Between September 2021 and November 2023, 50 eligible patients from 5 centers were enrolled and initiated the protocol-specified therapy (Fig. 1). Efficacy and safety were analyzed in all 50 as-treated patients. The median age was 64 years (range, 41–74), with 26 patients (52%) having a combined positive score (CPS) of PD-L1 below 1 (Table 1). Data cutoff occurred on April 1, 2024; the median follow-up period was 23.7 months (95% confidence interval [CI], 22.1–26.8).

Patients disposition in the intention-to-treat population

Among 50 as-treated patients, 14 (38%) discontinued initial treatment early (Fig. 1), mainly due to voluntary withdrawal (n = 8) and death (n = 2). The median initial therapy cycle was 5 (range, 1–6). During the initial phase, median relative dose-intensity were all 100% for benmelstobart (range, 33%–100%), anlotinib (range, 64%–100%), paclitaxel (range, 60%–100%), and cisplatin (range, 67%–100%). Mean treatment duration was 2.9 months (range, 0.03–4.2) for benmelstobart, 3.5 months (range, 0.03–4.8) for anlotinib, 3.0 months (range, 0.1–4.0) for paclitaxel, and 2.9 months (range, 0.03–3.9) for cisplatin. Thirty-six patients (72%) completed 4–6 cycles of initial therapy and proceeded to the maintenance phase. Maintenance treatment was discontinued in 19 patients (38%) because of disease progression (PD) (n = 9), death (n = 5), voluntary withdrawal (n = 3), poor compliance (n = 1), and investigator decision (n = 1). At cutoff date, 17 patients (34%) remained ongoing treatment.

After progression or withdrawal from the study treatment, 53.8% (7/13) of patients received subsequent treatments, with one patient (7.7%) undergoing third-line therapy. Among those who received subsequent treatments, the majority (6 [46.2%]) received anti-PD-1 therapy in later-line. Data on subsequent therapies are provided in supplementary Table S1.

Among 50 patients, 23 progression events or death occurred. The median progression-free survival (PFS) was 14.9 months (95% CI, 11.4-not estimable [NE]) according to Response Evaluation Criteria in Solid Tumors criteria version 1.1 (RECIST v1.1) and the estimated PFS rate at one year was 58.5% (95% CI, 41.9%–71.9%) (Fig. 2a). With 19 deaths, the median OS was not reached (NR; 95% CI, 13.2 months-NE); 1-year OS rate was estimated at 74.8% (95% CI, 59.8%–84.8%) (Fig. 2b).

Survival outcomes. a Kaplan-Meier plots of progression-free survival per RECIST version 1.1 criteria. b overall survival. Vertical lines denote censored patients

In the intention-to-treat (ITT) population, objective response was achieved in 36 patients (72.0%; 95% CI, 57.5%–83.8%), comprising 5 (10.0%) complete response (CR) and 31 (62.0%) partial response (PR) (Fig. 3a; Table 2). Target lesion size decreased from baseline in 45 (90.0%) of all patients (Fig. 3a). Of the 36 responders, durable responses lasting more than six months were observed in 30 patients (83.3%) (Fig. 3b), with a median duration of response (DoR) was 16.2 months (95% CI, 10.2–NE). Additional 6 patients (12.0%) were stable disease (SD), resulting in disease control in 42 (84.0%; 95% CI, 70.9%–92.8%) patients (Table 2). All six patients (100%) with SD showed initial tumor shrinkage from baseline (Fig. 3c). Among the 43 patients with post-baseline response assessments, the objective response rate (ORR) was 83.7% (95% CI, 69.3%–93.2%) and disease control rate (DCR) was 97.7% (95% CI, 87.7%–99.9%). Throughout the study, 13 patients experienced PD, with 12 (92.3%) developing oligo-progression (progression site 1–2) and 1 (7.7%) developing systemic progression (progression site ≥ 3).

Tumor response. a Waterfall plot of maximum percent change in tumor size from baseline in each patient as measured by Response Evaluation Criteria in Solid Tumors (version 1.1). * Patients whose PD status was documented due to the appearance of new lesions. b Swimmer-plot of time on treatment. c Longitudinal change in tumor size from baseline. # NE indicates patients with unevaluable post-baseline assessment or no post-baseline assessment available for tumor response. CR complete response, PR partial response, PD progressive disease, SD stable disease, NE not evaluable

PD-L1 CPS were available for 38 patients. Exploratory analyses based on PD-L1 CPS status indicated an ORR of 100% (4/4) for patients with CPS ≥ 10 and 67.6% (23/34) for those with CPS < 10; 75.0% (9/12) for CPS ≥ 1 and 69.2% (18/26) for CPS < 1 (Fig. 3a). Median PFS was not associated with PD-L1 CPS, irrespective of cutoff values (cutoff ≥ 1, 12.9 vs 14.9 months; ≥ 10, 12.9 months vs. NR) (supplementary Fig. S1). The median OS was NR in either subgroup, regardless of PD-L1 CPS (supplementary Fig. S2). Subgroup analyses of efficacy results (PFS or OS) based on other baseline characteristics (e.g., sex, age, metastasis, Eastern Cooperative Oncology Group [ECOG] performance status, prior surgery, or radiotherapy) are also available in supplementary Figs. S1 and S2. Patients without liver metastases exhibited significantly longer PFS (20.5 vs. 4.7 months; p < 0.0001) and OS (NR vs. 6.3 months; p < 0.0001) compared to those with liver metastases, while other characteristics had no correlations with PFS or OS (supplementary Figs. S1 and S2). Nevertheless, the significance based on liver metastases should be considered preliminary due to the limited number of cases.

Forty-six patients (92%) reported any grade AEs and 39 (78%) reported grade ≥3 events (Table 3). Common AEs included leukopenia (32 patients [64%]; of whom 12 [24%] with grade ≥3), neutropenia (29 [58%]; 22 [44%] with grade ≥3), anemia (28 [56%]; 3 [6%] with grade ≥3), hypertension (26 [52%]; 10 [20%] with grade ≥3), nausea (26 [52%]; none with grade ≥3), and anorexia (25 [50%]; 4 [8%] with grade ≥3).

Regarding treatment-related AEs (TRAEs), 46 patients (92%) experienced any grade events and 37 (74%) were grade 3 or worse (Table 3). Frequent TRAEs were leukopenia (31 patients [62%]; of whom 12 [24%] with grade ≥3), anemia (28 [56%]; 3 [6%] with grade ≥3), neutropenia (27 [54%]; 22 [44%] with grade ≥3), hypertension (26 [52%]; 10 [20%] with grade ≥3), nausea (26 [52%]; none with grade ≥3), and anorexia (25 [50%]; 4 [8%] with grade ≥3).

Immune-related AEs (irAEs) occurred in 24 patients (48%) with only 3 (6%) experiencing grade ≥3 irAEs (supplementary Table S2). The most common irAEs in at least 10% of patients were hypothyroidism (9 patients [18%]), diarrhea (7 [14%]), and alanine aminotransferase increased (5 [10%]). Seventeen patients (34%) required dose reduction of anlotinib or chemotherapy due to AEs, and 3 (6%) discontinued treatment secondary to toxicity.

During the treatment, three bleeding events (6%) were reported, including one grade 5 intracranial hemorrhage, one grade 1 epistaxis, and one grade 1 lower gastrointestinal hemorrhage, which was possibly attributed to anlotinib. The gastrointestinal hemorrhage did not lead to dose reduction or discontinuation of anlotinib and was managed successfully with supportive care. Two patients (4%) experienced grade ≥3 esophageal fistula; however, one was deemed unrelated to treatment. Another grade 5 esophageal fistula occurred in a patient who had previously undergone extensive radiotherapy (5400 cGy in 30 fractions) for a primary lesion in the cervical and upper thoracic segment of the esophagus and was considered possibly related to treatment according to the investigator’s assessment. Nineteen deaths (38%) occurred on treatment or during follow-up. Two deaths (4%) were attributed to non-disease-related causes (COVID-19) and 13 (26%) were attributed to PD. Four deaths (8%) occurred as a result of AEs, including one each with myelosuppression, pulmonary infection, intracranial hemorrhage, and esophageal fistula.