Pan‐cancer population pharmacokinetics and exposure‐safety and ‐efficacy analyses of atezolizumab in patients with high tumor mutational burden

Abstract We retrospectively investigated the pharmacokinetics and exposure‐efficacy/safety relationships of single‐agent atezolizumab based on tissue tumor mutational burden (tTMB) status (high vs low [≥16 vs <16 mutations/megabase]) in a pan‐tumor population from seven clinical trials. Data sources included the OAK, POPLAR, BIRCH, FIR, IMvigor210, IMvigor211, and PCD4989g studies; 986 of 2894 treated patients (34%) had TMB data. Exposure metrics were obtained using a prior two‐compartment intravenous‐infusion population‐pharmacokinetics model, merged with prognostic, biomarker, efficacy, and safety variables. Baseline demographic/clinical characteristics and prognostic factors were well balanced between patients with high (n = 175) and low (n = 811) tTMB. Exposure was similar in the high‐ and low‐tTMB subgroups, with no difference seen in the evaluable vs total treated populations. The objective response rate (ORR) was 29.7% vs 13.4%, complete response rate was 6.9% vs 3.2%, and median duration of response (95% CI) was 29.0 (18.6‐NE) months vs 15.9 (12.5‐20.5) months for patients with high‐tTMB vs low‐tTMB tumors, respectively. A flat exposure‐efficacy relationship was seen for ORR in patients with high‐tTMB based on the cycle 1 minimum atezolizumab concentration and area under the serum concentration time curve (AUC). A nonsignificant exposure‐safety profile was seen for grade 3/4 adverse events and adverse events of special interest based on the AUC of atezolizumab in the high‐tTMB population. tTMB is an additional predictive biological factor affecting response to atezolizumab, and quantitative investigations of atezolizumab exposure and relationships of exposure with safety and efficacy support the use of a 1200‐mg, every 3‐week regimen in a tumor‐agnostic high‐tTMB population.


Funding information
The studies were sponsored by Genentech, Inc/F. Hoffmann-La Roche Ltd. The sponsor was involved in the design of the studies; the collection, analysis, and interpretation of the data; and in the writing of the manuscript.

Abstract
We retrospectively investigated the pharmacokinetics and exposure-efficacy/safety relationships of single-agent atezolizumab based on tissue tumor mutational burden had TMB data. Exposure metrics were obtained using a prior two-compartment intravenous-infusion population-pharmacokinetics model, merged with prognostic, biomarker, efficacy, and safety variables. Baseline demographic/clinical characteristics and prognostic factors were well balanced between patients with high (n = 175) and low (n = 811) tTMB. Exposure was similar in the high-and low-tTMB subgroups, with no difference seen in the evaluable vs total treated populations. The objective response rate (ORR) was 29.7% vs 13.4%, complete response rate was 6.9% vs 3.2%, and median duration of response (95% CI) was 29.0 (18.6-NE) months vs 15.9 (12.5-20.5) months for patients with high-tTMB vs low-tTMB tumors, respectively. A flat exposure-efficacy relationship was seen for ORR in patients with high-tTMB based on the cycle 1 minimum atezolizumab concentration and area under the serum concentration time curve (AUC). A nonsignificant exposure-safety profile was seen for grade 3/4 adverse events and adverse events of special interest based on the AUC of atezolizumab in the high-tTMB population. tTMB is an additional predictive biological factor affecting response to atezolizumab, and quantitative investigations of atezolizumab exposure and relationships of exposure with safety and efficacy support the use of a 1200-mg, every 3-week regimen in a tumor-agnostic high-tTMB population.

| INTRODUC TI ON
Numerous clinical investigations of therapies that mobilize the immune system against cancer are underway, including those involving immune checkpoint inhibitors (ICIs), cell-or gene-based therapies, oncolytic viruses, vaccines, targeted therapies, and other novel modalities. 1,2 Programmed death-ligand 1 (PD-L1)and programmed death-1 (PD-1)-targeting ICIs prevent inhibitory signals to T cells, resulting in tumor rejection. These agents have expanded the therapeutic approaches in immuno-oncology, inspiring >2250 trials in ICIs since 2018. 3,4 Within this expansion, biomarker-based selection approaches, such as microsatellite instability/mismatch repair deficiency and PD-L1 expression, are being developed to help guide the selection of ICI-based therapies. 5,6 The US Food and Drug Administration (FDA) approval of ICI pembrolizumab for microsatellite instability-high/mismatch repair-deficient cancers illustrated a paradigm shift 7, paving the way for other biomarker-based tumor agnostic indications, including the mid-2020 accelerated FDA approval of pembrolizumab in previously treated TMB-high solid tumors. 8 TMB reflects the number of somatic mutations existing per coding area of a tumor genome. The number of mutations can vary across tumor type, and many mutagenic processes can drive high TMB, including but not limited to DNA replication infidelity, mismatch repair deficiency, environmental mutagens such as tobacco smoke and ultraviolet light, contaminated food pathogens, and aging. 9 Nonsynonymous mutations increase the number of tumor-specific neoantigens recognized by the immune system, thus, TMB is a proxy estimate of the neoantigen load of a tumor. 10 This process increases the number of tumor-infiltrating immune cells (IC) in the tumor microenvironment and bolsters cytotoxic T-cell responses. TMB was found to correlate with response to ICIs in a cross-study analysis of 27 tumor types 11 and in a prospective multicohort evaluation, 12 and correlations with overall survival have further reinforced the predictive value of TMB in many cancers. 11,[13][14][15] Given the association of TMB with response to ICIs, the substantial number of ongoing clinical trials surveying TMB as a potential biomarker is unsurprising. 13 Atezolizumab is an anti-PD-L1 monoclonal antibody that selectively targets PD-L1 to inhibit interaction with its receptors PD-1 and B7.1 to enhance T-cell responses. 16

| Studies contributing to the tTMB analysis
Seven studies evaluating the efficacy and safety of atezolizumab monotherapy were included in this analysis, as described in Only atezolizumab monotherapy studies were included to limit any potential bias from combination agents. PD-L1 status was evaluated using the VENTANA SP142 immunohistochemistry assay (Ventana Medical Systems, Tucson, Arizona).
The phase II/III studies each used a 1200-mg dose of atezolizumab every 3 weeks (q3w). The phase I study PCD4989g also included some patients treated with 10-, 15-, and 20-mg/kg doses of atezolizumab q3w, and these patients were also included in this analysis. Atezolizumab was administered by intravenous infusion on

| tTMB assessment
tTMB was evaluated by the FoundationOne hybrid-capture next-generation-sequencing assay (F1). Details on the F1 platform and TMBestimation algorithms were previously reported. [27][28][29] Briefly, the assay detects substitutions, insertion, deletion alterations, and copy number alterations in 324 genes using DNA from formalin-fixed paraffinembedded solid tumor specimens. The number of somatic mutations is quantified as mutations per megabase (mut/Mb) by removing polymorphisms and predicted drivers from all variants to provide somatic mutation count per Mb. The distribution of tTMB was observed to be a continuous variable, with a median tTMB of 7.9 mutations/Mb.
To determine an appropriate cutoff from the retrospective analysis, a tTMB-high cutoff was established based on balancing between a high response rate and a reasonable prevalence across a heterogenous set of tumor types (see Figure S1). A cutoff of ≥16 mutations/Mb was selected. Pooled response rates were evaluated at tTMB cutoffs from

| Pharmacokinetics (PK) sampling and analytical methods
In POPLAR, BIRCH, FIR, OAK, IMvigor210, and IMvigor211, PK sampling of atezolizumab occurred as follows: following infusion on day 1 of cycle 1; prior to infusion on day 1 of cycles 1, 2, 3, 4, 8, and 16; every eight cycles thereafter; at discontinuation; and 120 days after the last dose. The phase I study followed the same scheme as above, but additional samples were collected from the majority of patients at cycle 1 (24 hours, 72 hours, day 8, and day 15) and pre-dose (prior to infusion) at cycles 5, 7, 10, 12, and 14. Blood samples from patients were centrifuged at 1500-2000g for 15 minutes at 4°C. The serum samples were then stored at −60°C or less. Atezolizumab concentrations were quantified by enzyme-linked immunosorbent assay (ELISA), with a 60-ng/mL lower limit of quantification in human serum. The method for measuring atezolizumab in human serum was validated and included an inter-run and intra-run precision (%coefficient of variation [%CV]) of ≤4.59% and ≤4.12%, and inter-run and intra-run accuracy (%relative error) of −7.13% to 4.17% and −7.17% to 3.96%, respectively. The assay specifically detected atezolizumab in disease stage samples. No interference was observed from hemolysis, lipemia, and co-medications.

| Population-PK model and derivation of exposure metrics
A previously developed two-compartment population-PK (popPK) model of atezolizumab based on phase I (PCD4989g) data 32 was used in the PK analyses. According to the Phase I popPK model, the typical clearance (CL, in L/day) of atezolizumab for patient i was: where BWT = body weight (kg); ALBU = albumin (g/L); Tumor burden The popPK model was developed based on the Phase I study, which used a more intensive PK sampling schedule than the Phase II and III studies, from which mostly trough PK samples were collected. Therefore, the popPK model was not re-developed and the parameters were not re-estimated based on the pooled data.   Indeed, clinical development with pembrolizumab's microsatellite instability and TMB biomarkers in pan-tumor populations were previously supported by using ORR and DOR data. 7

| Exposure-safety analysis
Patients for whom both tTMB measurements and exposure data were available were included. Exposure-safety was assessed for grade 3/4 AEs and any-grade AESIs. Atezolizumab exposure levels were grouped based on quartiles of log-transformed AUC and displayed as described for the probability of response.

| Nomenclature of targets and ligands
Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guide topha rmaco logy.
org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY, 35 and are permanently archived in the Concise Guide to PHARMACOLOGY 2019/20. 36

| Patient demographics and clinical characteristics
The pooled data set of tTMB-evaluable patients comprised 986 pa- to-lymphocyte ratio and albumin and lactate dehydrogenase levels were balanced between both groups. The median tTMB was approximately three-fold lower in patients with low tTMB vs high-tTMB. The analysis populations are depicted in Figure 1. and more samples at C min .

| PopPK model
Goodness-of-fit plots were adequate for population predictions ( Figure S2A) of the pooled data without re-estimation of the popPK model parameters. CWRES were homogeneously distributed around 0, suggesting no bias in the predictions of high and low concentrations of atezolizumab. The pcVPC was performed using C max and C min atezolizumab data stratified by tTMB level, as shown in Figure S3.
The pcVPC plots suggested that the median, 5th, and 95th percentiles of observed C max and C min were within the prediction interval of the previously developed popPK model.
Having found that the popPK model described the data well, we  Figure S4).

| Exposure metrics
The model-derived geometric mean (%CV) cycle 1 C min of atezolizumab was 62.6 µg/mL (212%) in the overall treated population  Table 2). The large %CV associated with C min are related to the low exposure levels observed in a small number of patients. The AUC, C max , C min , and clearance were each similar between both groups, with a maximal difference observed between groups of <1%. Figure 2 illustrates the distribution of cycle 1 C min in patients with low tTMB vs high tTMB.
The median patient exceeded the target trough exposure of 6 µg/ mL 38 by more than 10-fold, regardless of tTMB status, with overlapping distributions between subgroups.  This observation-that patients with high tTMB were more likely to achieve an objective response with atezolizumab than patients with low tTMB (ratio of over two odds)-confirms the positive association and predictive value between ORR and tTMB status that has been observed in other meta-analyses following the use of ICIs. 39 The corresponding complete response (CR) frequencies were 3.2% (26 of 811 patients) and 6.9% (12 of 175 patients) in the low-tTMB and high-tTMB subgroups, respectively. DOR by tTMB status is shown in Figure S5. probability of achieving an objective response in patients with high tTMB was not significantly correlated to atezolizumab AUC at cycle 1; the exploratory P value was 0.751 ( Figure 3A). Similar results were found when cycle 1 C min was used ( Figure 3B, exploratory P = .998).

| TMB-efficacy and exposure-efficacy analysis
The mean change in tumor size is shown by tTMB status in Figure 4A and by cycle 1 C min exposure quartile in Figure 4B. More

| Exposure-safety analysis
The exposure-safety analysis was performed in high-tTMB patients with exposure data. Grade 3/4 AEs and all-grade AESIs occurred at an incidence of 56.9% (events in 167 patients) and 40.4% (events in 171 patients), respectively. The incidence of grade 3/4 AEs and AESIs by atezolizumab cycle 1 AUC is shown in Figure 3C and D.
Exposure metrics within the first treatment cycle were used rather than steady-state metrics to isolate potentially confounding factors on exposure, such as time-varying clearance. 26 There was no consistent trend of increased AE incidence with increased exposure for both grade 3/4 AEs (exploratory P = .812) and AESIs (exploratory P = .280) with atezolizumab AUC. Similarly, no trend was detected with C max or C min ( Figure S6).

| tTMB-response analysis
Additional exploratory analyses were conducted to evaluate longitudinal relationships of efficacy and safety by tTMB in the pooled data set. Overall, a steep relationship was observed between an increasing tTMB and the proportion of responders (patients with CR + PR) following treatment with atezolizumab, whereas the tTMB-response curve for grade 3/4 AE occurrence was flat and the tTMB-response curve for all-grade AESI occurrence was shallow ( Figure S7).

| Tumor types of responding patients with high tTMB
Response by tumor type for tTMB-evaluable responding patients is shown in Table S3. In patients with high tTMB, objective responses

| D ISCUSS I ON
As a result of widespread advances and expansion in immuno-oncology, the development of novel immunotherapy biomarkers to improve patient selection represents an ongoing challenge, given that not all patients derive benefit from ICIs. 40  The characterization of prognostic factors, dose selection, and exposure-response relationships is important in optimizing effective immunotherapies; yet, this area has been underinvestigated or has not been applied in a high-TMB tumor-agnostic indication. 42,43 Although tTMB is not expected to influence drug exposure, this is the first account of quantitative clinical pharmacology findings of an ICI evaluated by tTMB across multiple trials in patients with solid tumors. We observed that high tTMB is a potential positive predictive marker associated with increased clinical benefit following treatment with atezolizumab in diverse cancers-a finding similar to that in meta-analyses of other PD-L1/PD-1 agents. 44,45 Baseline prognostic factors that may lead to bias should also be explored. Patient characteristics and prognostic factors important in assessing clinical impact were generally well balanced between patients by tTMB status. In our study a larger percentage of men were observed in the high-TMB group-a result consistent with recent reports suggesting potential associations of higher TMB in male patients. 46 Additionally, baseline C-reactive protein (a factor known to be associated with immune-related AEs) was 24% higher at the median in patients with high-tTMB than in patients with low-tTMB. 47 No clear imbalance in other prognostic factors were observed apart from TC2/3 expression and CRP.
For this reason, logistic regression for efficacy and safety endpoints were performed using tTMB as the only predictor. in patients with high tTMB and in the entire tTMB-evaluable population with grade 3/4 AEs; there were some numerical differences indicating a slightly higher incidence of all-grade AESIs in the high-tTMB population, but these rates were comparable to those in patients treated with other anti-PD-L1/PD-1 agents. 48 We note that these observations might be related to improved efficacy in the high-tTMB population-slightly higher AESI incidences could normally be expected with longer treatment duration. No relationship between AEs and atezolizumab exposure was seen in the tTMB-high population, consistent with our knowledge of a flat exposure-safety profile of atezolizumab. 32,49 Our exploratory tT-MB-response results agree both with prior analyses that revealed there were higher response rates in patients with higher tTMB tumors and with reports of associations between immune-related AE reporting and tTMB score, 11,50 although other meta-analysis data reporting associations between TMB and efficacy did not find any associations between TMB and toxicity. 51 These observations will be evaluated prospectively in the MyPathway trial (ClinicalTrials. gov ID, NCT02091141). In patients treated with PD-L1 or PD-1 inhibitors, differences in immune-mediated AE frequencies could potentially be driven by T cells reacting to tumor antigens that are cross-reactive against wild-type protein in normal tissue. 52 Lastly, the exploratory tTMB-response assessments are independent of exposure at therapeutic doses based on our flat exposure-response findings.
Our analysis had several limitations in determining the potential for using an anti-PD-1/PD-L1 therapy across a tTMB-high pantumor type indication, these limitations may also be relevant to the broader field. Our analysis was completed retrospectively using existing clinical data. We limited our assessments in the study to the use of objective response, longitudinal SLD change, and DOR to evaluate the potential for TMB as a prognostic utility for ICI therapy.
A more comprehensive evaluation that included progression-free survival and overall survival could provide additional insights into the utility of TMB to guide ICI therapy in future assessments.
Currently, a standardized TMB cutoff that defines high mutational burden in any specific tumor type or across multiple tumor types does not exist. Moreover, differing cutoffs and algorithms are used with the multitude of TMB assessment platforms currently on the market. Determining a cutoff that can capture high mutational burden across a diverse set of tumor types is complex and highly dependent on tumor biology and the platform being used. 53,54 Also, it is possible that TMB values vary between tumor tissues, which could affect ICI treatment stratification. Recent findings report a bias for significantly higher TMB in metastatic tissue than in the primary tumor, although effectiveness in treatment benefit following ICI between both sources was considered comparable. 55 Overcoming these obstacles is vital, and several national and regional US initiatives to harmonize TMB for reliable and reproducible use as a clinical biomarker of response to ICIs are underway. 56,57 Paving the way toward broader use of tTMB will necessitate leveraging data across a combination of clinical trials, flexibility in approaches, and multidisciplinary efforts to further advance tTMB as a diagnostic, therapeutic, and predictive biomarker of ICI benefit. 58 The MyPathway trial will evaluate these prospectively and is adequately powered to address some of the limitations herein.
In summary, this article enhances our knowledge of complex predictive biological factors affecting response to atezolizumab.
The pooled analysis revealed a positive benefit-risk profile with higher response rates and longer duration of response achieved in patients with tTMB-high tumors that supports the use of a 1200mg, every-3-week regimen of atezolizumab in a tumor-agnostic high-tTMB population. Safety was consistent with the known safety profile of atezolizumab. Exposures of atezolizumab were in line with expectations, with no exposure-safety or exposure-efficacy relationships identified. Prospective investigations are warranted to expand the inquiry to larger populations across diverse tumor types.

ACK N OWLED G EM ENTS
The authors thank the patients and their families, without whom this study would not have been possible. The authors also thank the investigators and site staff as well as Dwayne Bracy of Navitas

PR I M A RY L A B O R ATO RY O F O R I G I N
The analyses conducted in this paper are based on pooled data from several studies that were overseen by a number of principal investigators not included as authors given the exploratory nature of these analyses.

DATA AVA I L A B I L I T Y S TAT E M E N T
Qualified researchers may request access to individual patient level data through the clinical study data request platform (https://vivli.