Atezolizumab plus Magrolimab, Niraparib, or Tocilizumab versus Atezolizumab Monotherapy in Platinum-Refractory Metastatic Urothelial Carcinoma: A Phase Ib/II Open-Label, Multicenter, Randomized Umbrella Study (MORPHEUS Urothelial Carcinoma)

Urothelial carcinoma (UC) is the 10th most diagnosed cancer in the world (1). A significant fraction of patients with UC develop metastatic disease and face a poor prognosis, with anticipated survival of only 3 to 6 months if left untreated (2). First-line treatment of metastatic UC (mUC) with platinum-based chemotherapy improves overall survival (OS) to 9 to 15 months, but most patients experience disease progression (2, 3). For these patients, immunotherapy is the standard of care. Second-line treatment with programmed death-ligand 1/programmed death-1 (PD-L1/PD-1) inhibitors leads to durable responses in some patients; however, a large proportion continue to progress (3). Thus, novel therapies are needed to enhance and/or reactivate antitumor immunity to prevent relapse in patients with mUC.

Ongoing research indicates that a series of stepwise events is necessary for the generation of an effective antitumor immune response (4). Each event is critical for an effective response and also susceptible to several evasion mechanisms. The MORPHEUS-UC study was designed to efficiently evaluate novel immunotherapy treatment combinations with the potential to overcome tumor immune evasion and treatment resistance (5). Patients with locally advanced or mUC whose disease progressed following a platinum-containing regimen were randomly assigned to receive the PD-L1 inhibitor atezolizumab in combination with either magrolimab (Hu5F9-G4), a monoclonal antibody that inhibits cluster of differentiation 47 (CD47); niraparib, a selective PARP-1/2 inhibitor; or tocilizumab, a monoclonal antibody that inhibits the IL6 receptor (IL6R). These agents were selected because they target mechanisms of tumor cell killing that are complementary to atezolizumab and thus may provide additive clinical benefit. Alternatively, or in addition, they may synergistically enhance atezolizumab-mediated antitumor effects by altering the tumor immune microenvironment (6–29). Atezolizumab monotherapy served as an internal control.

An adaptive, randomized, platform design was used to minimize the number of patients exposed to control treatment and provide flexibility to add or close treatment arms in response to emerging evidence. Here, we report biomarker, efficacy, safety, and pharmacokinetics results for atezolizumab plus magrolimab, niraparib, or tocilizumab compared with atezolizumab monotherapy in patients with platinum-resistant mUC.

Eligible patients were aged ≥18 years; had histologically documented, locally advanced (T4b, any N; or any T, N2-N3) or mUC (M1, stage IV) measurable by RECIST version 1.1; experienced disease progression during or following treatment with one platinum-containing regimen; and had an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0 or 1. Patients with mixed histologies had to have a dominant transitional cell pattern, and locally advanced bladder cancer had to be inoperable due to involvement of pelvic sidewall or adjacent viscera (clinical stage T4b) or bulky nodal metastasis (N2-N3). A representative tumor specimen suitable for determination of PD-L1 and/or additional biomarker status by central testing was required. Patients were required to have adequate hematologic, liver, and kidney function within 14 days of initiating study treatment as assessed by the following tests: absolute neutrophil, white blood cell, lymphocyte, and platelet counts; hemoglobin level; aspartate transaminase, alanine transaminase, and alkaline phosphatase levels; bilirubin, albumin, and partial thromboplastin levels; prothrombin time; and creatinine clearance (≥30 mL/min, calculated using the Cockroft-Gault formula). Negative serology for HIV, hepatitis B, and hepatitis C was also required at screening. Hematology, chemistry, coagulation, urinalysis, and thyroid function tests were required before each treatment cycle to evaluate patient safety.

MORPHEUS-UC (NCT03869190) is a global, phase Ib/II, open-label, multicenter, randomized, umbrella study in patients with locally advanced or mUC that has progressed during or following a platinum-containing regimen (Supplementary Fig. S1A). The study was conducted in full conformance with the International Conference on Harmonisation E6 guideline for Good Clinical Practice and the principles of the Declaration of Helsinki, or the laws and regulations of the country in which the research is conducted, whichever provided greater protection to the individual. All patients provided written informed consent for study participation. The study protocol, informed consent forms, and any information given to the patients were reviewed and approved by the institutional review board or ethics committee at each site.

MORPHEUS-UC was designed with the flexibility to open new treatment arms as new therapies became available, to close existing treatment arms that demonstrate minimal clinical activity or unacceptable toxicity, or to modify the patient population. Eligible patients were initially assigned to one of several treatment combinations or a control arm (stage 1); those who experienced loss of clinical benefit or unacceptable toxicity were eligible to continue treatment with a different regimen (stage 2). Here, we report stage 1 data from the treatment arms evaluating atezolizumab in combination with magrolimab, niraparib, or tocilizumab along with the control arm. Other treatment combinations and stage 2 data will be reported separately.

The study was designed to enroll approximately 15 patients per arm, with the potential to add additional patients if promising clinical activity was observed. The control arm remained open throughout the study as new treatment arms were added or closed; thus, the number of patients in the control arm was larger than in the experimental arm. Expansion of an experimental arm was gated on observing a clinically meaningful improvement in objective response rate (ORR) in the treatment arm relative to the internal control. The totality of efficacy [progression-free survival (PFS), OS] and safety data were also taken into account. Notably, given the small sample size, the study was designed to identify highly effective combinations with the potential to significantly outperform standard of care. The randomization ratio was dependent upon the number of experimental arms, with the stipulation that the likelihood of being allocated to the control arm was ≤35%; arm-specific exclusion criteria were accounted for during randomization. Patients were treated until unacceptable toxicity or loss of clinical benefit, as determined by the investigator after an integrated assessment of radiographic and biochemical data, local biopsy results (if available), and clinical status.

Baseline tumor measurements were derived from CT or MRI scans of the chest, abdomen, and pelvis (with contrast, unless contraindicated), as well as head and bone scans performed within 28 days of cycle 1 day 1. Tumor assessments were performed every 9 weeks (beginning on cycle 1 day 1) for the first 54 weeks, then every 12 weeks thereafter. Response was assessed by the investigator per RECIST 1.1. Baseline tumor tissue samples were required from all patients. Tumor tissue was also obtained from patients who discontinued treatment because of unacceptable toxicity or loss of clinical benefit, as determined by the investigator and if feasible. Blood samples were collected at various time points before and during study treatment to characterize the pharmacokinetic properties and immunogenicity of atezolizumab and other therapeutic agents, as well as blood-based biomarkers.

In the control arm, atezolizumab was administered at the approved dose of 1,200 mg i.v. on day 1 of each 21-day cycle (30). To achieve comparable drug exposure with 28-day cycles in the atezolizumab plus magrolimab arm, atezolizumab was administered at 840 mg i.v. on days 1 and 15 (31). A priming dose of magrolimab was administered at 1 mg/kg i.v. on cycle 1 day 1, followed by doses of 30 mg/kg i.v. on days 8, 15, and 22 of each 28-day cycle. In cycle 2, patients received magrolimab doses of 30 mg/kg i.v. on days 1, 8, 15, and 22. For all subsequent cycles, magrolimab 30 mg/kg i.v. was administered on days 1 and 15. This dose and schedule were chosen on the basis of the effects observed in nonclinical and clinical studies, pharmacokinetic–pharmacodynamic modeling results, and clinical safety data; a similar dosing scheme has since been used in a phase III clinical trial (NCT0431388; ENHANCE; ref. 32). In the atezolizumab plus niraparib arm, atezolizumab was administered at 1,200 mg i.v. on day 1 of each 21-day cycle, and niraparib was administered at 200 mg/day orally throughout each 21-day cycle. The niraparib dose was selected because it was the most commonly administered dose following dose modification in a phase III clinical trial (NCT01847274; NOVA) and was the recommended phase II dose in combination with a checkpoint inhibitor in patients with advanced triple-negative breast cancer (10, 33). In the atezolizumab plus tocilizumab arm, atezolizumab was administered at 840 mg i.v. on days 1 and 15 of each 28-day cycle, and tocilizumab was administered at 8 mg/kg i.v. on day 1 of each 28-day cycle, the approved dose for the treatment of rheumatoid arthritis (34, 35).

No formal stopping criteria were used in this study, although adverse events (AE) were regularly reviewed by an internal monitoring committee.

The primary efficacy endpoint was investigator-assessed ORR, defined as the proportion of patients with a complete response (CR) or partial response (PR) on two consecutive occasions ≥4 weeks apart, per RECIST 1.1. Key secondary efficacy endpoints included:

• 1. Investigator-assessed PFS, defined as the time from randomization to the first occurrence of disease progression per RECIST 1.1 or death from any cause (whichever occurs first). In the case that no progression or death was observed, PFS was censored at the last tumor assessment.

• 2. OS, defined as the time from randomization to death from any cause.

• 3. Investigator-assessed duration of response (DOR), defined as the time from the first occurrence of a documented objective response to disease progression or death from any cause (whichever occurs first), per RECIST 1.1.

• 4. Investigator-assessed disease control rate (DCR), defined as the proportion of patients with stable disease at ≥18 weeks or a CR or PR, per RECIST 1.1.

Key safety endpoints included the incidence, nature, and severity of AE and laboratory abnormalities, with severity determined according to the NCI Common Terminology Criteria for Adverse Events, version 4.0. The plasma or serum concentration of each agent was assessed at specified time points. Exploratory analyses were conducted to evaluate pharmacokinetics and immunogenicity.

The efficacy-evaluable population included all patients who received one or more dose of each drug for their assigned treatment regimen, and the safety-evaluable population included all patients who received any amount of study treatment. The control arm included all patients assigned to receive control treatment throughout the course of the study until the clinical cutoff date. Results were summarized by the treatment that patients actually received. This study was not designed to make explicit power and type I error considerations for a hypothesis test. Instead, the study was designed to obtain preliminary efficacy, safety, and pharmacokinetic data for atezolizumab treatment combinations in patients with locally advanced or mUC who progressed during or following a platinum-containing regimen.

ORR and DCR were calculated for each arm, along with 95% confidence intervals (CI) using the Clopper-Pearson exact method. Patients with missing or no response assessments were classified as nonresponders. DOR was derived for efficacy-evaluable patients who had a confirmed CR or PR. Median DOR, PFS, and OS were estimated using the Kaplan–Meier method, with 95% CI constructed using the Brookmeyer and Crowley method. For patients who did not have documented disease progression or death, PFS and DOR were censored at the last tumor assessment. If no postbaseline tumor assessment was available, the date of randomization was used instead. Patients alive at the OS analysis cutoff were censored for OS at the last date they were known to be alive. Pharmacokinetic parameters and the presence of antidrug antibodies were evaluated via descriptive statistics.

Decisions regarding further development of a treatment combination were informed by calculating the Bayesian posterior probability of the true difference in ORR between experimental and control arms. If the posterior probability was sufficiently high (e.g., >70%) that the ORR difference was greater than a threshold value (e.g., >10%), additional development could be warranted after taking into account the totality of available data for the specified treatment combination.

Baseline tumor tissue samples were collected at screening, preferably by means of a biopsy performed at study entry. If biopsy was not feasible, archival samples were used. PD-L1 expression was assessed in formalin-fixed, paraffin-embedded (FFPE) tissue using the SP263 IHC assay (Ventana Medical Systems; Tucson, AZ) and scored as tumor-associated positivity, defined as the number of staining cells/total tumor area. A 5% cutoff was used to differentiate PD-L1–positive and –negative tumors.

The presence and spatial distribution of CD8 T cells were assessed in FFPE tumor tissue using a duplex CD8-panCK IHC assay (CD8A/CD8B clone SP239; pan-cytokeratin clone AE1/AE3/PCK26; CellCarta). CD8 immune phenotypes were derived from density proportion scores that assessed the fraction of CD8-positive cells across the tumor area according to eight density bins (four intraepithelial bins, four intratumoral stromal bins).

For analysis of tumor mutational burden (TMB), tumor DNA extraction and preparation were done with CellCarta (Antwerp, Belgium). Foundation Medicine, Inc. (FMI; Cambridge, MA) performed sequencing library construction, hybridization capture, DNA sequencing, and genomic alteration detection. In addition to sample processing, FMI estimated the mutation burden for each sample using an algorithm that leverages genomic alterations detected by the targeted FoundationOne CDx test to extrapolate to the whole exome or genome. TMB was defined as the number of nonsynonymous somatic base substitutions and short insertions and deletions identified from coding regions within the FoundationOne CDx test, filtering out known or probable oncogenic driver mutations to reduce bias. All nonsynonymous mutations, including nonsense, missense, frameshift, splice-site, and nonstop changes, were considered. The resultant count was divided by the size of the genomic region of exonic sequence data interrogated to yield a resultant number of mutations per megabase.

Genes associated with the DNA damage repair (DDR) pathway were previously reported (36); these include BRCA1, BRCA2, RAD50, RAD51, RAD51B, RAD51C, RAD51D, RAD52, RAD54L, NBN, MRE11A, PALB2, BRIP1, FANCA, FANCC, FANCD2, BLM, ATM, ATR, MDC1, CHEK1, CHEK2, MLH1, MSH2, MSH6, PMS1, PMS2, ERCC2, ERCC3, ERCC4, ERCC5, MUTYH, RECQL4, PARP1, POLE, and POLD1. DDR gene alterations were estimated by targeted genomic profiling using the FoundationOne panel (T7 baitset). Samples with alterations (copy-number alterations, fusion/rearrangement and nonsynonymous short variants with known/likely functional impacts) of at least one of the DDR genes listed above were considered positive for DDR alterations.

Genomic loss of heterozygosity (LOH) was assessed with FMI's T5 next-generation sequencing assay. Briefly, LOH segments were inferred across the 22 autosomal chromosomes using the genome-wide aneuploidy/copy-number profile and minor allele frequencies of the >3,500 polymorphic SNPs sequenced in the FoundationOne assay. Using a comparative genomic hybridization-like method, we obtained a log-ratio profile of the sample by normalizing the sequence coverage obtained at all exons and genome-wide SNPs against a process-matched normal control. For each tumor, the percent LOH was computed as 100× the total length of nonexcluded LOH regions divided by the total length of nonexcluded regions of the genome. A prespecified cutoff of 14% or greater was considered to be high LOH.

For up-to-date details on Roche's Global Policy on the Sharing of Clinical Information and how to request access to related clinical study documents, see here: https://go.roche.com/data_sharing.

Between June 2019 and July 2021, 76 patients were randomized to the treatment arms of interest at 20 sites: atezolizumab plus magrolimab (n = 16), atezolizumab plus niraparib (n = 15), atezolizumab plus tocilizumab (n = 15), or atezolizumab control (n = 30; Supplementary Fig. S1B). Two patients discontinued prior to the first administration of study drug due to withdrawal of consent or meeting exclusion criteria for infection.

At the clinical cutoff date (April 8, 2022), 93.3% (n = 14) of patients in the atezolizumab plus magrolimab arm, 100% (n = 15) in the atezolizumab plus niraparib arm, 93.3% (n = 14) in the atezolizumab plus tocilizumab arm, and 82.8% (n = 24) in the atezolizumab control arm had discontinued study treatment. At this time, 68.8% (n = 11) of patients in the atezolizumab plus magrolimab arm, 73.3% (n = 11) in the atezolizumab plus niraparib arm, 80.0% (n = 12) in the atezolizumab plus tocilizumab arm, and 46.7% (n = 14) in the atezolizumab control arm had discontinued from the study (Supplementary Table S1). Death was the most common reason for study discontinuation across treatment arms (43%–80%).

Baseline patient demographics and characteristics are shown in Table 1, and the representativeness of the study population is described in Supplementary Table S2. The treatment arms were relatively well balanced for baseline characteristics; however, due to the small sample size, some differences in patient distribution were observed. Notably, compared with patients in the control arm, patients in the atezolizumab plus tocilizumab arm had less favorable disease presentation at baseline, with a higher median number of metastatic sites at enrollment (2 vs. 1), greater proportion of patients with liver metastases (33.3% vs. 13.3%), and a higher proportion of patients with ECOG PS of 1 (73.3% vs. 56.7%). Furthermore, 13.3% of patients in the atezolizumab plus tocilizumab arm had a Bellmunt score of 3, compared with 0% in all other treatment arms.

Biomarker expression was evaluated in baseline tumor tissue to assess potential predictive or prognostic significance (Fig. 1; Supplementary Fig. S2; Supplementary Fig. S3). PD-L1 status and tumor immune phenotype were evaluated in all patients (Fig. 1A), as these markers are known to be associated with response to atezolizumab (37). Relative to other treatment arms in this study, the atezolizumab plus tocilizumab and atezolizumab control arms had the highest prevalence of PD-L1–positive tumors (75.0% and 61.5%, respectively, vs. 46.2% for atezolizumab plus magrolimab and 50.0% for atezolizumab plus niraparib). A trend toward tumor shrinkage was observed in patients with PD-L1–positive tumors in the atezolizumab plus magrolimab, atezolizumab plus tocilizumab, and atezolizumab control arms but not in the atezolizumab plus niraparib arm (Fig. 1B–E). The atezolizumab control arm included the largest percentage of patients with inflamed tumor immune phenotype (40.0% for control vs. 14.3% for atezolizumab plus magrolimab, 38.5% for atezolizumab plus niraparib, and 27.3% for atezolizumab plus tocilizumab; Fig. 1A). Consistent with prior results, a trend was observed for shrinkage of inflamed tumors treated with atezolizumab monotherapy (Fig. 1E). Interestingly, a trend was also observed for shrinkage of immune-excluded tumors treated with atezolizumab and magrolimab (Fig. 1B).

LOH and DDR mutations (including BRCA1 and BRCA2) were evaluated in the atezolizumab plus niraparib and atezolizumab control arms because of their potential to predict PARP inhibition sensitivity (Fig. 1C and E; ref. 38). TMB was also evaluated because it has been shown to correlate with DDR mutations in some cancers and is known to be associated with increased efficacy of PD-L1 inhibitors (37, 39).

A trend was observed for shrinkage of high TMB tumors in the atezolizumab control arm, consistent with prior work (37). However, a significant fraction of samples could not be evaluated for LOH, DDR, and/or TMB due to samples not meeting assay requirements. The estimated prevalence of LOH and DDR mutations was 16.7% and 13.0%, respectively. None of the samples were positive for known BRCA1 or BRCA2 mutations. The limited number of responses in patients treated with atezolizumab plus niraparib (n = 1) and low biomarker prevalence impaired the ability to draw conclusions about the relationship between PARP inhibition efficacy and tumor mutational status (Fig. 1C). Interestingly, the sole patient who had a prolonged clinical response in the atezolizumab plus niraparib arm did not meet the threshold for LOH, DDR, or TMB positivity.

The relationship between baseline plasma IL6 levels and survival was evaluated in the atezolizumab control and atezolizumab plus tocilizumab arms (Supplementary Fig. S2). Consistent with prior results, a trend for worse survival probability was observed in patients with high plasma IL6 levels at baseline: median OS was 17.8 months in the IL6-high subgroup and not yet reached in the IL6-low subgroup for the atezolizumab monotherapy arm. Contrary to the therapeutic hypothesis being tested in this study, IL6 inhibition failed to improve clinical outcomes in patients with high IL6 levels at baseline: median OS was 5.0 months in the IL6-high subgroup and 8.5 months in the IL6-low subgroup for the atezolizumab plus tocilizumab arm. The less favorable disease presentation observed in this arm may at least partially explain why tocilizumab-treated patients had shorter OS overall than patients in the atezolizumab control arm.

Tocilizumab dosing regimens for oncology patients have not been established. Therefore, pharmacodynamic biomarkers (soluble IL6R and C-reactive protein) were evaluated in the atezolizumab plus tocilizumab arm (Supplementary Fig. S3). Soluble IL6R levels increased between cycle 1 day 1 and cycle 2 day 1 for 11 of 12 patients, indicating successful target engagement. Levels of C-reactive protein, a proinflammatory marker that is downstream of IL6 signaling (21, 40), decreased between cycle 1 day 1 and cycle 2 day 1 for nine of 12 patients, although the magnitude of the decrease was variable, and declines did not clearly correlate with clinical responses.

ORR, DCR, PFS, and OS were not improved with the addition of magrolimab, niraparib, or tocilizumab to atezolizumab. Best confirmed ORR was 26.7% (n = 4; 95% CI, 7.8–55.1) with atezolizumab plus magrolimab, 6.7% (n = 1; 95% CI, 0.2–32.0) with atezolizumab plus niraparib, and 20.0% (n = 3; 95% CI, 4.3–48.1) with atezolizumab plus tocilizumab. None of these values exceeded the ORR of 27.6% (n = 8; 95% CI, 12.7–47.2) observed in the control arm (Table 2; Supplementary Fig. S4). DCRs were 26.7% (n = 4; 95% CI, 7.8–55.1), 13.3% (n = 2; 95% CI, 1.7–40.5), 26.7% (n = 4; 95% CI, 7.8–55.1), and 48.3% (n = 14; 95% CI, 29.5–67.5), respectively. The response rate for patients with liver metastases at baseline was 0% for the atezolizumab plus magrolimab, atezolizumab plus niraparib, and atezolizumab control arms (of two, three, and four patients, respectively) and 20% (one of five patients) for patients in the atezolizumab plus tocilizumab arm.

The median duration of survival follow-up was 17.6 months in the atezolizumab plus magrolimab arm, 7.0 months in the atezolizumab plus niraparib arm, 6.3 months in the atezolizumab plus tocilizumab arm, and 13.0 months in the atezolizumab control arm (Supplementary Table S1). PFS data were mature, with 79% to 93% of patients reporting an event. Median PFS was 2.3 months (95% CI, 2.0–10.5) in the atezolizumab plus magrolimab arm, 2.2 months (95% CI, 2.0–3.0) in the atezolizumab plus niraparib arm, and 4.1 months (95% CI, 2.3–8.3) in the atezolizumab plus tocilizumab arm (Fig. 2A–C). In the atezolizumab control arm, median PFS was 6.1 months (95% CI, 2.6–10.2; Fig. 2). OS data were mature (>70% event rate) for the atezolizumab plus tocilizumab treatment arm, with lower event rates for all other arms (53.3%, 60%, and 41.4% for atezolizumab plus magrolimab, atezolizumab plus niraparib, and atezolizumab control arms, respectively). Median OS was 21.8 months [95% CI, 7.2–not estimable (NE)] in the atezolizumab plus magrolimab arm, 8.6 months (95% CI, 5.2–NE) in the atezolizumab plus niraparib arm, and 6.3 months (95% CI, 5.1–19.9) in the atezolizumab plus tocilizumab arm (Fig. 3A–C). Median OS was not reached (95% CI, 10.6–NE) in the atezolizumab control arm (Fig. 3).

Safety results, including treatment exposures, are summarized in Table 3. Patients in the atezolizumab control arm completed more treatment cycles than patients in any of the combination treatment arms (median eight cycles vs. three to four cycles, respectively). Nearly all patients (93.3%–100%) had at least one AE regardless of causality (Table 3). The most common AE (occurring in ≥10% of patients in any study arm) are reported in Supplementary Table S3. Grade 3/4 AE occurred in 66.7% (n = 10) of patients in the atezolizumab plus magrolimab arm, 60.0% (n = 9) of patients in the atezolizumab plus niraparib arm, 53.3% (n = 8) of patients in the atezolizumab plus tocilizumab arm, and 17.2% (n = 5) of patients in the atezolizumab control arm (Table 3); the majority of these events were grade 3. Grade 3/4 events were considered to be related to study treatment in 60.0% (n = 9) of patients in the atezolizumab plus magrolimab arm, 40.0% (n = 6) of patients in the atezolizumab plus niraparib arm, 13.3% (n = 2) of patients in the atezolizumab plus tocilizumab arm, and 3.4% (n = 1) of patients in the atezolizumab control arm. AE leading to any treatment discontinuation occurred in 20.0% (n = 3) of patients in the atezolizumab plus magrolimab arm, 13.3% (n = 2) of patients in the atezolizumab plus niraparib and tocilizumab arms, and no patients in the atezolizumab control arm (Supplementary Table S4). Serious AE occurred in 53.3% (n = 8) of patients in the atezolizumab plus magrolimab arm, 46.7% (n = 7 each) of patients in the atezolizumab plus niraparib and tocilizumab arms, and 31.0% (n = 9) of patients in the atezolizumab control arm. These events were treatment related in 33.3% (n = 5) of patients in the atezolizumab plus magrolimab arm, 26.7% (n = 4) of patients in the atezolizumab plus niraparib arm, and 3.4% (n = 1) of patients in the atezolizumab control arm; no treatment-related serious AE were reported in the atezolizumab plus tocilizumab arm (Table 3; Supplementary Table S5). One patient in the atezolizumab plus niraparib arm experienced a grade 5 AE of cardiorespiratory arrest (Supplementary Table S6). This patient had multiple cardiac risk factors.

Treatment-related AE led to dose modification or interruption in 40.0% (n = 6) of patients in the atezolizumab plus magrolimab arm, 33.3% (n = 5) in the atezolizumab plus niraparib arm, 26.7% (n = 4) in the atezolizumab plus tocilizumab arm, and 10.3% (n = 3) in the atezolizumab control arm. Treatment-related AE led to withdrawal from any treatment in 20.0% (n = 3) of the atezolizumab plus magrolimab arm and 13.3% (n = 2) of the atezolizumab plus niraparib arm.

Serum or plasma concentrations of each agent and the incidences of antidrug antibodies to atezolizumab or magrolimab are presented in Supplementary Table S7. Atezolizumab exposure was comparable across treatment arms and between regimens. Overall, findings were considered appropriate based on study protocol specifications and consistent with the clinical experience of the molecules to date.

The randomized phase Ib/II MORPHEUS-UC umbrella study was designed to accelerate the development of drug combinations that enhance the efficacy of immunotherapy in patients with mUC. Across treatment arms, no additive benefit was observed relative to the atezolizumab monotherapy control arm. These clinical results should be interpreted with caution, as this study was not powered for efficacy comparisons between arms, and the number of patients in each arm was small. In general, the treatment combinations were tolerable. Although a higher number of AE and serious AE were seen in the experimental arms compared with the control arm, the reported AE were consistent with the known safety profile of each agent. For atezolizumab, magrolimab, and niraparib, the doses used in this study were either approved by health authorities or based on clinical experience in phase II or III oncology trials. In contrast, the dose used for tocilizumab was based on the approved dose for rheumatoid arthritis because the optimal dosing regimen for oncology has not been established. Further dose exploration, balanced with safety and tolerability concerns, could be warranted as limited clinical activity was observed with atezolizumab plus tocilizumab in the current study.

mUC is associated with heterogeneous genetic mutations and diverse immunological responses to the presence of cancer cells. Accordingly, biomarker analyses were undertaken to generate hypotheses about specific subgroups that could potentially benefit from novel treatment combinations. Consistent with prior work (37), patients with PD-L1–positive, inflamed, or TMB-high tumors benefited from treatment with atezolizumab monotherapy. A trend was also observed for shrinkage of PD-L1–positive tumors treated with atezolizumab plus magrolimab or atezolizumab plus tocilizumab, which is likely to be at least partially explained by atezolizumab efficacy in this population. Interestingly, a trend was observed for increased efficacy of atezolizumab plus magrolimab in patients with immune-excluded tumors. Future work should evaluate whether enrichment of CD47 and/or tumor-associated macrophages could account for the apparent efficacy of PD-L1 and CD47 inhibition in this tumor type. Due to the small sample size, biomarker results require confirmation in a larger study.

PARP inhibitors have shown increased efficacy in patients with ovarian, breast, or prostate cancer whose tumors harbor DDR mutations, especially in BRCA1 and BRCA2 (38). However, for patients with UC, the biomarker(s) predictive of response to PARP inhibitors remain uncertain. We evaluated LOH, DDR mutations, and TMB in the current study to evaluate whether these markers could predict response to treatment with atezolizumab and niraparib. Of note, the prevalence of LOH and DDR mutations observed in the current study (16.6% and 13.0%, respectively) was comparable with prior observations (37, 41). Regrettably, limited clinical responses, low biomarker prevalence, and high assay failure rate made it difficult to evaluate predictive biomarkers in this population. These data are consistent with results from the ATLAS and BISCAY studies, which found limited clinical activity of PARP inhibitors in previously treated patients with mUC that did not appear to be correlated with LOH or DDR gene alterations (41, 42). However, secondary analyses from the BAYOU trial identified a potential role for PARP inhibition in untreated platinum-ineligible patients with mUC whose tumors contain homologous recombination repair gene mutations (43). It is unclear whether differences in patient population, investigational agents, specific genomic alterations defining biomarker-selected populations, or other factors explain this discrepancy.

Biomarker analyses revealed imbalances in the prevalence of PD-L1–positive and inflamed tumors in the atezolizumab control arm compared with combination treatment arms in this study. As anti–PD-L1 agents are known to have enhanced efficacy in these biomarker subpopulations, these differences may have influenced clinical outcomes in the current study. Notably, the atezolizumab control arm in this study substantially outperformed historical expectations. Compared with the phase III IMvigor211 trial conducted in the same patient population (44), best confirmed investigator-assessed ORR with atezolizumab monotherapy was approximately two times higher in this study (27.6% vs. 13.4%). Discrepancies in clinical outcomes between the internal control and historical data may partly be explained by differences in the prevalence of predictive biomarkers such as PD-L1 and tumor immune phenotype. Alternatively or in addition, differences in demographics or baseline characteristics may explain the superior efficacy observed with atezolizumab treatment in the current study.

These challenges highlight several limitations of the study. The small number of patients hinders interpretation in cases where differences emerge with historical data from larger trials, as it is difficult to determine whether an intratrial comparison with small sample sizes or an intertrial comparison with large sample sizes more closely approximates ground truth. As the study was designed to identify highly effective combinations, subtle benefits over standard of care could be missed with this approach when the number of patients per arm is small. The lack of stratification led to some imbalances in baseline characteristics and tumor biomarkers, which impede data interpretation. In addition, while treatment was randomly assigned, the study was not powered to evaluate clinical benefit in specific subpopulations (e.g., biomarker-defined populations) or for direct efficacy comparison between study arms.

Strengths of this study include its flexibility and controlled design, as well as substantial correlative biomarker analyses. The adaptability of the trial keeps the study relevant despite a changing treatment landscape. Enrollment of multiple experimental arms within one study, rather than across several, is an efficient way to test hypotheses in parallel and results in a reduction in the number of patients receiving control arm treatment. Further, the inclusion of a randomized internal control potentially improves the confidence of clinical signal detection in the experimental arms.

In conclusion, although the evaluated treatment combinations were generally tolerable, response rates in MORPHEUS-UC did not meet efficacy criteria for further development of these regimens in platinum-experienced patients with locally advanced or mUC.

A. Drakaki reports nonfinancial support and other support from UCLA during the conduct of the study and other support from Seagen, AstraZeneca, and Merck outside the submitted work. T. Powles reports personal fees from AstraZeneca, BMS, Exelixis, Incyte, Ipsen, MSD, Novartis, Pfizer, Seattle Genetics, Merck Serono, Astellas, Johnson & Johnson, Eisai, Roche, and MashUp Ltd.; grants from AstraZeneca, Roche, BMS, Exelixis, Ipsen, MSD, Novartis, Pfizer, Seattle Genetics, Merck Serono, Astellas, Johnson & Johnson, and Eisai; other support from Roche, Pfizer, MSD, AstraZeneca, and Ipsen outside the submitted work. A. Bamias reports other support from Roche during the conduct of the study and personal fees from Roche outside the submitted work. J. Martin-Liberal reports personal fees from Astellas; grants and personal fees from Bristol Myers Squibb, MSD, Novartis, Pierre Fabre, Pfizer, and Roche; personal fees from Sanofi and Highlight Therapeutics; and grants from Ipsen outside the submitted work. T. Friedlander reports personal fees from Astellas and Seagen outside the submitted work. D. Tosi reports other support from Roche during the conduct of the study. C. Gomez-Roca reports personal fees and other support from BMS and Roche/Genentech and personal fees from Macomics, Pharmamar, and Pierre Fabre outside the submitted work. F. Joly Lobbedez reports personal fees from Pfizer and Astellas outside the submitted work and travel support from Ipsen. D. Castellano reports other support from Pfizer, Janssen, Astellas, BMS, Roche, Ipsen, MSD, Merck, Gilead, AstraZeneca, Eisai, and Novartis outside the submitted work. I. Moreno-Candilejo reports personal fees from Ellipses Pharma Advisory Group outside the submitted work. A. Fléchon reports personal fees from Roche and personal fees and other support from Pfizer, Merck, Astellas, Janssen, and Gilead outside the submitted work. K. Yuen reports other support from Genentech, Inc. during the conduct of the study. K. DuPree reports employment by Roche/Genentech (i.e., salary and stock offerings). F. Young reports employment by Roche. F. Michielin reports personal fees from Roche during the conduct of the study and personal fees from Roche outside the submitted work. C.S. Shemesh reports other support from Genentech, Inc. outside the submitted work. E.E. Steinberg reports personal fees from Roche/Genentech during the conduct of the study. P. Williams reports personal fees from Genentech during the conduct of the study and has a patent for PCT/US2020/021738 pending to Genentech. J.L. Lee reports grants from Roche and Genentech during the conduct of the study; grants from Pfizer, Ipsen, Bristol Myers Squibb, MSD, Amgen, Seagen, and GI Innovation; grants and personal fees from Merck, AstraZeneca, Novartis, and Janssen; and personal fees from Astellas outside the submitted work. No disclosures were reported by the other authors.

A. Drakaki: Writing–review and editing. T. Powles: Conceptualization, resources, validation, investigation, visualization, writing–review and editing, concept and design, data analysis and interpretation, provision of study materials or patients, manuscript review and editing, final approval of manuscript. A. Bamias: Resources, data curation, validation, investigation, visualization, writing–review and editing, provision of study materials or patients, collection and assembly of data, manuscript review and editing, final approval of manuscript. J. Martin-Liberal: Resources, writing–review and editing, provision of study materials or patients, manuscript review and editing, final approval of manuscript. S.J. Shin: Resources, writing–review and editing, provision of study materials or patients, manuscript review and editing, final approval of manuscript. T. Friedlander: Conceptualization, resources, data curation, validation, investigation, visualization, writing–review and editing, concept and design, data analysis and interpretation, provision of study materials or patients, collection and assembly of data, manuscript review and editing, final approval of manuscript. D. Tosi: Resources, writing–review and editing, provision of study materials or patients, manuscript review and editing, final approval of manuscript. C. Park: Writing–review and editing, manuscript review and editing, final approval of manuscript. C. Gomez-Roca: Data curation, validation, investigation, visualization, writing–review and editing, data analysis and interpretation, collection and assembly of data, manuscript review and editing, final approval of manuscript. F. Joly Lobbedez: Resources, writing–review and editing, provision of study materials or patients, manuscript review and editing, final approval of manuscript. D. Castellano: Resources, data curation, validation, investigation, visualization, writing–review and editing, provision of study materials or patients, manuscript review and editing, final approval of manuscript. R. Morales-Barrera: Writing–review and editing. I. Moreno-Candilejo: Writing–review and editing. A. Fléchon: Writing–review and editing. K. Yuen: Data curation, validation, investigation, visualization, writing–review and editing, data analysis and interpretation, collection and assembly of data, manuscript review and editing, final approval of manuscript. D. Rishipathak: Conceptualization, resources, data curation, validation, investigation, visualization, writing–review and editing, data analysis and interpretation, manuscript review and editing, final approval of manuscript. K. DuPree: Conceptualization, resources, data curation, validation, investigation, visualization, writing–review and editing, concept and design, data analysis and interpretation, provision of study materials or patients, collection and assembly of data, manuscript review and editing, final approval of manuscript. F. Young: Writing–review and editing. F. Michielin: Data curation, validation, investigation, visualization, writing–review and editing, data analysis and interpretation, collection and assembly of data, manuscript review and editing, final approval of manuscript. C.S. Shemesh: Validation, investigation, visualization, writing–review and editing, data analysis and interpretation, manuscript review and editing, final approval of manuscript. E.E. Steinberg: Conceptualization, data curation, validation, investigation, visualization, writing–review and editing, data analysis and interpretation, manuscript review and editing, final approval of manuscript. P. Williams: Conceptualization, resources, data curation, validation, investigation, visualization, writing–review and editing, concept and design, data analysis and interpretation, collection and assembly of data, manuscript review and editing, final approval of manuscript. J.L. Lee: Resources, validation, investigation, visualization, writing–review and editing, data analysis and interpretation, provision of study materials or patients, manuscript review and editing, final approval of manuscript.

F. Hoffmann-La Roche Ltd. sponsored this study and was involved in the study design, collection, analysis, and interpretation of data, as well as the writing of the report. Authors had full access to all data, and the corresponding author had final responsibility for the decision to submit the results for publication.

The authors would like to acknowledge the patients and their families, investigators, and clinical study sites for the MORPHEUS-UC trial. Medical writing assistance for this report was provided by Kia C. E. Walcott, PhD, of Health Interactions, Inc., funded by F. Hoffmann-La Roche Ltd.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.