Trabectedin plus Durvalumab in Patients with Advanced Pretreated Soft Tissue Sarcoma and Ovarian Carcinoma (TRAMUNE): An Open-Label, Multicenter Phase Ib Study

About 70% of patients with cancer do not exhibit clinical benefit from single-agent programmed cell death protein 1/programmed death-ligand 1 (PD-1/PD-L1) inhibitors. Combining these agents with other therapeutic approaches may impact favorably the tumor microenvironment and could improve antitumor immune response and overcome primary resistance. Cytotoxic agents can have immunomodulatory effects which may represent opportunities of combination with immunotherapy.

Trabectedin is an alkylating agent which binds the minor groove of DNA and has effect on transcription regulation and DNA repair. This drug is approved in Europe and in the United States as a single agent for the management of patients with advanced L–soft tissue sarcoma (STS), such as leiomyosarcoma and liposarcoma (1, 2), and in Europe in combination with liposomal doxorubicin for patients with advanced ovarian carcinoma (3). Immune checkpoint blockers have shown modest efficacy in these two tumor types (4, 5), possibly as the result of the immunosuppressive action of tumor-associated macrophages (TAM; refs. 4, 6–9).

Trabectedin has an impact on tumor microenvironment, notably through inhibition of the production of cytokines such as CCL2 and IL6, as well as upregulation of expression of PD-1/PD-L1 in vivo, and a selective cytotoxic activity to human monocytes and TAM, in vivo and in treated patients (10). Preclinical data also suggest a strong synergy between trabectedin and anti–PD-1/PD-L1 antibodies in ovarian and sarcoma models, associated with an increase of intratumoral effector T-cell infiltrates and a decrease of immunosuppressive cell infiltrates in these models (11, 12). This led us to assess the combination of trabectedin with the anti–PD-L1 durvalumab in advanced or metastatic pretreated STS and ovarian carcinoma.

TRAMUNE is an open-label, multicenter, single-arm phase Ib trial combining a dose escalation phase, assessing three dose levels of trabectedin combined with durvalumab, followed by two expansion cohorts, in advanced or metastatic pretreated STS and ovarian carcinoma.

Main inclusion criteria were patients ≥18 years with: (i) histologically confirmed and centrally reviewed STS or ovarian carcinoma, (ii) no known germline or somatic BRCA mutation for ovarian carcinoma, (iii) locally advanced or metastatic disease, (iv) at least one line of chemotherapy in the palliative setting, including anthracyclines for STS and platinum salts for ovarian carcinoma, (v) measurable and progressive disease at inclusion according to RECIST v1.1, (vi) a performance status of 0–1, (vi) adequate hematologic, renal, metabolic, and hepatic functions. Complete inclusion criteria are given in the online protocol. All patients provided written informed consent. The trial was done in accordance with Good Clinical Practice and the Declaration of Helsinki. This study was approved by the Institutional Review Board of Institut Bergonié (Bordeaux, France) and is registered on ClinicalTrials.gov (NCT03085225).

The dose escalation phase followed a 3 + 3 classical design. In the dose escalation phase, three dose levels of trabectedin were assessed, 1 mg/m², 1.2 mg/m², and 1.5 mg/m² over 3 hours on day 1 after one single administration of dexamethasone 20 mg 30 minutes before infusion, in combination with durvalumab at fixed dose, 1,120 mg on day 2, every 3 weeks. In the expansion cohorts, patients were all treated at the MTD defined in the dose escalation phase. Treatment continued until disease progression assessed by local investigator, or unacceptable toxicity. Dose reductions were permitted, according to the protocol. Safety was assessed according to the NCI Common Terminology Criteria for Adverse Events version 4.03. Efficacy was assessed as per RECIST criteria version 1.1 every 6 weeks and after independent blinded central review of imaging.

The primary endpoint of the dose escalation phase was to assess safety and determine the dose-limiting toxicity, the MTD, and the recommended phase II dose of trabectedin combined with durvalumab. The primary endpoint of the expansion phase was to assess preliminary activity of trabectedin combined with durvalumab in terms of objective response rate (ORR), defined as the proportion of patients with complete or partial response as per RECIST v1.1 from study treatment initiation until its end. Secondary endpoints were best overall response, 3- and 6-month progression-free rate (PFR), 1-year progression-free survival (PFS) and 1-year overall survival (see Supplementary Methods for additional details).

Sequential tumor biopsies were performed at baseline and at cycle 2 to correlate tumor microenvironment features with patient outcome (see Supplementary Methods for additional details).

Dose-limiting toxicity was defined as an adverse event occurring during the first 21 days of treatment, at least possibly related to study treatment, and meeting one of these criteria: any grade 4 toxicity, grade 3 toxicity lasting >7 days, grade 4 neutropenia with fever, or grade >2 thrombocytopenia with bleeding. A minimum of 3 and a maximum of 6 patients were to be included in each dose level. The MTD of trabectedin was defined as the highest dose at which no more than 1 in 6 patients experienced a dose-limiting toxicity during the observation period of the first 21 days. A maximum of 20 patients were planned to be included.

Two independent expansion cohorts were planned at the recommended dose of trabectedin combined with durvalumab: cohort A in STS and cohort B in ovarian carcinoma. Each expansion cohort followed the first stage of a 2-stage Gehan design (13), assuming a 20% efficacy rate, a 5% false positive rate, and a 10% precision. Fourteen eligible and assessable patients were required in each cohort, with at least one objective response needed to consider the combination active. Fifteen patients were planned to be recruited in each cohort (see Supplementary Methods for additional details).

Complete study protocol is available online in the supplementary material. Qualified researchers may request access to individual patient level data by contacting the corresponding author.

Between October 2017 and November 2019, 40 patients were included in two centers, 9 patients in the dose escalation phase, 16 patients in the STS expansion cohort and 15 in the ovarian carcinoma expansion cohort (Fig. 1). Patient's characteristics are detailed in Table 1. Median number of previous lines was 3 (1–6) in the dose escalation phase, 1 (0–4) in the STS expansion cohort, and 3 (1–7) in the ovarian carcinoma expansion cohort (Fig. 1). Most patients with ovarian carcinoma were platinum–resistant/refractory (relapse less than 6 months after or progression during previous platinum-based therapy): 100% and 80% in the dose escalation phase and the expansion cohort, respectively. None had a known BRCA germline or somatic mutation.

In the dose escalation phase, 3 patients were treated at dose level 1 and 6 patients at dose level 2. There was one dose-limiting toxicity observed at dose level 2, a grade 4 alanine aminotransferase (ALAT) increase. The most frequent toxicities reported in the dose escalation phase were grade 1–2 fatigue, nausea, myalgia, neutropenia, and aspartate aminotransferase (ASAT) or ALAT increase. Eight patients had at least one grade 3–4 treatment-related adverse event, detailed on Supplementary Table S1. One [11%; 95% confidence interval (CI), 0.3–48.3] patient with ovarian carcinoma experienced a complete response at dose level 2 (Supplementary Table S2). This patient had received four lines in the platinum-refractory metastatic setting. Trabectedin 1.2 mg/m² was considered the MTD in combination with durvalumab 1,120 mg every 3 weeks, and the two dose expansion cohorts were opened at this dosing.

Thirty patients from the expansion cohorts were eligible for the safety analysis. 19 (63%) patients experienced a grade 3–4 and two (7%) a grade 5 treatment-related adverse events (Supplementary Table S3). There were eleven treatment-related serious adverse events reported, including two fatal febrile neutropenia: one in a pretreated STS patient who developed Candida pneumonitis and the other in a patient with pretreated ovarian carcinoma who developed Staphylococcus septicemia.

Median follow-up was 19.3 months (8–20). In the STS expansion cohort, the median number of cycles was 3 (1–27). Five patients discontinued the study for treatment-related toxicity: two for grade 2 and grade 3 left ventricular systolic dysfunction respectively, one for grade 3 thrombocytopenia, one for grade 4 CPK increase, and one for grade 5 febrile neutropenia. Among the 14 patients assessable for efficacy endpoints, 6 (43%) patients experienced tumor shrinkage, resulting in one partial response in a patient with leiomyosarcoma (Fig. 2A). The ORR was 7% (95% CI, 0.2–33.9; Supplementary Table S2). Eight (57%) patients had stable disease as their best response (Fig. 2B). The 3- and 6-month PFRs were 35.7% (95% CI, 12.8–64.9) and 28.6% (95% CI, 8.4–58.1), respectively. The 1-year PFS rate was 14.3% (95% CI, 2.3–36.6; Fig. 2C) and the 1-year overall survival rate was 56.3% (95% CI, 27.2–77.6).

In the ovarian carcinoma expansion cohort, the median number of cycles was 4 (1–18). One (7%) patient discontinued the study for treatment-related toxicity, which was a grade 5 febrile aplasia. Among the 14 patients eligible for efficacy endpoints, 6 (43%) patients experienced tumor shrinkage, resulting in three partial responses (Fig. 2D). The ORR was 21.4% (95% CI, 4.7–50.8; Supplementary Table S2). Five (36%) patients had stable disease as their best response (Fig. 2E). The 3- and 6-month PFRs were 42.9% (95% CI, 17.7–71.1) and 42.9% (95% CI, 17.7–71.1), respectively. The 1-year PFS rate was 7.1% (95% CI, 0.5–27.5; Fig. 2F) and the 1-year overall survival rate was 57.1% (95% CI, 28.4–78).

In the STS ancillary cohort, 20 and 13 patients had available tumor material from baseline and C2D8 biopsies, respectively. Four (20%) patients had expression of PD-L1 on tumor cells at baseline and 6 (46%) patients at C2D8. Baseline samples demonstrated a higher median density of CD163-positive cells than CD8-positive cells (Supplementary Table S4). Assessment of coexpression of CD163-positive and CD8-positive cell densities at baseline allowed to classify tumors into three categories according to different patterns of expression: CD8-low and CD163-low “immune desert” tumors, CD8-positive > CD163-positive “inflamed” tumors, and CD163-high “TAM-enriched” tumors (Supplementary Fig. S1). The two most progressive patients had a CD163-high “TAM-enriched” profile at baseline, and the only patient with a confirmed partial response had a PD-L1 positive and CD8-positive > CD163-positive “inflamed” tumor profile at baseline (Supplementary Fig. S1). There was no correlation between CD163-positive or CD8-positive density variations and best overall response (Supplementary Fig. S2). We observed a trend for longer median PFS in patients with baseline CD163-positive and CD8-positive cell density respectively below and above the median (Supplementary Fig. S3), although not statistically significant.

In the ovarian carcinoma ancillary cohort, 16 and 9 patients had available material from baseline and C2D8 biopsies, respectively. Six (37.5%) patients had expression of PD-L1 on tumor cells at baseline, and 4 (44.5%) patients at C2D8. Baseline samples demonstrated a higher median density of CD163-positive cells than CD8-positive cells (Supplementary Table S2). Assessment of coexpression of CD163-positive and CD8-positive cell densities allowed to identify a predominance of a CD163-high “TAM-enriched” profile among tumors at baseline and at C2D8 (Supplementary Fig. S4). The two responding patients with available material at baseline had a PD-L1 positive and CD8-positive > CD163-positive “inflamed” tumor profile (Supplementary Fig. S4). The patient with a complete response had no exploitable material but had consented to genomic profiling that retrieved a p53 mutation, a NF1 duplication, no microsatellites instability and a tumor mutational burden of 4/Mb. There was no correlation between CD163-positive or CD8-positive cell density variations and best overall response (Supplementary Fig. S2). We observed a significantly longer PFS for patients with ovarian carcinoma with PD-L1–positive tumor at baseline [6.8 months (95% CI, 0.9–13.7) vs. 1.3 months (95% CI, 0.9–4.3); log-rank test, P = 0.040, α = 5%; Supplementary Fig. S5] and with CD8-positive cell density above the median at baseline [5.7 months (95% CI, 1.2–9.9) vs. 1.2 months (95% CI, 0.9–1.3); log-rank test, P = 0.018, α = 5%; Supplementary Fig. S5], whereas baseline value or variation of CD163-positive cell density had no impact.

This is the first full report of a study investigating the combination of trabectedin with an immune checkpoint inhibitor in solid tumors. In this study of heavily pretreated patients, there were two toxic deaths, and, 20% and 13% of grade 3 and grade 4 neutropenia, respectively. We also observed 27% and 30% of grade 3 ASAT and ALAT increase, respectively, as well as 3% of grade 4 ALAT increase. Trabectedin toxicity in monotherapy has been extensively assessed in STS (1, 14–21). In the trials from Demetri and colleagues and Bui and colleagues, the discontinuation rates for toxicity were 6% and 20%, respectively, and included one fatal septic shock, one grade 3 LVEF drop, neutropenia and CPK increase. Grade 3–4 adverse event rates were similar to those retrieved in the this study, with 46% grade 3–4 neutropenia reported by Bui and colleagues with the 3-hour schedule at 1.3 mg/m², and 26% grade 3 and 21% grade 4 neutropenia reported by Demetri with the classical 24-hour infusion schedule at 1.5 mg/m² (15, 16). In the phase III from Blay and colleagues, 55% of the patients experienced grade 3–4 neutropenia and 49% received G-CSF. There was one fatal rhabdomyolysis (14). In the phase III from Demetri and colleagues, 43% of the patients had received more than two previous lines, and 13% discontinued treatment for toxicity. There were seven deaths related to trabectedin, mostly due to septic shock, rhabdomyolysis, and renal failure (1). Finally, in an expanded access program, 555 (31%) among 1,803 patients assessable for safety had a grade 3–4 adverse event and 23 deaths were related to trabectedin (20). Trabectedin has also been assessed in monotherapy in ovarian carcinoma. Krasner and colleagues reported a rate of grade 3–4 adverse event of 39% and one fatal left cardiac failure related to study drug (22), whereas Del Campo and colleagues reported 11% and 26% of grade 3 and 4 neutropenia, as well as two drug-related deaths from multiorgan failure with the 3-hour infusion schedule at 1.3 mg/m² (23).

The increase in ASAT/ALAT and a potential overlap in liver toxicity between the two drugs was a subject of concern in the protocol, and there was prespecified weekly monitoring of liver tests in both escalation and expansion phases. In our pooled expansion cohort analysis of patients treated with a 3-hour infusion at 1.2 mg/m², the rate of grade 3–4 ASAT/ALAT increase was in the range of previously published trials with the monotherapy, keeping in mind that an increase in liver toxicity has been observed with the 3-hour schedule compared with the 24-hour schedule in the literature. The trial by Bui and colleagues assessed trabectedin in STS given as 3-hour or 24-hour infusion. In the 3-hour infusion arm at 1.3 mg/m², the grade 3–4 ASAT and ALAT increase rates were 35% and 67%, whereas they were 22% and 49% in the 24-hour infusion arm (15). In the 3-hour infusion arm at 1.3 mg/m² from the trial by Del Campo and colleagues, the rates of grade 3 ASAT and ALAT increase were 17% and 53%, and grade 4 ASAT and ALAT increase 2% and 6%, respectively (23). In our study, these increases occurred during the 3-week interval between two cycles and were rapidly recovering, as typically retrieved with trabectedin. There was no immune-related hepatitis reported. Hematologic and liver toxicities observed in this study were therefore considered related to trabectedin use in a population of patients with pretreated metastatic disease. Discontinuation rate was however higher in the STS cohort than previously reported. Of note, 2 of 5 patients who discontinued treatment for treatment-related toxicity in this cohort had left ventricular dysfunction that was possibly related to trabectedin, in the context of previous use of anthracyclines, antiangiogenics, and preexisting cardiopathy. They had left ventricular ejection fraction drop that recovered with adequate management. There was no myocarditis reported. There was no apparent signal for an unexpected or significant additional toxicity related to the combination with durvalumab.

ORR and PFS rate we observed in the sarcoma expansion cohort were comparable with the ones reported in trabectedin single-agent studies, and there was no clear signal of synergistic activity in this unselected population (1, 14–16). Of note, preliminary data from the SAINT study assessing ipilimumab nivolumab and trabectedin at 1.2 mg/m² as first line in advanced STS have been reported at ASCO 2020. This phase II reached a promising ORR of 19.5% and a 6-month PFS rate of 50% with a good toxicity profile. Sarcoma heterogeneity and random overrepresentation of “cold” histologies and phenotypes in small samples from phase II immunotherapy trials have been reported previously, and can explain the results from this study (24).

There is no reliable biomarker identified yet to select for trabectedin efficacy. Some data suggest that trabectedin inhibits the production of cytokines such as CCL-22 and IL6 and decrease monocytes and TAM infiltrates, in vivo and in a small series of sarcoma patient samples (10, 25, 26). In this study, we observed high CD163-positive cell density at baseline. However, we did not observe a decrease in CD163-positive cell density under treatment, neither a relationship between CD163-positive cell density and response or PFS. PD-L1 expression and CD8-positive cell density were low in this unselected STS cohort, presuming rather “cold” tumors (27, 28). One explanation for the lack of synergy observed in this study may be the low rate of inflamed tumors at baseline. Main clinical research activity in STS is focusing on determining biomarkers to select the patients the more likely to respond to immunotherapy. In this regards, recent advances have been made, notably with the identification of tertiary lymphoid structures (TLS) as surrogate markers for PD-1/PD-L1 inhibition efficacy in STS (27). Assessing trabectedin and durvalumab in a more homogeneous selected population of TLS-positive STS may represent a relevant approach.

The vast majority of patients with ovarian carcinoma treated in our study had heavily pretreated platinum-resistant or platinum-refractory tumor, a setting of very poor prognosis with low response rates to salvage chemotherapy regimen and poor outcome (29–31). Patients with ovarian carcinoma considered as resistant or refractory to platinum have indeed low response rates (4%–23%) to salvage chemotherapy regimen, including pegylated liposomal doxorubicin, trabectedin, paclitaxel, topotecan or gemcitabine, and poor outcome (29, 30). Trabectedin alone reached an ORR of 18.2% and a median PFS rate of 3 months in this specific population (32). Trabectedin has also been assessed in combination with pegylated liposomal doxorubicin (3, 33). The combination showed no increase in median PFS versus pegylated liposomal doxorubicin alone in platinum–resistant/refractory patients (4 months vs. 3.7 months, respectively), leading to consider the combination not indicated in this population (3).

Although ovarian carcinoma has proven to be immunogenic (34, 35), no immunotherapy is approved to date and this holds true especially in the platinum-resistant setting. The tumor microenvironment of platinum-refractory ovarian carcinoma has been described as “cold”, with low infiltration of CD8-positive T cells (36), but also increased PD-L1 positivity, associated with poor prognosis (5, 31, 37, 38). Studies have also shown that platinum salts induce differentiation of macrophages into an M2 phenotype through secretion of IL6 and PGE2 by platinum-treated cells (39). TAM contribute to tumor progression in ovarian cancer models (40), and TAM targeting in combination with PD-1 checkpoint inhibitor has shown activity in several tumor models, including a murine model of ovarian cancer (11, 41). However, the first studies investigating anti–PD-L1/anti–PD-1 antibodies used as single agent in ovarian carcinoma showed response rates of 5% to 15% (5, 37, 42), and platinum sensitivity did not seem to impact response rates, from 5.6% in platinum-sensitive to 7.8% in platinum-resistant patients in the KEYNOTE trial. Combination of immune checkpoint inhibitors plus chemotherapy such pegylated liposomal doxorubicin has been specifically assessed in platinum-resistant ovarian carcinoma in a phase II and a phase III trials. In the phase II, 77% of patients had received two or less previous lines. The ORR of the combination was 26.1% (95% CI, 10.2–48.4), and the 6-month PFS reached a promising rate of 48.5% (95% CI, 27.1–66.9%; ref. 43). In the phase III conducted in a more heavily pretreated population, the combination arm led to an ORR of 13.3% (95% CI, 8.8–19), and a median PFS of 3.7 months versus 3.5 months for pegylated liposomal doxorubicin alone (5). In our heavily pretreated platinum–resistant/refractory ovarian carcinoma cohort, combining trabectedin to the anti–PD-L1 durvalumab reached a promising response rate of 21.4% (95% CI, 4.7–51) and a 6-month PFS of 42.9% (95% CI, 17.7–71.1). Interestingly, PD-L1 positivity was significantly associated with better PFS, as previously shown (5, 31). Other targets than the PD-L1 axis have been reported in ovarian carcinoma microenvironment, such as CSF1R, TLR8, or LAG-3 that are the focus of therapeutic approaches of interest (44).

Trabectedin activity involves transcription coupled nucleotide excision repair (NER) and homologous recombination repair (HR) and has demonstrated synthetic activity in NER-proficient and HR-deficient cells (45–48). The NER pathway is the main mechanism involved in repairing platinum DNA adducts (49), and NER-proficient cells have decreased sensitivity to platinum. On the other hand, up to 50% of high-grade serous and endometrioid ovarian carcinoma present an impairment in HR, or BRCAness phenotype (50). Trabectedin-specific mechanisms of action make it therefore particularly interesting in ovarian cancer with a BRCAness phenotype (46–48). Interestingly, BRCAness has been associated with increase in mutational burden, immune infiltrates, and PD-L1 expression (51, 52). Combination of anti–PD-1/PD-L1 with PARP inhibition has shown promising results in BRCA-mutated platinum-sensitive ovarian carcinoma (53), but also in the BRCA wild-type (WT) platinum-refractory setting (54, 55).

The interpretation of this study is limited by the lack of availability of omics data on patient samples apart from BRCA status.

The promising data we have obtained with trabectedin combined with durvalumab in a BRCA WT heavily pretreated platinum-refractory population, and the existing body of evidence for trabectedin and immune checkpoint inhibitors activity in BRCAness tumors in monotherapy suggest the combination deserves further assessment in BRCAness and BRCA WT ovarian carcinoma patients, in doublet and in association with PARP inhibitors.

A. Bessede reports other support from Explicyte outside the submitted work. J.-P. Guégan reports that he is an employee of Explicyte. J.-Y. Blay reports grants and personal fees from PharmaMar and grants and personal fees from AstraZeneca during the conduct of the study. I. Ray-Coquard reports personal fees from Roche, PharmaMar, AstraZeneca, Clovis, GlaxoSmithKline, Bristol Myers Squibb, Agenus, Mersana, ImmunoGen, MSD, EISAI, and Novartis outside the submitted work. A. Floquet reports grants from PharmaMar and AstraZeneca during the conduct of the study as well as personal fees from PharmaMar and AstraZeneca outside the submitted work. A. Italiano reports grants from AstraZeneca and nonfinancial support from PharmaMar during the conduct of the study as well as grants and personal fees from Bayer; grants from AstraZeneca, Bristol Myers Squibb, MSD, and Merck; grants and personal fees from Roche; and nonfinancial support from Epizyme outside the submitted work. No disclosures were reported by the other authors.

M. Toulmonde: Conceptualization, formal analysis, investigation, writing–original draft, writing–review and editing. M. Brahmi: Investigation, writing–review and editing. A. Giraud: Data curation, formal analysis, writing–review and editing. C. Chakiba: Investigation, writing–review and editing. A. Bessede: Data curation, investigation, methodology, writing–review and editing. M. Kind: Data curation, formal analysis, writing–review and editing. E. Toulza: Data curation, writing–review and editing. M. Pulido: Formal analysis, writing–review and editing. S. Albert: Data curation, writing–review and editing. J.-P. Guégan: Formal analysis, writing–review and editing. S. Cousin: Investigation, writing–review and editing. S. Mathoulin-Pelissier: Supervision, writing–review and editing. R. Perret: Formal analysis, writing–review and editing. S. Croce: Formal analysis, writing–review and editing. J.-Y Blay: Investigation, writing–review and editing. I. Ray-Coquard: Investigation, writing–review and editing. A. Floquet: Investigation, writing–review and editing. A. Italiano: Conceptualization, supervision, funding acquisition, investigation, methodology, writing–original draft, writing–review and editing.

This study was partly funded by PharmaMar and AstraZeneca. The funder of the study collaborated with academic authors on the study design but not on data collection, analysis, and interpretation. This study was sponsored by Institut Bergonié and partly founded by PharmaMar and AstraZeneca. J.-Y. Blay is supported by NetSARC+ (INCA & DGOS) LYRICAN (INCA-DGOS-INSERM 12563), InterSARC (INCA), LabEx DEvweCAN (ANR-10-LABX0061), PIA Institut Convergence François Rabelais PLAsCAN (PLASCAN, 17-CONV-0002), EURACAN (EC 739521), and RHU4 DEPGYN (ANR-18-RHUS-0009). We thank all patients, caregivers, and families who contributed to the study.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.