Introduction

Salivary gland carcinoma (SGC) is a relatively rare cancer that accounts for 0.3% of all cancers1. According to a histological classification of the World Health Organization, SGCs are classified into more than 20 histological types, each with different biological characteristics and prognosis2. Cytotoxic chemotherapy has been indicated for recurrent and/or metastatic (R/M) disease as a conventional systemic therapy, regardless of histopathological type3,4. While recent studies have revealed the efficacy of targeted therapy for SGC, including anti-androgen receptor (AR), HER2, and NTRK therapies5,6,7,8, the majority of patients with R/M SGC are not candidates for these targeted therapies.

The use of immune checkpoint inhibitors (ICIs) in SGC has been extensively studied over the past few years9,10,11,12,13,14. The latest NCCN guidelines recommend pembrolizumab, an anti-PD-1 antibody, for SGC with Microsatellite Instability (MSI)-High or Tumor Mutational Burden (TMB)-High8,15. In contrast, other studies note that the efficacy of ICIs is theoretically greater against tumors with high PD-L1 expression16,17,18. For head and neck squamous cell carcinoma, dosing is generally determined based on the results of the combined positive score (CPS) divided by evidence of therapeutic efficacy16. In SGC, the KEYNOTE-158 trial reported that PD-L1-positive patients (CPS ≥ 1) had a better response rate than PD-L1-negative patients, suggesting that CPS may be a biomarker for pembrolizumab in SGC9. However, the impact of PD-L1 expression on the efficacy of pembrolizumab in SGC patients remains controversial.

This study aimed to clarify the efficacy and safety of pembrolizumab monotherapy in R/M SGC in a real-world setting, and to measure CPS and tumor proportion score (TPS) as real values for PD-L1 expression to identify their optimal cut-off values for this treatment. Further, regarding treatment-related biomarkers, we investigated the association of patient background factors and hematological markers with clinical outcomes of pembrolizumab.

Results

Patient characteristics and treatment sequences

Twenty-seven patients with R/M SGC were treated with pembrolizumab monotherapy during the study period. No patients received pembrolizumab with chemotherapy concurrently. Patient characteristics are summarized in Table 1. Data cutoff date was April 30, 2021, and median follow-up period for all patients was 6.9 months (range, 2.1–58.8 months). The most common histopathological type was salivary duct carcinoma (SDC) (n = 15; 56%). Sixteen patients (59%) had received chemotherapy before pembrolizumab, all of whom had disease progression within 6 months prior to pembrolizumab initiation. Median follow-up period of patients with SDC was 6.9 months (range, 2.1–14.0 months). While 4 patients received pembrolizumab as the first-line treatment, 11 of 15 (73%) patients with SDC had received chemotherapy and then had disease progression prior to pembrolizumab monotherapy. Trastuzumab was administered to 6 patients with HER2-positive SGC.

Table 1 Characteristics of patients at baseline. ECOG Eastern Cooperative Oncology Group, NOS not otherwise specified, Tmab Trastuzumab, DTX docetaxel, PTX paclitaxel, CDDP Cisplatin, CBDCA carboplatin, FP fluorouracil–cisplatin, Cmab Cetuximab, SDC salivary duct carcinoma, CPS combined positive score, TPS tumor proportion score, NLR neutrophil–lymphocyte ratio, PLR platelet-lymphocyte ratio, LMR lymphocyte-monocyte ratio, mGPS modified Glasgow Prognostic Score.
Full size table

At the end of the study period, 10 patients (37.0%) continued to receive pembrolizumab whereas 17 patients (63.0%) discontinued treatment due to progressive disease (PD) (n = 16; 94.1%) or immune-related adverse event (irAE) (n = 1; 5.9%). The median number of cycles of pembrolizumab administered was 4 (range, 1–22) in all patients and 4 (range, 1–14) in patients with SDC. Twelve patients (44%), including 10 SDC patients, received 1 or more sequential chemotherapy regimens after pembrolizumab treatment.

Response and survival outcomes

The efficacy of pembrolizumab therapy is shown in Table 2. In all 27 patients, 1 (4%), 6 (22%), 5 (19%) and 15 (56%) patients showed CR, PR, stable disease (SD) and PD, respectively. Overall response rate (ORR) was 25.9% [95% confidence interval (95%CI), 11.1–46.3], and clinical benefit rate (CBR) and disease control rate (DCR) were 33.3% (95%CI, 16.5–54.0) and 44.4% (95%CI, 25.5–64.7), respectively. The Kaplan–Meier survival curves of progression-free survival (PFS) and overall survival (OS) of all patients are shown in Fig. 1. Median PFS and median OS were 2.6 months (95% CI, 1.9–not estimable [NE]) and not reached (NR) (95% CI, 11.5–NE), and 1-year PFS and OS were 30.3% (95% CI, 11.1–52.3) and 75.1% (95% CI, 33.2–92.8), respectively.

Table 2 Best overall response. SDC salivary duct carcinoma, AdCC adenoid cystic carcinoma.
Full size table
Fig.1
figure 1

Kaplan–Meier survival curves of progression-free survival (a) and overall survival (b) of all patients. Vertical lines show censored events, and the shaded regions represent 95% CIs.

Full size image

Response and survival outcomes by histological type

The therapeutic effects observed in 15 SDC patients were as follows: ORR, 26.7% (95%CI, 7.8–55.1); CBR, 33.3% (95%CI, 11.8–61.6); DCR, 40.0% (95%CI, 16.3–67.7); median PFS, 2.2 months (95% CI, 1.3–NE); and median OS, NR (95% CI, NE–NE) (see Supplementary Fig S1 online). None of the six patients with adenoid cystic carcinoma (AdCC) showed a response, and three of these AdCC patients died due to disease progression. Figure 2 shows images of the tumor before and during pembrolizumab monotherapy in a representative patient.

Fig.2
figure 2

Representative images of the tumor before and during pembrolizumab monotherapy in a patient with pancreatic metastasis of salivary duct carcinoma. (a) Pre-treatment of the patient with pancreatic metastasis. (b) Tumor shrinkage observed 63 days after the initiation of pembrolizumab monotherapy. (c) Tumor disappeared 154 days after the initiation of pembrolizumab monotherapy.

Full size image

Safety

All AEs reported are listed in Table 3. Nine patients (33.3%) experienced irAE during treatment. One patient (3.7%) showed grade 3 irAE diabetes. No grade 4 or greater irAEs were observed. One patient discontinued treatment due to the irAE of grade 2 uveitis.

Table 3 Adverse events.
Full size table

PD-L1 status

CPS and TPS are shown in Supplementary Table S1 and S2 online. In all 25 cases with numerical evaluation, median CPS and TPS was 5 (range, 0–180) and < 1% (range, 0–100), respectively. The optimal cut-off value of CPS and TPS against objective response was 15 and 25%, respectively (see Supplementary Fig S2 online). Using these cutoffs to divide patients into 2 groups, numbers of CPS-high (CPS-H, CPS ≥ 15) vs. CPS-low (CPS-L, CPS < 15) SGC, and TPS-high (TPS-H, TPS ≥ 25%) vs. TPS-low (TPS-L, TPS < 25%) SGC were 9 (36%) vs. 16 (64%), and 4 (16%) vs. 21 (84%), respectively. ORR of pembrolizumab monotherapy in the CPS-H SGC group was significantly better than that in the CPS-L SGC group (56% vs. 13%), and ORR in the TPS-H SGC group was also significantly better than that in the TPS-L SGC group (75% vs. 19%; Table 4). Median PFS of CPS-H SGC and CPS-L SGC was not reached and 2.3 months, respectively, while that of TPS-H SGC and TPS-L SGC was not reached and 2.2 months, respectively. There were no significant differences in PFS by CPS or TPS status (Fig. 3).

Table 4 Univariate analysis of predictive factors for overall response rate and progression-free survival in all patients (N = 27). CPS combined positive score, ECOG Eastern Cooperative Oncology Group, -H high, -L low, LMR lymphocyte-monocyte ratio, mGPS modified Glasgow Prognostic Score, NE not estimated, NLR neutrophil–lymphocyte ratio, NR not reached, ORR overall response rate, PFS progression-free survival, PLR platelet-lymphocyte ratio, SDC salivary duct carcinoma, TPS tumor proportion score. The optimal cut-off value of CPS and TPS against objective response was 15 and 25%, respectively.
Full size table
Fig.3
figure 3

Kaplan–Meier survival curves of SGC patients according to CPS (a) and TPS (b) on PFS. Vertical lines show censored events.

Full size image

Similarly, in 14 patients with SDC, median CPS and TPS was 10% (range, 1 – 100%) and < 1% (range, 0 – 70%), respectively. Numbers of CPS-H vs. CPS-L SDC and TPS-H vs. TPS-L SDC were 7 (50%) vs. 7 (50%), and 2 (14%) vs. 12 (86%), respectively. ORRs for pembrolizumab monotherapy between CPS-H SDC (43%) and CPS-L SDC (14%), and those between TPS-H SDC (50%) and TPS-L SDC (25%) showed no significant differences (Supplementary Table S3). Median PFS of CPS-H SDC and CPS-L SDC was 2.6 months and 2.0 months, respectively, and median PFS of TPS-H SDC and TPS-L SDC was 2.6 months and 2.2 months, respectively (Supplementary Table S3). Neither CPS nor TPS status significantly correlated with PFS in SDC patients (Supplementary Figure S4).

In 6 patients with AdCC, median CPS and TPS was low at 2.5 (range 0–5) and < 1% (range, 0–1%), respectively. SDC patients had significantly higher CPS than AdCC patients (Supplementary Figure S3). In one patient each with lymphoepithelial carcinoma and acinic cell carcinoma, CPS was high at 180 and 70, and TPS was 100% and 30%, respectively. One patient with adenocarcinoma was negative for CPS and TPS.

Impact of other biomarkers on clinical response

In all 27 patients, median neutrophil–lymphocyte ratio (NLR), platelet-lymphocyte ratio (PLR) and lymphocyte-monocyte ratio (LMR) was 3.3 (range 1.1–12.7), 194.1 (range 68.3–1331.3) and 3.1 (range 1.1–12.9), respectively (Table 1). The optimal cut-off value of NLR, PLR and LMR against objective response was 3.0, 232.4 and 4.5, respectively. We divided the patients accordingly into high and low groups to clarify the impact of NLR, PLR, and LMR on clinical outcomes. Seventeen (63%), 9 (33%), and 9 (33%) patients were categorized in the NLR-, PLR-, and LMR-high groups, and 10 (37%), 18 (67%), and 18 (67%) patients were categorized in the NLR-, PLR-, and LMR-low groups, respectively. Regarding the modified Glasgow Prognostic Score (mGPS), 24 (89%) and three (11%) patients were assigned scores of 0 and 1, respectively. No patient was assigned a score of 2.

Regarding hematological biomarkers, patients with higher PLR had a significantly worse PFS than those with lower PLR among all patients [Table 4, hazard ratio (HR) = 2.8; 95% CI, 1.0–7.8; p = 0.049].

Discussion

Here, we demonstrated the efficacy of immunotherapy with pembrolizumab in patients with R/M SGC, with an acceptable rate of irAE. Compared to patients with CPS-L and TPS-L SGC, patients with CPS-H and TPS-H SGC had significantly higher ORRs, respectively.

In our cohort, the ORR of pembrolizumab in patients with SGC and SDC was 25.9 and 26.7%, respectively. In previous clinical trials, the ORR of pembrolizumab monotherapy in SGC ranged from 4.6% – 11.5%8,9,13. Nivolumab has shown similar efficacy in R/M SGC10,12. The large differences in ORR between the previous and our present data are likely due to the abundance of SDCs with high PD-L1 expression, despite the difficulty of direct comparison given that many previous reports reported PD-L1 positivity only. We counted numerical scores for CPS and TPS in most patients, and determined optimal cut-off values of 15 and 25%, respectively. The ORRs of pembrolizumab in patients with R/M SGC with CPS-H and TPS-H were 55.6% and 75.0%, all of which were significantly higher than those with CPS-L and TPS-L, respectively. Higher CPS and TPS also tend to be associated with longer PFS in patients with R/M SGC treated with pembrolizumab. No patient with AdCC had high PD-L1 expression, and none responded to pembrolizumab. In contrast, although a good response was observed in two patients with SDC and myoepithelial carcinoma with no or low PD-L1 expression, a poor response was observed in the patient with the highest PD-L1 expression in SDC (Table S2). These results suggest that, despite their limitations, CPS and TPS may be useful biomarkers for the selection of pembrolizumab for SGC.

In addition to PD-L1, other factors in the tumor immune microenvironment (TIME) in SGC are important and can be targeted. The baseline TIME is heterogeneous among different histological types of SGC19,20,21: AdCC and myoepithelial carcinomas generally possess a low tumor mutational burden while SDC is reported to possess a high mutational burden and a disproportionately higher degree of immune infiltration, which is greater indeed than even those of lung adenocarcinoma and melanoma19. Similarly, heterogeneity exists in PD-L1 expression by histological type of SGCs. In particular, SDC express higher levels of PD-L1 than other subtypes of SGC20. In our recent analysis, 93 (53.1%) of 175 SDCs were positive for CPS (≥ 1), and elevated expression levels of PD-L1, CD8, FOXP3, PD1, CTLA4, and LAG3 in TIME were associated with poor prognosis21. Taken together with the treatment results of our present and previous studies10,14, SDC might respond better to ICIs than other types of SGC. Nevertheless, given that the response rate of ICIs in SGC patients is still unsatisfactory, the development of combination treatment with multiple ICIs and with an ICI and other types of antitumor drugs is warranted. In a phase 2 trial of nivolumab plus ipilimumab14, the ORR for non-AdCC SGC patients was 16%, suggesting promising efficacy in this cohort, especially for SDC. Results from ongoing trials of pembrolizumab plus docetaxel (NCT03360890) and pembrolizumab plus pemetrexed (NCT04895735) in R/M SGCs are awaited.

Hematological, inflammatory, and nutritional biomarkers predict prognosis in patients with diverse histological types of neoplasia treated with ICI22,23,24,25,26. These are standardized markers which are routinely available in daily practice. Among them, increased PLR was associated with worse prognosis in patients with non-small-cell lung cancer, urothelial carcinoma, and upper gastrointestinal cancer treated with pembrolizumab27,28,29,30, as the decreased number of lymphocytes reflects attenuated antitumor immunity while activated platelets protect tumor cells from immune elimination within the circulatory system and support the establishment of secondary lesions31. In the present study, the ORR of pembrolizumab in patients with R/M SGC with low PLR was 38.9%, which was higher than the 0% in those with high PLR. Patients with low PLR also had a significantly longer PFS. These results suggest that PLR might be a predictive marker of clinical outcome in patients with R/M SGC treated with pembrolizumab.

The frequency of irAE in patients with SGC in our cohort was consistent with previous reports and no unexpected toxicity was observed. Almost all patients were able to continue treatment without discontinuation due to irAEs, except for one patient with grade 2 uveitis.

Several limitations of this study should be addressed. First, the study was retrospective and the sample size was modest, which resulted in low statistical power and might have produced a degree of selection bias. This should be considered against the fact that markedly few patients with SGC are treated with pembrolizumab, and that we nevertheless successfully included 27 patients from multiple institutions in Japan. Second, the impact of HER2 status and HER2-targeted therapies on pembrolizumab therapy has not been fully investigated. Third, the follow-up period after pembrolizumab treatment was not long enough to assess OS. However, this study mainly addressed ORR and PFS, since these outcomes directly reflect the efficacy of pembrolizumab in SGC patients. Our future goal is to assess pembrolizumab’s impact on subsequent chemotherapy, which might affect time to second disease progression (PFS2), and OS32.

In conclusion, this study demonstrated the efficacy and safety of pembrolizumab monotherapy in patients with R/M SGC. We identified optimal cut-off values of CPS and TPS in predicting the clinical outcomes of this treatment, where CPS-H and TPS-H SGC had significantly better ORR than CPS-L and TPS-L SGC. We also showed that PLR was significantly associated with PFS in patients with SGC.

Patients and methods

This study was conducted under a retrospective cohort design in 27 patients with R/M SGC treated with pembrolizumab from March 2016 to April 2021 at seven facilities in Japan, following approval from the ethics committee of the participating institutions (Approval number of each institution: International University of Health and Welfare, Mita Hospital, 5–21-17; Nagoya City University Hospital, 70–00-0145; Tokyo Medical University Hospital, T2021-0142; Kitasato University Hospital, C21-092; Tokai University Hospital, 21R105; Niigata University Hospital, 2021–0086; Yokohama City University Hospital, B210500037). This study was also performed in accordance with the Declaration of Helsinki. Regarding consent, informed consent was waived by the Institutional Review Boards of all participating institutions. We provided information regarding the research plan via a web-based public release, and patients and their families were provided an opportunity to opt-out of the study under the policy of the Japanese government.

The charts of all 27 patients who received pembrolizumab monotherapy were retrospectively reviewed, and information on patient characteristics, treatment modality and clinical outcomes was collected. Twenty-five tumor samples were pathologically reviewed and evaluated by a pathologist with expertise in SGCs (T.N.). Carcinoma ex pleomorphic adenoma was classified according to individual carcinoma components. We excluded patients for whom therapeutic effect could not be evaluated.

Treatment and follow-up

Pembrolizumab was administered at 200 mg/body every 3 weeks. After the start of treatment, imaging examination by CT or MRI was performed every 6–9 weeks. The efficacy of pembrolizumab was evaluated with regard to ORR, CBR (defined as the percentage of patients who achieved CR, PR, or SD for at least 24 weeks), DCR (defined as the percentage of patients who achieved CR, PR, or SD regardless of duration), median PFS, and median OS. Treatment efficacy was evaluated according to Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1). PFS was defined as the time from the start of pembrolizumab to the diagnosis of PD. OS was defined as the period from the day of initiation of pembrolizumab to the day of death regardless of cause or the day of last follow-up. AEs were assessed according to CTCAE version 5.0 up to 30 days after the last administration of pembrolizumab. Administration of pembrolizumab was terminated when PD was confirmed, the patient experienced unacceptable adverse events, or at patient request. After pembrolizumab discontinuation, additional treatment with other agents was permitted.

Evaluation of biomarkers

To clarify the impact of PD-L1 expression on survival and response, we evaluated CPS and TPS, each categorized into two groups by optimum cut-off value. PD-L1 expression in samples was assessed using the PD-L1 IHC 22C3 pharmDx assay (Agilent Technologies, Carpinteria, CA, USA) and characterized by CPS and TPS through central pathological review (H.H.). CPS and TPS were defined as the number of PD-L1-positive cells (tumor cells, lymphocytes, and macrophages) divided by the total number of tumor cells × 100 and as the number of PD-L1-positive tumor cells divided by the total number of tumor cells × 100, respectively. Optimum cut-off values of CPS and TPS were evaluated using area under the receiver operating characteristic curve (AUROC) against objective response.

Additionally, we also examined associations between survival or response and age, sex, Eastern Cooperative Oncology Group (ECOG) performance status (PS), irAE, histopathological type (SDC or non-SDC), previous therapies, and hematological markers. As hematological markers, we evaluated NLR, PLR, LMR and mGPS as examined in peripheral blood just before the start of pembrolizumab treatment. These markers were categorized into two groups by optimum cut-off value evaluated using AUROC against objective response. Regarding mGPS, patients with both elevated C-reactive protein (CRP; > 1.0 mg/dL) and decreased albumin (Alb; < 3.5 g/dL) were assigned a score of 2; those with an elevated CRP (> 1.0 mg/dl) and non-decreased Alb (≥ 3.5 g/dl) were assigned a score of 1; and those with a non-elevated CRP (≤ 1.0 mg/dl) were assigned a score of 0.

Statistical analyses

OS and PFS were estimated by the Kaplan–Meier product-limit method and tested by the log-rank test in both the SGC and SDC groups. To evaluate the impact of clinical factors, including CPS and TPS, on PFS we estimated HRs and 95% CIs using univariate Cox proportional hazards models. Additionally, to evaluate the impact of these factors on objective response, we performed Fisher's exact test. All statistical analyses were performed using STATA version 16 (Stata Corp., College Station, TX, USA). All tests were two-sided, and p-values < 0.05 were considered statistically significant.