Alternative dosing regimens for atezolizumab: an example of model-informed drug development in the postmarketing setting

Abstract

Purpose: To determine the exposure-response (ER) relationships between atezolizumab exposure and efficacy or safety in patients with advanced non-small cell lung cancer (NSCLC) or urothelial carcinoma (UC) and to identify alternative dosing regimens.

Methods: ER analyses were conducted using pooled NSCLC and UC data from phase 1 and 3 studies (PCD4989g, OAK, IMvigor211; ClinicalTrials.gov IDs, NCT01375842, NCT02008227, and NCT02302807, respectively). Objective response rate, overall survival, and adverse events were evaluated vs pharmacokinetic (PK) metrics. Population PK-simulated exposures for regimens of 840 mg every 2 weeks (q2w) and 1680 mg every 4 weeks (q4w) were compared with the approved regimen of 1200 mg every 3 weeks (q3w) and the maximum assessed dose (MAD; 20 mg/kg q3w). Phase 3 IMpassion130 (NCT02425891) data were used to validate the PK simulations for 840 mg q2w. Observed safety data were evaluated by exposure and body weight subgroups.

Results: No significant ER relationships were observed for safety or efficacy. Predicted exposures for 840 mg q2w and 1680 mg q4w were comparable to 1200 mg q3w and the MAD and consistent with observed PK data from IMpassion130. Observed safety was similar between patients with a Cmax above and below the predicted Cmax for 1680 mg q4w and between patients in the lowest and upper 3 body weight quartiles.

Conclusion: Atezolizumab regimens of 840 mg q2w and 1680 mg q4w are expected to have comparable efficacy and safety as the approved regimen of 1200 mg q3w, supporting their interchangeable use and offering patients greater flexibility.

Keywords: Atezolizumab; Exposure–response; PD-L1; Population pharmacokinetics.

Conflict of interest statement

K.M. Morrissey, H. Patel, R. Zhang, B. Wu, H.P. Chan, S. Girish, J.Y. Jin, H.R. Winter, and R. Bruno are employed by Genentech, Inc. M. Marchand is employed by Certara Strategic Consulting. A. Mecke is employed by F. Hoffmann-La Roche, Ltd.

Figures

Fig. 1

Pooled exposure–response analyses of patients with locally advanced or metastatic NSCLC or UC. The proportions of responders are plotted vs (a) AUC or (b) Cmin at cycle 1. In part (a), for legibility, 1 extreme AUC value (> 15,000 μg.day/mL) is not displayed on the plot. Wald P values from logistic regression of the proportion of responders vs exposure are displayed. Gray solid lines and shaded areas represent the logistic regression slope model and 95% PI. Filled circles and error bars represent the proportions of responders in exposure quartiles and 95% CI; vertical lines are the limits of the exposure quartiles. Cross markings (×) represent response events (0: no, 1: yes). Triangle and two-headed arrows represent the mean exposure and exposure interval between the 10th and 90th percentiles, respectively, for patients receiving atezolizumab 1200 mg q3w. Cycle 1 AUC corresponds to the AUC during the first 3 weeks after treatment start and with PK parameters estimated based on cycle 1 data only. AUC area under the concentration–time curve, Cmin minimum (trough) serum atezolizumab concentration, CR complete response, n number of patients, NSCLC non-small cell lung cancer, PI prediction interval, PK pharmacokinetics, PR partial response, UC urothelial carcinoma

Fig. 2

Predicted OS HRs (atezolizumab vs comparator) by cycle 1 AUC quartiles for patients with median baseline covariates. Forest plots for OS HRs from (a) OAK (NSCLC) and (b) IMvigor211 (UC) are shown. Model-predicted HRs are shown as diamonds, with bars indicating 95% PIs (1000 replicates). AUC area under the concentration–time curve, HR hazard ratio, NSCLC non-small cell lung cancer, OS overall survival, PI prediction interval, UC urothelial carcinoma

Fig. 3

Pooled exposure–response analyses of safety in patients with locally advanced or metastatic NSCLC or UC. Indicated AE frequencies ([a, c] grade ≥ 3 AEs; [b, d] AESIs) are plotted vs (a, b) AUC or (c, d) Cmax at cycle 1. For legibility, 2 extreme AUC values (> 15,000 μg.day/mL) and 2 extreme Cmax values (> 1500 μg/mL) are not displayed on the plots. Wald P values from logistic regression of AE incidence vs exposure are displayed. Gray solid lines and shaded areas represent the logistic regression slope model and 95% PI. Filled circles and error bars represent AE proportion in exposure quartiles and 95% CI; vertical lines are the limits of the exposure quartiles. Cross markings (×) represent AE events (0: no, 1: yes). Triangle and two-headed arrows represent the mean exposure and exposure interval between the 10th and 90th percentiles, respectively, for patients receiving atezolizumab 1200 mg. Cycle 1 AUC corresponds to the AUC during the first 3 weeks after treatment start and with PK parameters estimated based on cycle 1 data only. AE adverse event, AESI adverse event of special interest, AUC area under the concentration–time curve, Cmax maximum serum atezolizumab concentration, n number of patients, NSCLC non-small cell lung cancer, PI prediction interval, PK pharmacokinetics, UC urothelial carcinoma

Fig. 4

Simulated atezolizumab exposure profiles for various dosing regimens. Geometric means are plotted. Shaded areas represent 90% PIs. PI prediction interval, q2w every 2 weeks, q3w every 3 weeks, q4w every 4 weeks

Fig. 5

Prediction-corrected VPC of atezolizumab data in TNBC (IMpassion130) using the phase 1 popPK model. Data are plotted on a semi-log scale. Two population-predicted concentrations n number of samples, PI prediction interval, popPK population pharmacokinetics, TNBC triple-negative breast cancer, VPC visual performance check