Clinical Significance of Tumor Markers for Advanced Thymic Carcinoma: A Retrospective Analysis from the NEJ023 Study

Simple Summary Advanced thymic carcinoma (ATC) is rare. Owing to its rarity, there is limited information on the prognostic factors, and the optimal serum tumor markers are also unknown. We conducted a multi-institutional retrospective study of patients with ATC. In this study, we collected data on patient characteristics, progression-free survival (PFS), overall survival (OS), and tumor marker values, and investigated the relationship between tumor marker values and PFS/OS. We found that the neuron-specific enolase (NSE) level may be a useful prognostic tumor marker for ATC, regardless of histology. The findings of the analysis limited to squamous cell carcinoma suggested that the NSE and squamous cell carcinoma antigen levels may be useful prognostic factors. Abstract The optimal tumor marker for predicting the prognosis of advanced thymic carcinoma (ATC) remains unclear. We conducted a multi-institutional retrospective study of patients with ATC. A total of 286 patients were treated with chemotherapy. Clinicopathological information, including serum tumor markers, was evaluated to determine the overall survival (OS) and progression-free survival (PFS). The carcinoembryonic antigen, cytokeratin-19 fragment, squamous cell carcinoma (SCC) antigen, progastrin-releasing peptide, neuron-specific enolase (NSE), and alpha-fetoprotein levels were evaluated. In the Kaplan–Meier analysis, the OS was significantly shorter in the patients with elevated NSE levels than in those with normal NSE levels (median, 20.3 vs. 36.8 months; log-rank test p = 0.029; hazard ratio (HR), 1.55; 95% confidence interval (CI), 1.05–2.31 (Cox proportional hazard model)); a similar tendency regarding the PFS was observed (median, 6.4 vs. 11.0 months; log-rank test p = 0.001; HR, 2.04; 95% CI, 1.31–3.18). No significant differences in the OS and PFS were observed among the other tumor markers. In both univariate and multivariate analyses of the patients with SCC only, the NSE level was associated with the OS and PFS. Thus, the NSE level may be a prognostic tumor marker for thymic carcinoma, regardless of histology.


Introduction
Thymic epithelial tumors, including thymoma and thymic carcinoma (TC), are rare intrathoracic malignancies that occur in the prevascular mediastinum [1]. TC is an even rarer malignancy, with an annual incidence of 0.13 cases/100,000 population, and accounting for 15% of all thymic epithelial tumors [2,3]. It has a propensity to invade the surrounding tissues and metastasize, and one-half to two-third of all patients with TC are diagnosed with locally advanced or metastatic disease [4][5][6]. In the stage classification for TC, both the Masaoka staging system [7] and World Health Organization (WHO) TNM classification [8] are used, although the Masaoka staging system is used more frequently. The 5-year survival rate of patients with TC is 90% for Masaoka stage 1-2 and 30% for stage [3][4]. This indicates that the prognosis of advanced thymic carcinoma (ATC) is poor as it progresses [9].
In patients with ATC, systemic chemotherapy is the standard of care, but the disease is so rare that an optimal regimen has not yet been established. Based on the results of small phase II trials and retrospective studies, combination chemotherapy, platinum-based doublets (e.g., carboplatin and paclitaxel [6,10], cisplatin or carboplatin and etoposide for thymic neuroendocrine carcinoma [11][12][13]), and other multidrug regimens (e.g., doxorubicin, cisplatin, vincristine, and cyclophosphamide [ADOC] [14,15] and cisplatin, doxorubicin, and vincristine [CAP] [16]) are selected for patients with ATC in clinical practice. The efficacy of these regimens are modest; however, the response to chemotherapy and outcomes vary considerably among patients. Therefore, predictive and prognostic biomarkers are required for ATC patients receiving chemotherapy.
Although the prognostic factors in patients with ATC treated with chemotherapy remain unclear, there have been a few reports on the prognostic factors in patients with TC. In these limited reports, the Masaoka stage [5,17,18], age [19], sex [5,20], race [5], Karnofsky performance status (PS) [18], histology [18], degree of resection [18,21], presence of radiotherapy [18,22], white blood cell count [23], and lactate dehydrogenase level [23] were reported as prognostic factors. However, most patients included in these studies had early stage diseases that can be treated locally, and data with advanced stages are limited.
We previously reported the results of a multi-institutional retrospective study (NEJ023 study) of patients with ATC, and found that using the Masaoka staging system and a history of volume reduction surgery may be a prognostic factor [31].
In this study, we collected clinicopathological information, including tumor markers, from the NEJ023 study, and analyzed the overall survival (OS) and progression-free survival (PFS). The study aimed to determine whether tumor markers can be prognostic factors and which of them is correlated with prognosis the most. This study was registered with the University Hospital Medical Information Network Clinical Trials Registry (identifier: UMIN000015649).

Patient Cohort
The details of the study design and results of first-line chemotherapy in ATC patients have been previously published [31]. In this observational multicenter study, we retrospectively reviewed the medical records of patients who received chemotherapy between 1995 and 2014. All institutions belonging to the North East Japan Study Group were invited to participate. The inclusion criteria for this study were (a) a histologic diagnosis of TC at local institution; (b) advanced-stage or recurrent disease treated with palliative-intent pharmacotherapy.

Data Analysis
Data were obtained from 324 consecutive patients from 40 institutions. Thirty-seven patients who did not meet the eligibility criteria and one patient with missing data were excluded. Finally, 286 patients who received first-line chemotherapy were included in the analysis. The following details were extracted from the medical records of the patients: age, sex, Eastern Cooperative Oncology Group PS, Masaoka-Koga stage [32], WHO TNM stage, histology [8], date of death or last follow-up, regimens of first-line chemotherapy (platinumbased doublet, other multidrug regimens or single agents), duration of chemotherapy, and efficacy of chemotherapy. Histologic subtypes were determined according to the 2004 WHO classification [8] in each institution. PFS data collection was not mandatory in this study; therefore, only cases for which data could be collected were analyzed.
Furthermore, we collected data on tumor markers before the start of first-line chemotherapy. In the recurrence cases, tumor markers taken at the time of recurrence were collected. In the setting of the study, there was no provision for mandatory measurement of tumor markers, and data on cases that were measured voluntarily were collected. The CEA, CYFRA, SCC antigen, NSE, ProGRP, and alpha-fetoprotein (AFP) levels were evaluated as the tumor markers. For each tumor marker, the normal levels were defined as follows, with reference [26,[33][34][35][36][37] to the values commonly used in other carcinomas: CEA level, ≤5.0 ng/mL; SCC antigen level, ≤1.5 ng/mL; CYFRA level, ≤3.5 ng/mL; ProGRP level, ≤80 pg/mL; NSE level, ≤10 ng/mL; and AFP level, ≤10 ng/mL; values above these limits were defined as elevated.
The PFS was evaluated by investigators based on the Response Evaluation Criteria in Solid Tumors version 1.1 [38].
The primary purpose of this study was to investigate tumor markers that correlate with OS, and for other purposes, the relationships between tumor markers and PFS were investigated. The effect of each tumor marker on the OS and PFS was examined using a univariate analysis, dividing patients according to normal and elevated values. Thereafter, when a significant result was obtained in the univariate analysis, we performed a multivariate analysis including the prognostic factors examined in the previous report [31].
Generally, the most common histological type is SCC, which accounts for 70-80% of the total cases [8,19]. An analysis limited to the patients with SCC only was performed in addition to the entire population. The institutional review boards of all participating institutions approved the protocol for this retrospective study.

Statistical Analysis
The OS was defined as the period from the start of first-line chemotherapy until the date of death from any cause. The PFS was defined as the period from the start of first-line chemotherapy until the date of progressive disease or death from any cause. The Kaplan-Meier method was used to estimate the OS and PFS curves. The log-rank test was used to evaluate the differences among the subgroups. Cox proportional hazard models were used to adjust for potential confounding factors. All categorical variables were analyzed using Fisher's exact test. Statistical significance was set at p < 0.05. All analyses were performed using JMP 16 for Windows (SAS Institute Japan Inc., Tokyo, Japan).

Patient Characteristics
The clinical characteristics of the 286 patients with ATC are presented in Table 1. Most patients had squamous histology (66.4%). Postoperative recurrence was noted in 53 patients (18.5%). For the first-line chemotherapy regimens, platinum-based doublets, other multidrug chemotherapies, and single agents were used in 178 (62.2%), 98 (34.3%), and 10 patients (3.5%), respectively. The most popular regimens were carboplatin/paclitaxel (70 patients) among the platinum-based doublets, ADOC (79 patients) among the other multidrug chemotherapies, and S-1 (4 patients) among the single agents. S-1 is an oral fluoropyrimidine agent containing the 5-fluorouracil prodrug tegafur and two enzyme inhibitors, which can reduce the adverse effects of tegafur. The tumor markers were assessed in 270 of the 286 patients (94.4%). In most cases, the number of tumor marker types measured was five (22.7%), with a median of four types.

Tumor Markers
The details of the tumor markers assessed and their median values are shown in Table 2. The most frequently assessed tumor marker before first-line chemotherapy was the CEA level (82.9%), followed by the CYFRA (73.1%), SCC antigen (66.4%), ProGRP (58.0%), NSE (51.4%), and AFP levels (37.4%). The data on the histological types limited to SCC only are also shown in Table 2.

Univariate Analysis of the Relationship between the OS/PFS and Each Tumor Marker
The median follow-up time was 55.5 months. The results of the comparison of the OS and PFS between the elevated-level group and the normal-level group for each tumor marker are shown in Table 3.
The OS of the elevated NSE level group was significantly shorter than that of the normal NSE level group (median, 20.3 vs. 36.8 months; log-rank test p = 0.029, hazard ratio (HR), 1.55; 95% confidence interval (CI), 1.04-2.31 (Cox proportional hazard model)) ( Figure 1a). The PFS with first-line chemotherapy was also significantly shorter in the elevated NSE level group than in the normal NSE level group (median, 6.4 vs. 11.0 months; log-rank test p = 0.001; HR, 2.04; 95% CI, 1.31-3.18) (Figure 1b). There were no significant differences in the OS and PFS among the other tumor markers. After stratification based on the NSE level, the patient characteristics were found to be well-balanced; these are summarized in Table S1.
The results of the analysis of the OS and PFS, in which the histological type was limited to SCC only, are shown in Table 4. Similar to the results of the previous analysis, the OS of the elevated NSE level group was significantly shorter than that of the normal NSE level group (median, 20.3 vs. 36.8 months; log-rank test p = 0.030; HR, 1.71; 95% CI, 1.05-2.80) (Figure 2a). The PFS with first-line chemotherapy was also significantly shorter in the elevated NSE level group than in the normal NSE level group (median, 5.9 vs. 11.0 months; log-rank test p = 0.001; HR, 2.52; 95% CI, 1.41-4.59) (Figure 2b). For the other tumor markers,  (Figure 3a,b); however, the PFS did not significantly differ between them.

Multivariate Analysis of the OS/PFS Including Factors and Tumor Markers
The results of the univariate and multivariate analyses of the OS using factors including the patient characteristics and NSE level are shown in  The results of the analysis of the OS and PFS, in which the histological type was limited to SCC only, are shown in Table 4. Similar to the results of the previous analysis, the OS of the elevated NSE level group was significantly shorter than that of the normal NSE level group (median, 20.3 vs. 36.8 months; log-rank test p = 0.030; HR, 1.71; 95% CI, 1.05-2.80) (Figure 2a). The PFS with first-line chemotherapy was also significantly shorter in the elevated NSE level group than in the normal NSE level group (median, 5.9 vs. 11.0 months; log-rank test p = 0.001; HR, 2.52; 95% CI, 1.41-4.59) (Figure 2b). For the other tumor markers, the OS of the elevated CYFRA and SCC antigen level group was significantly shorter than that of the normal CYFRA and SCC antigen level group (CYFRA level: median, 25.7 vs. 42.7 months; log-rank test p = 0.047; HR, 1.52; 95% CI, 1.01-2.33; SCC antigen level: median, 21.1 vs. 33.9 months; log-rank test p = 0.018; HR, 1.79; 95% CI, 1.08-2.87) (Figure 3a,b); however, the PFS did not significantly differ between them.    The results of the analysis of the OS and PFS, in which the histological type was limited to SCC only, are shown in Table 4. Similar to the results of the previous analysis, the OS of the elevated NSE level group was significantly shorter than that of the normal NSE level group (median, 20.3 vs. 36.8 months; log-rank test p = 0.030; HR, 1.71; 95% CI, 1.05-2.80) (Figure 2a). The PFS with first-line chemotherapy was also significantly shorter in the elevated NSE level group than in the normal NSE level group (median, 5.9 vs. 11.0 months; log-rank test p = 0.001; HR, 2.52; 95% CI, 1.41-4.59) (Figure 2b). For the other tumor markers, the OS of the elevated CYFRA and SCC antigen level group was significantly shorter than that of the normal CYFRA and SCC antigen level group (CYFRA level: median, 25.7 vs. 42.7 months; log-rank test p = 0.047; HR, 1.52; 95% CI, 1.01-2.33; SCC antigen level: median, 21.1 vs. 33.9 months; log-rank test p = 0.018; HR, 1.79; 95% CI, 1.08-2.87) (Figure 3a,b); however, the PFS did not significantly differ between them.

Discussion
In this study, we examined the association between tumor markers and prognosis using data from 286 patients with ATC. To our knowledge, this is the first study to investigate the relationship between tumor markers and prognosis. The results of this study indicate that the NSE level may be a prognostic factor in patients with ATC. In addition, the SCC antigen level may also be a prognostic factor in patients with SCC.
In this study, a correlation was found between the NSE level and the OS and PFS. We found that the pretreatment serum NSE level was an independent prognostic factor for ATC.
The NSE level is considered to be a useful diagnostic tumor marker for tumors of neural and neuroendocrine origins, such as neuroendocrine tumors, and tends to increase in neuroendocrine carcinoma [50,51]. In this study, the NSE level was significantly higher in the patients with neuroendocrine carcinoma than in those with SCC (p = 0.009, Wilcoxon rank sum test). This study included 37 cases of neuroendocrine tumors, and analysis was performed excluding these cases. The results were similar to those of the entire population (data were shown in Figure S1a,b and Tables S3-S5).
Among ATCs, SCC has a better prognosis than other histological types, except for basaloid carcinoma [8]; thus, SCC is considered to have a better prognosis than neuroendocrine carcinoma. We thought that a higher NSE level in patients with ATC reflects a phenotypic heterogeneity with a larger neuroendocrine component and may be related to worse prognosis. Several reports have suggested that the NSE level may be a negative prognostic marker for non-small-cell lung cancer for the same reason [28,52,53]. NSE is a key enzyme in glycolysis and plays an important role in aerobic glycolysis [54]. Cells with a high NSE expression are thought to proliferate more rapidly, and higher NSE levels may also indicate that the cancer progresses more rapidly with a poor prognosis.
Recently, next-generation sequencing (NGS) for ATC has become a popular topic. In TC, TP53 was one of the most frequent mutation genes, and the mutation frequency ranged from 7.7% to 26.7%; on the other hand, TP53 mutations were rare in thymomas [55]. Several studies have suggested that TP53 mutations are a negative prognostic factor for ATC [55,56]. TP53 mutation seems to be associated with a more aggressive behavior, as confirmed in an earlier report [57]. Another report mentioned that in ATC, p53 immunohistochemical expression, which is correlated with TP53 mutation, is correlated with 18 F-FDG uptake [58]. TP53 mutation is also thought to be associated with resistance to chemotherapy because of the involvement of multidrug resistance gene 1 (MDR1/ABCB1) [59]. Therefore, ATC with TP53 mutations is expected to have a poor prognosis. Although their study evaluated lung adenocarcinoma, Xu et al. reported that there was a correlation between serum NSE level elevation and polygenic mutations combined with TP53 mutations [60]. Thus, we assumed that TP53 mutations may be seen in ATC cases with high NSE levels, which may cause a poor prognosis.
Recently, immune check inhibitors (ICIs) for ATC have gained increasing attention, and many clinical trials have been reported [61][62][63]. Giaccone et al. reported the results of a phase II trial, in which pembrolizumab was administered to patients with ATC and NGS was performed. The most commonly mutated gene was TP53 (13 (36%) of 36 patients), and there was a correlation found between TP53 mutations and a shorter OS [61]. Although TP53 mutations may reduce the effect of ICIs in thymic cancer, there are reports of the opposite in lung cancer [64]; thus, further studies investigating the relationship with the NSE level are required.
The limitation of this study is its retrospective design without preregulation to measure the tumor markers, and a selection bias could have occurred. The generally permitted missing data ratio is less than 5%, and there were data losses more than 5% in this study, which may have affected the results. Since many tumor markers were examined, there was also a problem of statistical multiplicity. Moreover, there was an unbalanced number of cases between the high and normal levels of tumor markers other than the NSE and CYFRA levels. Therefore, it might be difficult to observe a significant difference in these tumor markers. Next, TC was diagnosed in each institution, and a central pathological judgment was absent. The previous reports mentioned that pathological diagnosis of thymic epithelial tumor is difficult, and the diagnosis concordance rate among pathologists is not high [10,65]. For this study, histological diagnosis should have been performed centrally, but this was difficult to carry out because some pathological specimens used for diagnosis were no longer stored, and were not available. The diagnoses and histological subtypes may not be as accurate as those in studies with a central pathological judgment, although this study was conducted at institutions with sufficient skills for the diagnosis and treatment of ATC. As some of tissue specimens were old and not stored any longer, the levels of NSE in tumor samples were not evaluated.

Conclusions
This study investigated whether tumor markers are useful prognostic factors in patients with ATC. The NSE level was found to be a prognostic factor for all histological types of ATC, while the SCC antigen level was found to be a prognostic factor for the SCC type.