Trilaciclib dose selection: an integrated pharmacokinetic and pharmacodynamic analysis of preclinical data and Phase Ib/IIa studies in patients with extensive-stage small cell lung cancer

Owing to the invasive nature of the PD marker collection (i.e. bone marrow aspiration), a semi-mechanistic PK/PD model based on animal PK/PD data and human PK data was developed to guide the selection of an optimal dose level for bone marrow aspiration in humans during dose escalation in the Phase I first-in-human study. The model was constructed using PK data from four species (mouse, rat, dog, and human) and PD data from three species (mouse, rat, and dog; Online Resources 1 and 2).

The PK/multi-compartment PD model was based on a myelosuppression model previously published by Friberg et al. [16], with modifications. Specifically, the Friberg model did not differentiate between the bone marrow stem cell and progenitor cell compartments, but combined these into a single proliferative cell compartment. In the model used in the current study, stem cells and progenitor cells were separated into two compartments, with experimental data measured for each compartment. Additionally, the current model further divided the cell cycles into G1-phase and S-phase compartments, which is critical to capture trilaciclib’s mechanism of arresting cells in the G1 phase. Another important deviation from the Friberg model was the mechanism postulated for the homeostasis of the peripheral mononuclear cells (PMNs). We postulated that the feedback mechanism has a set point target and that deviations from that target would accelerate (in the case of too few PMNs) or decelerate (in the case of too many PMNs) the transition of G1-phase cells to S-phase cells in the stem cell population.

The model was constructed in two stages. First, preclinical PK data (e.g. doses, routes, and schedule of administration, observed concentrations) were compiled and modeled by allometric scaling. A sequential PK/PD modeling approach (population PK parameters and data approach [17]) was then employed in which a semi-mechanistic PD model with various physiological compartments was created, and specific assumptions entered (Online Resource 3). Traditional forward addition/backward elimination was used for PK modeling. Forward addition/backward elimination consists of adding “features” to the model (one at a time) until the data support no more parameters, then eliminating features (one at a time) until all remaining effects are statistically significant by likelihood ratio test, or until diagnostic plots suggest that the model is less predictive. Many parameters in this complex model could not be estimated and were fixed either to the literature values or values selected on the basis of discussions with domain experts and sensitivity analysis. In preclinical animal models, the PD analysis modeled bone marrow cells positive for 5′-ethynyl-2′-deoxyuridine (EdU) and peripheral neutrophil counts simultaneously. Although the preclinical animal models could determine the G1 phase, they could not differentiate S, G2, and M phases; cells in these phases will be referred to as “cycling” throughout this manuscript.

A combined systems biology and empiric modeling approach was taken, largely based on an understanding of the biology of hematopoiesis and the mechanism of action of trilaciclib, with literature values used when needed, and domain expert estimates used when literature values were not available. The multi-compartment PD model (Fig. 1a) was built to capture the effect of trilaciclib on hematopoiesis, with cell proliferation originating at the hematopoietic stem cell (HSC), differentiating into progenitor cells, and ultimately culminating in the release of a mature neutrophil into peripheral circulation. An additional regulatory mechanism was used to describe a “persistence effect” on the delayed transition from G1-phase stem cells to cycling stem cells even after concentrations of trilaciclib had fallen to below sub-therapeutic levels observed in preclinical studies, which may be related to the time required for the biosynthesis of required proteins, or other processes required prior to transition to the S/G2/M phase. The effect of trilaciclib was modeled as a maximal effective concentration (E_max) in plasma on the transition rate constant from G1 to cycling for HSPCs. The effect of 5-fluorouracil (5-FU) was modeled as removing (killing) cells while HSPCs are cycling. Owing to model complexity, only a minority of parameters could be estimated. The remaining parameters were either fixed using literature-based values or reliance on the model to find the best fit.

Simulations were performed for total cycling bone marrow cells across different dose levels from 4 to 384 mg/m². The PK model for trilaciclib was fixed to the final model from the PK analysis, and an additional 5-FU model was fixed to the literature values [18]. The literature estimate for the half-life of 5-FU is 8–20 min. The PD of 5-FU cytotoxicity, an inhibitor of thymidylate synthase [19], suggests that an indirect model [20] may be appropriate and, therefore, that PD activity may persist after the concentration has fallen. In addition, a key consideration in interpreting the result of this analysis was to prevent the release of the G1-phase cells into S phase until there was confidence that the cytotoxicity of the 5-FU was eliminated. Thus, a conservative model for 5-FU activity was sought. Neither data nor literature was available to support such a model. Therefore, a modification of the literature estimate of a half-life of 8–20 min was made, and the pharmacologic activity half-life of 5-FU was estimated at 42 min (elimination rate constant of 1/h). The 5-FU PD model then consisted of a first-order elimination model with a rate constant of 1/h. Volume could not be estimated, as no concentration data were available, and was effectively fixed to 1.0. The final PD model fixed values obtained from model fits for the transit constant for stem cells at 0.5/h (i.e. mean residence time [MRT] for cycling stem cells was estimated to be approximately 2 h; Online Resource 3). The ratio of G1-phase to cycling cells was set to 1% (stem cells spend 1% of their time cycling [21]), and the total MRT for bone marrow progenitors was set to 120 h. Under the assumption of 120 h for the total bone marrow progenitor life span (in line with reported transit time from progenitor cells to peripheral PMNs [22, 23]), the cell cycle through five G1–S/G2/M-phase cycles, with 16 h in the G1 phase and 8 h cycling per cycle, was consistent with animal data [24]. These values could not be estimated and were fixed; however, by testing feasible ranges for these parameters in the context of a sensitive model (i.e. the goodness of the model fits is sensitive to model parameters), parameter values with the best fit for the observed data were selected. In the 5-FU model, the elimination rate constant was fixed to 1/h, and the concentration resulting in 50% of maximum effect (EC₅₀) for 5-FU, as identified in an earlier model (in-house modeling results), was fixed at 0.000334 ng/mL. The E_max for 5-FU (fraction of cells that could be killed at a very high concentration) was fixed to 1.0. In the final model, between-subject variability (log normal in all cases) was included on central volume of distribution, clearance, and the scale parameter for PMNs. Random between-species variability was included on central volume of distribution and clearance. The residual error for both bone marrow cells and PMNs was log normal.

Model diagnostics consisted of standard PK/PD plots, including basic goodness of fit, individual PK and PD predictions and observed values versus time, conditional weighted residuals versus time, and visual predictive checks (VPCs; Online Resources 4 and 5) [25].

Parameters from emerging human PK data from Phase I dose escalation were incorporated into the PK/PD model to simulate the relative percentage of total bone marrow cells cycling or arrested in the G1 phase and the relative percentage of peripheral neutrophils over time, following a single dose of trilaciclib. NONMEM version 7.3 (ICON PLC, Hanover, MD, USA) was used for parameter estimations [26]. XPOSE version 4 [27] was used for figures, and Perl-speaks-NONMEM version 3.62 [28] used for VPC calculation.

Details of the G1T28-1-01, G1T28-02, and G1T28-03 study designs have been reported previously [5, 11, 14].

Briefly, Study G1T28-1-01 (NCT02243150) was a single-center, single-dose, first-in-human study of trilaciclib in healthy human volunteers [11]. Healthy male and/or female subjects aged 18–60 years, with a body weight of ≥ 50 kg and a body mass index of 18–32 kg/m², were enrolled. The study included a double-blind, randomized, placebo-controlled single ascending dose (SAD) part in six 4- to 8-subject cohorts (3:1; trilaciclib vs placebo [6, 12, 24, 48, 96, or 192 mg/m²]; 30-min intravenous [IV] infusion), and an open-label 12-subject cohort who received a single 192 mg/m² dose to confirm the BED on the basis of the totality of available safety, PK, and PD data. The PD activity of trilaciclib was assessed in an ex vivo phytohemagglutinin (PHA)-stimulated lymphocyte proliferation assay (SAD and BED cohorts), and by evaluation of HSPC proliferation in bone marrow aspirates (BED cohort only). To assess the anti-proliferative activity of trilaciclib on lymphocytes, blood samples were obtained from subjects in the BED cohort and the SAD 96 and 192 mg/m² cohorts pre-dose and at specified times after the end of infusion, and then treated with PHA for 48 h ex vivo to stimulate lymphocyte proliferation. Following a 1-h exposure to EdU, cells were harvested, processed, and analyzed by flow cytometry to measure the percentage of proliferation through EdU + incorporation in CD45 + /CD3 + cells (T lymphocytes). A single bone marrow aspirate at various time points relative to IV dosing of trilaciclib was obtained from all subjects enrolled in the BED cohort to determine the effect of trilaciclib on the cell cycle phases (i.e. G1 or cycling) of various bone marrow progenitor lineages, on the basis of expression of cell-surface markers. HSPC proliferation was measured before dosing (n = 5), 24 h after trilaciclib dosing (n = 3), or 32 h after trilaciclib dosing (n = 4), using flow cytometry [11].

G1T28-02 (NCT02499770) was a multicentre Phase Ib/IIa study in patients with newly diagnosed ES-SCLC [5]. Eligible patients were aged ≥ 18 years and had histologically or cytologically confirmed ES-SCLC, measurable disease per Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1, an Eastern Cooperative Oncology Group (ECOG) performance status of 0–2, and adequate organ function. Part 1 was a Phase Ib, open-label, dose-finding portion, in which patients received trilaciclib 200 or 240 mg/m², followed by Part 2, a Phase IIa, randomized, double-blind, placebo-controlled, expansion portion at the recommended Phase II dose (RP2D). All patients (Parts 1 and 2) received carboplatin (area under the concentration–time curve [AUC] 5 mg min/mL) on Day 1 and etoposide (100 mg/m²) on Days 1–3, and IV trilaciclib (or placebo in Part 2) was administered once daily before chemotherapy. The primary objective of the study was to define the trilaciclib RP2D (Part 1) and to assess the safety and tolerability of trilaciclib in combination with E/P (Parts 1 and 2). Secondary objectives included analysis of the PK profiles of trilaciclib and of E/P (Part 1), and assessment of trilaciclib efficacy (Parts 1 and 2), including thorough evaluation of common myelosuppression endpoints. In the dose-finding Part 1 portion, blood samples for PK analysis were collected at prespecified time points on Cycle 1 Day 1 and Cycle 1 Day 3 (200 mg/m² cohort: all patients; 240 mg/m² cohort: optional).

G1T28-03 (NCT02514447) was a multicentre Phase Ib/IIa study in patients with previously treated ES-SCLC [14]. Patients were aged ≥ 18 years and had histologically or cytologically confirmed ES-SCLC that had progressed during or after prior first- or second-line chemotherapy, measurable disease per RECIST version 1.1, an ECOG performance status of 0–2, and adequate organ function. Part 1 was a Phase Ib, open-label, dose-finding portion. Part 2 was a Phase IIa, randomized, double-blind, placebo-controlled portion. In Part 1, trilaciclib (200, 240, or 280 mg/m²) was administered once daily prior to chemotherapy on Days 1–5 of each 21-day topotecan cycle (IV topotecan [0.75, 1.0, 1.25, or 1.5 mg/m²]; Days 1–5). In Part 2a, patients were randomized 2:1 to receive trilaciclib 240 mg/m² prior to topotecan 0.75 mg/m², or placebo prior to topotecan 1.5 mg/m² using the same schedule as that in Part 1. Part 2b evaluated trilaciclib 240 mg/m² prior to topotecan 1.5 mg/m² using the same schedule as that in Part 1. Study objectives included assessment of dose-limiting toxicities (DLTs; Part 1), safety and tolerability (Parts 1 and 2), PK (mainly Part 1), and myelosuppression efficacy (Parts 1 and 2) of trilaciclib in combination with topotecan. In Part 1 of the study, blood samples for PK analysis were collected at prespecified time points on Cycle 1 Day 1 and Cycle 1 Day 4 for all patients.

All studies were designed and conducted in compliance with the principles of the Declaration of Helsinki and the Good Clinical Practice guideline of the International Council for Harmonisation. Study protocols and study-related materials were approved by the institutional review board or independent ethics committee of each investigational site. Written informed consent was obtained from each patient before the initiation of study procedures.

Previously reported preclinical data and PK/PD modeling were used to estimate the dose level at which a maximal decrease in the percentage of cycling HSCs would be expected, to be further assessed using bone marrow sampling in healthy human volunteers. Data obtained using murine HSCs indicated that a single dose of trilaciclib induced a dose-dependent, reversible cell cycle arrest of all hematopoietic cell types, and that concentrations resulting in the maximum (~ 50%) decrease from baseline in the relative percentage of total bone marrow cycling cells were effective in decreasing the percentage of cycling HSPCs so that almost 100% of HSPCs were arrested in the G1 phase [11]. Similar dose-dependent effects on G1 arrest of total bone marrow cycling cells (~ 50% maximum decrease from baseline) were observed in dogs [13], although results on specific cell types were not available owing to study limitations. Since hematopoietic development is well-conserved across species, trilaciclib’s CDK4/6 inhibition in humans was expected to be similar to that in mice and dogs, and mouse and dog data infer that a dose level resulting in an approximate 50% decrease in total bone marrow cycling cells would result in an almost 100% decrease in the percentage of cycling HSPCs, representing a therapeutically effective dose.

The final PD model (Fig. 1a) suggested an EC₅₀ value of 5.9 ng/mL for trilaciclib, which predicts an EC₉₀ (therapeutically active dose) of approximately 50 ng/mL. Model simulation of bone marrow arrest predicted that a trilaciclib dose of ≥ 192 mg/m² would induce a 40–50% decrease from baseline in total bone marrow proliferation, with a return to baseline at approximately 48 h (Fig. 1b). The level of decrease was close to the maximum effect observed in animal studies. In addition, per simulation, a higher dose of 384 mg/m² showed a similar PD effect. Therefore, 192 mg/m² was selected as the dose to obtain bone marrow aspirates from healthy subjects enrolled in the first-in-human study (G1T28-1-01).

Of the 33 subjects enrolled in the SAD part of the study, 24 received trilaciclib (6 mg/m², n = 3; 12 mg/m², n = 3; 24 mg/m², n = 3; 48 mg/m², n = 3; 96 mg/m², n = 6; 192 mg/m², n = 6), and nine received placebo. Twelve additional subjects, who were enrolled to confirm the BED, received trilaciclib at a dose of 192 mg/m². Demographic characteristics, published previously [11], were comparable among the seven trilaciclib groups and the placebo group. Geometric mean (gMean) maximum plasma concentration (C_max)and total systemic exposure [AUC from time 0 to infinity (AUC_0-inf)] increased dose dependently following a single 30-min IV infusion of trilaciclib over the dose ranges tested (0–192 mg/m²), with AUC_0-inf increasing slightly more than dose proportionally. Mean (range) AUC_0-inf for both 192 mg/m² groups combined (n = 18) was 3029 (2379–3869) ng h/mL (Online Resource 6).

At doses of 96 and 192 mg/m², trilaciclib demonstrated dose-dependent inhibition of the proliferation of CD45 + /CD3 + lymphocytes (a surrogate of CDK4/6 inhibition in hematopoietic cells) with a maximum mean inhibition of 37.2% and 60%, respectively, at 4 h after end of infusion (Fig. 2a). This indicated that the 96 mg/m² dose was on the ascending part of the dose–response curve and the percentage of inhibition was suboptimal. Peripheral lymphocyte proliferation started to recover from 8 h after end of infusion; however, the variability makes it hard to conclude if it has fully recovered at the last sampling time point 24 h after end of infusion. In bone marrow aspirates collected following infusion of 192 mg/m² trilaciclib in the BED cohort, a clear increase in the percentage of cells in G1 arrest was observed at 24 and 32 h post-dose for most bone marrow progenitor subsets (Fig. 2b). For HSPCs, the percentage of cells in the G1 cell cycle phase increased over time, with a maximum increase observed at 32 h post-dose. At 192 mg/m², trilaciclib achieved almost 100% G1 arrest for most of the cell types, with the exception of erythrocyte and megakaryocyte lineages (Fig. 2b). By comparison, analysis of total bone marrow demonstrated an approximate 40% arrest at 24 h, with partial recovery at 32 h (Fig. 2c), which was in line with the level of decrease in total cycling bone marrow cells and cycling HSPC observed in the preclinical studies. In addition, it was consistent with the PK/PD model predictions. Given that the arrest for most of the cell types reached almost 100%, it suggests that the 192 mg/m² dose reached the plateau part of the dose–response curve. Thus, it was assumed that a further increase in the trilaciclib dose would result in a diminishing return for efficacy and could result in deeper and more prolonged neutropenia.