Population pharmacokinetic and exposure-response analyses from ALTA-1L: Model-based analyses supporting the brigatinib dose in ALK-positive NSCLC

Abstract

The ALK in Lung Cancer Trial of brigAtinib in First Line (ALTA-1L) compared brigatinib versus crizotinib in anaplastic lymphoma kinase (ALK) inhibitor-naive patients with ALK+ non-small cell lung cancer (NSCLC). A population pharmacokinetic (PK) model was used to estimate brigatinib exposures for exposure-efficacy and exposure-safety analyses in ALTA-1L. A previously developed population PK model for brigatinib was applied to estimate brigatinib PK parameters. Relationships between static (time-independent) and dynamic (time-varying) exposure metrics and efficacy (progression-free survival [PFS], objective response rate [ORR], and intracranial ORR [iORR]) and safety outcomes (selected grade ≥2 and grade ≥3 adverse events [AEs]) were evaluated using logistic regression and time-to-event analyses. There were no meaningful differences in brigatinib PK in the first-line and second-line settings, supporting use of the previous population PK model for the first-line population. Exposure-response analyses showed no significant effect of time-varying brigatinib exposure on PFS. Brigatinib exposure was not significantly related to ORR, but higher exposure was associated with higher iORR (odds ratio: 1.13, 95% confidence interval: 1.01-1.28, p = 0.049). Across the observed median exposure (5th-95th percentile) at steady state for 180 mg once daily, the predicted probability of iORR was 0.83 (0.58-0.99). AEs significantly associated with higher exposure were elevated lipase (grade ≥3) and amylase (grade ≥2). Time to first brigatinib dose reduction was not related to exposure. These results support the benefit-risk profile of first-line brigatinib 180 mg once daily (7-day lead-in dose at 90 mg once daily) in patients with ALK+ NSCLC.

FIGURE 1

(a) Histogram comparing individual CL/F estimates between ALTA‐1L and ALTA. To further justify the ability of the model to accurately predict exposure in patients in ALTA‐1L, distributions of post hoc CL/F estimates were compared for patients in ALTA and ALTA‐1L. CL/F estimates were similar between the trials, confirming that the Bayesian re‐estimation approach was an appropriate method to obtain individual CL/F and post hoc exposure estimates for patients in ALTA‐1L. (b) Brigatinib exposure (AUC) following 180 mg once daily stratified by covariates of interest. The relationship between covariates (continuous: age, body weight, ALT, AST, bilirubin, and eGFR; categorical: sex and race) and model‐based exposure estimates was explored using linear regression models with each covariate as a predictor. Comparison of post hoc brigatinib AUC confirmed the lack of clinically meaningful effects of the covariates on brigatinib systemic exposure compared with overall exposure variability. Black vertical line and values at the base of the figure refer to the predicted AUC of brigatinib in a typical patient with baseline albumin of 41 g/dL. Black shaded bar illustrates the 5th to 95th percentile exposure range across the entire population expressed as a ratio relative to the reference exposure in a typical patient. Blue shaded bar represents the influence of baseline albumin on exposure. Lines above the blue shaded bar represent the different groups for the covariates listed at the left. For continuous covariates, the two dots represent the ratio of exposure of the 95th percentile covariate value versus the median and 5th percentile versus the median. For categorical covariates, dots represent the ratio of exposure of the categories versus the reference (most common) category. The horizontal bars on the two sides of the dots represent the corresponding 90% CI. A, Asian; ALT, alanine aminotransferase; AST, aspartate aminotransferase; AUC, area under the concentration‐time curve; BILI, bilirubin; CI, confidence interval; CL/F, apparent oral clearance from the central compartment; eGFR, estimated glomerular filtration rate; F, female; M, male; W, White

FIGURE 2

Exposure‐efficacy analyses. (a) Kaplan‐Meier probability of PFS by simulated brigatinib exposure quartiles. To evaluate the relationship between brigatinib exposure and PFS, a static exposure metric of time‐averaged AUC between the last two disease assessment scans preceding progression or censoring was used. PFS KM estimates plotted by exposure quartiles suggested that patients with higher exposure had faster onset and higher incidence of disease progression than those with lower exposure. Values for the crizotinib arm of the study are superimposed; however, no exposure values were available for crizotinib. For median PFS values, NA indicates that the probability of having no disease progression or death has not yet gone beyond 0.50 and hence the median survival time cannot be determined. ^aSimulated exposure metric is time‐averaged AUC between the last two disease assessment scans preceding progression for PFS or censoring. Observed incidence and model‐predicted probability of (b) ORR and (c) iORR as a function of brigatinib exposure. The relationships between ORR and iORR and brigatinib exposure were analyzed using the static exposure metric of time‐averaged AUC between the last two disease assessment scans preceding best confirmed response. The probability of response was plotted against predicted exposure values, and probabilities were calculated by observed exposure quartiles or tertiles. Exposure–clinical response relationships were characterized by logistic regression models, which did not show a significant relationship between the probability of achieving ORR and time‐averaged brigatinib AUC between the last two disease assessment scans preceding the best confirmed objective response. In contrast, time‐averaged brigatinib AUC between the last two disease assessment scans preceding best confirmed intracranial response was a statistically significant predictor of iORR in patients with brain metastases at baseline. Dotted curves represent the 95% CI of the logistic regression model prediction. The horizontal black line separated by vertical black solid lines denotes the brigatinib exposure range in each quartile (ORR) and tertile (iORR). Black dots (vertical lines) represent the observed proportion of patients (95% CI) in each quartile (ORR) and tertile (iORR). n/N is the number of patients with events/total number of patients in each quartile (ORR) and tertile (iORR). Grey open circles represent observed individual data. AUC, area under the concentration‐time curve; CI, confidence interval; KM, Kaplan‐Meier; iORR, intracranial objective response rate; NA, not available; Obs, observed; ORR, objective response rate; PFS, progression‐free survival

FIGURE 3

Exposure‐safety analyses. Observed incidence and predicted probability of (a) grade ≥3 lipase increase and (b) grade ≥2 amylase increase as a function of brigatinib exposure. The relationship between time‐averaged AUC across days 8 to 14 of cycle one and AE probability was examined using logistic regression models. The analysis demonstrated a statistically significant relationship between exposure and grade ≥3 lipase increase and grade ≥2 amylase increase. (c) Kaplan‐Meier estimates for time to first brigatinib dose reduction stratified by time‐averaged AUC quartiles. To explore the relationship between brigatinib exposure and dose reductions, KM plots of time to first brigatinib dose reduction were generated for brigatinib exposure (time‐averaged AUC to the first occurrence of a dose reduction) quartiles. No discernible effect of brigatinib exposure on time to first brigatinib dose reduction was noted. Values for the crizotinib arm of the study are superimposed; however, no exposure values were available for crizotinib. Dotted curves represent the 95% CI of the logistic regression model prediction. The horizontal black line separated by vertical black solid lines denotes the brigatinib exposure range in each quartile. Black dots (vertical lines) represent the observed proportion of patients (95% CI) in each quartile. n/N is the number of patients with events/total number of patients in each quartile. Grey open circles represent observed individual data. AE, adverse event; AUC, area under the concentration‐time curve; CI, confidence interval; KM, Kaplan‐Meier; Obs, observed