Introduction

Immune checkpoint inhibitors (ICIs), such as programmed death 1 (PD-1)/programmed death-ligand 1 (PD-L1) antibodies, have markedly changed the treatment strategies for advanced non-small cell lung cancer (NSCLC) (Borghaei et al. 2015; Brahmer et al. 2015) and various other cancers. Among the most important benefits of ICIs are their long-lasting tumor-inhibiting effects and potentially curative action (Reck et al. 2019). However, the rate of long-term inhibition is only ~ 10% (Borghaei et al. 2015), and the establishment of predictive biomarkers is important for ICI therapy. Only the PD-L1 tumor proportion score (TPS) in tumor tissue samples is an established clinical biomarker for NSCLC (Reck et al. 2016), although there are some conflicting data (Carbone et al. 2017; Zhang et al. 2020). Pembrolizumab has become a standard therapy in previously untreated NSCLC patients with PD-L1 ≥ 50%, where more than half of the patients were still alive after 30 months of treatment with pembrolizumab in the KEYNOTE-024 trial (Reck et al. 2019). Combination treatment with ICI and platinum doublet chemotherapy was shown to confer a significant survival benefit compared to chemotherapy alone in multiple recent phase III clinical trials and has become a standard therapy in advanced NSCLC regardless of PD-L1 expression (Gandhi et al. 2018; Socinski et al. 2018). However, it remains unclear whether pembrolizumab alone or ICI in combination with chemotherapy is better for NSCLC patients with PD-L1 ≥ 50%. PD-L1 is also an important biomarker when determining the treatment for patients intolerant to platinum doublet chemotherapy.

PD-L1 expression was determined in tumor tissue samples, and the use of cytology samples was not permitted in the KEYNOTE-024 trial (Reck et al. 2016). Therefore, the clinical significance of PD-L1 expression in cytology samples remains unclear. However, tissue samples appropriate for determining PD-L1 levels are not always available in clinical practice, whereas the acquisition of cytology samples is minimally invasive and could be available even in patients whose histology sample is unavailable. Therefore, determining the significance of PD-L1 expression in the cytological samples is very important for clinical practice.

This study was performed to investigate the clinical significance of PD-L1 expression in cytological samples, including cell block (coelomic fluid) and endobronchial ultrasound-guided (EBUS) transbronchial needle aspiration (TBNA) samples, in advanced NSCLC.

Materials and methods

Patients

This branch study analyzed our data from the Okayama Lung Cancer Study Group-Immunotherapy Database (OLCSG-ID), which was established to allow various analyses of ICI efficacy for NSCLC (Ichihara et al. 2020a, b). OLCSG-ID includes the medical records for NSCLC patients who received PD-1/PD-L1 antibody monotherapy at nine institutions between December 2015 and May 2018.

Response assessment

Responses were revaluated by each investigator according to the Response Evaluation Criteria In Solid Tumors (RECIST) version 1.1; the iRECIST criteria, which are often used in ICI studies, were not employed.

Statistical analysis

Overall survival (OS) was evaluated from the beginning of ICI therapy to the day of death from any cause, and progression-free survival (PFS) was evaluated from the beginning of ICI therapy to the day of disease progression or death from any cause using the Kaplan–Meier method. Multivariate analysis was performed using the Cox proportional hazards model. Statistical analyses were performed using STATA software (version 11.0; StataCorp, College Station, TX, USA). In all analyses, p < 0.05 was taken to indicate statistical significance.

Results

Patient characteristics

We retrospectively investigated 265 NSCLC patients in whom PD-L1 expression had been determined. We excluded one patient whose PD-L1 expression was determined from an unknown sample, and therefore finally analyzed the data of 264 patients. The patients were divided into two groups: those whose PD-L1 TPS was determined in cell block using coelomic fluid or EBUS-TBNA samples (cytology group), and those in whom tissue samples were analyzed (histology group). The patient characteristics are listed in Table 1. The cytology and histology groups consisted of 55 and 209 patients, respectively. There were no significant differences in demographic characteristics between the two groups, except for the disease stage. The rate of postoperative recurrence was significantly higher in the histology group, suggesting that resected tumor samples were likely used for the PD-L1 testing in this group. The majority of the patients were male, had a good performance status (PS), were smokers, and had adenocarcinoma histology.

Table 1 Patients characteristics
Full size table

The proportions of patients with a PD-L1 TPS of 0%, 1–49%, and ≥ 50% were 13, 26, and 61%, respectively, in the cytology group, and 22, 30, and 48%, respectively, in the histology group (Fig. 1).

Fig. 1
figure 1

Patient rates according to programmed cell death ligand 1 (PD-L1) expression. a Proportions of patients with a PD-L1 tumor proportion score (TPS) of 0%, 1–49%, and ≥ 50% in cytology samples and b histology samples

Full size image

Response to ICIs according to PD-L1 expression

The objective response rate (ORR) was determined according to PD-L1 expression in both the cytology and histology groups (Fig. 2). PD-L1 expression predicted response to ICI treatment in both the cytology and histology groups (cytology group, TPS < 50 vs. ≥ 50%: 17.6 vs. 48.4%; histology group: 18.1 vs. 42.6%).

Fig. 2
figure 2

Objective response rate (ORR) to immune checkpoint inhibitors (ICIs). The ORR to ICIs in the cytology and histology groups are shown

Full size image

Higher PD-L1 expression in cytological specimens predicted PFS

We determined the PFS according to PD-L1 expression in both groups. In the cytology group, a higher PD-L1 expression level predicted significantly better PFS, with cutoff values of both 50 and 1% (PD-L1 ≥ 50 vs. < 50%: median PFS, 6.5 vs. 1.9 months, hazard ratio [HR], 0.43, log-rank test p = 0.008, Fig. 3a; PD-L1 ≥ 1 vs. < 1%: median PFS, 4.2 vs. 1.5 months, HR, 0.18, log-rank test p < 0.001, Fig. 3b). Similar results were obtained in the histology group (PD-L1 ≥ 50 vs. < 50%: median PFS, 8.1 vs. 3.2 months, HR, 0.49, log-rank test p < 0.001, Fig. 3c: PD-L1 ≥ 1 vs. < 1%, median PFS, 6.5 vs. 2.5 months, HR, 0.50, log-rank test p < 0.001, Fig. 3d). We also determined OS. In the cytology group, PD-L1 expression with a cutoff value of 50% was not significantly associated with OS (median OS, not reached vs. 12.4 months, HR, 0.40, log-rank test p = 0.393, Fig. 4a), while that with a cutoff value of 1% was significantly associated with OS (median OS, not reached vs. 3.9 months, HR, 0.27, log-rank test p = 0.023, Fig. 4b). In the histology group, higher PD-L1 expression predicted better OS with cutoff values of both 50 and 1%, although the difference with a cutoff value of 1% was not significant (PD-L1 ≥ 50 vs. < 50%: 28.4 vs. 16.1 months, HR, 0.59, log-rank test p = 0.018, Fig. 4c; PD-L1 ≥ 1 vs. < 1%: 20.4 vs. 16.1 months, HR, 0.67, log-rank test p = 0.102, Fig. 4d).

Fig. 3
figure 3

Progression-free survival (PFS) according to PD-L1 expression. a PFS in the cytology group using a PD-L1 expression cutoff value of 50%. b PFS in the cytology group using a PD-L1 expression cutoff value of 1%. c PFS in the histology group using a PD-L1 expression cutoff value of 50%. d PFS in the histology group using a PD-L1 expression cutoff value of 1%

Full size image
Fig. 4
figure 4

Overall survival (OS) according to PD-L1 expression. a OS in the cytology group using a PD-L1 expression cutoff value of 50%. b OS in the cytology group using a PD-L1 expression cutoff value of 1%. c OS in the histology group using a PD-L1 expression cutoff value of 50%. d OS in the histology group using a PD-L1 expression cutoff value of 1%

Full size image

Discussion

Some previous studies have reported a correlation of PD-L1 expression between tissue and cytology samples (Capizzi et al. 2018; Wang et al. 2018; Gagné et al. 2019). Therefore, it has been assumed that using cytology to determine PD-L1 can predict the efficacy of ICIs. However, there have been no studies on the direct correlation between PD-L1 expression in cytology samples and the efficacy of ICIs. In this study, we demonstrated that the PD-L1 expression level in cytological samples predicted the efficacy of ICIs.

Consistent with a previous study (Herbst et al. 2016), the proportions of patients with a PD-L1 TPS < 1%, 1–49%, and ≥ 50% were roughly equivalent in our histology group. In our cytology group, the proportion of NSCLC patients with a PD-L1 TPS ≥ 50% was high, accounting for more than 60% of the patients (Fig. 1). One plausible reason for this is that cytology samples may be more amenable to PD-L1 staining. However, this seems unlikely, because PD-L1 expression in cytology samples was reported to be equivalent to that in histology samples (Capizzi et al. 2018; Wang et al. 2018; Gagné et al. 2019). Another possible reason is that, in the group with cytology samples (which require the presence of coelomic fluid or lymph node swelling), inflammation may have been more severe with higher levels of PD-L1 expression.

With regard to the efficacy of ICIs, patients with a PD-L1 TPS < 1% in our cytology group showed an especially short response (all within 3 months of starting ICI therapy; Fig. 3). This is an important finding; although further studies are required for confirmation, PD-L1 < 1% in cytology samples may be a good marker to exclude ICI therapy, while PD-L1 < 1% in histology samples cannot fully exclude ICI (because some of the NSCLC cases still showed a long-term response to ICI therapy in a previous study) (Borghaei et al. 2015). Malignant pleural effusion is known to be a poor prognostic factor (Morgensztern et al. 2012), which may partly explain why such patients in our cytology group, with a TPS < 1%, showed poor outcomes.

In contrast to previous clinical trials of ICI (Herbst et al. 2016), some of the analyses conducted in the present study, such as the cytology sample analyses, indicated that, when the cutoff value was 50%, PD-L1 did not predict OS (Fig. 4a). In the present study, there was large variation among the patients in terms of regimen numbers prior to ICI therapy and OS was calculated from the start date of ICI therapy in all cases. Therefore, the OS data were heterogeneous, which probably reduced the differences in survival according to PD-L1 expression level.

This study had some limitations. First, the cell block preparation process was not standardized among institutions, which may have led to bias in the PD-L1 staining results. Second, we did not investigate the concordance between PD-L1 expression in cytology and that in histology using paired samples. Paired samples were not available in most patients, because cytology samples were used for PD-L1 examination only when histology samples were unavailable. Third, different from previous studies (Mok et al. 2019), tissue-determined PD-L1 with a 1% cut-off value did not predict OS (Fig. 4d), which is most likely to be due to many censored cases with an immature follow-up period, which would obscure the difference. Finally, this was a retrospective analysis of data that were heterogeneous in terms of patient cohorts and follow-ups, such that the results cannot be considered definitive and should therefore be interpreted cautiously.

In conclusion, PD-L1 expression in cytological samples, including cell block (coelomic fluid) samples, can predict the efficacy of ICI for NSCLC.