Abstract
Background/Aim: Programmed cell death ligand 1 (PD-L1) expression is a predictive marker for immunotherapy effects in advanced non-small cell lung cancer (NSCLC), but its association with patient characteristics or specimens is controversial. We aimed to retrospectively analyze the association of PD-L1 expression with clinicopathological features of NSCLC patients. Materials and Methods: The PD-L1 expression and clinicopathological features of NSCLC patients were assessed from January 2017 to June 2017 in the Tokyo Metropolitan Cancer and Infectious Diseases Centre, Komagome Hospital were reviewed (n=108). Results: For PD-L1 expressions of 0% and >1%, multivariate analysis showed that lymph node sample results were associated with positive PD-L1 expression. Archival samples and high serum carcinoembryonic antigen (CEA) levels were associated with negative PD-L1 expression. Sample preservation time and CEA levels correlated with PD-L1 expression. Conclusion: Nodal metastasis, sample preservation time and CEA levels were associated with PD-L1 expression in NSCLC.
Lung cancer is one of the most common cancers with poor prognoses (1). A combination therapy of platinum doublet and third-generation anticancer drugs has been considered to be the standard therapy for advanced non-small cell lung cancer (NSCLC) without any gene mutations [epidermal growth factor receptor (EGFR) mutation, ALK rearrangement and ROS1 mutation] (2). However, in recent years, immune checkpoint inhibitors have changed the treatment of NSCLC.
Programmed cell death ligand 1 (PD-L1) is a ligand of the PD-1 protein, that is expressed on immune cell membranes. PD-L1 has an important role in cancer growth by inducing local immune tolerance to tumors. Nivolumab and pembrolizumab have been developed to provide anti-PD-1 antibody release immune tolerance and activate the immune function to cancer.
PD-L1 expression in tumors is considered to be a predictive biomarker of the benefit of immunotherapy on the basis of clinical trial results (3-8). PD-L1 expression is assessed quantitatively by the tumor proportion score (TPS) after immunohistochemistry (IHC) analysis. The American Food and Drug Administration has approved 4 IHC methods (22C3, 28-8, SP142 and E1L3N) for assessing PD-L1 expression.
Although some studies including meta-analyses have been conducted to detect clinical and pathological features related to PD-L1 expression, the associations between clinical and pathological features in patients and PD-L1 expression remain controversial. Past studies have reported many features related to PD-L1 expression. Male sex (9, 10), smoker (9, 11, 12), squamous cell carcinoma (9, 13), cavitation (14), nodal metastasis (9, 15, 16) surgically resected samples (17), CD8-positive T-cells in tumors (16), EGFR mutation (15, 18), ALK rearrangement (18) and K-RAS mutation (18, 19) have been associated with positive PD-L1 expression in various studies, but the findings have not been consistent. It is possible that these differences were caused by different assays and antibodies used in the previous studies or by unknown factors. Therefore, we conducted this retrospective study to identify new clinical and pathological features related to PD-L1 expression and their relationships with PD-L1 expression in NSCLC patients.
Materials and Methods
Patients and samples. We reviewed the clinical records of all NSCLC patients whose samples had been submitted to assess PD-L1 expression from January 2017 to June 2017 in the Tokyo Metropolitan Cancer and Infectious Diseases Centre, Komagome Hospital. A total of 108 patients were identified, and their PD-L1 expression levels and clinical and pathological features were analysed. All sample types were included in this study. Our study followed the World Medical Association Declaration of Helsinki recommendations and was approved by the Institutional Review Board of Tokyo Metropolitan Cancer and Infectious Diseases Centre, Komagome Hospital.
Patient's features and immunohistochemistry. The following background factors were examined: age, sex, past history, drug history, smoking history, performance status, histology, TNM stage, sample preservation time, organ origin, diagnostic procedure [transbronchial biopsy (TBB), surgical operation, endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA)], EGFR mutation, anaplastic lymphoma kinase rearrangement, serum carcinoembryonic antigen (CEA) level, serum squamous cell carcinoma (SCC) antigen level, serum cytokeratin 19 fragment (CYFRA) level, neutrophil/lymphocyte ratio (NLR), C-reactive protein (CRP) and tumor cavitation. A heavy smoker was defined as a former smoker who had smoked >30 pack-years or current smoker. Metformin and statin were selected as drug history factors because they have been reported to be associated with tumor immunity or prevention of carcinogenicity (20, 21). The histologic pattern of the sample was assessed by >2 pathologists in accordance with the International Association for Lung Cancer/American Thoracic Society/European Respiratory Society. Samples preserved for >30 days were defined as archival samples. We investigated all blood samples taken on the day nearest to when the tumor specimen was taken. On the basis of values used in previous studies (22-26), the following cutoff values were used: CEA, 5.0 ng/ml; SCC, 1.5 ng/ml; CYFRA, 3.5 ng/ml; CRP, 0.5 mg/dl; and NLR, 3. Two independent respirologists assessed tumor cavitation by chest computed tomography performed on the day nearest to when the specimen was taken. We compared the clinical and pathological features between the patients with a TPS of 0% and those with a TPS of >1%. A PD-L1 IHC 22C3 pharm Dx kit was used to assess all samples. Two other pathologists assessed PD-L1 expression independently by central evaluation, which is required by insurance companies in Japan.
Statistical analysis. We performed univariate analysis to identify clinical factors associated with PD-L1 expression by using Fischer's exact test. Variables from previous study reports and from our univariate analysis were selected for multivariate analysis using multiple logistic regression analysis. If possible, analysis of correlations between PD-L1 expression and variables were performed by using Pearson's correlation coefficient or Spearman's rank correlation coefficient. All statistical analyses were performed by using EZR (Saitama Medical Centre, Jichi Medical University, Saitama, Japan), which is a graphical user interface for R that is a modified R commander designed for biostatistics. When p-value was lower than 0.05, differences were considered statistically significant.
Results
Patient characteristics and univariate analysis. The patient's characteristics and univariate analysis results for the 108 patients who submitted their specimens to assess PD-L1 expression from January 2017 to June 2017 are shown in Table I. Of these 108 patients, 35 (32%) were TPS negative and 71 (65%) were TPS-positive. Two samples were assessed as inadequate because of an insufficient number of cells. There were 52 patients >70 years old. There was a greater tendency towards TPS positivity with increasing age, but the difference in TPS positivity was not significant between older and younger patients. There was no significant difference in the number PD-L1 expression status between males and females. Histories of diabetes mellitus and malignant disease were the most common previous illnesses among the patients. There were no differences in PD-L1 status between these 2 factors. Autoimmune disease was observed in 4 patients. All tumors of autoimmune disease patients were TPS-positive, but the differences in TPS positivity between these patients and others were not significant. Four patients received metformin treatment, and 13 patients received statin treatment; neither of these factors were related to TPS status. Among heavy smokers, there was a tendency towards more patients to be TPS-positive. Performance status was not related to PD-L1 expression. Adenocarcinoma was significantly associated with TPS negativity (p=0.033). In contrast, SCC was significantly associated with TPS positivity (p=0.038). Cancer stage was not related to TPS. There were more archival samples for the patients with >N1 stage than for the patients with N0 stage (the archival sample/recent sample ratio was 43/28 for patients with >N1 stage and 16/19 for patients with N0 stage). Archival samples were significantly associated with TPS negativity (p=0.012). In addition, lymph node samples tended to be associated with TPS positivity. EBUS-TBNA, TBB, CRP >0.5 ng/ml, and tumor with cavitation tended to be associated with TPS positivity also. High serum CEA level was significantly associated with TPS (p=0.040). Eleven patients had EGFR mutation-positive tumors and were not associated with TPS status. Three patients had ALK rearrangement and all samples were TPS-positive; however, ALK rearrangement was not significantly associated with PD-L1 expression.
Multivariate analysis and correlation between TPS and clinical factors. We selected heavy smoker, adenocarcinoma, archival sample, CEA >5 ng/ml, lymph node sample, CRP >0.5 mg/dl and cavitation as variables. Squamous cell, EBUS-TBNA and necrosis were excluded because they exhibited multicollinearity. The analysis results are shown in Table II. Lymph node, archival sample and CEA >5.0 ng/ml were significantly associated with TPS.
Patient characteristics and univariate analysis results.
Correlations among sample preservation time and CEA level with TPS. We determined the correlations among TPS and sample preservation time and CEA level by Spearman's rank correlation coefficient because these 2 factors showed non-regular distributions. Sample preservation time and serum CEA level were negatively correlated with TPS (sample preservation time: r=−0.366; p=0.0001; Figure 1A; serum CEA level: r=−0.209, p=0.032; Figure 1B).
Discussion
In this study, we found that lymph node sample, serum CEA level and archival sample were associated with PD-L1 expression in NSCLC. EGFR mutation and ALK rearrangement were not associated with PD-L1 expression status in the multivariate or univariate analyses. These results may have been affected by the small number of patients with EGFR mutations and ALK rearrangements in our study. High serum CEA level and archival sample were identified as independent factors associated with our study. Correlations between TPS and 2 significant factors identified in multivariate analysis were observed, which were consistent with the results of our multivariate analysis.
Correlations between TPS and sample preservation time or CEA level. *Spearman's rank correlation coefficients were determined.
PD-L1 expression has been found to be predictive of the benefit of immunotherapy on advanced NSCLC and poor prognosis of NSCLC patients (27). On the other hand, PD-L1 is affected by inflammatory cytokines, such as interferon gamma (28). A previous study found a discrepancy between PD-L1 expression in the primary tumor and in ≤38% of nodal metastases (16). Heterogeneity of PD-L1 expression in tumors has been reported in previous studies (29-31). PD-L1 expression was assessed by several kinds of methods. Although PD-L1 expression is assessed by IHC, previous studies report PD-L1 mRNA which assessed by real time chain reaction was useful for assessing PD-L1 expression (32). However, previous study reported the differences of assessment of PD-L1 expression between each antibodies of IHC (33). On the basis of these findings, PD-L1 is considered a dynamic biomarker, and exact assessment of PD-L1 expression is difficult. TPS is not considered to be a biomarker reliably predictive of the treatment effect of anti-PD-1 antibodies, such as EGFR mutation, which is considered to be a biomarker reliably predictive of a promising EGFR-TKI treatment effect. Therefore, identification of clinical factors related to PD-L1 expression in lung cancer that are predictive of a beneficial effect of treatment for NSCLC would be useful clinically.
Multivariate analysis result.
Nodal metastasis has been reported to be associated with positive PD-L1 expression in lung cancer and various other cancers. (e.g. head and neck, gastric, hepatocellular, renal cell, oesophageal, pancreatic and ovarian) (34) On the other hand, the present study showed that N stage was not associated with PD-L1 expression. We recognised that these results were caused by a higher number of samples with an N stage >1 that had been preserved for >30 days rather than by the N stage.
One previous report showed that archival samples showed lower PD-L1 expression than that of recent samples (11). On the other hand, previous study (6) showed no difference in treatment effects for archival or recent samples used to identify TPS-positive patients. As in the previous study, archival samples preserved for >30 days were more frequently evaluated as TPS negative than were recent samples. Therefore, sample preservation time was correlated with TPS values in our study. These results suggest that samples should be submitted earlier to obtain more accurate results for TPS status and that TPS status in archival samples is more likely to give false negatives. In contrast, it is unlikely that an archival sample assessed as TPS-positive will give a false positive. The reason for these differences is thought to be related to time-related sample deterioration.
CEA is a glycoprotein normally produced only during the fetal period in healthy humans. Serum CEA level increases in patients with various cancers. NSCLC, especially adenocarcinoma, can produce high serum CEA levels in patients (22). Although high serum CEA level was associated with negative TPS in our study, the underlying mechanism is unknown. The relationship between tumors producing CEA and PD-L1 expression has not been previously investigated in clinical or basic studies. Consequently, this is the first study to report the association between serum CEA level and PD-L1 expression. Regardless of histology, serum CEA level was found to be associated with TPS status in multivariate analysis that included histological results. CEA-producing cancers with any histology may have lower PD-L1 expression.
There were some limitations in this study that should be considered. Selection bias may have occurred because this was a retrospective single-centre study. Second, we could not analyse the relationship between the treatment effect of immune checkpoint inhibitors and PD-L1 expression status, although a previous study reported that PD-L1 expression was considered to be a dynamic biomarker that may change with each clinical phase (35-36). One study reported efficacy and safety of rebiopsy for assessing PD-L1 expression (37). Future, rebiopsy for assessing PD-L1 expression may be required. Previous studies have reported that older lung cancer patients and EGFR mutation-positive patients had inadequate effects of treatment with immune checkpoint inhibitors (5, 6). Consequently, investigation of the relationships among the effect of immune checkpoint inhibitor treatment, PD-L1 expression and other factors, such as mutation burden, will be required. Third, the rationale for selection of the variables used in the multivariate analysis should be discussed. Smoking status, histology, nodal metastasis and cavitation have been reported in previous studies, and they showed tendencies towards being associated with PD-L1 expression in our study (9, 11, 12-16). CRP and CEA have never been previously reported as being associated with PD-L1 expression. However, we included these variables because they showed non-negligible tendencies towards associations with PD-L1 expression in our univariate analysis.
Conclusion
Lymph node sample, high serum CEA level and archival samples were found to be associated with PD-L1 expression. Nodal metastasis in lung cancer was also found to be predicted by PD-L1 expression. Serum CEA level and sample preservation time correlated with TPS for PD-L1 assessed by using the IHC 22C3 method. We also found that PD-L1 expression could be underestimated in archival samples, and that serum CEA level more accurately correlated with PD-L1 expression in NCSLC.
Acknowledgements
This study was not funded by any organization.
Footnotes
Conflicts of Interest
The Authors declare no conflicts of interest associated with this manuscript.
- Received November 15, 2017.
- Revision received December 11, 2017.
- Accepted December 12, 2017.
- Copyright© 2018, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.