Talazoparib Formulation Bridging in Cancer Patients—Challenges and the Critical Role of Model‐Informed Drug Development in Approval Despite Failed Bioequivalence

ABSTRACT

Talazoparib is a poly(ADP‐ribose) polymerase inhibitor approved for the treatment of breast and prostate cancer. Commercialization of a soft gelatin capsule (SGC) formulation developed post‐approval required a bioequivalence (BE) and food effect (FE) study to bridge SGC with the initial commercial hard capsule (HC) formulation. Study execution and meeting BE criteria are challenging due to high variability in C_max, potentially higher C_max for SGC based on dissolution data, and the need to perform BE/FE assessment at steady state in cancer patients. Model‐informed drug development (MIDD) was used to facilitate an efficient/feasible study design. Semi‐mechanistic pharmacokinetic (PK)/pharmacodynamic (PD) modeling and simulations showed that AUC, instead of C_max, drove hematologic events, the main side effects of talazoparib. This supported a BE study powered for AUC equivalence only. Population PK simulation showed that following a 28‐day treatment in the first period, 14 days in subsequent periods is sufficient for steady‐state BE/FE assessments. Study results showed AUC met BE criteria while C_max was 37% higher for SGC relative to HC, which was deemed not clinically significant based on the PK/PD model. FE on SGC formulation was consistent with FE on HC formulation reported previously. The safety profile of the two formulations was generally consistent with the known safety profile. The totality of data (AUC equivalence, lack of impact of C_max on safety, observed safety data) supported bridging of the two formulations although C_max failed to meet BE criteria. MIDD was critical in study design optimization and supported approval of the SGC formulation.

Trial Registration: ClinicalTrials.gov Identifier: NCT04672460

Study Highlights.

• What is the current knowledge on the topic?

  • ○
    After the initial approval of talazoparib, a new soft gelatin capsule (SGC) formulation was developed and bridging with the initial commercial hard capsule (HC) formulation was needed. For post‐approval formulation bridging, the demonstration of both AUC and C_max equivalence in a bioequivalence (BE) study and evaluation of food effect (FE) are generally required. However, there are multiple challenges in executing the BE and FE study and meeting the BE criteria for both AUC and C_max.
• What question did this study address?

  • ○
    This manuscript summarized the strategies used to bridge between the talazoparib SGC formulation and the initial commercial HC formulation, including the study design optimization for the BE and FE study through model‐informed drug development (MIDD), and interpretation of the study results based on PK/PD modeling and simulation.
• What does this study add to our knowledge?

  • ○
    MIDD supported a more logistically feasible and smaller study design. The totality of data (AUC equivalence, lack of impact of C_max on safety endpoints, and observed safety data) supported the bridging of talazoparib SGC formulation with initial commercial HC formulation although C_max failed BE criteria. In this New Drug Application approval, MIDD played a critical role in enabling a realistic study design in patients with cancer and mitigation of the failed C_max equivalence.
• How might this change drug discovery, development, and/or therapeutics?

  • ○
    This work showcases how MIDD played a critical role in study design optimization, mitigation of regulatory risk and data interpretation to drive clinical decision‐making. The principle and framework shown in this work are applicable to other drugs or therapeutic areas as long as the exposure metric driving efficacy and safety is well characterized with existing data.

1. Introduction

Talazoparib is a potent poly(ADP‐ribose) polymerase inhibitor that has been approved in the US, Europe, and other countries as monotherapy at 1 mg QD for the treatment of patients with germline BRCA‐mutated HER2‐negative locally advanced or metastatic breast cancer. Subsequently, talazoparib 0.5 mg QD in combination with enzalutamide was approved for the treatment of homologous recombination repair (HRR) gene‐mutated metastatic castration‐resistant prostate cancer.

The initial commercial formulation of talazoparib was an immediate‐release hard capsule (HC) administered orally with or without food [1]. To facilitate greater flexibility in batch size production, a talazoparib liquid‐filled soft gelatin capsule (SGC) formulation has been developed. Talazoparib has low/moderate permeability and high solubility at 1 mg QD. As shown in Figure 1, the reference formulation is not considered rapid dissolving (defined as ≥ 85% dissolved in ≤ 15 min) [2]. In addition, the formulation change is considered level III per guidance from the FDA Immediate Release Scale‐up and Post Approval Change Expert Working Group [3]. Therefore, the biowaiver requirements were not met. A bioequivalence (BE) and food effect (FE) study was warranted to bridge this new SGC formulation with the initial HC commercial formulation.

FIGURE 1.

Comparison of the dissolution profile of the talazoparib soft gel capsule with hard shell capsule batches in 0.01M HCl with paddles at 75 rpm.^a HCl, hydrochloric acid; rpm, revolutions per minute. ^aThe dissolution media for the hard shell capsules include 0.2% sodium dodecyl sulfate to aid the dispersion of the drug from the excipients.

There are multiple challenges to executing the BE and FE study and meeting the BE criteria. First, due to the pharmacokinetic (PK) and safety properties of talazoparib, there are operational challenges of the BE study (Table 1). These factors have a significant impact on the feasibility of study conduct and meeting the development timeline.

TABLE 1.

Safety and PK properties of talazoparib and impact on study design.

Drug properties	Challenges	Impact on study design
Clastogenic [1]	BE study cannot be conducted in healthy volunteers	BE/FE study population is limited to patients with cancer who may benefit from talazoparib treatment
Long elimination half‐life (~90 h) [1]	Unethical to conduct a single‐dose crossover study in patients with cancer as it would require too long of a treatment washout period, which could negatively impact disease control for these patients	BE/FE need to be assessed after multiple doses at steady state, and BE/FE assessment period is lengthy Lengthy BE/FE assessment duration means higher rate for dropout and dose modification, which increased total enrollment number to have enough PK‐evaluable patients
Large total variability for AUC (> 30%) and Cmax (> 40%) [1]	Precludes the option of a parallel study as it would have required too large a sample size (more than 100 PK‐evaluable patients per arm)	Crossover design is needed, which means long treatment duration as each patient needs > 1 treatment period
Large intra‐patient variability for Cmax: 40%	Large sample size if the BE study is powered for Cmax equivalence	Replicate crossover design is needed to reduce sample size, which means more periods and thus longer BE assessment duration
Low intra‐patient variability for AUC: 17%		Reasonable sample size for crossover design to show AUC equivalence

Abbreviations: AUC, area under the curve; BE, bioequivalence; C_max, maximum observed plasma concentration; FE, food effects; PK, pharmacokinetic.

In addition, the in vitro dissolution data suggested a potentially higher maximum concentration (C_max) with the SGC formulation. As shown in Figure 1, the in vitro dissolution for the SGC formulation is faster than various batches of the HC formulation, which could potentially translate to a higher absorption rate and C_max. Therefore, there is a risk for failing BE criteria for C_max. Based on FDA guidance [4], when BE is not demonstrated the sponsor should demonstrate that the differences in rate and extent of absorption do not significantly affect the safety and efficacy based on available dose–response or concentration‐response data. Therefore, the clinical relevance of C_max vs. the area under the curve (AUC) on efficacy and safety needs to be evaluated based on existing clinical experience.

The New Drug Application (NDA) submission for the talazoparib SGC formulation was approved by the US FDA in March 2024 [5]. This manuscript summarizes the strategies used to bridge between the talazoparib SGC formulation and the initial HC formulation. This example showcases how model‐informed drug development (MIDD) played a critical role in study design optimization (endpoint selection, reducing sample size, shortening the treatment duration) and, most importantly, supported commercialization of the SGC formulation despite failed C_max equivalence by demonstrating lack of impact of increased C_max on safety endpoints using a semi‐mechanistic PK/PD model for each hematologic safety endpoint.

2. Materials and Methods

2.1. BE Study Design Optimization

Different study design options, including 2‐period crossover and full‐replicated and partial‐replicated crossover, were considered (Figure S1). The sample size (number of PK‐evaluable patients and total enrollment) and treatment duration were calculated for different designs suited to meet BE criteria for steady‐state C_max and AUC based on an assumed true ratio of 1.1 for both AUC and C_max, and intra‐patient variability of 40% for C_max and 17% for AUC, which were obtained from a talazoparib single‐dose drug–drug interaction study using the HC formulation [6]. The sample size provided 90% power that the 90% confidence intervals (CIs) for the test/reference ratio of the primary endpoint(s) fell within the 80%–125% acceptance interval for BE. Sample size for replicated design was based on European Medicines Agency requirements, which are more conservative than the requirements of the US FDA [7]. The percentage of PK‐evaluable patients after a specific treatment period was based on the estimated dropout rate due to disease progression and dose modification/discontinuation due to AEs from clinical experience in patients with breast cancer.

The impact of varying duration of each treatment period on the accuracy of the estimated difference in AUC between test and reference formulation was evaluated by simulation using a population PK model [8]. The goal was to have a long enough treatment period to allow adequate assessment of changes in bioavailability between different treatment periods, while being as short as possible to minimize patient drop‐out due to disease progression or toxicity, thereby maximizing the PK evaluability rate and lowering the total patient enrollment. The impact of shortening the treatment period on total sample size was evaluated.

Due to the larger sample size required to demonstrate C_max equivalence, which makes completing the BE study within the development timeline less feasible, and the potential of failing C_max equivalence based on dissolution data, the clinical relevance of increased C_max on safety was evaluated. If increased C_max was deemed not to have an impact on the safety profile of talazoparib, this may not only allow a BE study design that is powered only for AUC equivalence, which would require a much smaller sample size due to the lower intra‐patient variability of AUC, but also mitigate the risk of failed C_max in demonstrating therapeutic equivalence when AUC meets BE criteria. Myelosuppression‐related adverse events (AEs) are the main side effects of talazoparib (Grade ≥ 3 anemia, neutropenia, and thrombocytopenia in 39%, 21%, and 15% of patients who received talazoparib 1 mg QD, respectively) and the most frequent AEs that lead to talazoparib dose modification [1]. A semi‐mechanistic population PK/PD model (Friberg model) [9] was developed to describe the relationship between talazoparib exposure and each of these hematologic endpoints using data from 466 patients enrolled in Phase 1, 2, and 3 studies PRP‐001, 673‐201, and 673‐301 in which talazoparib monotherapy 0.025 to 1.1 mg QD was administered [10]. All 3 models used linear treatment effect on proliferation of progenitor cells. These PK/PD models and simulations were used to evaluate the impact of increase in C_max and/or AUC on hemoglobin, neutrophil, and platelet levels. Different scenarios (e.g., higher C_max with similar AUC, or higher AUC with minor increase in C_max) were simulated by manually changing the interplay between the structural model parameters. For example, a scenario with increase of C_max by ~50% than population's typical and same AUC was simulated by increasing Ka, and adjusting Q, V2 and V3; and a scenario with increase of AUC by 25% and a minor increase in C_max than population's typical was simulated by decreasing CL (see footnote of Figure 3). Details about the PK/PD model development, evaluation and simulation are available in FDA multi‐disciplinary review of this submission [10]. Population PK simulations were performed using a published model [8].

FIGURE 3.

Simulated PK and PD profiles after treatment with talazoparib 1 mg once daily (a) PK profile, (b) PD profile, hemoglobin count, (c) PD profile, neutrophil count, (d) PD profile, platelet count. ^aPatient 1 represents a population‐typical talazoparib PK profile (reference); patient 2 has 22% higher C_max and the same AUC (at steady state vs patient 1); patient 3 has a 52% higher C_max and the same AUC (at steady state vs patient 1); patient 4 has a 17% higher C_max and 25% higher AUC (at steady state vs patient 1). ^bAdapted from figures in previous publication [10].

2.2. BE/FE Study Design and PK and Safety Assessments

Study 1037 was a Phase 1, open label, 2‐sequence, crossover study to establish BE of the talazoparib SGC formulation to the HC formulation after multiple dosing under fasted conditions in patients with advanced solid tumors, and to evaluate the impact of food on the PK of the SGC formulation. A study schema is presented in Figure 2. The duration of each period was determined by talazoparib PK characteristics and simulation results described in the Results section. The first 2 periods were for BE assessment under fasted conditions. Patients were asked to fast for 2 h before and 2 h after drug administration, except for overnight fasting prior to PK assessment days. Period 3 was for FE assessment and was included as a fixed sequence after the first 2 periods to have at least 12 PK‐evaluable patients. Patients were asked to start breakfast within 30 min before drug administration, and the breakfast needed to be a high‐fat/high‐calorie meal on PK assessment days. Patients were allowed to roll over to the maintenance phase after criteria were met (See Appendix S1) or completing the FE assessment. The study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines after approval from the Ethics Committees and Institutional Review Boards at each study site. All patients provided written informed consent.

FIGURE 2.

Study schema. D1, Day 1; D21, Day 21; D28, Day 28; P1, period 1; P2, period 2; P3, period 3; PK, pharmacokinetic; QD, once daily.

On the day before the last treatment day of each period, a pre‐dose PK sample was collected. On the last treatment day of each period, serial blood samples were collected pre‐dose and at 0.5, 0.75, 1, 1.5, 2, 4, 6, 8, and 24 h post‐dose to determine the plasma concentrations of talazoparib. Samples were analyzed for plasma talazoparib concentration using a validated liquid chromatography coupled to tandem mass spectrometry bioanalytical method, with a validated analytical range of 25.0 to 25,000 pg/mL.

PK parameters were calculated for each PK‐evaluable patient. AUCs were calculated using the linear/log trapezoidal method, using Open NCA Version 2.5.7. C_max and C_trough were observed directly from the data.

BE of PK parameters was determined by constructing 90% CIs around the estimated difference between the test and reference treatments using a mixed‐effects model based on natural log‐transformed data, with sequence, period, and treatment as fixed effects and patient within sequence as a random effect. The adjusted mean differences, as a percentage, and 90% CIs for the differences were exponentiated to provide estimates of adjusted geometric mean ratio (GMR) for test/reference and 90% CIs for ratios.

Safety assessments included collection of AEs, serious AEs, vital signs, 12‐lead electrocardiogram, and laboratory test assessments.

Definition of PK evaluability and safety analysis set is described in Appendix S1.

3. Results

3.1. Relevance of Higher C_max on Safety

The longitudinal hemoglobin, platelet, and neutrophil changes were adequately described by a semi‐mechanistic myelosuppression PK/PD model in patients after talazoparib treatment [10]. The model was used to simulate hemoglobin, platelet, and neutrophil profiles after continuous 1 mg QD dosing of talazoparib when talazoparib C_max or AUC was increased (Figure 3). Subject 1 represented the population's typical PK profile and had Grade 1 or no hematological AEs (hemoglobin level > 100 g/L, platelet count > 75.0 × 10e9/L, and neutrophil count > 1.5 × 10e9/L). Subjects 2 and 3 had the same AUC as subject 1 but had 22% and 52% higher C_max than subject 1, respectively. Subjects 2 and 3 had similar hemoglobin, platelet, and neutrophil profiles compared with subject 1 despite higher C_max, indicating that up to a 52% increase in C_max had no significant impact on myelosuppression. However, subject 4, who had 17% higher C_max and 25% higher AUC, had significantly lower hemoglobin, platelets, and neutrophils than subjects 1, 2, and 3, indicating that AUC is the driver for myelosuppression. Thus, the simulation results showed that limited increase of C_max would not have significant impact on the safety profile of talazoparib when AUC remains the same. This result supports a BE study design powered for AUC equivalence only.

3.2. Optimization of the Duration of Each Treatment Period

PK simulations showed that for patients with slow clearance, 28 days of daily dosing was sufficient for talazoparib to reach steady state in Period 1. However, the subsequent treatment periods can be much shorter and still sufficiently characterize the difference in bioavailability between formulations, as shown in Table 2. Table 2 shows the PK simulation as the HC formulation is dosed daily from Day 1 to Day 28 of Period 1, followed by the SGC formulation from Day 29 to Day 56. Details of the simulation are in the table footnote. Following a 28‐day QD dosing cycle for Period 1, the AUC ratio of test/reference obtained on Day 42 (Period 2, Day 14) was similar to that obtained on Day 56 (Period 2, Day 28). The AUC ratio was similar to the true difference in bioavailability (regardless of increase/decrease in bioavailability) for the second formulation. This observation applies to patients with typical population clearance and those with lower 5th percentile clearance (slower elimination and longer half‐life). Therefore, following 28‐day treatment in the first period, 14‐day continuous dosing in the following periods is sufficient for steady‐state BE and FE assessments. The impact of shortening treatment duration for Periods 2 and 3 on sample size is shown in Table 3. For the same study design (full replicate crossover) with the same number of PK‐evaluable patients, shortening the treatment duration for Periods 2–5 from 28 days to 14 days means shortening total treatment duration from 140 days to 84 days, which would result in fewer dropouts and a higher PK‐evaluable rate, thus significantly decreasing the total enrollment required.

TABLE 2.

Impact of PK sample collection days on the determination of change in bioavailability.

Simulation scenarios	PK sampling days	Population typical CL (6.37 L/h)	Population lower 5 percentile CL (4 L/h)
		AUC0–24 ratio
10% increase in F1 (A)	Day 42 (P2D14)	1.099	1.098
	Day 56 (P2D28)	1.100	1.103
5% increase in F1 (B)	Day 42 (P2D14)	1.050	1.050
	Day 56 (P2D28)	1.050	1.053
10% decrease in F1 (C)	Day 42 (P2D14)	0.902	0.907
	Day 56 (P2D28)	0.900	0.903
5% decrease in F1 (D)	Day 42 (P2D14)	0.951	0.955
	Day 56 (P2D28)	0.950	0.953

Note: Simulation was conducted as the original formulation dosed daily from Day 1 to Day 28 of Period 1, followed by the new formulation from Day 29 to Day 56 (i.e., Day 1 to Day 28 of Period 2). The new formulation was assumed to have 10% or 5% higher F1, or 10% or 5% lower F1 compared with the old formulation, respectively, in the simulation scenarios. PK sampling was collected on Days 28, 42, and 56. The AUC on Days 42 and 56 were compared to that of Day 28 to obtain AUC ratios, which should have been close to the true difference, 1.1, 1.05, 0.9, and 0.95, for scenarios A, B, C, and D, respectively. Population typical CL and lower 5 percentile CL were used to represent typical patients and patients with lower CL (slower elimination and longer half‐life).

Abbreviations: AUC, area under the curve; AUC_0–24, area under the plasma concentration‐time profile from time 0 to 24 h post‐dose; CL, clearance; D, day; F1, bioavailability; P, period; PK, pharmacokinetic.

TABLE 3.

Study designs to establish AUC equivalence only, or both AUC and C_max equivalence between two formulations.

Primary endpoint(s) to meet BE	Design	Duration	Evaluable patients, n	Non‐evaluable rate	Total enrollment, n
AUC only	2‐period crossover	42 days (2‐period) BE +14 days FE a	32	43%	56
Cmax and AUC	2‐period crossover	42 days (2‐period) BE +14 days FE a	164	43%	290
	Partial replicate crossover c	56 days (3‐period) BE +14 days FE a	58	54%	126
	Full replicate crossover c	70 days (4‐period) BE +14 days FE a	40	65%	114
	Full replicate crossover c	112 days (4‐period) BE + 28 days FE b	40	75%	160

Abbreviations: AUC, area under the curve; BE, bioequivalence; C_max, maximum observed plasma concentration; FE, food effects.

^a The duration of the first period was 28 days and the duration of subsequent periods was 14 days.

^b The duration of each period was 28 days.

^c The sample size for replicate crossover designs was calculated using the reference‐scaled average bioequivalence approach.

3.3. Study Design Optimization: Impact on Study Duration and Sample Size

Different crossover designs powered for AUC equivalence only or both AUC and C_max equivalence were considered (schema shown in Figure S1). The most efficient design to show BE for both AUC and C_max was a full replicate crossover design with ~114 total enrollment and > 84 days of treatment for PK evaluation (Table 3). In contrast, demonstration of AUC equivalence only required a 2‐period crossover design with a more feasible sample size of 56 total enrollment (32 PK‐evaluable) and a shorter study duration than a replicate design (Table 3).

Therefore, the optimal design was determined to be a 2‐period crossover schema, with the sample size calculation based on AUC equivalence only, while the impact of formulation change on C_max would be evaluated based on the test/reference ratio. Period 2 of the BE assessment and the FE period (Period 3) would be 14 days, shorter than Period 1 (28 days). However, due to feedback from the regulatory agency, the actual length in the finalized protocol for Periods 2 and 3 was 21 days (Figure 2). Therefore, the total enrollment turned out to be higher than 56 patients, but the relative advantage of the optimal design compared to other design options remained true.

3.4. BE Study PK Assessment Results

A total of 73 patients were assigned to the study treatment: 35 to Sequence 1 and 38 to Sequence 2. A total of 53 patients who had a primary PK parameter of total exposure (AUC_0–24) or C_max in at least 1 treatment period where PK‐evaluable criteria were met were included in the PK‐evaluable population. The PK‐evaluable population included 40 patients who received SGC treatment and 47 patients who received HC treatment, all of whom were included in the BE assessment. Twenty‐two PK‐evaluable patients were included in the SGC FE assessment.

A statistical summary of the BE assessment for AUC_0–24 and C_max of the SGC versus the HC formulation under fasting condition and FE assessment are shown in Table 4. The SGC formulation met the prespecified BE criteria for AUC compared to the initial HC formulation when both formulations were administered at 1 mg QD to steady state under fasted conditions, as the 90% CI of the GMR of steady‐state AUC_0–24 fell within the BE limits of 80%–125%. Steady‐state C_max was approximately 37% higher for the SGC formulation compared to HC.

TABLE 4.

Statistical summary of log‐transformed plasma talazoparib PK parameters (AUC_0–24 and C_max; pharmacokinetic‐evaluable population).

Parameter (unit)	Adjusted (least‐square) geometric means		Ratio (test/reference) of adjusted means a	90% CI for ratio a
	Test	Reference
Soft gel capsule (fasted) vs. commercial hard capsule (fasted)
AUC0–24 ng. hour/mL	186.8	177.6	105.16	(99.00–111.70)
Cmax ng/mL	20.33	14.88	136.62	(125.05–149.27)
Ctrough ng/mL	4.047	4.245	95.33	(89.02–102.09)
Soft gel capsule (fed) vs. soft gel capsule (fasted)
AUC0–24 ng. hour/mL	152.6	173.5	87.97	(81.82–94.58)
Cmax ng/mL	11.18	19.19	58.26	(51.55–65.84)
Ctrough ng/mL	3.650	3.680	99.20	(89.37–110.11)

Note: Values were back transformed from the log scale.

Abbreviations: AUC, area under the curve; AUC_0–24, area under the plasma concentration–time profile from time 0 to 24 h post‐dose; CI, confidence interval; C_max, maximum observed plasma concentration; C_trough, pre‐dose plasma drug concentration; PK, pharmacokinetic.

^a The ratios (and 90% CIs) are expressed as percentages.

Daily administration of SGC formulation under fed conditions did not affect talazoparib AUC_0–24 at steady state relative to administration under fasted conditions, as the 90% CI of the GMR fell within the BE limits of 80% to 125%. However, administration of talazoparib SGC under fed conditions decreased talazoparib steady‐state C_max by approximately 42% compared with fasted conditions.

Even though C_trough comparison is generally not part of the BE assessment, C_trough equivalence was also evaluated (Table 4). Steady‐state C_trough was comparable between the two formulations, as the GMR fell within the BE limits of 80% to 125%.

Descriptive statistics of PK parameters by treatment group are presented in Table S2. Intra‐patient variability of AUC_0–24 and C_max at steady state for talazoparib were 15.0% and 22.7%, respectively. Total variability in talazoparib exposure was comparable between the two formulations under fasted conditions (geometric CV% for AUC_0–24 and C_max ranged from 32% to 40%). Steady‐state talazoparib exposures were generally achieved prior to the planned intensive PK sampling collection days for each treatment as shown in Figure S2, which is consistent with population PK prediction discussed earlier.

3.5. BE Study Safety Results

All treated patients (N = 73) were included in the safety analysis set. Baseline demographic and patient characteristics of the safety analysis population are shown in Table S1.

Talazoparib 1 mg QD with the HC formulation (administered under fasted conditions) and the SGC formulation (administered under fasted or fed conditions) was generally tolerated in patients with advanced solid tumors in this study. Due to the nature of the crossover design, only treatment‐emergent adverse events (TEAEs) observed in Period 1 can be completely attributed to the respective formulation. A comparison of safety results from Period 1 is presented in Tables S3 and S4. There were no clinically significant differences between the two formulations in the frequencies of all‐grade TEAEs, Grade 3/4 TEAEs, or SAEs. Treatment discontinuation due to TEAEs, dose reduction/interruption due to TEAEs, incidences of the most frequent all‐grade TEAEs (anemia, neutropenia, thrombocytopenia, diarrhea, nausea, and vomiting), incidences of the most frequent Grade ≥ 3 TEAEs (anemia, neutropenia, thrombocytopenia, and fatigue) were similar between the two formulations. The TEAEs were generally consistent with the known safety profile of talazoparib. No new safety signals were identified. This comparison of AEs has limitations due to the small number of patients (n = 38 for HC; n = 35 for SGC), the low prevalence of the specific TEAEs, and the length of the treatment periods (28 days for Period 1).

4. Discussion

MIDD addressed the logistic and operational challenges of evaluating steady‐state BE for talazoparib, which has a long elimination half‐life and high intra‐patient and total variability in C_max, in patients with cancer. The optimized design was a two‐way, two‐period crossover for BE assessment, followed by a period to assess FE (Figure 2). The sample size calculation for BE assessment was based on AUC equivalence only, as supported by the lack of impact of C_max on safety endpoints demonstrated by PK/PD modeling and simulation. Simulation results based on the population PK model enabled us to shorten the treatment duration of Periods 2 and 3, thus further reducing the number of patients needed to be enrolled. The study results showed that the SGC formulation met the prespecified BE criteria for AUC compared to the initial HC formulation. However, the C_max for the SGC formulation was 37% higher than the HC formulation; PK/PD modeling demonstrated that this increase in C_max is not likely to increase the incidence of AEs (anemia, thrombocytopenia, neutropenia) when the AUC remains the same. Safety data did not show any clinically significant difference in the safety profile between the two formulations. Based on the totality of data, including AUC equivalence, the lack of impact of C_max on safety endpoints, and the observed safety data, bridging between the two formulations was established and the SGC formulation was approved by the FDA for commercialization.

Overall, the study design is in alignment with FDA and ICH guidance [11]. Due to a few non‐conventional design components (e.g., sample size calculation based on AUC equivalence only and uneven duration of each treatment period), the team sought pre‐alignment on the study design with regulatory agencies before protocol finalization. After completion of the BE/FE study, the sponsor requested an MIDD meeting with the FDA to obtain agreement for submission based on AUC equivalence only (as C_max failed to meet the equivalence criteria), supported by PK/PD modeling and simulations which demonstrated the lack of clinical significance of a higher C_max on all three hematologic AEs (thrombocytopenia, anemia, and neutropenia) [10]. The evaluation of BE for a generic drug in an Abbreviated New Drug Application [12] is different from that of a post‐approval product change in an NDA. NDA review considers the totality of information available in the submission using the principles of BE, exposure‐response evaluations, and clinical trial results [13]. The approval of the talazoparib SGC formulation is an example of the totality of data being considered in the NDA review.

Clinical pharmacology studies for most poly(ADP‐ribose) polymerase (PARP) inhibitors were conducted in patients with advanced solid tumors, such as the olaparib renal impairment and hepatic impairment studies [14, 15], olaparib FE study [16], rucaparib hepatic impairment study [17], and niraparib BE/FE study [18]. These were conducted to evaluate PK after single‐dose administration of the drug, instead of steady‐state PK, due to shorter half‐life of these drugs than talazoparib. However, clinical pharmacology evaluations for talazoparib need to be conducted after multiple doses at steady‐state as it is unethical to conduct a single‐dose crossover study in patients with cancer; the necessary long treatment washout period could negatively impact disease control for these patients. There are multiple challenges in conducting a multiple‐dose BE study in patients with cancer, as shown in Table 1, making it more difficult to achieve formulation bridging for talazoparib than other PARP inhibitors. MIDD was used to support a more logistically feasible and smaller study design, by providing justification for powering the study on AUC equivalence only and shorter treatment duration for Periods 2 and 3. Compared with the most efficient traditional BE study design (a replicated crossover study powered for both C_max and AUC equivalence with equal length of treatment duration for each period), this optimized study design requires a smaller sample size (56 vs. 160 total enrollment) and shorter treatment duration (56 vs. 140 days) (Table 3), which translates to a significant saving in the development timeline and cost. We were able to shorten the study conduct from 2 years to 1 year. In comparison, the niraparib BE study enrolled 179 patients with cancer; 64 and 66 were evaluable for BE in each sequence [18].

Periods 2 and 3 can be condensed to 14 days because the majority of the exposure on Day 14 of Periods 2 and 3 is contributed by the formulation administered in that period, with minor contribution to the exposure by the formulation administered in the previous period. Such limited contribution from the previous formulation would not have an impact on characterizing the change in bioavailability as shown in Table 2. Therefore, even though it could take up to 28 days for patients with slow clearance to reach steady state in Period 1, the 14‐day treatment duration in Periods 2 and 3 is sufficient.

Due to the lengthy treatment duration, dose interruptions, missed doses, and dosing outside of the nominal time are inevitable. It would be infeasible to replace a patient if any of the above occur, especially considering the difficulty in enrolling eligible patients with cancer (those who may benefit from talazoparib treatment and are willing to participate in the trial). Therefore, modeling and simulation were performed to determine the level of deviation allowed without compromising the quality of PK assessment, and to set the criteria for PK evaluability and completion of a treatment period, as described in Appendix S1.

In addition to MIDD, other considerations were applied to ensure successful execution of the study. For example, the eligibility criteria were designed to ensure that patients could stay on treatment long enough to achieve steady‐state PK, and have stable renal function for elimination, since talazoparib is eliminated mainly through renal excretion. More details are in Appendix S1. Realistic considerations for patients with cancer were given to the requirements for fasting and fed status on non‐site visit days, as described in the Methods. In Period 3, there was no restriction on the fat and caloric content for the breakfast on non‐PK sampling days. The breakfast on PK assessment days was a high‐fat/high‐calorie meal, as physiological conditions induced by a high‐fat meal generally provide the greatest effect on gastrointestinal physiology and the maximum effect on the systemic availability of a drug [19]. It should also be noted that high‐fat food only had an impact on C_max but not AUC or C_trough. Therefore, whether talazoparib was taken with high‐fat food or not on non‐PK sampling days should not have an impact on PK of talazoparib on PK assessment day. Real‐time access to patient daily dosing record via eDiary is important to assess the PK evaluability, which determines if a patient can move on to the next treatment period, needs to repeat/extend the current treatment period, or can move on to maintenance phase.

The previous FE study on the HC formulation showed that a high‐fat, high‐calorie meal decreased C_max by 46%, delayed T _max from 1 h to 4 h, and did not affect AUC. As AUC, but not C_max, was considered a driver for efficacy, the decrease in C_max under fed conditions was not expected to impact efficacy. Therefore, talazoparib HC can be taken with or without food [1]. The result of FE assessment for the SGC formulation showed a 42% decrease in C_max in the fed state. This is consistent with the food effect observed with the HC formulation, supporting that the SGC can be taken with or without food.

In conclusion, the totality of data (AUC equivalence, lack of impact of C_max on safety, and observed safety data) supported the bridging of the talazoparib SGC formulation with the initial HC formulation although C_max failed BE criteria. In this NDA approval, MIDD played a critical role in enabling a realistic study design in patients with cancer and the mitigation of the impact of failed C_max equivalence. The principles and framework shown in this work are applicable to other drugs or therapeutic area as long as the exposure metric driving efficacy and safety is well characterized with existing data.

Author Contributions

C.C.G., D.W., and X.G. wrote the manuscript; C.C.G., D.W., Y.Y., A.P., M.E., and J.H. designed the research; C.C.G., X.G., Y.W., H.S., L.D., and J.H. performed the research; C.C.G., Y.Y., X.G., Y.W., A.P., S.J., and J.H. analyzed the data.

Conflicts of Interest

The authors are employees of Pfizer and may hold Pfizer stock/stock options.

Supporting information

Appendix S1: psp470157‐sup‐0001‐AppendixS1.docx.

Wang D., Guo C. C., Gao X., et al., “Talazoparib Formulation Bridging in Cancer Patients—Challenges and the Critical Role of Model‐Informed Drug Development in Approval Despite Failed Bioequivalence,” CPT: Pharmacometrics & Systems Pharmacology 15, no. 1 (2026): e70157, 10.1002/psp4.70157.

Data Availability Statement

Upon request, and subject to review, Pfizer will provide the data that support the findings of this study. Subject to certain criteria, conditions and exceptions, Pfizer may also provide access to the related individual de‐identified participant data. See https://www.pfizer.com/science/clinical‐trials/trial‐data‐and‐results/ for more information.