizing aggregated data from five clinical trials, encompassing a study population of Asian and Caucasian patients with ET and healthy volunteers. A sequential modeling strategy was employed to evaluate the PK‐PD with respect to key hematologic markers including platelet and white blood cell counts. Hematologic changes were effectively modeled using sigmoidal indirect‐response models. Individual exposure metrics were subsequently simulated by a target‐mediated drug disposition model and then applied in E‐R analyses for reductions in the allele burden of MPN driver mutations JAK2V617F and CALR, and safety outcomes. Simulations indicated no significant disparities between ethnicities in ropeginterferon alfa‐2b exposure or the extent of hematologic response. A correlation was observed between drug exposure and a decrease in JAK2V617F allele burden at 6 and 12 months of treatment. Exposure–safety assessments identified a risk of exposure‐related adverse events, including reversible anemia. Taken together, the data provided a solid PK‐PD framework and insights into the E‐R relationships for ropeginterferon alfa‐2b in the treatment of ET.

Introduction

Essential thrombocythemia (ET) is a Philadelphia chromosome‐negative myeloproliferative neoplasm (MPN), characterized by elevated platelet count, thrombocytosis‐related symptoms, and risks of thrombosis and transformation to myelofibrosis or acute myeloid leukemia. ¹ , ² ET is associated with MPN driver mutations in genes encoding Janus kinase 2 (JAK2): predominantly the point mutation JAK2V617F (50%–60%); calreticulin (CALR, 20%–25%); and myeloproliferative leukemia virus oncogene (MPL, 3%–4%), respectively. ¹ Thrombotic risk is a major clinical feature of ET, with the risk of a vascular event increasing over time. ³ Major thrombotic events involve cerebrovascular, coronary, and peripheral arterial circulation. In combination with microvascular occlusion, the vascular syndrome is characterized by a wide range of symptoms. ⁴ A large cohort analysis showed that 17.8% of patients with ET experienced thrombosis before or at the time of ET diagnosis, with individual risk influenced by factors including age, history of thrombosis, presence of cardiovascular risk factors, and JAK2 mutational status. ⁵ , ⁶

Current treatment options for ET are limited and lack newer therapies that not only manage the thrombotic risk and symptom burden but also inhibit disease progression. Low‐risk patients with ET usually receive low‐dose aspirin. High‐risk, symptomatic low‐risk, or intermediate‐risk patients are typically prescribed hydroxyurea as an off‐label, first‐line cytoreductive therapy to lower platelet counts and thrombotic risk while taking aspirin. ¹ However, up to 20%–30% of patients develop resistance or intolerance to hydroxyurea, further increasing the risk of disease progression and reducing survival. ⁷ , ⁸ , ⁹ , ¹⁰ Hydroxyurea has potential side effects, including anemia, neutropenia, leg ulcers, gastrointestinal symptoms, and non‐melanoma skin cancers. ¹¹ Anagrelide is an approved treatment for ET in the United States. It reduces platelet counts and the risk of thrombosis. Since its approval in 1997, it remains the latest drug specifically approved for ET in the United States. Its side effects include palpitations, headaches, and diarrhea and are potentially associated with fibrotic progression. ¹²

Interferon α (IFN‐α) has demonstrated clinical efficacy in controlling thrombocytosis and leukocytosis in patients with MPN, including ET. ¹³ , ¹⁴ , ¹⁵ Human IFN‐α acts via binding to the type 1 IFN receptors and shares similar biological properties to its family member IFN‐β, which induces apoptosis and inhibits tumor cell‐cycle progression accompanied by senescence and a loss of tumorigenesis. ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ IFN‐α therapy promotes myelofibrosis‐free survival, event‐free survival, and overall survival in polycythemia vera (PV); it may be postulated to have a similar effect in ET, as ET and PV are related MPNs. ²⁰ , ²¹ , ²² IFN‐α therapy has been recommended by international treatment guidelines for ET management and was used as an off‐label treatment for ET. PEG‐IFN‐α‐2a, a polyethylene glycol‐conjugated (PEGylated) IFN‐α‐2a, has recently received approval for the treatment of ET and PV in Europe based on the Myeloproliferative Disorders Research Consortium trials. ²³ , ²⁴ Ropeginterferon alfa‐2b, a mono‐PEGylated proline‐IFN‐α‐2b given once every 2 weeks, showed significant hematologic and molecular responses in the randomized Phase‐3 study SURPASS‐ET in patients with ET who were resistant or intolerant to hydroxyurea. ²⁵ , ²⁶ It is currently under regulatory evaluation for approval in ET in the United States, Japan, and other regions.

Ropeginterferon alfa‐2b is currently approved for the treatment of PV globally, including Europe and the United States. ²⁷ In PV, ropeginterferon alfa‐2b treatment showed a dose‐related reduction of allele burden (variant allele frequency, VAF) of the MPN driver mutation JAK2V617F. ²⁸ , ²⁹ Reduction in JAK2V617F VAF is increasingly recognized as a clinically meaningful treatment effect in MPNs, as it is associated with lower thrombotic risk and improved event‐free and overall survival. ³⁰ , ³¹ , ³² , ³³ In PV, ropeginterferon alfa‐2b treatment leads to durable complete hematologic response (CHR) with higher PK exposure correlating with an increased likelihood of achieving CHR. ²⁸ , ³⁴ Achieving CHR by concurrent normalization of white blood cell (WBC) and platelet counts is also clinically important, as leukocytosis (alongside thrombocytosis) is a risk factor for thrombosis and disease progression. ³⁵ , ³⁶ , ³⁷

For ET treatment, the exposure–response (E‐R) relationship of ropeginterferon alfa‐2b regarding MPN driver mutations and hematologic parameters has not been reported. In this study, we constructed population pharmacokinetics‐pharmacodynamics (PopPK‐PD) models to characterize the PK‐PD and E‐R relationships of ropeginterferon alfa‐2b in patients with ET. Covariate effects were assessed based on five clinical studies in healthy Asian and Caucasian volunteers and patients with ET. The primary objective was to precisely characterize the PK‐PD and E‐R relationships of ropeginterferon alfa‐2b in patients with ET, helping to optimize treatment strategies and maximizing therapeutic efficacy while ensuring patient safety.

Methods

Clinical Trial Design and Study Participants

The PopPK‐PD and E‐R models were developed and evaluated using pooled data from five clinical studies. All studies were conducted in accordance with the Declaration of Helsinki and approved by the respective institutional review boards. Written informed consent was obtained from all participants.

The analysis integrated data from three studies in healthy volunteers and two studies in patients with ET. Ropeginterferon alfa‐2b was administered subcutaneously in all clinical studies. A09‐102 was a single‐center, randomized double‐blind, active‐controlled, single‐dose escalation study in 48 healthy Caucasian subjects. Subjects were assigned to six sequential cohorts, each testing one dose level of ropeginterferon alfa‐2b. In every cohort, two subjects were administered a single subcutaneous dose of peginterferon alfa‐2a (180 µg), while the remaining six received a single injection of ropeginterferon alfa‐2b at doses, including 24, 48, 90, 180, 225, or 270 µg. ³⁸ A17‐101 was a randomized, open‐label, single‐dose escalation study in 40 healthy Chinese adults. Subjects received single doses of ropeginterferon alfa‐2b at doses ranging from 90 to 270 µg or peginterferon alfa‐2a at 180 µg. ³⁹ A17‐102 was a dose‐escalation study in 48 healthy Japanese and Caucasian participants. Subjects received single doses of ropeginterferon alfa‐2b at doses ranging from 100 to 300 µg. ⁴⁰ SURPASS‐ET was a randomized, multicenter, open‐label Phase‐3 study of ropeginterferon alfa‐2b versus anagrelide as second‐line therapy in patients with ET. Ropeginterferon alfa‐2b was administered using a higher initial dose with an accelerated titration regimen (HIDAT), starting at 250 µg, titrating to 350 µg at Week 2, and 500 µg from Week 4; doses could be adjusted according to tolerability and safety. Anagrelide was given orally according to its labeled dosage. ²⁵ , ²⁶ EXCEED‐ET was a Phase 2, single‐arm, multicenter study of ropeginterferon alfa‐2b in patients with ET conducted in the United States and Canada. Patients received ropeginterferon alfa‐2b under the HIDAT regimen. ⁴¹

PK Assessments

The time points for PK blood sample collections were identical for Phase 1 studies: pre‐dose and 1, 3, 6, 9, 12, 24, 36, 48, 72, 96, 120, 144, 168, 192, 240, 288, 336, 504, and 672 h post‐dose. Samples for SURPASS‐ET were collected at Week 0 and Week 12, at the following time points: pre‐dose and 48, 96, and 168 h. Trough concentrations were also measured at each visit. For EXCEED‐ET, trough PK samples were collected from patients 2 weeks after dose administration and prior to the administration of the next dose on that day. Participants with at least one measurable PK concentration after the first ropeginterferon alfa‐2b administration were included in the PopPK analysis.

Efficacy and Safety PD Assessment

VAF reductions of the MPN driver mutations were assessed as the molecular efficacy endpoint. They were assessed after treatment of 6 months and 1 year: Months 6 and 12 for SURPASS‐ET and Months 7 and 13 for EXCEED‐ET, respectively. JAK2V617F and CALR VAFs were assessed in the E‐R analyses. MPL VAF was not included as only nine patients had its positivity at baseline. Blood parameters, including WBC and platelet counts, measured every 2 weeks for patients with ET, were analyzed as hematologic PD markers in the PK‐PD analyses.

Safety endpoints were focused on Grade 3 or higher treatment‐emergent adverse events (TEAEs) defined by NCI‐CTCAE v5.0, TEAEs leading to treatment discontinuation, and TEAEs leading to dose reduction or interruption. Other safety endpoints included any grade of TEAEs with incidence greater than 20%, including increased alanine aminotransferase (ALT), increased aspartate aminotransferase (AST), anemia, fatigue, headache, pruritus, and diarrhea.

PopPK‐PD Models

Structure Model

The structural basis for PopPK analyses was a target‐mediated drug disposition (TMDD) model, which was constructed previously, ²⁹ , ⁴² with the structure shown in Figure 1a. The binding affinity of ropeginterferon alfa‐2b to its target (K_D, ng/mL), maximum binding capacity (R_tot, ng/mL), first‐order rate of elimination of the drug–target complex (k_int, 1/h), and first‐order rate constant for degradation (k_deg, 1/h) were estimated. The zero‐order rate constant for target synthesis (k_syn, in 1/h) was the product of R_tot and k_deg. Differences in target turnover over time were described by allowing R_tot to decrease from an initial starting value (R_tot0, in ng/mL) to a new steady‐state value (R_{tot, SS}, in ng/mL) with an estimated first‐order rate constant (k_dec, in 1/day).

Figure 1.

Schematic diagram of PopPK model (a) and PK‐PD model for platelet and white blood cell counts (b). CMT, compartment; CL, clearance; C_S, total serum ropeginterferon alfa‐2b concentrations; EPHP, intercept of platelet count reduction; EPP, intercept of platelet count reduction on the logit; EPHW, intercept of white blood cell count reduction; EPW, intercept of white blood cell count reduction on the logit; HI_max,P, maximum reduction due to ropeginterferon alfa‐2b for platelet count on logit; IC₅₀, the serum ropeginterferon alfa‐2b concentration resulting in 50% of I_max; IC_50,P, the serum ropeginterferon alfa‐2b concentration resulting in 50% of I_max,P; IC_50,W, serum ropeginterferonferon alfa‐2b concentration resulting in 50% of I_max,W; I_max, maximum reduction due to ropeginterferon alfa‐2b; I_max,P, maximum reduction due to ropeginterferon alfa‐2b for platelet count; I_max,W, maximum reduction due to ropeginterferon alfa‐2b for white blood cell count; K_a, absorption rate constant; K_D, binding affinity of ropeg to its target; k_deg, first‐order rate constant for degradation; k_in, zero‐order rate constant for production; k_int, first‐order rate of elimination of the drug–target complex; k_syn, zero‐order rate constant for target synthesis; PLT, platelet count; SC, subcutaneous administration; WBC, white blood cell count; V_C, volume of serum compartment.

It was further linked to indirect response (turnover) models for PD endpoints. Sequential modeling was used to construct separate PopPK‐PD models to characterize the magnitude and time course of platelet and WBC count reductions.

Basic turnover PK‐PD models were used to characterize the effect of serum ropeginterferon alfa‐2b concentrations on the reduction of platelets and WBC counts in patients with ET (Figure 1b). The parameters estimated included initial values of platelet (PLT₀) or WBC (WBC₀), the intercept EP, and a first‐order elimination rate constant (k_out, in 1/day) for each endpoint. The effect of total serum ropeginterferon alfa‐2b concentrations (C_S) was modeled as inhibiting each endpoint using an I_max function, estimating the maximum reduction due to ropeginterferon alfa‐2b (I_maxs), the serum ropeginterferon alfa‐2b concentration resulting in 50% of I_max (IC₅₀, in ng/mL), and the Hill coefficient of I_max function ($\gamma$).

Statistics Model

Random effects models were constructed by accounting for both inter‐individual variability (IIV) and residual unexplained variability (RUV). IIV was modeled using a log‐normal distribution, and potential correlations among random effects were investigated and included where appropriate. RUV, assumed to follow a normal distribution, was characterized using either an additive, proportional, or combined error structure.

Covariate Model

Covariates consisted of demographic factors, laboratory measurements, and baseline JAK2V617F allele burden, as summarized in Table 1. Missing values for continuous covariates were imputed using the population median. The relationships between parameters and covariates were evaluated through a stepwise procedure involving forward inclusion and subsequent backward elimination.

Table 1.

Demographics and Baseline Characteristics of Study Populations

	A09‐102 (N = 32)	A17‐101 (N = 27)	A17‐102 (N = 35)	EXCEED‐ET (N = 88)	SURPASS‐ET (N = 91)	Overall (N = 273)
Continuous Covariates, median [range] Missing, n (%)
Age (years)	31.5 [20.0, 44.0]	30.0 [22.0, 43.0]	27.0 [19.0, 49.0]	57.0 [23.0, 84.0]	61.0 [21.0, 80.0]	45.0 [19.0, 84.0]
Weight (kg)	80.0 [58.2, 94.4]	71.1 [51.4, 84.4]	68.5 [51.9, 105]	76.9 [51.7, 153]	60.0 [41.5, 107]	70.0 [41.5, 153]
Body mass index (kg/m2) Missing	25.1 [20.8, 29.6] 0 (0%)	24.4 [19.1, 29.5] 0 (0%)	22.6 [19.0, 28.5] 0 (0%)	26.0 [19.7, 49.7] 18 (20.5%)	23.4 [16.9, 36.6] 0 (0%)	24.3 [16.9, 49.7] 18 (6.6%)
Body surface area (m2) Missing	1.98 [1.61, 2.19] 0 (0%)	1.82 [1.49, 2.08] 0 (0%)	1.81 [1.52, 2.37] 0 (0%)	1.87 [1.48, 2.58] 18 (20.5%)	1.64 [1.32, 2.20] 0 (0%)	1.80 [1.32, 2.58] 18 (6.6%)
Alanine aminotransferase (U/L)	20.5 [8.00, 77.0]	18.6 [8.70, 40.5]	17.0 [9.00, 30.0]	19.5 [7.00, 66.0]	20.0 [8.00, 62.0]	19.0 [7.00, 77.0]
Aspartate aminotransferase (U/L)	19.5 [12.0, 39.0]	18.9 [12.1, 31.6]	18.0 [13.0, 27.0]	23.0 [12.0, 48.0]	22.0 [12.7, 42.0]	22.0 [12.0, 48.0]
Alkaline phosphatase (U/L) Missing	71.5 [39.0, 107] 0 (0%)	77.0 [45.0, 128] 0 (0%)	79.0 [52.0, 115] 0 (0%)	67.0 [38.0, 135] 1 (1.1%)	81.0 [21.0, 327] 0 (0%)	73.0 [21.0, 327] 1 (0.4%)
Albumin (g/dL) Missing	4.00 [4.00, 4.00] 31 (96.9%)	4.70 [4.20, 5.20] 0 (0%)	4.00 [3.50, 4.80] 0 (0%)	4.40 [3.40, 5.10] 0 (0%)	4.40 [3.70, 5.30] 0 (0%)	4.40 [3.40, 5.30] 31 (11.4%)
Total bilirubin (mg/dL)	0.581 [0.268, 1.59]	0.819 [0.345, 1.43]	0.643 [0.292, 1.35]	0.500 [0.200, 1.40]	0.579 [0.200, 1.77]	0.585 [0.200, 1.77]
JAK2V617F (%) Missing	0 [0, 0] 0 (0%)	0 [0, 0] 0 (0%)	0 [0, 0] 0 (0%)	0 [0, 77.6] 6 (6.8%)	31.7 [0, 97.9] 0 (0%)	0 [0, 97.9] 6 (2.2%)
White blood cell counts (109/L)	5.38 [4.00, 9.13]	6.30 [3.98, 8.74]	5.40 [3.30, 7.40]	7.47 [2.70, 14.0]	11.6 [7.20, 44.7]	7.70 [2.70, 44.7]
Platelet counts (109/L)	229 [126, 324]	259 [170, 333]	246 [164, 405]	796 [440, 2080]	942 [397, 2140]	638 [126, 2140]
Hematocrit (%)	43.5 [38.4, 48.5]	42.6 [36.2, 52.0]	44.0 [39.0, 50.0]	41.6 [29.9, 50.2]	42.1 [32.6, 52.8]	42.3 [29.9, 52.8]
Hemoglobin (g /L)	15.4 [12.8, 16.7]	14.6 [11.6, 17.5]	14.9 [13.2, 16.8]	13.9 [9.30, 17.2]	13.6 [10.3, 16.8]	14.2 [9.30, 17.5]
Serum creatinine (mg/dL)	0.850 [0.660, 1.03]	0.750 [0.480, 1.00]	0.890 [0.630, 1.10]	0.860 [0.480, 2.25]	0.810 [0.470, 1.40]	0.840 [0.470, 2.25]
Creatinine clearance (mL /min)	115 [76.9, 152]	112 [80.3, 146]	107 [70.1, 153]	93.7 [37.4, 204]	79.6 [40.1, 167]	98.9 [37.4, 204]
Categorical Covariates, n (%) Missing, n (%)
Sex
Male	32 (100%)	15 (55.6%)	35 (100%)	32 (36.4%)	44 (48.4%)	158 (57.9%)
Female	0 (0%)	12 (44.4%)	0 (0%)	56 (63.6%)	47 (51.6%)	115 (42.1%)
Race
Caucasian	29 (90.6%)	0 (0%)	17 (48.6%)	68 (77.3%)	0 (0%)	114 (41.8%)
Other	3 (9.4%)	0 (0%)	0 (0%)	13 (14.8%)	5 (5.5%)	21 (7.7%)
Asian	0 (0%)	27 (100%)	18 (51.4%)	7 (8.0%)	86 (94.5%)	138 (50.5%)
ECOG Score
0	32 (100%)	27 (100%)	35 (100%)	63 (71.6%)	68 (74.7%)	225 (82.4%)
1	0 (0%)	0 (0%)	0 (0%)	21 (23.9%)	22 (24.2%)	43 (15.8%)
2	0 (0%)	0 (0%)	0 (0%)	2 (2.3%)	1 (1.1%)	3 (1.1%)
Missing	0 (0%)	0 (0%)	0 (0%)	2 (2.3%)	0 (0%)	2 (0.7%)
Participants
Healthy subjects	32 (100%)	27 (100%)	35 (100%)	0 (0%)	0 (0%)	94 (34.4%)
Patients with ET	0 (0%)	0 (0%)	0 (0%)	88 (100%)	91 (100%)	179 (65.6%)
Anti‐Drug Antibody
Negative	23 (71.9%)	27 (100%)	29 (82.9%)	79 (89.8%)	85 (93.4%)	243 (89.0%)
Positive	9 (28.1%)	0 (0%)	6 (17.1%)	9 (10.2%)	6 (6.6%)	30 (11.0%)

n, number of participants.

Forward inclusion: For the first round, each parameter–covariate combination was sequentially added to the base model. Based on the log‐likelihood ratio test, the effect of a covariate was significant when the objective function value (OFV) decreased by more than ${\chi }_{\alpha ,\mathrm{\Delta }\mathrm{df}}^2$ (degree of freedom equals Δdf; α equals 0.01) from the reference model. The parameter–covariate combination with the largest decrease in OFV was added to the base model in this round. This served as the reference model for the next round, in which other significant parameter–covariate combinations were tested. The insignificant combinations were abandoned. This process was repeated until all parameter–covariate combinations were insignificant. The model with all significant parameter–covariate combinations was called the full model.

Backward elimination: In the first round, each parameter–covariate combination was sequentially removed from the full model. Based on the log‐likelihood ratio test, the effect of a covariate was significant when the OFV increased by more than ${\chi }_{\alpha ,\mathrm{\Delta }\mathrm{df}}^2$ (degree of freedom equals Δdf; α equals 0.001) from the reference model. The parameter–covariate combination with the smallest increase in OFV was removed from the full model. It then served as the reference model for the next round, in which other parameter–covariate combinations were tested. This process was repeated until all parameter–covariate combinations were significant to develop the final model. Continuous covariates were modeled using Equation (1). Categorical covariates were modeled differently from continuous covariates, as Equation (2) demonstrates by taking gender as an example.

$${\mathrm{P}}_{\mathrm{i}}={\mathrm{P}}_{\mathrm{TV}}\times {\left(\frac{\mathrm{COV}}{\mathrm{CO}{\mathrm{V}}_{\mathrm{med}}}\right)}^{\theta }$$ (1)

$$\left\{\begin{array}{c}{\mathrm{P}}_{\mathrm{i}}\left(male\right)={\mathrm{P}}_{\mathrm{TV}}\\ {\mathrm{P}}_{\mathrm{i}}\left(female\right)={\mathrm{P}}_{\mathrm{TV}}\times \theta \end{array}\right.$$ (2)

*P_TV is the population typical value of the parameter; P_i, subpopulation parameter; COV, continuous covariate; COV_med, median value of the continuous covariate; and θ, influence coefficient of the covariate on the parameter.

Model Evaluation

The visual predictive check (VPC) provides a graphical assessment of how well model predictions align with observed data, evaluating both typical values and variability. Using the final model and its estimated parameters, 1000 simulated datasets were produced via Monte Carlo simulation. The median, 5th, and 95th percentiles of the simulated concentration distributions were then computed and compared visually when superimposed on the actual observations.

Model‐Based Simulations

Based on the constructed PK‐PD models, simulations were performed to investigate the impact of ethnicity on ropeginterferon alfa‐2b exposure and hematological efficacy.

Exposure–Efficacy Analysis

Data

The observed JAK2V617F and CALR VAF changes at Month 6 or 7 and Month 12 or 13 of ropeginterferon alfa‐2b treatment in SURPASS‐ET and EXCEED‐ET informed the exposure–efficacy model development. Ropeginterferon alfa‐2b exposure metrics were derived from simulations based on actual dosing records and individual post hoc (empirical Bayes) PK parameters obtained from the PopPK model constructed in this study. In the efficacy analysis, average concentration (C_{avg, 0–6 or 7 months} and C_{avg, 0–12 or 13 months}) of patients from Month 0 to Month 6 or 7 and Month 12 or 13 was selected to approximate the actual average drug exposure of patients from the start of treatment to Month 6 or 7 and Month 12 or 13. Across the two studies, 12 patients (7.1%) had a treatment duration of less than 6 or 7 months, while 11 patients (6.5%) received treatment for 6 to 7 months up to 12 or 13 months. All remaining patients completed the full 12‐ or 13‐month treatment. For patients with a treatment duration shorter than 6 or 7 months and 12 or 13 months, the average concentration was calculated from the first dose through 2 weeks after the actual last dose.

Before model development, exploratory analyses of the E‐R relationship were performed. For changes in JAK2V617F and CALR VAFs from baseline, scatterplots between exposure and efficacy were used to explore whether there was a significant positive trend between exposure and JAK2V617F or CALR VAF reduction. Statistical significance level was derived from an F‐test. If clear relationships between exposure and efficacy were found by exploratory analyses with clinical significance, model development was performed.

Model Development

Linear or non‐linear regression was used to quantify the relationship between exposure and JAK2V617F or CALR VAF reduction. Different parameterizations were also evaluated (Equations 3 and 4).

$$\begin{array}{c}\hfill \mathrm{Allele}\mathrm{burden}\mathrm{reduction}\mathrm{from}\mathrm{basline}=\\ \hfill {\beta }_0+{\beta }_1·\mathrm{Exposur}{\mathrm{e}}_{\mathrm{i}}+{\beta }^{\mathrm{T}}·{\mathrm{X}}_{\mathrm{i}}\end{array}$$ (3)

$$\begin{array}{c}\hfill \mathrm{Allele}\mathrm{burden}\mathrm{reduction}\mathrm{from}\mathrm{basline}=\\ \hfill {\mathrm{E}}_0+\frac{{\mathrm{E}}_{\max }·\mathrm{Exposur}{\mathrm{e}}_{\mathrm{i}}}{\mathrm{E}{\mathrm{C}}_{50}+\mathrm{Exposur}{\mathrm{e}}_{\mathrm{i}}}+{\beta }^{\mathrm{T}}·{\mathrm{X}}_{\mathrm{i}}\end{array}$$ (4)

$\ast {\beta }_0$ and ${\mathrm{E}}_0$ represent the baseline effect; ${\beta }_1$, scalar parameter representing the effect of exposure; ${\mathrm{E}}_{\max }$, maximum effect; $\mathrm{E}{\mathrm{C}}_{50}$, exposure at half maximum effect; and ${\beta }^{\mathrm{T}}$, vector parameter representing effect of possible covariates ${\mathrm{X}}_{\mathrm{i}}$, respectively.

After establishing the base model, an assessment was conducted to investigate potential covariate effects on the treatment response, if drug exposure was identified as a significant predictor of efficacy (P < .05). The covariates examined matched those included in the PopPK analysis, and the model was developed using a forward‐selection procedure with a statistical significance threshold set at P = .01.

To assess the alignment between model predictions and actual observations, a predicted regression curve along with its 95% confidence interval was plotted against the observed responses across the range of ropeginterferon alfa‐2b exposures. This visual comparison served to evaluate whether the model's predicted range was generally consistent with the distribution range of the observed data.

Exposure–Safety Analysis

Data

The exposure simulation method was consistent with exposure–efficacy analysis; however, alternative exposure metrics were used. The average concentrations throughout the treatment, C_{avg, throughout the treatment} (from time 0 to 2 weeks after the last dose), were simulated and calculated. Boxplots were plotted to explore whether exposure in patients who experienced AE was significantly higher than that in patients who did not experience AE. Statistical significance level was derived from a two‐sided t‐test. Furthermore, probability plots of AE occurrence stratified by quantiles of exposure were developed to investigate the trend of AE occurrence in patients with increasing exposure.

Model Development

Logistic regression was used to quantitatively assess the relationship between exposure and AE occurrence. Multiple parameterizations were considered based on the E‐R relationship observed in exploratory analysis, such as linear and maximum effect (E_max) models (Equations 5 and 6):

$$\begin{array}{ccc}\hfill \mathrm{logit}\left({\mathrm{P}}_{\mathrm{i},\mathrm{AE}}\right)& =& \log \left(\frac{{\mathrm{P}}_{\mathrm{i},\mathrm{AE}}}{1-{\mathrm{P}}_{\mathrm{i},\mathrm{AE}}}\right)={\beta }_0\hfill \\ & & +{\beta }_1·\mathrm{Exposur}{\mathrm{e}}_{\mathrm{i}}+{\beta }^{\mathrm{T}}·{\mathrm{X}}_{\mathrm{i}}\hfill \end{array}$$ (5)

$$\begin{array}{ccc}\hfill \mathrm{logit}\left({\mathrm{P}}_{\mathrm{i},\mathrm{AE}}\right)& =& \log \left(\frac{{\mathrm{P}}_{\mathrm{i},\mathrm{AE}}}{1-{\mathrm{P}}_{\mathrm{i},\mathrm{AE}}}\right)={\mathrm{E}}_0\hfill \\ & & +\frac{{\mathrm{E}}_{\max }·\mathrm{Exposur}{\mathrm{e}}_{\mathrm{i}}}{\mathrm{E}{\mathrm{C}}_{50}+\mathrm{Exposur}{\mathrm{e}}_{\mathrm{i}}}+{\beta }^{\mathrm{T}}·{\mathrm{X}}_{\mathrm{i}}\hfill \end{array}$$ (6)

* ${\beta }_0$ and ${\mathrm{E}}_0$ represent the baseline logit; ${\beta }_1$, scalar parameter representing the effect of exposure on the logit; ${\mathrm{E}}_{\max }$, maximum effect on the logit; $\mathrm{E}{\mathrm{C}}_{50}$, exposure at half maximum effect; and ${\beta }^{\mathrm{T}}$, vector parameter representing the effect of possible covariates ${\mathrm{X}}_{\mathrm{i}}$ on the logit, respectively.

Other modeling procedures were consistent with those in the development of the exposure–efficacy model.

Software and Data Analysis

The PopPK‐PD model was developed and applied using NONMEM (version 7.5.0, ICON Development Solutions, Ellicott City, MD, USA), a nonlinear mixed‐effects modeling software. Automation and post‐processing tasks were handled by Perl‐speaks‐NONMEM (version 5.2.6). During the entire PopPK‐PD model development process, parameter estimation was conducted using the first‐order conditional estimation (FOCE) method, incorporating the η‐ε interaction option. For data analysis, graphical visualization, and E‐R model development, R software (version 4.2.2, R Foundation for Statistical Computing, Vienna, Austria) was utilized.

Results

Study Population

A total of 94 healthy volunteers and 179 patients with ET with available PK data from the Phase 2 and Phase 3 clinical studies were included in the PopPK model development. All 179 patients with ET were included in the subsequent PK‐PD and E‐R analyses. The baseline demographics and disease characteristics are summarized in Table 1. Most of the subjects were Caucasian and Asian: 114 (42%) and 138 (51%), respectively.

PopPK Model

Table 2 presents the parameter estimates of the final PopPK model obtained using the FOCE method. The fixed and random effect parameters were estimated with high precision, as indicated by relative standard errors below 15%. The RUV followed a normal distribution and was expressed using a combined error model consisting of proportional and additive residual error components. Body mass index (BMI) was included as a covariate of clearance. Clearance was higher in subjects with higher BMI. Vc/F for ET patients was higher than that of healthy subjects. Collectively, the predictions from both the population and individual models accurately captured the observed data. Goodness‐of‐fit plots demonstrated a uniformly distributed, random scatter of points around the line of identity. Furthermore, the conditional weighted residuals showed no discernible trends when assessed across the evaluated concentration ranges and time points (Figure S1). The VPC figure (Figure S2) indicates the predictive ability of the model for describing the ropeginterferon alfa‐2b concentrations.

Table 2.

Parameter Estimates of Final PopPK Model

Model Parameter (units)	Estimate	RSE (%)	Shrinkage (%)
CL/F (L/day) at median BMI of 24 kg/m2 CL/F‐BMI power	0.606 1.03	2.68 14.8	‐
Vc/F (L) Fold‐increase in Vc/F for ET patient	1.88 2.007	8.37 11.9	‐
ka (1/day)	0.123	3.82	‐
Tlag (h)	0.651	2.09	‐
Rtot,0 (ng/mL)	0.17	15.9	‐
Rtot, SS (ng/mL)	0.0142	13.3	‐
kint (1/h)	0.12	19.8	‐
kdeg (1/h)	0.579	7.63	‐
KD (ng/mL)	0.0455	32.2	‐
kdec (1/day)	0.0203	2.7	‐
ω for CL/F	0.312	3.87	‐
ω for Vc/F	0.867	5.79	‐
ω for ka	0.34	7.25	15.1
ω for Rtot,0	1.36	7.28	10.3
ω for kint	1.67	7.26	33.7
σ of proportional residual error	0.209	0.637	20.6
σ of additional residual error (ng/mL)	0.155	5.18	23

BMI, body mass index; CL/F, apparent clearance; ET, essential thrombocythemia; K_a, absorption rate constant; K_D, binding affinity of ropeg to its target; k_int, first‐order rate of elimination of the drug‐target complex; k_deg, first‐order rate constant for degradation; k_dec, first‐order rate constant of decline in R_tot (binding capacity) over time to a new steady‐state value; RSE, relative standard error; R_tot0, baseline maximum binding capacity; R_{tot, SS}, steady‐state maximum binding capacity; T_lag, absorption lag time; V_C/F, apparent volume of serum compartment.

Shrinkage (%), calculated as $100·(1-\frac{\mathrm{sd}(\widehat{\eta })}{\mathrm{SD}})$, in which $\mathrm{sd}(\widehat{\eta })$ is the standard deviation of maximum a posteriori probability Bayes estimates of individual random effects. The proportional and additional residual errors represent RUV (residual unexplained variability).

PK‐PD Model

Table 3 shows parameter estimates of the final popPD model of platelet and WBC using the FOCE method. Both fixed and random effects of these three models were precisely estimated, with RSEs for fixed effect parameters less than 15%. Baseline JAK2V617F allele burden and race were significant covariates on WBC₀. The VPC figures (Figure S3A,B) indicate excellent predictive ability of the model for describing the platelet and WBC counts.

Table 3.

Parameter Estimates of PopPK‐PD Model

Model	Model Parameter	Estimate	RSE (%)	Shrinkage (%)
WBC	kout (1/h)	0.07	1.92	‐
	EPW	0.147	1.65	‐
	WBC0 (109/L)	7.67	1.65	‐
	Imax,W	0.493	1.76	‐
	IC50,W (ng/mL)	7.09	1.79	‐
	γ	0.923	2.1	‐
	JAK2 on WBC0	0.178	15.2	‐
	Asian on WBC0	1.34	14	‐
	ω for kout	1.26	8.85	19.6
	COV for IIV_kout and IIV_EPW	−0.604	29.9	‐
	ω for EPW	1.3	9.44	24.9
	COV for IIV_EPW and IIV_WBC0	0.163	27.8	‐
	ω for WBC0	0.305	5.74	7.57
	COV for IIV_WBC0 and IIV_HImax,W	0.0539	48.1	‐
	ω for IIV_ HImax,W	0.832	8.04	25.3
	COV for IIV_IC50,W and IIV_HImax,W	0.376	31.8	‐
	ω for IC50,W	0.702	20.1	48.3
	σ of proportional residual error	0.162	0.552	5.06
	σ of additional residual error (109/L)	0.0000146	0	5.06
PLT	kout (1/h)	0.033	1.73	‐
	PLT0 (109/L)	884	1.67	‐
	Imax,P	0.598	1.26	‐
	IC50,P (ng/mL)	7.02	1.86	‐
	EPP	0.121	1.86	‐
	γ	1.86	2.02	‐
	ω for kout	0.632	6.17	10.1
	ω for PLT0	0.353	6.55	0.837
	COV for IIV_PLT0 and IIV_HImax,P	0.0998	22.7	‐
	ω for HImax,P	0.589	8.04	17.9
	ω for IC50,P	1.02	9.3	24
	ω for EPP	1.5	11.5	31.9
	σ of proportional residual error	0.116	0.541	6.4
	σ of additional residual error (109/L)	0.103	3.67	6.4

γ, Hill coefficient; COV, covariance; EPW, empirical fractional scaling parameter incorporated on the baseline while blood cells production term; EPP, empirical fractional scaling parameter incorporated on the baseline platelet cells production term; HI_max,P, maximum reduction due to ropeg for platelet count on logit; HI_max,W, maximum reduction due to ropeg for white blood cell count on logit; IC_50,P, the serum ropeg concentration resulting in 50% of I_max,P; IC_50,W, serum ropeg concentration resulting in 50% of I_max,W; k_out, first‐order elimination rate constant; PLT, platelet counts; PLT₀, initial value of platelet count; RSE, relative standard error; WBC, white blood cell counts; WBC₀, initial value of white blood cell counts.

Shrinkage (%), calculated as $100·(1-\frac{\mathrm{sd}(\widehat{\eta })}{\mathrm{SD}})$, in which $\mathrm{sd}(\widehat{\eta })$ is standard deviation of maximum a posteriori probability Bayes estimates of individual random effects.

Model‐Based Simulations

Based on the constructed PopPK model of ropeginterferon alfa‐2b, individual exposures under the HIDAT regimen were simulated. The median exposures for different racial populations were all in the range of 80%–125% of the population median, indicating different ethnic backgrounds had minimal effect on ropeginterferon alfa‐2b exposure parameters (Figure S4 and Table S1).

Based on the constructed PK‐PD model of WBC, a virtual population was generated based on typical values and IIV and given the HIDAT regimen to simulate the WBC profile of populations of different ethnicities and baseline JAK2V617F VAF. Simulated WBC profiles are shown in Figure 2. Asians and higher baseline JAK2V617F VAF lead to a higher baseline WBC. However, in both Asians and Caucasians, WBC can decrease rapidly to below target (9.5 × 10⁹/L) soon after the administration of ropeginterferon alfa‐2b, with simulated median times to reach the target being 3.1 days (Caucasians) and 11.8 days (Asians). A 1‐week variance represents a minor deviation when considered over the course of the 1‐year therapy duration.

Figure 2.

Simulated median white blood cell counts for different populations. Three panels are white blood cell count profiles under baseline JAK2V617F 10th, 50th, and 90th percentile for the modeling population. The blue solid line represents the predicted median of white blood cell counts of Asian, while the red solid line represents the predicted median of white blood cell counts of Caucasian. The blue shaded area represents the 90% prediction interval for white blood cell counts of Asian, while the red shaded area represents the 90% prediction interval for white blood cell counts of Caucasian. WBC, white blood cell counts; JAK2, JAK2V617F allele burden.

Exposure–Efficacy Analysis

In two studies conducted in ET patients, 87 and 80 patients underwent JAK2V617F VAF measurements at Months 6 or 7 and 12 or 13, respectively. For CALR VAF, 22 and 18 patients underwent measurements at Months 6 or 7 and 12 or 13, respectively.

Both scatterplots (Figure S5A,B) of C_avg for JAK2V617F VAF reduction at Months 6 or 7 and 12 or 13 from baseline suggested significant positive linear trends between exposure and efficacy (P = .008 and P = .035 for Months 6 or 7 and 12 or 13, respectively). No efficacy plateau was observed. However, no positive correlation was observed between C_avg and CALR VAF (Figure S5C,D). Then, treatment effects on JAK2V617F VAF reduction at Months 6 or 7 and 12 or 13 from baseline were modeled using two linear regression models for ropeginterferon alfa‐2b exposure on JAK2V617F molecular response. For both models, baseline JAK2V617F allele burden was identified as a significant covariate that significantly improved model fit. The final model showed that JAK2V617F VAF reduction from baseline increased with higher exposure and was dependent on baseline JAK2V617F VAF (Figure 3a,b). Table 4 presents parameter estimates of the two linear regression models. Both predicted linear regression curves and the observed JAK2V617F reduction from baseline in the range of ropeginterferon alfa‐2b exposure showed good alignment between observed and predicted data (Figure 3c,d).

Figure 3.

Exposure–response relationship for JAK2V617F reduction from baseline and occurrence of anemia. (a, b) Predicted linear regression curve and observed JAK2V617F allele burden reduction from baseline in the range of ropeginterferon alfa‐2b exposure at Month 6 or 7 and Month 12 or 13, respectively. Solid blue lines are the linear fit lines of linear regression models. Blue shading is the 95% prediction interval of the models. Open black circles are the observed change from baseline in JAK2V617F allele burden at Month 6 or 7 and Month 12 or 13, respectively. Gray boxes under each plot represent the exposure distribution of each study. (c,d) Predicted linear regression curve under ropeginterferon alfa‐2b exposure and baseline JAK2V617F allele burden at Month 6 or 7 and Month 12 or 13, respectively. Blue, yellow, and green solid lines are the first quantile, median, and third quantiles for baseline JAK2V617F allele burden, respectively. Gray boxes under each plot represent exposure distribution; red vertical dashed lines represent median lines of exposure under the given dosing regimen. (e) Predicted logistic regression curves and observed anemia occurrence for each ropeginterferon alfa‐2b exposure quantile. Open blue circles represent AEs observed in patients. Solid black circles are the observed probability of AEs at the median of each exposure quantile interval. The error bars are standard deviations—calculated as sqrt (P*(1 − P) *N⁻¹), where P is the probability of AEs, and N is the number of patients in each quantile interval. Solid blue lines are fitted to model predictions, and the shaded blue area is the 95% confidence interval of the predicted values. Gray boxes under each plot represent the exposure distribution of each study. (f) Predicted logistic regression curve of anemia occurrence under ropeginterferon alfa‐2b exposure and baseline hemoglobin. Blue, yellow, and green solid lines are first, median, and third quantiles for baseline hemoglobin, respectively. Gray boxes under each plot represent exposure distribution; red vertical dashed lines represent median lines of exposure under the given dosing regimen. C_{avg,0 to 6 or 7 months}, average concentration of participants from 0 to 6 or 7 months_; C_{avg,0 to 12 or 13 months}, average concentration of participants from 0 to 12 or 13 months_; C_{avg, throughout the treatment}, average concentration throughout the treatment; JAK2V617F, JAK2V617F allele burden; N, number of patients in a subgroup; SD, standard deviation.

Table 4.

Parameter Estimates of Exposure–Response Models

Model	Parameter	Estimate (Standard Error)	P Value
Linear regression of JAK2V617F at Month 6 or 7	Intercept, ${\beta }_0$	2.45 (2.41)	.310545
	Exposure effect, ${\beta }_1$	−0.127 (0.054)	.021201
	Baseline JAK2 effect ${\beta }_2$	−0.079 (0.038)	.043368
Linear regression of JAK2V617F at Month 12 or 13	Intercept, ${\beta }_0$	1.94 (3.807)	.610909
	Exposure effect, ${\beta }_1$	−0.147 (0.082)	.077276
	Baseline JAK2 effect ${\beta }_2$	−0.103 (0.058)	.082337
Logistic regression of occurrence of anemia	Intercept, ${\beta }_0$	5.20 (1.77)	.00336
	Exposure effect, ${\beta }_1$	0.024 (0.0106)	.02155
	Baseline hemoglobin effect ${\beta }_2$	−0.522 (0.1306)	<.001

P value was based on z‐test.

Exposure–Safety Analysis

The relationship between exposure and incidence of safety indicators of interest was explored in this study. A summary of AE occurrence is presented in Table S2. A total of 179 ET patients from SURPASS‐ET and EXCEED‐ET were included in the exposure–safety population (Table 1). Of the safety indicators, there was a positive association of anemia development with average exposure of ropeginterferon alfa‐2b. As the graphical results (Figure S6) show that the incidence of anemia increased linearly with increasing exposure, the exposure–safety relationships were subsequently modeled using logistic regression.

A logistic regression model for C_{avg, throughout the treatment} and anemia indicated a statistically significant E‐R relationship for this AE. Results of the final model showed that anemia increased with increasing exposure (Figure 3e). The predicted logistic regression curve and observed occurrence of anemia increased in each ropeginterferon alfa‐2b exposure quantile, showing good alignment between observed and predicted data. Baseline hemoglobin level was further identified as having a significant covariate effect. The lower baseline hemoglobin level was correlated with a higher rate of anemia (Figure 3f).

Discussion

The PK‐PD models show consistent reductions in both platelet and WBC counts into target ranges, supporting the clinical utility of ropeginterferon alfa‐2b in controlling established thrombotic risk factors in ET patients, including leukocytosis and thrombocytosis. Furthermore, the model, characterized by sigmoidal indirect‐response mechanisms, successfully captured the E‐R relationships for these key PD markers after administration of ropeginterferon alfa‐2b. These models incorporated parameters accounting for initial values, thereby enhancing the fit by accommodating potential variability in initial measurements, which could stem from previous therapies or inherent disease‐related fluctuations. For the indirect effects model regarding platelets, the estimated typical initial count was 884 × 10⁹/L. Ropeginterferon alfa‐2b was estimated to maximally reduce k_in by 65%, with corresponding IC_{50, P} estimated to be 7.02 ng/mL. For the final indirect effects model for WBC, the estimated typical initial count was 7.67 × 10⁹/L, with ropeginterferon alfa‐2b estimated to maximally reduce k_{in, W} 57% and corresponding IC_{50, W} estimated at 7.09 ng/mL. In previous PK‐PD modeling studies involving patients with PV, it was shown that initial platelets were lower than those in patients with ET (477 × 10⁹/L), while initial WBCs were higher (11.4 × 10⁹/L). ²⁹ The model parameters reflected differences in initial blood cell counts across different disease types. Regarding IC₅₀, values reported in prior studies of patients with PV were higher than those observed in the present study (IC_50,W = 152 ng/mL and IC_50,P = 72.4 ng/mL). ²⁹ This discrepancy may be attributed to the inclusion of both HIDAT and slow titration regimens in previous PV models, whereas the current study only had patients with ET under the HIDAT regimen and covered the full dynamic profile of PD response from baseline to plateau. The data features deliver better parameter identifiability. As a result, the estimated IC₅₀ was closer to the true intrinsic value, hence showing a lower numerical value.

In this study, the PK‐PD model accounted for measured covariates and modeled random effects. The simulation results indicated that Asian patients took approximately 1 week longer than Caucasian patients to first achieve a WBC < 9.5 × 10⁹/L, a difference which might not be clinically significant. The underlying cause might be due to the difference in baseline WBC levels, which were higher on average in Asian than in Caucasian patients.

JAK2V617F VAF is a known risk factor for thrombosis and cardiovascular disease in MPN, being observed in patients with PV or ET. ³⁰ , ³¹ , ³² , ⁴³ , ⁴⁴ For the JAK2V617F VAF decrease, average ropeginterferon alfa‐2b concentration was a significant predictor at both Month 6 or 7 and Month 12 or 13. Furthermore, baseline JAK2V617F VAF was a covariate for the VAF decrease post ropeginterferon alfa‐2b treatment. No ethnic difference was observed. The model suggests that higher exposure to ropeginterferon alfa‐2b effectively reduces JAK2V617F VAF and therefore eliminates neoplastic clones harboring the JAK2V617F mutation. Our findings highlight the requirement for maintaining a higher PK exposure to support the disease‐modifying action of ropeginterferon alfa‐2b. In this study, an E‐R relationship for the CALR VAF was not clearly defined even though ropeginterferon alfa‐2b treatment led to notable CALR molecular responses. ²⁵ It is likely to be due to a need for a larger sample size in establishing an E‐R relationship, rather than a lack of an exposure‐related effect. Further E‐R analyses with more patient samples or after a longer time of treatment are planned for the CALR VAF decrease.

The nonlinear PK often observed with interferons, attributable to target binding, is a characteristic feature previously documented for IFN‐β. ⁴⁵ , ⁴⁶ In this analysis, the TMDD model structure and significant covariates included align with our priordata. ²⁹ , ⁴² Besides, the effect of patients with ET on the apparent volume of distribution was also included as a covariate; however, we found that this covariate had a limited impact on exposure. Model‐based simulation showed that ropeginterferon alfa‐2b exposure was not significantly different across ethnic groups.

In the exposure–safety analysis, we found that the incidence of anemia increased with increasing exposure, suggesting a risk of anemia in ET patients with higher exposure. However, only five cases of Grade 3 anemia without more severe levels were observed, and most anemia cases were reversible and manageable. Although the exposure‐related anemia risk does not appear to significantly impact the benefit‐to‐risk ratio, our finding supports close monitoring of anemia at high doses of ropeginterferon alfa‐2b treatment in patients with ET.

We previously analyzed the E‐R relationships of the HIDAT versus slow titration regimen of ropeg in patients with PV. The results showed that the incidence of manageable elevated ALT and AST levels increased with exposure in the dose titration phase. ²⁹ In the present study, all patients with ET received the HIDAT regimen, and the exposure metrics represented the overall 1‐year treatment period of individual patients under the same dosing regimen. The analyses reflect differences in the inter‐patient exposure variabilities associated with the response. Our results indicate that ALT and AST elevations were not significantly associated with individual patient exposure levels among patients with ET.

This PopPK‐PD and E‐R analysis in healthy subjects and patients with ET has several limitations. No correlation was observed between the change from baseline in CALR VAF and ropeginterferon alfa‐2b exposure in the exposure–efficacy analysis, likely due to the small sample size. Furthermore, since most CALR‐mutation‐positive patients in the analysis were Caucasian from the EXCEED‐ET study, an evaluation of potential ethnic differences was not feasible. In the PK‐PD analysis, WBC and platelet counts were not monitored after treatment discontinuation or interruption. Accordingly, our PK‐PD analyses were unable to capture the duration of efficacy and post‐treatment rebound.

Conflicts of Interest

Albert Qin serves as the chief medical officer of PharmaEssentia. Lucia Masarove: advisory board participation with Cogent, GSK, MorphoSys, and PharmaEssentia. Zhijian Xiao, Jie Jin, and Lei Zhang declared no competing interests in this work. Harinder Gill: consulting fees/speaker fees from AbbVie, Bristol Myers Squibb, GSK, MSD, Novartis, PharmaEssentia Corporation, and Pfizer; advisory board for Abbvie, Bristol Myers Squibb, GSK, Novartis, PharmaEssentia Corporation, and Pfizer; research support from Celgene, MSD, Novartis, and PharmaEssentia Corporation; conference support from Bristol Myers Squibb, MSD, Novartis, Otsuka, PharmaEssentia, and Pfizer. Brandi N Reeves received honoraria/consulting fees from PharmaEssentia, Pint Pharma, CTI BioPharma, GSK, Incyte, and Sobi. Firas El Chaer is serving as a consultant for SPD Oncology, Amgen, CTI BioPharma, AbbVie, MorphoSys, and Association of Community Cancer Centers; received Clinical Trial Grant Support (PI) to the University of Virginia from Amgen, BMS, Celgene, SPD Oncology, Sanofi, Bristol Myers Squibb, FibroGen, PharmaEssentia, BioSight, MEI Pharma, Novartis, and Arog Pharmaceuticals; and a travel grant from DAVA Oncology. Keita Kirito received honoraria for lectures, presentations, speaking engagements, manuscript writing, and educational events from AbbVie G.K., Novartis Pharma K.K., PharmaEssentia Japan K.K., Sanofi K.K., and Takeda Pharmaceutical Co., Ltd. Norio Komatsu is a board member of and has received funding for medical writing and article processing charges from PharmaEssentia Japan K.K. and received research funding from Meiji Seika Pharma Co., Ltd. and consulting fees from Torii Pharmaceutical Co., Ltd. Ghaith Abu‐Zeinah received honoraria for speaking engagements from PharmaEssentia and research support provided to the institution from PharmaEssentia; is a consultant for PharmaEssentia, Takeda, and Silence Therapeutics. Kazuya Shimoda is a consultant for Sierra Oncology, Inc. and reports research funding from AbbVie G.K., Chugai Pharmaceutical Co., Ltd., Daiichi Sankyo Co., Ltd., Eisai Co., Ltd., Kyowa Kirin Co., Ltd., Mochida Pharmaceutical Co., Ltd., Nippon Kayaku Co., Ltd., Otsuka Pharmaceutical Co., Ltd., PharmaEssentia Japan K.K., Shionogi & Co., Ltd., Sumitomo Dainippon Pharma Co., Ltd., Taisho Pharmaceutical Co., Ltd., and Takeda Pharmaceutical Co., Ltd. and honoraria from Takeda Pharmaceutical Co., Ltd., and Novartis Pharma K.K., and is a member of an advisory committee for AbbVie. Oleh Zagrijtschuk is employed by PharmaEssentia USA. Toshiaki Sato is employed by PharmaEssentia Japan K.K. Daoxiang Wu is employed by PharmaEssentia Biotech (Beijing) Ltd. Haoqi Chen, Lei Wu, Xia Su, Yucheng Gao, and Jie Cui are employed by Pharmaron Clinical Services, Co. Ruben A Mesa: consulting fees from Incyte, PharmaEssentia, CTI, AbbVie, GSK, BMS, and Genentech.

Funding

The research was supported by PharmaEssentia Corporation.

Supporting information

Supplemental Information

Data Availability Statement

Data are available from PharmaEssentia upon reasonable request. Some data may not be available because of privacy or ethical considerations.