Document resource
WHAT IS ALREADY KNOWN ON THIS TOPIC In the phase Ib SAFFRON-103 trial, combination therapy with the multityrosine kinase inhibitor, sitravatinib, plus the anti-programmed cell death protein-1 (anti-PD-1) antibody, tislelizumab, had promising antitumour activity with no new safety signals in patients with locally advanced or metastatic non-small cell lung cancer (NSCLC) who progressed after anti-PD-1/anti-programmed death-ligand 1 (PD-L1) therapy. The phase III SAFFRON-301 trial was performed to confirm the efficacy signals observed with sitravatinib plus tislelizumab in SAFFRON-103.WHAT THIS STUDY ADDS Combination therapy with sitravatinib plus tislelizumab had an unclear efficacy benefit with an increased risk of serious and fatal pulmonary haemorrhage in patients with squamous and non-squamous NSCLC who progressed on or after anti-PD-(L)1 therapy and platinum-based chemotherapy.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY Continued investigation of combination therapy with sitravatinib plus tislelizumab in patients with NSCLC is not warranted.Introduction Checkpoint inhibitor-based immunotherapy that targets the anti-programmed cell death protein-1 (PD-1) or anti-programmed death-ligand 1 (PD-L1) axis with or without chemotherapy has become a first-line standard-of-care for patients with locally advanced/metastatic non-small cell lung cancer (NSCLC) without actionable genetic alterations. 1–8 Despite initial improvement in some patients, many patients with NSCLC have tumours that are treatment refractory or develop resistance following an initial response to anti-PD-1 or PD-L1-based treatment.9 Patients with NSCLC without a targetable oncogenic driver who progress during/after platinum-based chemotherapy and anti-PD-(L)1 therapy have limited treatment options; these include (depending on histology) single-agent chemotherapy (docetaxel, pemetrexed, nab-paclitaxel or gemcitabine), or docetaxel in combination with an antiangiogenic agent (nintedanib or ramucirumab).1 3 10–17 Modest activity combined with tolerability issues creates a substantial unmet need for efficacious and safe treatments for the treatment of second-line and beyond disease in patients with prior exposure to anti-PD-(L)1 agents.10Strategies to improve outcomes and overcome resistance to anti-PD-(L)1 agents are being actively investigated. One rational approach involves combining anti-PD-(L)1 therapy with vascular endothelial growth factor (VEGF)/VEGF receptor (VEGFR) inhibitors that target the molecular and cellular mechanisms of resistance to anti-PD-(L)1 agents.9 18 19 VEGF/VEGFR inhibition with both small molecule tyrosine kinase inhibitors (TKIs) and monoclonal antibodies (mAbs; ie, bevacizumab or ramucirumab) has been studied in combination with anti-PD-(L1) therapy for the treatment of NSCLC.19 In preclinical models, inhibitors targeting multiple tyrosine kinases, including VEGF (eg, sitravatinib), can stimulate the immune system and stimulate tumour shrinkage.20Early phase clinical trials provided encouraging data supporting the theoretical and preclinical framework for using combination therapy with anti-PD-(L)1 agents plus TKIs or mAbs targeting VEGF and other tyrosine kinases. In a phase Ib/II trial, combination therapy with lenvatinib plus pembrolizumab had promising clinical activity, with an overall response rate (ORR) of 33% at week 24, in patients with histologically and/or cytologically confirmed metastatic NSCLC that progressed after therapy or for which standard therapy was not available.21 In the phase II COSMIC-021 study, treatment with cabozantinib in combination with atezolizumab was associated with improved clinical outcomes (ORR, 19%) compared with cabozantinib alone (ORR, 6%) in patients with non-squamous stage IV NSCLC who progressed on prior anti-PD-(L)1-based therapy.22 A randomised phase II trial of ramucirumab and pembrolizumab versus standard-of-care in patients with squamous and non-squamous NSCLC previously treated with immunotherapy met its primary endpoint of improved overall survival (OS) in the ramucirumab and pembrolizumab arm.23 These encouraging early phase trials justified the performance of phase III trials investigating the same combination therapies in patients with NSCLC previously exposed to anti–PD-(L)1 agents. These phase III trials include LEAP-008 (pembrolizumab plus lenvatinib vs docetaxel), CONTACT01 (cabozantinib plus atezolizumab vs docetaxel) and Pragmatica-Lung (ramucirumab plus pembrolizumab vs standard-of-care).24–26Sitravatinib (MGCD516) is an oral spectrum-selective receptor TKI that targets TAM receptors (TYRO3/AXL/MERTK) and split tyrosine kinase domain-containing receptors (VEGFR2/KIT).27 28 TAM receptor tyrosine kinases, primarily found on antigen-presenting cells like macrophages, help maintain immune homeostasis by suppressing inflammation.29 Their overexpression is associated with tumour metastasis, including in lung cancer. Targeting TAM receptors encourages M1 macrophage polarisation and promotes an antitumour immune microenvironment.20 In addition, targeting VEGFR2 and KIT reduces regulatory T cells and myeloid-derived suppressor cells, lessening immunosuppression and enhancing the effectiveness of PD-(L)1 inhibitors.30 31 Tislelizumab is an anti-PD-1 humanised monoclonal antibody that binds to PD-1 with high affinity and was designed to minimise binding of fragment crystallisable region receptors on macrophages in order to abrogate antibody-dependent phagocytosis, a mechanism of T-cell clearance and potential resistance to anti-PD-1 therapy.32 33 By binding to PD-1, tislelizumab prevents PD-1 from interacting with its ligands, PD-L1 and PD-L2.32 Combination therapy with sitravatinib and tislelizumab was tested in the phase Ib SAFFRON-103 trial which enrolled patients with various solid tumours.34 Sitravatinib plus tislelizumab had promising antitumour activity in patients with metastatic non-squamous/squamous NSCLC who were refractory/resistant to anti-PD-(L)1 therapy; the mOS was 10.1 and 10.5 months in patients with non-squamous and squamous histology, respectively.34 The phase III SAFFRON-301 study was therefore initiated to confirm the efficacy signal observed in SAFFRON-103.Here, we report the results of SAFFRON-301 (ClinicalTrials.gov, NCT04921358), an open-label randomised phase III study of sitravatinib in combination with tislelizumab versus chemotherapy in patients with locally advanced/metastatic NSCLC that progressed on or after platinum-based chemotherapy and anti-PD-(L)1 therapy.Methods Study design and participants SAFFRON-301 was an open-label, randomised, multicentre, phase III clinical trial ( online supplemental files 2–5). Eligible patients had histologically or cytologically confirmed locally advanced or metastatic (stage IIIB/IIIC or stage IV) NSCLC and disease progression following platinum-based chemotherapy and anti-PD-(L)1 antibody, with the anti-PD-(L)1 antibody administered in combination with, or sequentially before or after platinum-based chemotherapy (online supplemental figure 1). Other eligibility criteria included two or fewer lines of prior systemic therapy for locally advanced and unresectable or metastatic disease, at least one measurable lesion per Response Evaluation in Solid Tumours (RECIST) V.1.1,35 and Eastern Cooperative Oncology Group performance status of ≤1. Patients with known EGFR-sensitising or BRAF-sensitising mutation, or ALK or ROS1 rearrangement, squamous NSCLC with central cavitation, or NSCLC with haemoptysis (>50 mL/day), or with radiologic evidence of tumours invading or abutting major blood vessels were ineligible for the study. Detailed inclusion/exclusion criteria are shown in the online supplemental file.SP210.1136/bmjonc-2025-000890.supp2Supplementary dataSP310.1136/bmjonc-2025-000890.supp3Supplementary dataSP410.1136/bmjonc-2025-000890.supp4Supplementary dataSP510.1136/bmjonc-2025-000890.supp5Supplementary dataSP110.1136/bmjonc-2025-000890.supp1Supplementary dataRandomisation and masking After obtaining informed consent, study personnel accessed the Interactive Response Technology system to randomise treatment assignment (1:1). Patients were stratified by histological subtype (non-squamous vs squamous), PD-L1 expression (<1% tumour cell (TC) vs ≥1% TC; patients whose tissues were unevaluable for PD-L1 expression were included in the <1% TC group) and race (Asian vs non-Asian). PD-L1 expression was determined by the Ventana SP263 assay (Roche Diagnostics, North America).Treatment and assessments Patients received sitravatinib 100 mg orally once a day in combination with tislelizumab 200 mg intravenously once every 3 weeks or docetaxel 75 mg/m 2 intravenous once every 3 weeks. Cycles were 21±3 days unless a delay was medically necessary. Tumour assessments were conducted by CT and/or MRI every 6 weeks (±7 days) from cycle 1 day 1 in the first 12 months, and at 9-week±7 day intervals thereafter. Tumour assessment and responses were determined by blinded Independent Review Committee (IRC) and investigator using RECIST V.1.1.35 Patients were treated until disease progression, intolerable toxicity, death or withdrawal of consent, whichever occurred earlier. Crossover to the other treatment group was not permitted. Patients were contacted every 3 months (±14 days) to assess survival during follow-up.Outcomes Primary endpoints The dual primary endpoints were OS (time from randomisation to death from any cause) and IRC-assessed progression-free survival (PFS; time from randomisation to the first occurrence of disease progression as determined by the IRC based on RECIST V.1.1, 35 or death from any cause, whichever occurred first).Secondary endpoints Secondary endpoints included investigator-assessed PFS, IRC-assessed ORR (proportion of patients with partial response (PR) or complete response (CR)), duration of response (DoR; time from the first occurrence of a documented objective response to the time of the first occurrence of disease progression, or death from any cause, whichever occurred first), and disease control rate (DCR; proportion of patients whose best overall response was CR, PR, or stable disease) as determined based on RECIST V.1.1. 35Statistical analysis The intention-to-treat (ITT) analysis set included all randomised patients, analysed according to their randomised treatment group. The safety analysis set included all patients who received at least one dose of any study drug. We planned for approximately 420 patients to undergo 1:1 randomisation to receive sitravatinib plus tislelizumab or docetaxel alone, as determined based on the primary endpoints of OS and PFS per IRC with type I error allocation (one-sided) of 0.024 and 0.001, respectively. Approximately 289 OS events were planned for the final analysis of OS, to have a power of 85% with an alpha of 0.024 under the HR assumption of 0.7. Subgroup analyses for OS were prespecified in the statistical analysis plan. Approximately 332 PFS events per IRC were expected to occur at the final analysis of PFS, to have a power of 87% with an alpha of 0.001 under the HR assumption of 0.63. One interim analysis was planned to occur when approximately 197 of the targeted 289 OS events (68.0%) were documented in the ITT analysis set. No efficacy interim analysis of PFS was planned; final analysis of PFS was to be performed at the time of the interim analysis of OS. Due to early termination of the study, the preplanned efficacy interim analysis was aborted. Formal hypothesis testing was not performed.A treatment-emergent adverse event (TEAE) was defined as an adverse event (AE) with an onset date or a worsening in severity from baseline (pretreatment) on or after the first dose of study drug and up to 30 days following study drug discontinuation or initiation of new anticancer therapy, whichever occurred first. Treatment-related treatment-emergent AEs (TRAEs) included TEAEs considered by the investigator to be related to a study drug or with missing assessment of the causal relationship. AEs were classified based on Medical Dictionary for Regulatory Activities Version 25.0 and were graded based on National Cancer Institute-Common Terminology Criteria for Adverse Events Version 5.0.36 Immune-mediated AEs (imAEs) were of special interest. Recommendations for diagnostic evaluation and management of imAEs were based on European Society for Medical Oncology and American Society of Clinical Oncology guidelines.37 38 ImAEs were identified from all AEs with an onset date or a worsening in severity from baseline (pretreatment) on or after the first dose of any drug and up to 90 days from the last dose of any drug, regardless of whether the patient started new anticancer therapy.Results Patient population Patients were enrolled at 61 sites in China and Australia. The first patient was dosed on 27 July 2021. Because the study was terminated early, enrolment was not completed, and efficacy data are premature. The last visit for the last patient was completed on 20 December 2023 (data cut-off).Overall, 377 patients were randomised (n=187, sitravatinib+tislelizumab; n=190, docetaxel); of these, 14 patients did not receive treatment (1 (0.5%) patient in sitravatinib+tislelizumab arm; 13 (6.8%) patients in docetaxel arm) (figure 1). One patient randomised to sitravatinib+tislelizumab who did not receive treatment had a significant decline in Eastern Cooperative Oncology Group performance status due to increasing pain since screening. Of the 13 patients who did not receive docetaxel, 12 patients withdrew from treatment due to patient decision, and 1 patient was not treated because of difficult venous access; this patient subsequently deteriorated with elevated bilirubin and was then not eligible for study treatment. Median study follow-up time was 8.0 months (range, 0.4–25.3 months) and 7.6 months (range, 0–25.1 months) for sitravatinib+tislelizumab and docetaxel, respectively. By the data cut-off date, all treated patients had discontinued treatment, mostly due to progressive disease; all patients discontinued from the study, mostly due to death or study termination by sponsor (figure 1). Baseline demographics and disease characteristics, including randomisation stratification factors (histological subtype, PD-L1 expression and race), were generally well balanced between arms (table 1).Figure 1Patient disposition. ITT, intention-to-treat; Sitra, sitravatinib; TIS, tislelizumab.Table 1Patient demographics and baseline characteristics (ITT population)Sitravatinib+tislelizumab (n=187)Docetaxel (n=190)Total (N=377)Age, years Median63.063.063.0 Min, max29, 7634, 7929, 79Age group, n (%) <65 years106 (56.7)112 (58.9)218 (57.8) ≥65 years81 (43.3)78 (41.1)159 (42.2)Sex, n (%) Male153 (81.8)151 (79.5)304 (80.6) Female34 (18.2)39 (20.5)73 (19.4)Race*, n (%) Asian175 (93.6)177 (93.2)352 (93.4) White12 (6.4)13 (6.8)25 (6.6)ECOG performance status, n (%) 047 (25.1)46 (24.2)93 (24.7) 1140 (74.9)144 (75.8)284 (75.3)Histology*, n (%) Squamous96 (51.3)97 (51.1)193 (51.2) Non-squamous91 (48.7)93 (48.9)184 (48.8)PD-L1 expression status*, n (%) ≥1%68 (36.4)70 (36.8)138 (36.6) <1%106 (56.7)107 (56.3)213 (56.5) Not evaluable13 (7.0)13 (6.8)26 (6.9)Smoking status, n (%) Current21 (11.2)24 (12.6)45 (11.9) Former110 (58.8)108 (56.8)218 (57.8) Never56 (29.9)58 (30.5)114 (30.2)Number of lines of prior systemic therapy, n (%) 1142 (75.9)140 (73.7)282 (74.8) 245 (24.1)50 (26.3)95 (25.2)Type of last line systemic therapy, n (%) Anti–PD-(L)1 monotherapy8 (4.3)13 (6.8)21 (5.6) Chemotherapy1 (0.5)1 (0.5)2 (0.5) Anti–PD-(L)1 in combination with chemotherapy150 (80.2)156 (82.1)306 (81.2) Other†28 (15.0)20 (10.5)48 (12.7)*Value per EDC.†Other includes therapies not listed as above, such as anti-VEGF antibody or antiangiogenic therapy in combination with anti–PD-(L)1 and/or chemotherapy.ECOG, Eastern Cooperative Oncology Group; EDC, electronic data capture; ITT, intention-to-treat; Max, maximum; Min, minimum; PD-L1, programmed cell death protein-ligand1; VEGF, vascular endothelial growth factor.Efficacy Primary efficacy endpoints At data cut-off, the median follow-up time for OS was 11.7 months and 11.4 months in the sitravatinib+tislelizumab and docetaxel arms respectively, with deaths reported in 48.7% and 42.1% of patients, respectively. Median OS was 11.5 months (95% CI 9.4 to 14.6) and 11.4 months (95% CI 9.9 to 15.0) in the sitravatinib+tislelizumab and docetaxel arms, respectively, with no survival benefit (HR 1.02 (95% CI 0.75 to 1.39)) ( figure 2A). IRC-assessed PFS was numerically longer in the sitravatinib+tislelizumab compared with the docetaxel arm: 4.4 months (95% CI 4.0 to 5.7) versus 2.9 months (95% CI 2.6 to 4.2; HR 0.82 (95% CI 0.62 to 1.07)) (figure 2B).Figure 2Overall survival (OS) and progression-free survival (PFS). (A) OS. (B) IRC-assessed PFS. (C) Investigator (INV)-Assessed PFS. IRC, Independent Review Committee; SITRA, sitravatinib; TIS, tislelizumab.Secondary efficacy endpoints Investigator-assessed median PFS was 4.4 months (95% CI 4.0 to 5.6) and 2.9 months (95% CI 2.6 to 4.1) in the sitravatinib+tislelizumab and docetaxel arms, respectively (HR 0.64 (95% CI 0.50 to 0.83)) ( figure 2C). IRC-assessed ORR was similar in both arms (12.3% (95% CI 8.0% to 17.9%) vs 12.6% (95% CI 8.3% to 18.2%) for sitravatinib+tislelizumab and docetaxel, respectively) (table 2). There were no CRs in either group; PR rates were similar in both arms (sitravatinib+tislelizumab, 12.3% vs docetaxel, 12.6%). IRC-assessed DCR was 69.5% (95% CI 62.4% to 76.0%) and 53.7% (46.3% to 60.9%) in the sitravatinib+tislelizumab and docetaxel arms, respectively (table 2); IRC-assessed DoR was 6.0 months and not reached in the sitravatinib+tislelizumab therapy and docetaxel arms, respectively (table 2).Table 2Disease response by IRC (ITT analysis set)Response categorySitravatinib+tislelizumab (n=187)Docetaxel (n=190)Best overall response, n (%) Complete response0 (0.0)0 (0.0) Partial response23 (12.3)24 (12.6) Stable disease102 (54.5)74 (38.9) Non-CR/non-PD5 (2.7)4 (2.1) Progressive disease33 (17.6)51 (26.8) Could not be determined24 (12.8)37 (19.5)Overall response rate, n (%)23 (12.3)24 (12.6) 95% CI (%)*8.0 to 17.98.3 to 18.2Disease control rate130 (69.5)102 (53.7) 95% CI (%)*62.4 to 76.046.3 to 60.9 Median duration of response (95% CI), months†6.0 (3.2, 8.5)NR (5.4, NE)Complete response and partial response were confirmed per RECIST V.1.1.*The 95% CI was estimated using the Clopper-Pearson method.†Medians were estimated using the Kaplan-Meier method with 95% CIs estimated using the Brookmeyer and Crowley method with log–log transformation.CR, complete response; IRC, Independent Review Committee; ITT, intention-to-treat; NE, not estimable; NR, not reached; PD, progressive disease; RECIST, Response Evaluation Criteria in Solid Tumours.Subgroup analyses Consistent with the overall population analysis, there were no clear improvements in OS favouring sitravatinib+tislelizumab over docetaxel in any prespecified subgroup ( online supplemental figure 2A).In a post hoc subgroup analysis for PFS, the HR with sitravatinib+tislelizumab versus docetaxel numerically favoured investigational arm in patients with squamous histology (HR 0.59 (95% CI 0.40 to 0.88)), in patients aged <65 years (HR 0.66 (95% CI 0.47 to 0.92)), and in multiple other subgroups with specific clinical pathology characteristics (online supplemental figure 2B). However, given the limited number of patients in each subgroup, the results should be interpreted with caution.Safety The median duration of exposure was 4.1 months (range, 0.0–24.3) and 4.1 months (range, 0.5–24.8) for sitravatinib and tislelizumab, respectively, with median relative dose intensities of 83.3% (range, 7.7–100.0) and 96.9% (range, 40.0–108.6). Median duration of exposure to docetaxel was 2.1 months (range, 0.1–25.8) with a median relative dose intensity of 95.3% (range, 40.9–103.7).Overall, 183 (98.4%) and 162 (91.5%) patients in the sitravatinib+tislelizumab and docetaxel arms, respectively, had TEAEs (table 3 and online supplemental table 2). The incidence of grade 3 TEAEs was higher in the sitravatinib+tislelizumab than the docetaxel arm (121 (65.1%) patients vs 100 (56.5%) patients) (table 3 and online supplemental table 2). Grade ≥3 TEAEs occurring in ≥5% of sitravatinib+tislelizumab arm were hypertension (25 (13.4%)), pneumonia (17 (9.1%)), palmar-plantar erythrodysesthesia syndrome (12 (6.5%)), and hypokalaemia (10 (5.4%)) (table 3). Grade ≥3 TEAEs occurring in ≥5% of docetaxel arm were white blood cell count decreased (52 (29.4%)), neutrophil count decreased (51 (28.8%)), pneumonia (15 (8.5%)), neutropenia (13 (5.3%)) and febrile neutropenia (10 (5.6%)) (table 3). An overview of TRAEs and TRAEs by preferred term is shown in online supplemental tables 2 and 3, respectively.Table 3Treatment-emergent adverse events with an incidence ≥10% (safety analysis set)Preferred termSitravatinib+tislelizumab (n=186)Docetaxel (n=177)Any graden (%)≥Grade 3n (%)Any graden (%)≥Grade 3n (%)Patients with ≥1 TEAE with incidence ≥10%183 (98.4)121 (65.1)162 (91.5)100 (56.5) Aspartate aminotransferase increased111 (59.7)6 (3.2)18 (10.2)3 (1.7) Alanine aminotransferase increased97 (52.2)5 (2.7)17 (9.6)4 (2.3) Diarrhoea83 (44.6)9 (4.8)26 (14.7)5 (2.8) Blood creatine phosphokinase MB increased74 (39.8)5 (2.7)3 (1.7)0 (0.0) Palmar-plantar erythrodysesthesia syndrome68 (36.6)12 (6.5)0 (0.0)0 (0.0) Anaemia61 (32.8)7 (3.8)84 (47.5)5 (2.8) Hypoalbuminaemia58 (31.2)1 (0.5)41 (23.2)0 (0.0) Hypertension56 (30.1)25 (13.4)6 (3.4)2 (1.1) Weight decreased54 (29.0)8 (4.3)15 (8.5)1 (0.6) Decreased appetite53 (28.5)2 (1.1)29 (16.4)2 (1.1) Hypothyroidism46 (24.7)0 (0.0)0 (0.0)0 (0.0) Hyponatraemia45 (24.2)9 (4.8)18 (10.2)2 (1.1) Blood lactate dehydrogenase increased41 (22.0)0 (0.0)8 (4.5)0 (0.0) Platelet count decreased38 (20.4)8 (4.3)22 (12.4)3 (1.7) Hypokalaemia36 (19.4)10 (5.4)15 (8.5)4 (2.3) Proteinuria34 (18.3)0 (0.0)13 (7.3)0 (0.0) Rash32 (17.2)1 (0.5)10 (5.6)0 (0.0) Vomiting31 (16.7)0 (0.0)19 (10.7)0 (0.0) Constipation30 (16.1)0 (0.0)26 (14.7)0 (0.0) Haemoptysis29 (15.6)5 (2.7)19 (10.7)1 (0.6) Nausea29 (15.6)0 (0.0)25 (14.1)0 (0.0) Gamma-glutamyltransferase increased27 (14.5)5 (2.7)13 (7.3)0 (0.0) Hypocalcaemia27 (14.5)0 (0.0)13 (7.3)0 (0.0) Blood alkaline phosphatase increased26 (14.0)1 (0.5)10 (5.6)1 (0.6) Cough26 (14.0)0 (0.0)28 (15.8)0 (0.0) Pneumonia26 (14.0)17 (9.1)25 (14.1)15 (8.5) Dysphonia25 (13.4)0 (0.0)0 (0.0)0 (0.0) Blood creatine phosphokinase increased24 (12.9)2 (1.1)6 (3.4)0 (0.0) SARS-CoV-2 test positive23 (12.4)1 (0.5)10 (5.6)0 (0.0) Blood creatinine increased20 (10.8)0 (0.0)8 (4.5)0 (0.0) Blood thyroid stimulating hormone increased20 (10.8)0 (0.0)1 (0.6)0 (0.0) Fatigue20 (10.8)4 (2.2)17 (9.6)0 (0.0) Insomnia20 (10.8)1 (0.5)14 (7.9)0 (0.0) Abdominal pain upper19 (10.2)2 (1.1)3 (1.7)0 (0.0) Hypertriglyceridaemia19 (10.2)5 (2.7)5 (2.8)1 (0.6) Hyperuricaemia19 (10.2)0 (0.0)8 (4.5)0 (0.0) Fever19 (10.2)0 (0.0)12 (6.8)1 (0.6) Hyperglycaemia15 (8.1)0 (0.0)18 (10.2)0 (0.0) White blood cell count decreased11 (5.9)0 (0.0)66 (37.3)52 (29.4) Neutrophil count decreased10 (5.4)1 (0.5)59 (33.3)51 (28.8) Lymphocyte count decreased9 (4.8)4 (2.2)19 (10.7)5 (2.8) Alopecia5 (2.7)0 (0.0)59 (33.3)0 (0.0)Adverse events were classified based on MedDRA V.25.0.Adverse event grades were evaluated based on NCI-CTCAE V.5.0.Patients with multiple events for a given preferred term were counted only once at the worst severity for each preferred term.Events were sorted by decreasing frequency of preferred term in the sitravatinib+tislelizumab any grade group.MB, muscle-bound; MedDRA, Medical Dictionary for Regulatory Activity; NCI-CTCAE, National Cancer Institute-Common Terminology Criteria for Adverse Events; TEAE, treatment-emergent adverse event.45 (24.2%) and 15 (8.5%) patients in the sitravatinib+tislelizumab and docetaxel arms, respectively, had a TEAE leading to treatment discontinuation (online supplemental table 2). The most common TEAEs leading to discontinuation in the sitravatinib+tislelizumab arm were haemoptysis (4 (2.2%)), fatigue (3 (1.6%)), and immune-related lung disease (3 (1.6%)), whereas those in the docetaxel arm included pneumonitis (2 (1.1%)), cancer pain (2 (1.1%)), peripheral sensory neuropathy (2 (1.1%)) and pleural effusion (2 (1.1%)).The incidence of any grade (91 (48.9%) vs 18 (10.2%)) and ≥grade 3 (18 (9.7%) vs 4 (2.3%)) imAEs was higher in the sitravatinib+tislelizumab versus docetaxel arm, respectively (online supplemental table 4). The most commonly reported imAEs of any grade by category in the sitravatinib+tislelizumab and docetaxel arms, respectively, were immune-mediated endocrinopathies (hypothyroidism) (47 (25.3%) vs 0%), immune-mediated skin adverse reaction (37 (19.9%) vs 11 (6.2%)), and immune-mediated pneumonitis (pneumonitis and immune-mediated lung disease) (9 (4.8%) vs 4 (2.3%)) (online supplemental table 4). Immune-mediated pneumonitis that was ≥grade 3 occurred in 6 (3.2%) and 3 (1.7%) patients in the sitravatinib+tislelizumab and docetaxel-treated arms, respectively (online supplemental table 4).Overall, 83 (44.6%) and 66 (37.3%) patients in the sitravatinib+tislelizumab and docetaxel arms, respectively, had serious TEAEs (online supplemental tables 2 and 5). The most common serious TEAEs in the sitravatinib+tislelizumab were pneumonia (16 (8.6%)), death (6 (3.2%)), and haemoptysis (5 (2.7%)) whereas those in the docetaxel arm were pneumonia (17 (9.6%)), neutropenia (8 (4.5%)), neutrophil count decreased (8 (4.5%)), white blood cell count decreased (8 (4.5%)), febrile neutropenia (7 (4.0%)) and pleural effusion (6 (3.4%)) (online supplemental table 5). Of patients experiencing serious haemoptysis, 4 (2.2%) and 0 patients in the sitravatinib+tislelizumab and docetaxel arms, respectively, had events assessed by the investigator as being related to study treatment (online supplemental tables 6 and 7).TEAEs leading to death were more common in the sitravatinib+tislelizumab compared with the docetaxel arm (15 (8.1%) vs 5 (2.8%) patients); the most common in the sitravatinib+tislelizumab arm were: unexplained death (4 patients (2.2%) vs 1 (0.6%)), hemoptysis (3 (1.6%) vs 0%), and pneumonia (3 (1.6%) vs 2 (1.1%)) (online supplemental table 8). All other TEAEs leading to death occurred in one patient each in either treatment arm or in both arms (pneumonitis) (online supplemental table 8). The Independent Data Monitoring Committee reviewed and concluded that the fatal haemoptysis events were confounded by squamous central NSCLC in two patients and pre-existing haemoptysis at baseline in one patient. Serious TRAEs and TRAEs leading to death are shown in online supplemental tables 6 and 8, respectively.SAFFRON301 was terminated due to an imbalanced rate of serious or fatal pulmonary haemorrhage and an unfavourable overall risk–benefit assessment in the sitravatinib+tislelizumab arm.Discussion Patients with advanced NSCLC who progress following chemotherapy and/or checkpoint inhibitor-based immunotherapy targeting PD-1/PD-L1 have limited efficacious treatment options. 10 Combining checkpoint inhibitors with agents that target molecular and cellular mechanisms of resistance to checkpoint inhibitors is a rational approach to overcoming resistance.18 The SAFFRON-301 investigated whether sitravatinib plus tislelizumab can help to overcome resistance to anti-PD-(L)1 agents and improve the clinical outcomes in patients with metastatic or unresectable locally advanced NSCLC who had disease progression following platinum-based chemotherapy and anti-PD-(L)1 antibody. At median follow-up times for OS of 11.7 months (sitravatinib+tislelizumab) and 11.4 months (docetaxel), there was no clear OS advantage favouring sitravatinib+tislelizumab over docetaxel. Due to early termination of the study, the number of OS events (171) did not reach the predefined threshold for interim analysis (197); thus, the OS data were immature. IRC-assessed and investigator-assessed PFS were numerically longer with sitravatinib+tislelizumab combination therapy compared with docetaxel, with HRs of 0.82 and 0.64, respectively. While both PFS assessments are exploratory due to the early termination of the study, the investigator-assessed result appears more favourable, which may be a potential bias in open-label trials.SAFFRON-301 was terminated prematurely due to an unfavourable risk–benefit assessment. There was a numerical imbalance of both serious and fatal haemoptysis between the sitravatinib+tislelizumab and docetaxel arms (serious: n=5 (2.7%) vs n=1 (0.6%), fatal: n=3 (1.6%) vs 0%, respectively). Thus, pulmonary haemorrhage was identified as an important risk of sitravatinib when used in combination with tislelizumab, especially in this population of patients with locally advanced or metastatic squamous and non-squamous NSCLC.The reasons for an increased risk of pulmonary haemorrhage with sitravatinib+tislelizumab in SAFFRON-301 are unclear. A risk of fatal haemoptysis was not previously identified in SAFFRON-103, a phase Ib study of sitravatinib+tislelizumab in locally advanced/metastatic squamous and non-squamous NSCLC, although treatment-related haemoptysis leading to study drug discontinuation was reported in 2 of 122 (1.6%) patients in the combination therapy arm.34 Patients with NSCLC are at increased risk of pulmonary haemorrhage due to the disease pathogenesis itself.39 In a retrospective analysis of patients with lung cancer (n=877), massive and fatal haemoptysis occurred in 29 (3.3%) patients; this was significantly associated with cavitation, squamous histology and bronchial tumours. Non-lethal haemorrhage occurring in 140 patients (15.9%) was not cell-type specific.40 In SAFFRON-301, a higher percentage of patients who were randomised to the docetaxel arm never received their assigned treatment compared with the sitravatinib+tislelizumab arm (6.8% vs 0.5%). As patients who did not receive study treatment had a very limited follow-up time in the survival analysis, the study results should be interpreted with caution in view of this imbalance observed between treatment arms.Published literature indicates that histology may play a role in serious/fatal bleeding events. However, because squamous tumours tend to be cavitated and centrally located, it is unclear if histology is an independent risk factor for serious haemorrhage or is a surrogate marker.39 41 Of the five serious or fatal TEAEs of haemoptysis cases in the sitravatinib+tislelizumab arm in SAFFRON-301, 3 patients had squamous disease, and the other 2 patients had non-squamous histology. Notably, the current study had a higher proportion of patients with squamous histology (51%) compared with similar trials, such as LEAP-008 (27% in lenvatinib+pembrolizumab arm, 31% in docetaxel arm), CONTACT-01 (26% in atezolizumab+cabozantinib arm, 24% in docetaxel arm), and SAPPHIRE (0%).24 25 42 Based on available data, no definitive conclusion can be made regarding a potential role of histology in the serious/fatal pulmonary haemorrhage events observed in SAFFRON-301 in the sitravatinib+tislelizumab arm. Because the primary tumour location (central or peripheral) was not prospectively collected in this study, we were not able to analyse if the tumour location contributed to the increased occurrence of haemoptysis in sitravatinib+tislelizumab; this is a key limitation of the study.There are limited data regarding the safety risk of pulmonary haemorrhage in patients treated with multi-kinase inhibitors. However, the use of bevacizumab and other anti-VEGF agents in some groups of patients with NSCLC resulted in clinically significant or fatal pulmonary haemorrhage.39 In AVF0757g, a phase II trial of bevacizumab plus carboplatin/paclitaxel in patients with NSCLC, 9% of bevacizumab-treated patients experienced life-threatening pulmonary haemorrhage; squamous cell histology, tumour necrosis and cavitation, and disease location close to major blood vessels were associated with major haemoptysis.41 A meta-analysis of randomised controlled trials revealed that unlike ramucirumab, bevacizumab significantly increased the risk of all-grade and high-grade pulmonary haemorrhage in patients with lung cancer.43 In REVEL, the incidence of overall/≥grade 3 pulmonary haemorrhage was 7%/1% and 6%/1% among patients treated with ramucirumab plus docetaxel and placebo plus docetaxel, respectively, in patients with non-squamous NSCLC compared with 10%/2% and 12%/2% in patients with squamous histology.44 In LEAP-008, where NSCLC patients were enrolled from a very similar treatment setting to SAFFRON-301, treatment-related fatal haemoptysis was reported for 2/181 patients who received lenvatinib+pembrolizumab, and for 1/47 patients who received lenvatinib monotherapy; no fatal haemoptysis was reported from the docetaxel arm in the same trial.24The lack of an OS benefit with combination therapy in SAFFRON-301 is consistent with the phase III LEAP-008, CONTACT-01 and SAPPHIRE trials where addition of a TKI to a PD-(L)1 inhibitor failed to meet the primary endpoint of improved OS compared with docetaxel in patients with advanced NSCLC that progressed on or after a PD-(L)1 inhibitor and platinum-containing chemotherapy; however, cross-trial comparisons should be cautiously interpreted.24 25 42 Moreover, the current data strongly suggest that alternative treatment strategies to that of combining a TKI with a PD-(L)1 inhibitor should be considered. Resistance to checkpoint inhibitors is a complex, multimechanism process. In addition to the immunosuppressive tumour microenvironment and TAM receptor activation, coinhibitory checkpoints, defects in antigen processing or neoantigen loss, oncogenic signalling pathways and tumour-mediated immune suppression all contribute to resistance.45 46 Treatments with novel and diverse mechanisms of action as well as customised therapies are likely needed to overcome resistance to PD-L(1) based therapy in patients with NSCLC who were previously treated with checkpoint inhibitors with or without platinum-based chemotherapy.In summary, an unclear efficacy benefit in combination with an increased risk of serious and fatal pulmonary haemorrhage in the combination therapy arm led to early termination of SAFFRON-301. As such, the efficacy data should be interpreted with caution due to early termination and thus immature survival data. These efficacy data are, however, consistent with the lack of survival benefit in other phase III trials of addition of a TKI to a PD-(L)1 inhibitor.