Patient Population
Between November 16, 2017, and December 31, 2018, 32 eligible patients with advanced unresectable or metastatic BTCs were enrolled, of which 7 patients were refractory to gemcitabine- or cisplatin-based chemotherapy and 25 patients were chemotherapy- naive (fig.1). All enrolled patients, including 1 (3%) patient with regional unresectable disease, 6 (19%) with metastatic disease, and 25 (78%) with recurrent disease (defined as patients who had regional relapsed disease or distant metastases after complete resection or locoregional or systemic therapies), were administered at least one cycle of nivolumab plus gemcitabine and cisplatin combination therapy (Table 1). Patients who did not meet inclusion criteria or were participating other trials were excluded (n=9). At the time of data cutoff (September 31, 2019), all patients in cohort A and cohort B were eligible for safety analyses, of which 6 in cohort A and 21 in cohort B were qualified for efficacy analyses, 5 patients discontinued treatment within the first cycle due to rapidly deteriorated tumor-related complications (n=4) or adverse events unrelated to study drugs (n=1). Detailed baseline demographics and characteristics of all enrolled patients were summarized in Table 1. The median age was 60 years (range 27 -69). 14 patients (44%) had target lesion larger than or equal to 5 cm. Liver metastases were detected in 28 patients (88%) , while abdominal lymphatic metastases in 21 patients (66%). PD-L1 status was evaluable in 26 tumor samples (81%), of which 12 (37%) were detected positive PD-L1 expression and 14 (44%) negative.
Treatment-Related Toxicity
Safety data from cohort A and cohort B were summarized and analyzed together. Of the 32 enrolled patients, all patients experienced at least one treatment-related adverse event. The most frequent adverse events were nausea in 29 patients (91%), neutropenia in 26 patients (81%), fatigue in 21 patients (66%), thrombocytopenia in 20 patients (62%), and anemia in 19 patients (59%) (Table 2). The most common grade 3 or higher treatment-related adverse events were thrombocytopenia, reported in 18 patients (56%), and neutropenia in 7 patients (22%). Other severe adverse events included elevated alanine aminotransferase in 1 patient (3%, grade 3), elevated aspartate aminotransferase in 1 patient (3%, grade 4), elevated lipase in 1 patient (3%, grade 3), hyponatremia in 1 patient (3%, grade 3), and hypertension in 2 patients (6%, grade 3). 1 (3%) patient had immune-related adverse event (rash, grade 1). There were no treatment-related deaths at the time of analysis.
Clinical response and biomarkers
After a median follow-up of 12.8 months (95% CI, 10.8-14.8), 27 response-evaluable patients received a median of 4 cycles nivolumab plus gemcitabine and cisplatin combination therapy (IQR, 3-6). 15 (55.6%) patients in total achieved confirmed objective response, including 5 (18.6%) CR and 10 (37%) PR (Table 3, fig. 2A). The disease control was achieved in 25 patients (92.6%), including 10 (37%) patients who had SD as their best response. Radiological changes of each response-evaluable patient were summarized in Additional file 2. In cohort A, 6 of 7 patients who were refractory to gemcitabine- or cisplatin-based regimens were response evaluable, of which 1 patient achieved CR and 1 patient obtained PR, the ORR and DCR was 33.3%, 83.3%, respectively. In cohort B, 13 of 21 chemotherapy-naive patients (61.9%) achieved CR or PR and the proportion of patients with disease control was 95.2%. Responses were ongoing at the time of data cutoff in 2 patients with CR and 1 patient with PR (fig. 2B, fig. 2C). Analysis of 27 response evaluable patients found that PD-L1 expression level could not be used as a biomarker for predicting clinical response (p=0.395, Additional file 2: fig. S1A). Whole-exome sequencing was performed on patients’ biopsied tumor samples and their paired peripheral-blood mononuclear cells, which were allocated to respond group (CR+PR) and non-respond group (SD+PD) according to their clinical response. TMB and TNB were generally low in this study (Additional file 2: fig. S2). However, the median value of TMB, TNB, and fitness was higher in respond group than that in non-respond group, while the median value of heterogeneity was lower in respond group, of which fitness had statistical difference (p= 0.041, fig. 3A). Mutations of RYR2, MUC4, APOB were detected only in samples from respond group (fig. 3B). We did exploratory analysis to study the association between the activation of peripheral T cells and clinical antitumor activity. Evaluation of T cells in peripheral blood showed that the baseline percentage of CD3+ cells in responders were higher than those in non-responders (p=0.046, Additional file 2: fig. S3A). The proportion of HLA-DR+CD3+ cells in patients’ peripheral blood increased after the start of the combination therapy, especially in patients with objective response (p=0.009). However, statistic difference was not observed between responders and non-responders (Additional file 2: fig. S3B and 3C). The association between change of peripheral serum cytokines and chemokines at baseline (C1D0) and C3D0 (the day before the first dose of the 3rd cycle) and clinical response was also assessed. The concentration of serum sFasL and Granzyme A were higher in non-responders than responders after 2 cycles of combination treatment (p = 0.042 and 0.048), while the concentration of IL-2, IL-18, sFasL and CCL2 dropped significantly in responders than non-responders (P=0.036, 0.047, 0.012 and 0.042, Additional file 2: fig. S4A and4B).
PFS and biomarkers
Median PFS in this study was 6.1 months (95% CI, 3.4 -8.2), and the proportion of patients who were progression free at 6 months and 12 months were 51.9 % (95% CI, 31.9-68.6) and 18.5% (95% CI, 6.8-34.8), respectively (fig. 2D). Comparison between cohort A and cohort B showed chemotherapy-naïve patients could obtain longer median PFS, however, there was no statistical difference. Further analysis found that patients who were administered more than 4 cycles of combination treatment had longer PFS (p= 0.024), and PD-L1 expression status could not been established as a biomarker in predicting PFS (p= 0.125, Additional file 2: fig. S1B). We also analyzed the impact of TMB, TNB, and fitness on PFS in this study. However, there was no correlation between the above four biomarkers and PFS (fig. 3C; Additional file 2: fig. S5). Analysis of peripheral serum cytokines found that patients whose concentration of IFN-γ decreased following the combination therapy could obtain longer PFS (p=0.033, fig. 3D), similar association was found between the decrease of MCP-1 and PFS (p=0.019, fig.3D).
OS and biomarkers
Median OS was 8.5 months (95% CI, 5.0-12.5), the 12-month OS rate and 18-month OS rate were 33.3% (95% CI, 16.8- 50.9) and 24.7% (95% CI, 10.2- 42.4), respectively (fig. 2D). There was no statistical difference between the median OS from cohort A and that from cohort B. 4 cycles or more combination therapy was a parameter that could be correlated with longer OS (HR 0.595, [95% CI, 0.398-0.89], p=0.012; Additional file 2: fig. S1C), while the correlation between PD-L1 expression and OS was not established (p=0.499, Additional file 2: fig. S1C). Whole-exome sequencing results showed that 1.37Neos/Mb as cutoff value of TNB in this study could be a prognostic biomarker, and patients with TNB of greater than 1.37Neos/Mb had significant longer OS (p=0.048, fig. 3C). Analysis of serum cytokines detected the concentration of sFASL or IFN-γ dropped significantly in patients with longer OS (p=0.00076, p=0.032; fig. 3E). The change of Granulysin, MCP-1, IL-17a, IL-23, TNF-α and Granzyme B in serum following the combination therapy had no statistical influence on OS ( fig. 3E; Additional file 2: fig. S6).