Pharmacokinetic study of capivasertib and the CYP3A4 substrate midazolam in patients with advanced solid tumors

This phase 1, multicenter (5 sites in the USA), open-label, fixed-sequence study (NCT04958226) involved patients with advanced solid tumors. The study comprised two parts (parts A and B). Part A was conducted to evaluate the effect of capivasertib monotherapy on the PK of midazolam, and consisted of a screening period of up to 28 days and three treatment periods, all occurring in the first half of cycle 1 (29 days). Treatment period 1 was midazolam alone (1 mg) administered on day 1 of cycle 1, treatment period 2 was days 2–7 of cycle 1 with capivasertib 400 mg BID dosing on an intermittent schedule (4 days on/3 days off), and treatment period 3 was days 8–15 of cycle 1 with midazolam (1 mg) administered on day 8 and day 12 (corresponding to the 3rd day off and 4th day on capivasertib, respectively; Fig. 1). In part A, at the investigator's discretion, patients with estrogen receptor-positive breast cancer could also receive intramuscular fulvestrant 500 mg on day 1 and day 15 of cycle 1, and day 1 of each 28-day cycle thereafter; other anticancer agents (except luteinizing hormone-releasing hormone [LHRH] analogs for patients with breast or prostate cancer) and radiotherapy were avoided.

Patients who completed part A without disease progression or unacceptable toxicity, and who were considered by the investigator as likely to benefit from further capivasertib treatment, entered part B (treatment period 4) of the study, in which the efficacy of capivasertib (with or without standard-of-care anticancer treatment, fulvestrant, paclitaxel, docetaxel, abiraterone, or olaparib), was evaluated as an exploratory endpoint. Patients continued to receive capivasertib 400 mg BID dosing on an intermittent schedule of 4 days on/3 days off (per 28-day cycle) until disease progression, unacceptable toxicity, withdrawal of consent, or any other reason for discontinuation of treatment (Fig. 1).

The study was conducted in accordance with the Declaration of Helsinki, the International Council for Harmonisation Good Clinical Practice guideline, and all applicable local regulations. The institutional review boards or independent ethics committees of all investigational sites approved the protocol, and all patients provided written consent.

Adult patients (≥ 18 years of age) with body mass index (BMI) 18–32 kg/m², and with documented evidence of locally advanced inoperable or metastatic solid tumors suitable for capivasertib treatment, were included in the study. Documented tumor alterations in PI3K/AKT pathway genes (by local testing) were recorded where available, but were not required for study inclusion after a protocol amendment (31 March 2022). Patients were required to have an Eastern Cooperative Oncology Group (ECOG)/World Health Organization (WHO) performance status (PS) of 0 or 1, with no deterioration over the previous 2 weeks, a minimum life expectancy of 12 weeks, and at least one lesion (measurable and/or non-measurable) that was accurately assessed at baseline by computed tomography (CT)/magnetic resonance imaging (MRI), or X-ray and was suitable for repeated assessment. Female patients were not lactating, not pregnant, and were not of childbearing potential (or, if they were of childbearing potential, had to abstain from sexual activity or use two forms highly effective contraceptive methods from screening until 4 weeks after discontinuation of study drug). Male patients had to abstain from sexual activity or use barrier contraceptive methods from screening until 16 weeks after discontinuation of study drug.

Exclusion criteria included: previous allogeneic bone marrow or solid organ transplant, major surgery, or radiotherapy with a wide field of radiation within 4 weeks prior to first dose of capivasertib; radiotherapy with a limited field of radiation for palliation within 2 weeks prior to study intervention initiation; unresolved toxicity from prior therapies of grade > 2; and any unresolved toxicity that may interfere with PK assessment. Patients with inadequate bone marrow reserve or organ function, including renal impairment (creatinine > 1.5 × upper limit of normal, concurrent with creatinine clearance < 50 mL/min), patients with cardiac ejection fraction < 50% or outside the institutional range of normal, and patients with clinically significant abnormalities in glucose metabolism (insulin treatment needed or glycated hemoglobin (HbA1c)  > 7.5%) or blood lipid profiles, were excluded; patients with alanine aminotransferase and aspartate transaminase > 2.5 × upper limit of normal, if no demonstrable liver metastases, or > 5 × upper limit of normal in the presence of liver metastases, as well as patients with total bilirubin > 1.5 × upper limit of normal were also excluded. Elevated alkaline phosphatase was not an exclusion criterion if, due to the presence of bone metastasis and liver function, it was otherwise considered adequate. Those with spinal cord compression or symptomatic brain metastases, or severe or uncontrolled systemic diseases (including uncontrolled hypertension, active bleeding diatheses, or clinically significant active infections) were also excluded.

Patients could not have received potent or moderate inhibitors or inducers of CYP3A4 within the 2 weeks (or 5 half-lives, whichever was longer) prior to the first dose of study intervention (3 weeks for St. John’s wort, and 4 weeks for enzalutamide), and could not have received sensitive substrates of CYP3A4 with a narrow therapeutic window within 1 week prior to the first dose of study intervention. Patients also could not have received any investigational agent from a previous study within 30 days of the first dose of study drug, or any chemotherapy, immunotherapy, immunosuppressant medication (other than corticosteroids), or anticancer agents (except LHRH analogs in patients with breast or prostate cancer) within 2 weeks (or 5 half-lives, whichever was longer) of the first dose of study intervention. Nitrosourea or mitomycin C within 6 weeks of the first dose of study intervention was not permitted.

In part A, midazolam (and thus capivasertib, when given concomitantly) was given in the morning after an overnight fast to minimize PK variability; on the other days, capivasertib was dosed in a fasted state from at least 2 h prior to the dose to at least 1 h post-dose, where possible. In part B, capivasertib was given with or without food at the investigator’s discretion.

Dose modifications for midazolam were not permitted. A maximum of two dose reductions for capivasertib were permitted on the intermittent schedule (4 days on/3 days off), the first dose reduction from 400 to 320 mg BID and the second from 320 to 200 mg BID; dose re-escalations were not allowed. Substantial acute toxicities (any-grade intolerable AEs, grade 1 or 2 clinically significant, intolerable AEs, or grade ≥ 3 AEs related to capivasertib) were managed as medically indicated, including with temporary suspension of capivasertib.

Concomitant administration of drugs known to prolong the QT interval was restricted, or patients were closely monitored if such drugs were considered essential to patient management. Substrates for CYP3A4, multidrug and toxin extrusion 1 (MATE1), or organic cation transporter 2 (OCT2) transport were either avoided or used cautiously, and statins were avoided or used at capped doses. In addition to the prior/concomitant therapies documented in the exclusion criteria, patients avoided herbal supplements (e.g., St. John’s wort) and the ingestion of foods and beverages known to modulate CYP3A4 during the study; grapefruit juice and Seville oranges (known CYP3A4 inhibitors) were prohibited from 7 days prior to initiating treatment period 1 until after the completion of PK sample collection in part A.

Additional restrictions were implemented in part A of the study: other anticancer agents (except LHRH analogs for patients with breast or prostate cancer, and fulvestrant) and radiotherapy were avoided, although radiation for palliation at focal sites was permitted. The use of midazolam was prohibited from at least 14 days before the first on-study administration of midazolam until the collection of the last PK sample for either treatment.

To assess the full PK profile of midazolam, 2 mL blood samples were collected at 0 (pre-dose), 0.25, 0.5, 0.75, 1, 1.5, 2, 3, 4, 6, 8, and 12 h post-midazolam dose on days 1, 8, and 12 of cycle 1, corresponding to midazolam alone on day 1 (treatment period 1), midazolam without capivasertib on day 8 (3rd day off capivasertib; treatment period 3), and midazolam with capivasertib on day 12 (4th day on capivasertib; treatment period 3). Single 2-mL 24-h post-dose samples were also collected on days 2, 9, and 13 of the same cycle.

Blood was sampled for capivasertib PK analysis on day 9 (one pre-capivasertib morning dose sample), day 12 (full PK profile: 0 [pre-capivasertib morning dose], 0.5, 1, 2, 4, 6, 8, and 12 h post-capivasertib morning dose), and day 13 (one 12-h post-day 12 capivasertib evening sample) of cycle 1, again by obtaining 2-mL blood samples.

Study samples were received frozen with dry ice, and stored at –60 to –80 °C. Analysis of plasma concentrations of midazolam was performed by Labcorp Early Development Laboratories Inc. using a liquid chromatography–mass spectrometry method that was fully validated according to US Food and Drug Administration bioanalytical guidelines. The samples were extracted using supported liquid extraction (SLE). Specifically, 50 µL of plasma was diluted with 25 µL of the internal standard (midazolam-d4) and 200 µL water: ammonium hydroxide (95:5). The diluted sample (200 µL) was loaded onto a Biotage Isolute^® SLE + cartridge 200 µL and subsequently eluted with 1000 µL of methyl tert-butyl ether. The resulting extract was dried under nitrogen at 40 °C and then reconstituted in 150 µL of methanol: water: formic acid (50:50:0.1). Chromatographic separation was performed using a Shimadzu Prominence 20-series high-performance liquid chromatography system equipped with a Mac-Mod ACE C18 column (50 × 2.1 mm dimensions, 5 µm). The liquid chromatography separations were carried out at 30°C with a flow rate of 0.600 mL/min. Initially, a 0.25-min hold at 50% Mobile Phase B was maintained, followed by a linear gradient from 50 to 95% B over 1.25 min. Mass spectrometry detection was performed using a Sciex API 5500™ triple quadrupole tandem mass spectrometry system operating in positive mode. The following multiple reaction monitoring transitions were utilized: midazolam (326.3 → 291.3) and internal standard (330.3 → 295.1). Quantitation was based on linear regression with a 1/x^2 weighting factor. The method’s validated concentration range was 0.0500–50.0 ng/mL midazolam with intra-batch variation of ±4.5% bias and ±6.3% relative standard deviation (RSD) and inter-batch variation of ±3.0% bias and ±5.2% RSD observed during validation. The results of an incurred sample reproducibility assessment performed for midazolam samples during the sample analysis phase of the study, showed a relative percentage difference within ±20% for 98.8% of the samples assessed. Plasma concentrations of capivasertib were determined using high-performance liquid chromatography with tandem mass spectrometric detection, using validated analytical methods as described previously [12].

All PK parameters were determined from plasma concentrations using non-compartmental methods with Phoenix^® WinNonlin^® Version 8.3.5 (Certara, Princeton, NJ). PK parameters assessed for midazolam in the absence of capivasertib (day 1 of cycle 1; treatment period 1) and during capivasertib treatment on an intermittent 4 days on, 3 days off schedule (day 8 and day 12 of cycle 1; treatment period 3) included: maximum observed plasma (peak) drug concentration (C_max); area under the concentration–time curve from time zero to infinity (AUC_inf); area under the plasma concentration–time curve from zero to the last quantifiable concentration (AUC_last); time to reach maximum observed plasma drug concentration (t_max); and half-life associated with terminal slope of a semi-logarithmic concentration–time curve (t½_λz).

PK parameters assessed for capivasertib included: C_max, area under the plasma concentration–time curve in the dose interval (AUC_τ), t½_λz, and t_max of capivasertib and its glucuronide metabolite, apparent total body clearance of capivasertib from plasma after oral administration (CL/F), and trough plasma concentration (C_trough) of capivasertib and its glucuronide metabolite on days 9 and 13 of cycle 1.

Safety and tolerability were evaluated in terms of AEs, which were classified and graded according to the National Cancer Institute (NCI) Common Terminology Criteria for Adverse Events (CTCAE) v5.0; vital signs; 12-lead electrocardiogram (ECG); and clinical laboratory assessments, including clinical chemistry and hematology parameters, urinalysis, HbA1c levels, and lipid profiles. Patients were followed for safety for at least 30 days after the last dose of capivasertib.

Efficacy was evaluated as an exploratory endpoint by investigator assessment, and per Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1. Tumor assessments were performed by CT and/or MRI scans at screening and then every 8 weeks (± 7 days; earlier if clinically indicated) for 2 years, and then every 12 weeks until disease progression. Radiographic bone scans were performed at screening and repeated as clinically indicated, depending on tumor type. Efficacy endpoints were: PFS, defined as time from start of capivasertib dosing (day 2 of cycle 1) until disease progression; objective response rate (ORR), defined as the proportion of patients with measurable disease at baseline who have a confirmed complete or partial response; duration of response (DoR), defined as time from the date of the first documented response until the date of documented progression or death in absence of progression; and clinical benefit rate (CBR) at 24 weeks, defined as the percentage of patients who have a confirmed complete or partial response, or who have stable disease for at least 23 weeks from the start of capivasertib dosing (day 2 of cycle 1).

The number of patients included in the study was selected to gain sufficient precision for the 90% CI of the least squares (LS) geometric means ratio (GMR) of midazolam AUC_inf and C_max, while exposing as few patients as possible to study drugs and procedures. Based on the estimates of the within-patient coefficient of variation (CV) of 0.3 for log AUC_inf and 0.35 for C_max, a sample size of 14 evaluable patients was projected to provide relative precision (ratio between the upper and lower limits of the 90% CI) of 1.6 and 1.7 for AUC_inf and C_max, respectively, with a probability of 80%.

Plasma midazolam or capivasertib PK parameters were analyzed for the PK analysis set, consisting of all patients who received at least one dose of midazolam and capivasertib and had at least one reportable post-dose plasma concentration. Patients were also required to have day 1 and day 8 or Day 12 PK data to be included in statistical analyses. Patients with protocol deviations, dose interruptions or reductions, or AEs that could have significantly impacted the PK of capivasertib and/or midazolam were considered non-evaluable.

Statistical analyses of C_max and AUC_inf were performed using a mixed-effects model following a natural logarithmic transformation of the PK parameters, with fixed effect for day and random effect for patient. LS geometric means of treatment effect and the corresponding two-sided 95% CIs, as well as ratios of LS geometric means together with two-sided 90% CIs of test treatment (corresponding to the 3rd day off or 4th day on capivasertib; day 8 or day 12 of cycle 1) and reference treatment (midazolam alone; day 1 of cycle 1), were estimated. AUC_inf or C_max ratios with 90% CIs between 0.80 and 1.25 indicated no interaction between midazolam and capivasertib. A ratio of LS geometric means of AUC_inf in the ranges of ≥ 1.25 to < 2, ≥ 2 to < 5, or ≥ 5 indicated a weak, moderate, or strong CYP3A4 inhibitor classification, respectively, and a ratio of LS geometric means of AUC_inf in the ranges of ≥ 0.5 to < 0.8, ≥ 0.2 to < 0.5, or < 0.2 indicated a weak, moderate, or strong CYP3A4 inducer classification, respectively. The remaining PK parameters were summarized using descriptive statistics.

Safety analyses were performed on the full analysis set, which consisted of all patients who received at least one dose of midazolam and/or capivasertib, irrespective of protocol adherence or continued participation in the study. Efficacy analyses were also performed on the full analysis set. Summary and/or descriptive statistics were used to analyze baseline demographic, safety, and efficacy data; PFS and DoR were estimated using Kaplan–Meier methodology.

All statistical computations were performed using SAS^® 9.4 (SAS Institute, Cary, NC). There was no imputation of missing data.

A total of 28 patients were enrolled in the study, of whom 21 received study treatment in part A and were included in the full analysis set and the PK analysis set (7 screening failures did not receive study treatment; see Online Resource 1: Supplementary Fig. 1). Of the 21 patients who received study treatment in part A, 20 completed part A and 19 went on to part B and received capivasertib monotherapy with or without standard-of-care anticancer treatment. There were four deaths during the study: one in part A (treatment period 2), and three in part B. All patients who died had discontinued capivasertib prior to death, because of clinical disease progression; none of the deaths were deemed by the investigator to be possibly related to study treatment.

Although permitted in part A, no patients with estrogen receptor-positive breast cancer received concomitant fulvestrant in the study. Fifteen patients were deemed to have had at least one important protocol deviation (Online Resource 1: Supplementary Table 1). Following exclusions due to important protocol deviations and the presence of quantifiable (> 0.05 ng/mL) pre-dose concentrations for midazolam in some PK profiles (prior to cycle 1, day 1 and at cycle 1, day 8 and day 12), a total of 18 patients were included in the PK statistical analysis.

The demographic and baseline characteristics of the 21 patients included in the full analysis set are summarized in Table 1. Most patients were female (66.7%), and the median (range) age was 65 years (38–77). Most patients were of White race (90.5% [not Hispanic/Latino 81%]) and had an ECOG PS of 0 (61.9%). Most patients entered the study with metastatic disease (85.7%), and the most common primary tumor sites were the colon (23.8%), uterus (19%), and lung (14.3%). Most patients (95.2%) had tumors harboring PI3K/AKT pathway alterations. Patients had received a median of 3 (range 1– 9) prior anticancer therapies.

Analysis was conducted with a data cut-off of 11 May 2023. In part A, the median total treatment duration was 11 (range 2–21) days, and the median actual treatment duration was 8 (range 2–9) days. The total number of midazolam doses administered over the course of the study was 58: 21 in treatment period 1 (midazolam only on day 1 of cycle 1) and 37 in treatment period 3 (midazolam administered on day 8 and day 12 of cycle 1). Two patients missed one dose of midazolam during treatment period 3, and one patient withdrew from the study prior to treatment period 3 (withdrawal due to death).

The median total capivasertib treatment duration across part A and part B was 60.0 (range 3–228) days, and the median actual capivasertib treatment duration was 31.0 (range 3–127) days. All patients experienced at least one AE during the study. The most frequently reported AEs by treatment period are shown in Table 3. A lower incidence of AEs occurred during treatment period 1 (single midazolam dose) compared with treatment periods 2 (capivasertib alone) and 3 (capivasertib + midazolam). The most frequently reported AEs (frequency > 25% of patients overall) included diarrhea (76.2%), nausea and fatigue (both 33.3%), and hyperglycemia and anemia (both 28.6%).

In total, 14 (66.7%) patients experienced grade ≥ 3 AEs. During part A, the rate of grade ≥ 3 AEs was higher in treatment period 3 (capivasertib) than in treatment periods 1 or 2 (Table 3). Twelve patients (57.1%) experienced grade 3 AEs; only rash (group term n = 3; 14.3%) and fatigue and decreased appetite (both n = 2; 9.5%) were reported in more than one patient (in part B). One patient (4.8%) experienced a grade 4 AE (thrombocytopenia) during treatment period 2; this was not considered related to either capivasertib or midazolam. One patient (4.8%) experienced a grade 5 AE (sepsis) in part B; the death was not considered related to either capivasertib or midazolam. A total of five serious AEs (SAEs) were reported in the study (n = 1 in treatment period 2, n = 1 in treatment period 3, and n = 3 in part B). Only one SAE (treatment period 3) was considered related to study treatment (rash considered related to both capivasertib and midazolam).

Five patients (23.8%) required a capivasertib dose reduction due to AEs: n = 1 in treatment period 3 and n = 4 in part B. Fourteen patients (66.7%) required a capivasertib dose interruption due to AEs (n = 6 patients in treatment period 3, n = 7 patients in part B, and n = 1 patient with an event in treatment period 3 and part B). Three patients experienced AEs leading to permanent discontinuation of capivasertib. Only one patient discontinued capivasertib because of an AE considered related to capivasertib and midazolam (fatigue during part B).

Although some fluctuations or other abnormalities were observed in laboratory results, vital signs, physical examinations, or ECG during part A of the study, these were expected, given the advanced nature of the patients’ disease and previous clinical experience with capivasertib and midazolam. There were no reports of QT prolongation in the study.

There was a total of 15 PFS events at the time of analysis, and median PFS (95% CI) was 3.4 months (1.8–5.4). Fifteen patients had measurable disease at baseline and at least one post-baseline scan; response to treatment over time for these patients is shown in Fig. 4. The ORR (95% CI) was 9.5% (1.2–30.4). Two patients had a confirmed partial response to treatment: one with endometrial adenocarcinoma and phosphoinositide-3-kinase regulatory subunit 2 (PIK3R2)  alteration, and one with ovarian carcinoma and PIK3CA and PTEN alterations. The CBR (95% CI) at 24 weeks was 14.3% (3.1–36.3); three patients met the criteria for clinical benefit. Median DoR (95% CI) was 3.8 months (3.6–not calculable).