Efficacy, safety, and prognostic factors of PD-1 inhibitors combined with lenvatinib and Gemox chemotherapy as first-line treatment in advanced intrahepatic cholangiocarcinoma: a multicenter real-world study

This multicenter retrospective study assessed the efficacy and safety of PD-1 inhibitors combined with lenvatinib plus Gemox chemotherapy as first-line therapy for advanced ICC at the Peking Union Medical College Hospital (PUMCH) and The Fifth Medical Center of PLA General Hospital (PLAGH). A total of 104 patients with advanced ICC who received a PD-1 inhibitor combined with lenvatinib and Gemox chemotherapy were enrolled in this study from June 2020 to September 2022. The primary eligibility criteria were histologically confirmed intrahepatic cholangiocarcinoma and at least one measurable tumor lesion according to the RECIST v1.1 criteria [19]. Among the initial 104 patients, 15 had combined use of one cycle, 19 did not receive triple combined regimens, 12 did not have measurable target lesions, and 5 had other additional malignant tumors (Fig. 1). Finally, 53 patients were enrolled in this study, 30 and 23 were enrolled in the PUMCH and PLAGH groups, respectively. Information on age, sex, Eastern Cooperative Oncology Group (ECOG) performance status, Child–Pugh score, carbohydrate antigen 19-9 (CA19-9), carcinoembryonic antigen (CEA), hepatitis B virus (HBV) infection, maximum tumor diameter, tumor burden score (TBS), differentiated histology, tumor-node metastasis classification (TNM) stage, site of metastases, PD-L1 expression, and type of PD-1 inhibitors were compiled and recorded (Table 1). The TBS was calculated based on the maximum tumor size and the number of tumors in the liver [20, 21].

Information regarding the dates of initiation and completion of treatment, initial dose, radiological evaluation, laboratory data, and adverse events (AEs) during treatment was systematically collected. Lenvatinib was administered orally at a dose of 12 mg (for patients with body weight ≥ 60 kg) or 8 mg (for patients with body weight < 60 kg) once a day. Anti-PD-1 antibodies were administered at a fixed dose of 200 mg (240 mg for toripalimab) or a fixed dose of 3 mg/kg body weight every 3 weeks. The Gemox chemotherapy regimen was administered as 1 g/m² of gemcitabine on days 1 and 8, 100 mg/m² of oxaliplatin on day 1, and every 3 weeks by IV injection for six cycles.

The clinical objective response was measured using the RECIST v1.1 criteria [19] and evaluated by professional radiologists at PUMCH and PLAGH. Computed tomography (CT) and magnetic resonance imaging (MRI) were performed to assess treatment response. The primary endpoints were OS and PFS, whereas the secondary endpoints were the ORR, disease control rate (DCR), clinical benefit rate (CBR) and safety. CBR was defined as the proportion of patients with a radiologically confirmed objective response (complete response [CR] or partial response [PR]) or stable disease (SD) for > 6 months [22]. Safety were recorded by physical examination, laboratory evaluation, and electronic medical records or collected by the investigators using the Common Terminology Criteria for Adverse Events (version 5.0) as a reference [23].

Whole sections from formalin-fixed, paraffin-embedded tumor specimens were subjected to immunohistochemistry. For each tissue slice, 5-μm-thick sections were selected and placed on glass slides. The primary antibody used was anti-PD-L1, followed by the addition of secondary antibodies to all sections, including the negative control slides. Evaluation of PD-L1 expression was performed by independent pathologists who were blinded to the clinicopathological data, including therapeutic response and survival time. PD-L1 positivity or overexpression was defined as > 5% positive expression in tumor cells.

The cutoff date for analysis was December 30, 2022 in this study. Survival curves were estimated using the Kaplan–Meier method, and the log-rank test were used to analyzed the comparison groups. Hazard ratios (HRs) of each clinical factors for PFS and OS were estimated using the Cox proportional hazard model. For comparisons of individual variables, the t-test, Mann–Whitney U test, χ² test, and Fisher’s exact test were performed as appropriate. Results with two-tailed p-values < 0.05 were considered statistically significant. Statistical analyses were performed using R-4.2.0 (https://​www.​r-project.​org/​) and the SPSS 25 software.

We screened 104 patients with advanced ICC who were treated in PUMCH and PLAGH from June 2020 to September 2022; 51 patients were excluded from the study (Fig. 1). Finally, we included 53 patients with advanced ICC who received PD-1 inhibitors combined with lenvatinib and Gemox chemotherapy as first-line treatment. The demographic and baseline characteristics of the 53 patients are summarized in Table 1. At the time of initial treatment, the median age was 58 years, with 41.5% of the patients being over 60 years old and 37.7% being women. We observed that 28 (52.8%) patients had an ECOG performance status of 0, and 32 (60.4%) patients had Child–Pugh score A. At baseline, the median CA19-9 level was 210 U/mL; 56.6% of the patients had a level > 200 U/mL. The median CEA level was 4.2 ng/mL; 41.5% of the patients had a level > 5 ng/mL. We observed that 10 patients (18.9%) had a history of HBV infection. The median maximum tumor diameter was 5.4 cm, and 54.7% of the patients had a level > 5 cm. At baseline, 22 patients (41.5%) had TBS > 8. In total, 17 (32.1%) patients had poorly differentiated histology, 23 (43.4%) had TNM stage III, and 17 (32.1%) had positive PD-L1 expression. Further, we observed that before treatment, most patients had metastatic tumors in the liver (36/53, 67.9%), lymph nodes (34/53, 64.2%), lungs (7/53, 13.2%), and bones (4/53, 7.5%). Among the 53 patients who had received different types of PD-1 inhibitors, 29 (54.7%) were treated with the toripalimab regimen, 11 (20.8%) with the tislelizumab regimen, 7 (13.2%) with the camrelizumab regimen, and 6 (11.3%) with the pembrolizumab regimen.

The median duration of treatment with PD-1 inhibitors combined with lenvatinib and Gemox chemotherapy was 8.07 (interquartile range: 5.3–11.6) months. The treatment duration for all the patients is shown in Fig. 2a. The median follow-up time was 13.7 (95% confidence interval (CI): 12.9–17.2) months for all participants in our cohort. All patients underwent a complete radiological evaluation. Overall, we identified 36 (67.9%) patients with decreased tumor sizes from baseline (Fig. 2b). Interestingly, we observed that 28 (52.8%) patients achieved an objective response, including 3 (5.7%) who showed CR and 25 (47.1%) who showed PR. We observed that 22 (41.5%) patients exhibited SD, whereas 3 (5.7%) exhibited PD. Consistently, we observed that the overall radiologically confirmed ORR was 52.8% (95% CI: 39.7–65.6%), and DCR was 94.3% (95% CI: 84.6–98.1%) (Fig. 2b, Table 2).

We investigated the survival outcomes of the enrolled patients. For the entire cohort, we observed that the median OS was 14.3 (95% CI: 11.3–NR) and the median PFS was 8.63 (95% CI: 7.17–11.6) months (Fig. 2c, d, Table 2). The 6 months and 12 months OS were 90.0% (95% CI: 82.1–98.7%) and 59.6% (95% CI: 46.6–76.3%), respectively (Table 2). The 6 months and 12 months PFS were 73.3% (95% CI: 61.8–86.9%) and 31.3% (95% CI: 19.6–49.9%), respectively (Table 2). We further determined CBR in all patients. We observed that the CBR in all 53 patients was 75.5% (95% CI: 62.4–85.1%) (Fig. 2b, Table 2).

Six patients underwent conversion surgery after triple combination treatment, with a patient ID of 16, 22, 31, 41, 47, and 50, respectively (Fig. 2a, Table S1). Four patients achieved PR, and two achieved CR before conversion surgery. A median of 5 cycles of triple therapy was administered before conversion surgery (Table S1). One patient (patient ID 16) had disease progression 1.2 months after undergoing conversion surgery and changed treatment regimens. The remaining 5 patients had no disease progression at the time of the last follow-up and were receiving maintenance therapy with a PD-1 inhibitor plus lenvatinib.

Twelve potential prognostic variables for PFS and OS were first selected using univariate Cox analysis, including age, sex, ECOG PS, Child–Pugh score, CA19-9, CEA, HBV infection, maximum tumor diameter, TBS, differentiated histology, TNM stage, and PD-L1 expression (Table 3). In the univariate Cox analysis, TBS (≥ 8 vs. < 8; HR: 3.5; 95% CI: 1.6–7.4; P = 0.001), TNM stage (IV vs. III; HR: 4; 95% CI: 1.8–8.8; P < 0.001), and PD-L1 expression (positive vs. negative; HR: 0.18; 95% CI: 0.071–0.44; P < 0.001) were different for PFS, and ECOG PS (0 vs. ≥ 1; HR: 0.33; 95% CI: 0.15–0.74, P = 0.007), CA19-9 (≥ 200 vs. < 200; HR: 3.8; 95% CI: 1.6–8.9; P = 0.003), TBS (≥ 8 vs. < 8; HR: 4.8; 95% CI: 2.1–11; P < 0.001), differentiated histology (poor vs. moderate + well; HR: 2.8; 95% CI: 1.1–6.7; P = 0.024), TNM stage (IV vs. III; HR: 2.6; 95% CI: 1.1–6.1; P = 0.029), and PD-L1 expression (positive vs. negative; HR: 0.044; 95% CI: 0.0095–0.2; P < 0.001) were different for OS. Six factors that differed in the univariate Cox analysis, including ECOG PS, CA19-9, TBS, differentiated histology, TNM stage, and PD-L1 expression, were further subjected to multivariate Cox analysis. In the multivariate Cox analysis, TBS (≥ 8 vs. < 8; HR: 3.86; 95% CI: 1.32–11.29; P = 0.014), TNM stage (IV vs. III; HR: 6.69; 95% CI: 2.198–20.34; P < 0.001), and PD-L1 expression (positive vs. negative; HR: 0.27; 95% CI: 0.077–0.92; P = 0.037) were different for PFS, and TBS (≥ 8 vs. < 8; HR: 6.31; 95% CI: 1.659–24.0; P = 0.007) and PD-L1 expression (positive vs. negative; HR: 0.11; 95% CI: 0.018–0.7; P = 0.019) were different for OS (Table 3, Fig. 3a).

Subgroup analyses of PFS and OS for three potentially prognostic variables, including TBS, TNM stage, and PD-L1 expression, were performed in the entire cohort. When we stratified patients according to TBS, the Kaplan–Meier survival curve and log-rank test analysis demonstrated that patients with TBS < 8 had a longer median PFS (10.7 vs. 6.6 months, P < 0.001; Fig. 3b) and a longer median OS (16.93 vs. 9.07 months, P < 0.001; Fig. 3c) than those with TBS ≥ 8. When we stratified patients according to the TNM stage, the Kaplan–Meier survival curve and log-rank test analysis showed that patients with TNM stage III had a longer median PFS (13.17 vs. 7.17 months, P < 0.001; Fig. 3d) and a longer median OS (16.9 vs. 11.3 months, P = 0.024; Fig. 3e) than those with TNM stage IV. When we stratified patients according to PD-L1 expression, the Kaplan–Meier survival curve and log-rank test analysis revealed that patients with positive PD-L1 expression had a longer median PFS (13.17 vs. 6.87 months, P < 0.001; Fig. 3f) and a longer median OS (NR vs. 9.57 months, P < 0.001; Fig. 3g) than those with negative PD-L1 expression.

AEs were reported in all 53 patients (100%) throughout the study (Table S4, Fig. 4a). However, we did not detect grade 5 AE. Regarding severe AEs (SAEs), 41.5% (22/53) of the patients had ≥ grade 3 AEs, and only 1.9% (1/53) experienced grade 4 AEs (myelosuppression). The most common AEs (of any grade) were fatigue (31/53, 58.5%), myelosuppression (14/53, 26.4%), and decreased appetite (12/53, 22.6%). Most AEs that occurred during combination immunotherapy were not fatal, well-tolerated, and controlled. Particularly, the most common grade 3 or 4 SAEs were fatigue (8/53, 15.1%), myelosuppression (7/53, 13.2%), abdominal pain (4/53, 7.5%), hypertension (4/53, 7.5%), and bilirubin elevation (4/53, 7.5%).

We compared the AEs that occurred in the different PD-1 inhibitor groups, and there was no statistically significant difference in grade 3–4 AEs between the different PD-1 inhibitor groups (Table S2, Fig. 4b). In the toripalimab group, the most common grade 3–4 AEs were fatigue (17.2%, 5/29) and bilirubin elevation (10.3%, 3/29). In the tislelizumab group, the most common grade 3–4 AEs were fatigue (27.3%, 3/11), myelosuppression (18.2%, 2/11), and hypertension (18.2%, 2/11). In the camrelizumab group, the most common grade 3–4 AEs were myelosuppression (14.3%, 1/7), decreased appetite (14.3%, 1/7), and skin rashes (14.3%, 1/7). In the pembrolizumab group, the most common grade 3–4 AEs were myelosuppression (33.3%, 2/6) and abdominal pain (33.3%, 2/6). After careful treatment, we discovered that all observed AEs were controllable.