Elevated preoperative CEA is associated with subclinical nodal involvement and worse survival in stage I non-small cell lung cancer: a systematic review and meta-analysis

Background The standard for clinical staging of lung cancer is the use of CT and PET scans, however, these may underestimate the burden of the disease. The use of serum tumor markers might aid in the detection of subclinical advanced disease. The aim of this study is to review the predictive value of tumor markers in patients with clinical stage I NSCLC. Methods A comprehensive search was performed using the Medline, EMBASE, Scopus data bases. Abstracts included based on the following inclusion criteria: 1) adult ≥18 years old, 2) clinical stage I NSCLC, 3) Tumor markers (CEA, SCC, CYFRA 21-1), 4) further imaging or procedure, 5) > 5 patients, 6) articles in English language. The primary outcome of interest was utility of tumour markers for predicting nodal involvement and oncologic outcomes in patients with clinical stage I NSCLC. Secondary outcomes included sub-type of lung cancer, procedure performed, and follow-up duration. Results Two hundred seventy articles were screened, 86 studies received full-text assessment for eligibility. Of those, 12 studies were included. Total of 4666 patients were involved. All studies had used CEA, while less than 50% used CYFRA 21-1 or SCC. The most common tumor sub-type was adenocarcinoma, and the most frequently performed procedure was lobectomy. Meta-analysis revealed that higher CEA level is associated with higher rates of lymph node involvement and higher mortality. Conclusion There is significant correlation between the CEA level and both nodal involvement and survival. Higher serum CEA is associated with advanced stage, and poor prognosis. Measuring preoperative CEA in patient with early stage NSCLC might help to identify patients with more advanced disease which is not detected by CT scans, and potentially identify candidates for invasive mediastinal lymph node staging, helping to select the most effective therapy for patients with potentially subclinical nodal disease. Further prospective studies are needed to standardize the use of CEA as an adjunct for NSCLC staging.


Introduction
Lung cancer is the leading cause of cancer related death. Optimal treatment for non-small cell lung cancer (NSCL C) is dependent on accurate clinical staging to determine the extent of disease [1][2][3]. Patients with lymph node involvement have worse prognosis and may be candidates for neoadjuvant treatment prior to surgical resection. The standard for clinical staging of NSCLC is the use of computed tomography (CT) and positron emission tomography (PET) scans. However, these can underestimate the burden of the disease [4][5][6][7]. The median sensitivity and specificity of PET-CT for detection of mediastinal nodal disease is 80 and 88%, respectively [8]. False negatives on imaging studies result in understaging patients who might have benefited from invasive mediastinal staging or neoadjuvant therapy [7,9,10]. This had led researchers to investigate the use of biomarkers to increase the sensitivity of clinical staging and allow proper treatment selection. These markers include carcinoembryonic antigen (CEA), squamous cell carcinoma antigen (SCC), and cytokeratin fragment antigen (CYFRA 21-1) [4,[11][12][13]. To our knowledge this is the first systematic review about the use of tumor markers in patients with clinical stage I NSCLC for predicting lymphatic spread.

Methods
A comprehensive search was performed for articles published on non-small cell lung cancer and tumor markers using the Medline, EMBASE, Scopus data bases. Search terms included "non-small cell lung cancer" or NSCLC or lung adenocarcinoma, AND carcinoembryonic antigen or squamous cell carcinoma antigen or cytokeratin fragment antigen, AND stage I or stage IA or early stage". Literature was limited to human studies in the English language. Abstracts and titles were screened for inclusion by two reviewers (AN and JL). Non-relevant articles based on their abstract were not included for full-text evaluation. Abstracts were then further screened based on the following inclusion criteria: 1) adult patient ≥18 years old, 2) primary non-small cell lung cancer (stage I or stage IA or early stage), 3) Tumor markers (Carcinoembryonic antigen (CEA), cytokeratin fragment antigen (CYFRA 21-1) and squamous cell carcinoma antigen (SCC)), 4) any further imaging such as positron emission tomography (PET) scan or procedure such as mediastinoscopy or endobronchial ultrasound (EBUS) or further treatment, 5) studies including > 5 patients, 6) articles in English language. Exclusion criteria included non-English studies, abstracts only, and duplicates. The primary outcome of interest was utility of tumour markers for predicting pathological tumor invasiveness in patients with clinical stage I NSCLC. Secondary outcomes included subtype of lung cancer, follow-up duration, procedure performed, smoking status, and region of publication. Meta-analysis was performed to determine the following: death within 5 years, and lymph node involvement. This study was conducted and the results are presented according to PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines.
In addition, two independent reviewers assessed the risk of bias using the Newcastle-Ottawa Scale (NOS) for evaluating the quality of the included studies. We rated the quality of the studies (good, fair and poor) according to the guidelines of the NOS. A "good" quality score required 3 or 4 points in selection, 1 or 2 points in comparability, and 2 or 3 points in outcomes. A "fair" quality score required 2 points in selection, 1 or 2 points in comparability, and 2 or 3 points in outcomes. A "poor" quality score reflected 0 or 1 point in selection, or 0 points in comparability, or 0 or 1 point in outcomes (Table 1).

Study selection
Preliminary literature search yielded 270 articles after duplicates were removed. All these 270 studies were screened. Eighty-six studies received full-text assessment for eligibility. Of those, 12 studies were included in the final systematic review (Fig. 1).

Basic demographics
Twelve studies with 4666 patients were included for systematic review. The majority of these studies were retrospective (10/12) and most were conducted in Japan (8/ 12). The mean age of the subjects was 65.3 ± 3.2 years, 2602 were males, and 2064 were females ( Table 2). The mean follow-up period was 48.86 months. Nine studies, involving 3842 patients reported smoking status, in which 2003 were smokers (52.1%).

Reporting of tumor markers and tumor characteristics
All the studies included investigated CEA, while 4 studies also used CYFRA 21-1, and 2 studies used SCC. Given there was a low number of studies looking at tumour markers other than CEA, we excluded the   analysis of these markers from our review. The cut-off value for abnormal CEA differed among the studies, ranging from 2.5 to 10 ng/mL (  (Fig. 3). Further subanalysis revealed that high CEA had higher odds of positive N2 that is 3.61 times that of low CEA (95% CI 1.73 to 7.53, p = 0.0006). This analysis included 2 studies and 1085 patients (Fig. 4). Heterogeneity was low (chi2 = 0%, p = 0.53).

Risk of bias assessment
Most of the studies included had good or fair quality score, only 3 studies their quality score was poor. Those whose scored poor on their quality lost some point on follow up.

Discussion
Accurate preoperative staging of NSCLC is integral for appropriate treatment plan. The main treatment for patients with stage I NSCLC is surgery. Unfortunately, due to limited sensitivity of preoperative imaging, up to 30% of patients with stage I NSCLC may have positive N2-N3 lymph nodes at the time of resection [8,14,15]. A meta-analysis of 20 studies showed that mediastinal lymph node staging using CT scan had 57% sensitivity and 82% specificity [16]. Similarly, Cerfolio et al. showed that 7 of 17 patients with cN1 (41%) were found to have positive N2 after lymph node sampling by mediastinoscopy or endoscopic ultrasound with fine needle aspiration (EUS FNA), although N2 involvement was excluded initially by PET/CT scan [17]. The ability to detect subclinical nodal involvement prior to surgery could allow identification of cN0 patients who might benefit from invasive staging, while patients with low CEA levels could conceivably be spared invasive staging if they would otherwise qualify for reasons such as large tumour size or central tumour. Better preoperative staging should result in improved treatment selection, as patients may benefit from neoadjuvant chemotherapy or chemoradiotherapy rather than upfront surgery. In patients with early stage I NSCLC, in whom lymph sampling versus lymph dissection is controversial, the use of CEA could also identify patients in need of more aggressive lymphadenectomy [18][19][20]. Whether the increased risk of mortality found in this meta-analysis is completely attributable to the increased rate of nodal involvement can not be determined from the available studies' data, but identification of patients with poor prognosis related to high CEA may also allow for a more tailored approach to post-resection surveillance and patient counselling. Due to the limited ability of preoperative imaging CT or PET-CT to detect mediastinal lymph nodes disease, interest in serum biomarkers in lung cancer is growing. The most frequently studied tumor marker is carcinoembryonic antigen. All histological types of lung cancer can produce CEA and a role for its use in lung cancer screening and staging was first proposed in the 1970s [21][22][23][24][25]. Recently, studies have demonstrated the usefulness of CEA in patients with NSCLC for postoperative follow up, response to chemotherapy, recurrence, and prognosis. High CEA level has been correlated with advanced disease and poor prognosis. Serum CEA measurement is a simple, non-invasive, inexpensive test. In such case, patients with high CEA level might benefit from lymph node sampling by mediastinoscopy or endoscopic ultrasound with fine needle aspiration (EUS FNA) [18,[25][26][27]. As shown in our meta-analysis CEA level is correlated with lymph node involvement, and further sub-analysis did reveal that higher CEA associated with positive N2.
There is a discrepancy in the cut-off value of CEA ranges from 2.5 to 10 ng/mL and is attributable to the different techniques used for measurement such as radioimmunoassay (RIA) and enzyme immunoassay [21,28]. Further studies are needed to standardize the cutoff value. Our meta-analysis about lymph nodes involvement and death within 5 years was limited to few studies because most did not mention the specific details needed to conduct the analysis. Other limitations in our study include the following: the majority of included studies were retrospective, done in a single country (Japan) and many lacked specific lymph nodes details (N0, N1, N2, N3). In addition, we excluded non-English articles. As such we recommend a prospective study using CEA preoperatively to accurately correlate the level of CEA with risk of lymph nodes metastasis, and to determine the cut-off value of CEA.

Conclusion
There is significant correlation between the CEA level and both nodal involvement and survival. Higher level of CEA is associated with advanced stage, and poor prognosis. Performing preoperative CEA in patient with early stage NSCLC might help to identify patients with more advanced disease which is not detected by imaging, and potentially identify patients for invasive mediastinal lymph node staging, helping to select the most effective therapy for patients with potentially subclinical nodal disease. Further prospective studies are needed to standardize the use of CEA as an adjunct for NSCLC staging.