Population Pharmacokinetic Models of Venetoclax in Hematologic Malignancies: A Systematic Review

Introduction

Direct stimulation of the mitochondrial apoptotic pathway is a new therapeutic strategy in the current targeted therapy of hematologic malignancies. B cell lymphoma-2 (BCL-2) family is a key apoptotic factor in the endogenous apoptotic signaling pathway, which inhibits apoptosis and plays an important role in tumorigenesis and development.¹ Many hematological tumors show the overexpression of the anti-apoptotic protein BCL-2,^2–4 such as acute myeloid leukemia (AML), acute lymphoblastic leukemia (ALL), chronic lymphocytic leukemia (CLL) and multiple myeloma (MM), so BCL-2 inhibitors have a broad application prospect in hematological malignancies. Venetoclax (VEN) is the world’s first approved oral BCL-2 inhibitor targeting the apoptotic pathway in tumors. By binding directly to BCL-2, VEN displaces and releases pro-apoptotic proteins, ultimately leading to apoptosis of malignant cells.^5–7 This mechanism of action targets BCL-2 protein and is innovative and unique. The Food and Drug Administration (FDA) granted an accelerated approval to it.⁸ Currently, FDA has approved the use of VEN in primary and relapsed or refractory (R/R) CLL/small lymphocytic leukemia (SLL), as well as newly diagnosed AML in patients who are age 75 years or older, or who have comorbidities that preclude the use of intensive induction chemotherapy. VEN has been approved for marketing in China by the end of 2020 and entered the national health insurance catalog in early 2023. However, due to the short period of time since its approval, there is little experience in clinical application. Despite its widespread use, consensus on its standardized clinical application and patient management has not been reached.

The pharmacokinetic (PK) characteristics of VEN are established.^9–11 A peak plasma concentration (Cmax) is achieved in 5 to 8 hours following oral administration.¹² However, the safe range of blood concentrations and monitoring protocols are currently inconclusive. Based on available studies, several factors affect the pharmacokinetics of VEN, including body weight, age, and ethnicity.¹³ After low- and high-fat meals, VEN exposure was higher than when fasting.¹⁴ In addition, VEN, which is highly bound to plasma proteins, is primarily metabolized by cytochrome P450 (CYP) 3A4.¹⁵ VEN is a substrate for CYP3A4 and P-gp, so it may be affected by P-gp inhibitors such as rifampicin. In clinical studies involving patients with AML, an observed relationship between VEN exposure and a higher probability of response, as well as certain safety events, was identified through an analysis of efficacy and safety using exposure-response methodology. Freise et al¹⁰ indicated that a daily dosage of 400 mg of VEN is associated with a high (>80%) probability of achieving an objective response in R/R CLL/ SLL patients and with a minimal chance of increasing neutropenia as well as infection.

A variety of VEN population pharmacokinetic (PPK) models^11,15–19 have been developed for patients with different types of hematological malignancies. However, there is no study that has summarized and evaluated these PPK models yet. Selecting an appropriate PK model and identifying important covariates which affect VEN has becoming very challenging. Consequently, existing models’ strategies should be better used, and new models’ development should take additional considerations. It is the first systematic investigation of the PPK models of VEN that has been conducted to our knowledge. The purpose of our study was to provide a comprehensive analysis of the published studies on PPK, investigate clinical determinants influencing the PK of VEN, and identify areas that require further investigation.

Method

Search Strategy

Using PubMed and Embase, an electronic literature search was conducted to identify PPK analyses of VEN for the entire period from the beginning to September 2023. Search terms used to identify the relevant PPK analyses on VEN are as follows: venetoclax, Venclyxto, BCL-2 inhibitor, ABT-199, population pharmacokinetic, PPK, nonlinear mixed effect, NONMEM, etc. Further, additional relevant materials were identified by conducting a thorough examination of all relevant lists of references.

Inclusion Criteria and Exclusion Criteria

Those studies included in this systematic review were (1) conducted on humans, (2) based on the use of VEN as the treatment, (3) providing PPK analyses of VEN, and (4) using at least one type of PPK analysis method. In contrast, the publications that (1) were reviews or methodology studies or (2) excluded the model development process were excluded.

Data Extraction

To interpret the results, a reviewer extracted information about the study design, baseline characteristics of the population, and PPK analyses. A reviewer-extracted information related to the study design, population baseline characteristics, and PPK analyses, which were critical to interpretation of the results. To minimize errors, the second reviewer independently verified the data extraction. Standard data collection forms were used to extract the following information from eligible studies: study design (eg, number of patients and samples, data source, dosage range, methods of concentrations determination), patient population characteristics (eg, sex, age, weight, race), and PPK analyses (eg, pharmacokinetic model, sample collection, tested covariates, methods of screening covariates, covariates of the final model and their relationship with pharmacokinetic parameters and model evaluation).

Quality Analysis

In each included study, two authors independently assessed its methodological quality using the National Institutes of Health (NIH) Study Quality Assessment Tool for Case Series Studies.²⁰ Methodologists and the NIH developed this tool based on quality assessment methods and concepts that can be applied to nonrandomized studies and case series, which are commonly used in systemic reviews including observational studies.²¹

Covariate Effects

Using R software (version 4.3.1), forest plots were used to summarize the effects of significant covariates on CL/F and V₂/F in each study. Due to the fact that most published PPK analyses used only the final model, results from this manuscript are based on the covariate effects estimated from the final PPK models.

Results

Study Characteristics

Among 380 unique titles screened, in total, six analyses were selected for further analysis (Figure 1). Six of the analyses had a good quality rating according to the NIH study quality assessment tool (Table S1). A summary of the included analyses is shown in Table 1. Among the six analyses included, a median (range) of 274 (73–3016) patients were used for the development of the PPK model, with five analyses^{11,16,18,19,22} involving more than 200 subjects. All analyses were based on clinical trials, and three analyses did not state NCT numbers.^11,16,22 Among the analyses that evaluated the model, three used three different evaluation methods.^15,18,19 Quantification of VEN concentrations in serum was performed using liquid chromatography methods with tandem mass spectrometric detection (LC-MS).^{11,15,17–19}

Table 1 Characteristics of the Included Analyses [n = 6]

Figure 1 Study selection process.

A summary of the gender, age, and body weight of the patients can be found in Table 2. In the included analyses, a wide range of races (White, Black, Asian, Hispanic, etc.) and multiple disease types (CLL, ALL, AML, NHL, SLE, DLBC, etc.) were represented. There is one analysis that female subjects with SLE were only analyzed in terms of VEN pharmacokinetics.¹⁷ The study populations received variable VEN dosing regimens (10–1200mg, po), and it is commonly necessary to ramp up the VEN dose to 400 mg per day over a five-week period.

Table 2 Studies Design and Characteristics of the Population

Population Pharmacokinetic Analyses

For each analysis, Table 3 summarizes selected structural models, pharmacokinetic parameters, IIV, residual error, and evaluation of PPK models. An industry-standard software program, NONMEM, was used to develop all PPK models.

Table 3 Modeling Strategies and PK Parameters of Published PPK Analyses of VEN

All the publications described VEN’s pharmacokinetics using a two-compartment model. The median (range) estimated apparent clearance (CL/F), inter-compartmental clearance (Q/F), apparent volume of distribution of the central compartment (V₂/F) and apparent volume of distribution of the peripheral compartment (V₃/F) were 446 L/Day (360–469 L/day), 98.8 L/Day (94.5–158 L/day), 114.5 L (37–118 L), and 121 L (116–217 L). In the final model, five analyses were used to describe the IIV of CL/F and V₂/F. The median (range) IIV of CL/F reported was 39.5% (37.3–45.7%) and V₂/F was 46.7% (42.3–507.2%). There were five analyses^{11,15–17,19} that used the proportional residual error model to explain random residual variability. The additive error was applied in three analyses,^11,15,19 which were 3.05*10,⁻⁷ 3.03*10⁻⁷ and 3.07*10⁻⁷. Two analyses^18,22 lacked information about residual error. An internal evaluation was almost conducted on the final models of the included analyses, and only one analysis was not known. A number of methods were commonly used, including visual predictive checks (VPC),^11,15–19 bootstrapping,^11,17 normalized prediction distribution errors (NPDE)^15,18,19 and diagnostic plots.^15,16,18,19

Information about the pharmacokinetic sampling is summarized in Table 4. Five analyses showed the schedule of pharmacokinetic sampling,^{11,15,17–19} which are subtly different according to single-dose, multiple-dose, and multiple-drug administration. Most samples of analyses were collected at 2, 4, 6, and 8h post-dose. Only one analysis¹⁷ mentioned non-blood samples, it described that urine samples were collected on the 7th post-dose study day of cycle 1 in the multiple-dose part of the study over intervals of 0–4, 4–8 and 8–24 h. Besides, none of these analyses has reported cerebrospinal fluid (CSF) samples.

Table 4 Pharmacokinetic Sampling of Published PPK Analyses of VEN

Covariate Modeling

The main purpose of most PPK analyses was to identify the potential covariates that affect VEN’s PK. Table 5 summarizes the covariate screening process and the covariates incorporated into the final models.

Table 5 List of Tested and Included Covariates in the PPK Models of VEN

In all the publications, among the covariates assessed for inclusion were demographic factors (body weight, sex, age, race and region), laboratory values [renal and hepatic function, creatinine clearance (CrCL), bilirubin (BIL), aspartate aminotransferase (AST), albumin (ALB) and alanine aminotransferase (ALT)], disease characteristics [population, Binet stage, Eastern Cooperative Oncology Group performance status (ECOG PS), Cumulative illness rating scale score, serum ß2-microglobulin levels, B symptoms, mutational status, cytogenetic factors and so on], co-administrations (with moderate and strong CYP3A inhibitors, OATP1B1 transporter inhibitors, bortezomib, navitoclax, daratumumab, dexamethasone, carfilzomib, bendamustine, azacitidine, decitabine, rituximab and obinutuzumab), dose and food.

In the above covariates, the stepwise covariate method including forward addition and backward elimination is usually used to identify the covariates. As significant covariates on CL/F, the effects of moderate and strong CYP3A inhibitors (5/5),^{11,15,18,19,22} rituximab co-administrations (2/4)^11,15 and OATP1B1 transporter inhibitors (4/5)^15,16,18,19 were frequently reported. In addition, covariates associated with CL/F, such as population,¹⁶ bortezomib co-administration,¹⁶ carfilzomib co-administration¹⁶ and bendamustine co-administration,¹⁶ have been reported in a limited amount of studies. The effect on V₂/F were commonly described by population (4/5)^11,16,18,19 and sex (5/5)^{11,15,16,18,19} as the covariates. Covariates affecting F₁ have rarely been investigated, only food (3/3),^15,18,19 dose (2/3)^18,19 and Asian race¹⁶ (1/4).

Covariate Effects

A quantitative analysis was performed on the effects of all covariates that were included in the final model (Figures 2 and 3). It is interesting that all covariates are categorical. The most studied covariates that significantly influenced CL/F were CYP3A inhibitors, OATP1B3 inhibitors and rituximab co-administration. Three analyses^11,15,16 showed that moderate and strong CY3A inhibitors were estimated to decrease the CL/F, with the latter to a greater extent. Two analyses^19,22 only provided estimated value, 95% Confidence Interval is missing. A minimal increase (7% and 21%) in the CL/F of VEN was observed after rituximab co-administration. An investigation of other covariates, such as OATP inhibitor co-administration, was conducted only in one study, and CL/F was also considered to be significantly affected by co-administration of bendamustine, carfilzomib, and bortezomib. The combined administration of OATP inhibitors would result in a 17% decrease in CL/F. It was estimated that co-administration of carfilzomib, bortezomib and bendamustine increased CL/F by 56%, 48%, and 33%, respectively.

Figure 2 The effects of included categorical covariates on CL/F of VEN. The horizontal lines indicate categorical covariate effects (with 95% confidence intervals [CI]). The typical value of clearance in each study was considered to be 1. The effect of each covariate for CL/F is displayed by the ratio of CL/F in the range of each covariate to the typical CL/Fvalue (Data from Jones201611, Deng201915, Gong202316, Samineni2022b19). Abbreviations: C3AHIB, CYP3A inhibitor indicator variable (2 = moderate, 3 = strong); RTX, rituximab; BTZ, bortezomib; CFZ, carfilzomib; BEN, bendamustine.

Figure 3 The effects of included categorical covariates on V2/F of VEN. Categorical covariate effects (95% confidence interval [CI]) are represented by open symbols (horizontal lines). The typical value of V2/F in each study was considered to be 1. The effect of each covariate for V2/F is displayed by the ratio of V2/F in the range of each covariate to the typical V2/F valuee (Data from Jones201611, Deng201915, Gong202316). Abbreviations: CLL, chronic lymphocytic leukemia; SLL, small lymphocytic lymphoma; NHL, non-Hodgkin’s lymphoma; Heme, hematological malignancies.

Three analyses^11,15,16 found that the V₂/F of VEN has a decrease in SEX, female subjects were estimated to have 28–32% lower V₂/F. Population was also an influential covariate with an increase in V₂/F. In Gong et al’s study, patients with non-hematological conditions or healthy subjects had 46% lower V₂/F than subjects with hematological malignancies.¹⁶ And compared to healthy subjects, CLL/SLL and NHL subjects have larger V₂/F.

Discussion

nOver the past few years, there has been growing interest in exploring the PK of VEN, and plenty of studies have sought to identify the sources of variability in VEN through PPK and apply the model’s predictive ability to clinical situations. This is the first review which summarizes on the PPK modeling of VEN, which is a potent and selective inhibitor of BCL-2. According to date, there have been six analyses of VEN’s PPK model published so far in patients with healthy or non-hematological conditions, and hematological conditions. All analyses described the PK of VEN using the two-compartment PK model with first-order absorption and elimination. The most commonly used covariates in the final PPK models were moderate and strong CYP3A inhibitors, OATP1B3 inhibitors, and rituximab co-administration for CL/F, and population and sex for V₂/F. Typically, models are evaluated using two or three evaluation methods, such as VPC, NPDE, bootstrap analysis, and diagnostic plots. This analysis may provide a uniform platform for selection of covariates for PPK analyses across studies.

In most of these publications, pharmacokinetic samplings were blood samples. Most samples of analyses were collected at 2, 4, 6, and 8h post-dose. For the real-world analyses, it is too difficult to obtain dense blood samples. Gradually increase the dose of VEN to 400 mg per day over 5 weeks according to a weekly ramp-up schedule.²³ Before administering VEN, Kobayashi et al collected blood samples and evaluated the minimum plasma concentration (Cmin).²⁴ The first-in-human study including 116 R/R CLL/SLL patients evaluated the pharmacokinetic profile of VEN and reported that the peak VEN level was reached six to eight hours after the first dose of VEN monotherapy.²⁵ In accordance with its pharmacokinetics, peak plasma concentrations of VEN are reached 5–8 hours after oral administration.²⁶ In previous study, it took the median time of 6 hours for VEN to reach plasma concentration, while the approximate t_1/2 was 18 hours.²⁷ From the existing research, we proposed to collect steady state trough concentration (C0, after 5 days of regular dosing at a constant dose, 0–0.5 hours before the next dose, as close as possible to 0 hour before dosing.) and steady state peak concentration (C6, 6 hours after the next dose after 5 days of regular dosing at a constant dose.) for therapeutic drug monitoring (TDM) in clinical work. However, there is no clinically referable range for TDM, which needs to be further explored in follow-up study.

Among these analyses, only Minocha et al mentioned urine samples that would be collected over the 0–4, 4–8 and 8–24 h intervals on study day 7 post-dose in cycle 1 in the multiple-dose part of the study.¹⁷ None of analyses mentioned CSF samples. There is little evidence that VEN is effective for patients with hematological malignancies and central nervous system (CNS) involvement, although some observations have been made in cases of a CLL patient,²⁸ an AML patient²⁹ and an acute promyelocytic leukemia (APL) patient³⁰ suggest that blood-brain barrier is crossed by VEN, which reaches potential therapeutic concentrations. In addition, Badawi et al reported the pharmacokinetics of VEN in plasma and CSF samples from pediatric patients with R/R malignancies from a Phase 1 study, which demonstrates VEN’s ability to cross into the central nervous system.³¹ In light of these findings, VEN may have a role to play for patients with CNS complications. However, there is a need for further research in order to determine its utility in improving clinical outcomes.

In all PPK studies of VEN, CYP3A inhibitors were tested as covariates in five analyses, and all included them in the final model. Inhibitors of CYP3A4 may affect the levels of VEN, as it is a substrate for to CYP3A4. In terms of clinical consideration, PPK studies could be conducted on concomitant administration of CYP3A inhibitors of moderate and strong potency. In their study, Gong et al¹⁶ found that multiple moderate CYP3A inhibitors used in combination cause a greater decrease in clearance than a single moderate inhibitor. In a study using the PBPK approach,³² the results were as follows: C_max or AUC_∞ was not affected by weak CYP3A inhibitors or inducers, but VEN exposure was decreased by moderate and strong CYP3A inducers. Moderate and strong CYP3A inhibitors were estimated to increase VEN AUC_∞. If potential VEN dose adjustments are required, this should be considered. Brüggemann et al suggested that the use of strong CYP3A4 inhibitors on the ramp-up phase of VEN is contraindicated, thereafter, reduce the dose by at least 75%.³³ When used in the ramp-up phase, moderate CYP3A4 inhibitors should be reduced by 50%, after that, VEN dose should be reduced by at least 50%.³³

The concomitant use of rituximab has a negligible effect on VEN PK and is not anticipated to have any clinical significance. Moreover, the co-administration of carfilzomib, bendamustine, and bortezomib was estimated to increase CL/F in one model,¹⁶ which is new findings. This interaction may help patients in multiple myeloma who use VEN plus carfilzomib and bortezomib.³⁴ Since venetoclax disposition pathways are not expected to be impacted by bortezomib and carfilzomib, the mechanism for the effect of these drugs is unclear.

Among the five publications in the VEN PPK, the factor population was also frequently tested. However, only one analysis reported that population had a significant influence on CL/F, and three analyses reported it had an influence on V₂/F. In addition to AML and CLL/SLL, clinical studies of VEN are currently underway in a number of areas, such as myelodysplastic syndrome,³⁵ multiple myeloma,^36,37 and large B-cell lymphoma.³⁸ This is a problem that needs to be paid attention in the follow-up research.

Three analyses^15,18,19 tested the effect of food on F₁, and as a categorical covariate, it was included in all three final models. Among adult females in good health, VEN achieved C_max 4h after administration and had a half-life of 16.4–17.2h.¹⁴ Food administration resulted in a significantly higher Cmax than fasting administration, approximately 2h later.¹³ Patients should be recommended to take VEN with food based on the results of these studies. When dissolution limited absorption occurs, food consumption can increase bioavailability because fat content of food and bile secretion may increase solubility. When administered with low-, moderate-, and high-fat meals, VEN was 2.99-, 3.77-, and 4.25-fold more bioavailable than when administered fasting state,¹¹ respectively. Therefore, it is advised to administer VEN with food, regardless of fat content, in order to ensure adequate and consistent bioavailability. The instructions recommend taking it within 30min after a meal.²³ Four analyses investigated the effect of race on the PK of VEN, only one¹⁶ of which included it in the final model. It showed that Asian was estimated to have 61% higher F₁. Future research is needed to determine whether VEN requires dose adjustment for Asian population, especially Chinese.

There have been developed and validated sensitive LC-MS/MS methods for the determination of VEN in human plasma and CSF.³⁹ We could determine its effective therapeutic concentration range, study the blood–brain barrier permeability of VEN and develop PPK models to determine optimal VEN dosing regimens required to achieve targeted CSF exposures. Although VEN is approved for adult patients, it is also used for pediatric patients in practical clinical applications. Currently, it is being investigated to expand its use to pediatric patients, but more robust data are required. According to recent researches, it appears that VEN is well tolerated by pediatric patients and is safe.^40,41 There is still a need for further studies to determine the optimal dose, duration of treatment, and proper combination as well as to assess its long-term safety. We could develope PPK models to estimate the use of VEN in Chinese children.

Our analyses were limited. Firstly, it is important to note that this review was not able to include all model information due to a lack of details in some publications. Then, clinical trial data were used for PPK analyses of VEN, and the study population was mostly representative of the target population. There is a need for some real-world data-based studies. Thirdly, because of the publications included were all in English, relevant studies published in other language might have been missed out.

Conclusion

VEN is becoming more and more widely used in the clinical practice. We proposed to explore reference ranges for TDM in clinical work and highlighted the potential role of VEN for patients with CNS complications, which need further investigate. One hand this review provides a detailed description of the parameters for the construction of VEN models, their key features, and the established covariate relationships. The covariates CYP3A inhibitors, OATP1B1 transporter inhibitors, sex, population, and food might always be included in the development of a PPK model. This can also allow for a deeper understanding of the pharmacokinetics of VEN, inform the construction of other PPK models for VEN, and identify areas where additional research is needed to facilitate the application of PPK models. On the other hand, doctors and pharmacists could use the summarized information, such as inhibitors of CYP3A4 may affect the levels of VEN, the co-administration of carfilzomib, bendamustine, and bortezomib was estimated to increase CL/F and food administration resulted in a higher Cmax than fasting administration, to make clinical recommendations for the patients.