Correlation of Thyroid Transcription Factor-1 Expression with EGFR Mutations in Non-Small-Cell Lung Cancer: A Meta-Analysis

Objectives: This meta-analysis investigated the relationship between thyroid transcription factor-1 (TTF-1) expression and epidermal growth factor receptor (EGFR) mutations in non-small-cell lung cancer (NSCLC) to clarify whether TTF-1 can be a potential surrogate marker for EGFR mutation status in advanced NSLCL. Methods: A systematic searching of databases, including PubMed, EMBASE, Cochrane Library, and Google Scholar, was performed to identify studies assessing the correlation of TTF-1 expression with EGFR mutations. From 17 studies, 9764 patients were included in the combined analysis of odds ratio (OR) for the correlation between TTF-1 expression and EGFR mutations. Results: Compared with NSCLCs showing negative TTF-1 expression, tumors harboring TTF-1 overexpression showed a significantly higher rate of EGFR mutations (OR = 5.19, 95% confidence interval: 3.60–7.47, p < 0.00001). This correlation was observed in both subgroups of East Asian (OR = 4.33, 95% CI: 3.46–5.41, p < 0.00001) and European patients (OR = 4.64, 95% CI: 1.41–15.28, p < 0.01). In addition, TTF-1 expression was significantly associated with EGFR mutations in exon 19 (OR = 4.63, 95% CI: 2.89–7.41, p < 0.00001) as well as exon 21 (OR = 3.16, 95% CI: 1.04–9.60, p = 0.04). Conclusions: This meta-analysis demonstrates a significant correlation between TTF-1 expression and EGFR mutations in patients with NSCLC. The status of TTF-1 expression may be a biomarker to guide anticancer treatment in patients with NSCLC and unknown EGFR mutation status.


Introduction
Lung cancer is the second most common malignancy in both genders worldwide [1]. It still remains the leading cause of cancer-associated deaths [1,2], although systemic chemotherapy or immune checkpoint inhibitors can significantly improve prognosis for patients with advanced non-small-cell lung cancer (NSCLC) [3][4][5]. For patients with epidermal growth factor receptor (EGFR)-mutant NSCLC, targeted therapy with tyrosine kinase inhibitors (TKIs) can significantly prolong survival [6,7].
Thyroid transcription factor-1 (TTF-1) is a regulatory transcription factor for tissue-specific genes [8]. It is expressed in the thyroid, forebrain, or lungs, playing a physiologic role during their development. In the normal lungs, TTF-1 helps to maintain functions of terminal respiratory unit cells [9]. High TTF-1 expression by immunohistochemistry (IHC) has been observed in 70-90% of primary lung adenocarcinomas (ADCs), while almost all squamous cell carcinomas are negative for TTF-1 IHC. Therefore, it has been considered a specific marker of ADCs of the lung. In addition,

Publication Search Strategy
This meta-analysis was done according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [23]. A systematic search of the databases including PubMed, EMBASE, Cochrane Library, and Google Scholar (up to December 2018) was performed to identify studies assessing the correlation of TTF-1 expression with EGFR mutations. The search used a combination of the following terms: "epidermal growth factor receptor" or "EGFR" AND "mutation" AND "thyroid transcription factor-1" or "TTF-1" AND "non-small-cell lung cancer" or "NSCLC" or "lung cancer." All of the relevant articles identified by the related article function were also included in the analysis. The references reported in the identified articles were also reviewed to complete the search process.

Eligibility Criteria
Eligible studies should meet the following inclusion criteria: (i) patients with pathologically confirmed NSCLC; (ii) analysis of EGFR mutations in exons 19 and 21; (iii) IHC test for TTF-1 expression in lung cancer tissue; (iv) the use of adequate IHC methods and criteria for positive TTF-1 staining; and (v) prospective or retrospective cohort studies assessing the correlation of TTF-1 expression with EGFR mutations.

Article Review and Data Extraction
Two authors (D.R.C. and B.H.) independently searched the databases and extracted data from the selected studies. The following data were extracted from each study: the first author, year of publication, study design, inclusion period, country, sample size, histology, disease stage, TTF-1 expression status, IHC criteria for positive expression, EGFR mutation status, and detecting method.

Quality Assessment
The methodological quality of included studies was scored based on the Newcastle-Ottawa System (NOS) with the score range of zero to nine [24]. Studies having a score ≥ six were considered to have a high quality.

Statistical Analysis
The strength of the association between TTF-1 expression and EGFR mutations was shown as odds ratios (ORs) with 95% confidence intervals (CIs). If the study did not report the OR or 95% CI directly, we calculated them from available data by using the Engauge Digitizer software. The heterogeneity of the individual ORs was estimated using the chi-squared test with significance being set at p < 0.1. The total variation among studies was estimated by anI 2 inconsistency test, where I 2 > 50% was considered to indicate significant heterogeneity. If there was heterogeneity among studies, we used the random effect model based on the DerSimonian-Laird method to pool the OR. Otherwise (p ≥ 0.1 and I 2 ≤ 50%), the fixed effect model based on the Mantel-Haenszel method was selected. Subgroup analyses were planned according to the ethnicity and mutational types. The sensitivity analysis was performed to detect the influence of individual trials on the pooled results by removing one trial each time. Forest plots were produced to show a summary estimate of the combined results of all the studies. Each square represented the OR point estimate, and its size was proportional to the weight of the study. The location of the diamond represented the estimated effect size, and its width reflected the precision of the estimate.
The potential publication bias was assessed by visual inspection of the funnel plot [25]. For quantitative analyses, Egger's test and Begg's tests were performed using the statistical software packages R [26]. Statistical significances were considered when a p-value was less than 0.05.

Medicina2019, 54, x FOR PEER REVIEW 3 of 13
The strength of the association between TTF-1 expression and EGFR mutations was shown as odds ratios (ORs) with 95% confidence intervals (CIs). If the study did not report the OR or 95% CI directly, we calculated them from available data by using the Engauge Digitizer software. The heterogeneity of the individual ORs was estimated using the chi-squared test with significance being set at p < 0.1. The total variation among studies was estimated by anI 2 inconsistency test, where I 2 > 50% was considered to indicate significant heterogeneity. If there was heterogeneity among studies, we used the random effect model based on the DerSimonian-Laird method to pool the OR. Otherwise (p ≥ 0.1 and I 2 ≤ 50%), the fixed effect model based on the Mantel-Haenszel method was selected. Subgroup analyses were planned according to the ethnicity and mutational types. The sensitivity analysis was performed to detect the influence of individual trials on the pooled results by removing one trial each time. Forest plots were produced to show a summary estimate of the combined results of all the studies. Each square represented the OR point estimate, and its size was proportional to the weight of the study. The location of the diamond represented the estimated effect size, and its width reflected the precision of the estimate.
The potential publication bias was assessed by visual inspection of the funnel plot [25]. For quantitative analyses, Egger's test and Begg's tests were performed using the statistical software packages R [26]. Statistical significances were considered when a p-value was less than 0.05.
The EGFR mutations were usually detected by polymerase chain reaction (PCR) and direct DNA sequencing methods, or by amplification refractory mutation system (ARMS) analysis. Most of the studies examined four kinase domains in exons 18-21; however, some screened for EGFR mutations in only exons 19 and 21 [21,31,36].
The EGFR mutations were usually detected by polymerase chain reaction (PCR) and direct DNA sequencing methods, or by amplification refractory mutation system (ARMS) analysis. Most of the studies examined four kinase domains in exons 18-21; however, some screened for EGFR mutations in only exons 19 and 21 [21,31,36].

Subgroup Analysis According to the Ethnic Group
When we performed the subgroup analysis according to the region, the combined ORs were 4.33 (95% CI: 3.46-5.41, p < 0.00001, fixed-effects model) for patients in East Asia ( Figure 3A) and 4.64 (95% CI: 1.41-15.28, p < 0.01, random-effects model) for patients in Europe ( Figure 3B).

Subgroup Analysis According to the Ethnic Group
When we performed the subgroup analysis according to the region, the combined ORs were 4.33 (95% CI: 3.46-5.41, p < 0.00001, fixed-effects model) for patients in East Asia ( Figure 3A) and 4.64 (95% CI: 1.41-15.28, p < 0.01, random-effects model) for patients in Europe ( Figure 3B).

Publication Bias
Visual inspection of the funnel plot for ORs showed symmetry, indicating that there was no substantial publication bias ( Figure 6). Egger's and Begg's tests also indicated no evidence of substantial publication bias, with p-values of 0.172 and 0.152, respectively.

Publication Bias
Visual inspection of the funnel plot for ORs showed symmetry, indicating that there was no substantial publication bias ( Figure 6). Egger's and Begg's tests also indicated no evidence of substantial publication bias, with p-values of 0.172 and 0.152, respectively.

Publication Bias
Visual inspection of the funnel plot for ORs showed symmetry, indicating that there was no substantial publication bias ( Figure 6). Egger's and Begg's tests also indicated no evidence of substantial publication bias, with p-values of 0.172 and 0.152, respectively.

Discussion
Over the past two decades, several major progresses have been made toward the personalized treatment for patients with NSCLC. The first breakthrough was the discovery of EGFR mutations. TKIs are small molecular agents targeting EGFR mutations that have revolutionized the treatment of NSCLC, leading to improved survival in patients with advanced or metastatic EGFR-mutant tumor [6,7]. Therefore, it is essential to screen for EGFR mutations before introducing anticancer treatment for patients with advanced NSCLC.
Studies evaluating the prognostic value of TTF-1 expression in NSCLC reported TTF-1 expression to be an independent predictor of survival [10,11,43,44]. In the North East Japan 002 study, investigators reported that the rate of EGFR mutations was higher in patients with ADC positive for TTF-1 expression [6]. In particular, Asians, women, and non-smokers revealed a higher rate of EGFR mutations and TTF-1 positive expression. Based on these findings, some researchers hypothesized that TTF-1 expression is significantly correlated with EGFR mutations in patients with lung ADC [34].
In the current meta-analysis, we combined the data from 17 studies, including 9764 patients, to determine whether positive TTF-1 expression by IHC can be a surrogate marker for EGFR mutation status in NSCLC. The results revealed that EGFR mutations were significantly correlated with TTF-1 overexpression in patients with NSCLC (OR = 5.34, 95% CI: 3.54-8.04, p < 0.00001). Both high TTF-1 expression and EGFR mutations have been known to be significantly associated with Asian, women, and non-smokers. However, our subgroup analysis according to the ethnicity indicated that the correlation of TTF-1 expression and EGFR mutations was observed in European patients (OR = 4.64, 95% CI: 1.41-15.28, p < 0.01) as well as East Asian (OR = 4.33, 95% CI: 3.46-5.41, p < 0.00001). In addition, in the subgroup analysis according to the gender, the significant correlation between TTF-1 expression and EGFR mutations was observed not only in female (OR = 4.87, 95% CI 2.27-10.45, p < 0.0001), but also in male patients with NSCLC (OR = 3.34, 95% CI 1.43-6.82, p = 0.0009).
Among EGFR mutations, deletion in exon 19 and L858R mutation in exon 21 are common, accounting for more than 90% of all EGFR mutations. Some studies conducted subgroup analysis according to EGFR mutation subtype and reported that positive TTF-1 expression was only significantly correlated with EGFR mutation in exon 21, and not on exon 19 [34,38]. However, Zhao et al. also observed a significant correlation of TTF-1 positivity with EGFR mutation in exon 21 in Chinese patients with lung ADC [38]. In the current meta-analysis, the pooled data from four studies [30,34,38,41] indicated a significant relationship between TTF-1 expression and EGFR mutation in both exon 19 (OR = 4.63, 95% CI: 2.89-7.41, p < 0.00001) and exon 21 (OR = 3.16, 95% CI: 1.04-9.60, p = 0.04).
Several studies reported a high negative predictive value (88.2-97%) of TTF-1 for the presence of activating EGFR gene mutations [21,28,35,40]. In this study, only 9.8% (176 of 1801) of TTF-1-negative NSCLC cases had EGFR mutations. Overall, the negative predictive value of TTF-1 for EGFR mutations was 90.2%. However, when including only patients in Europe, where the rate of EGFR mutations is low, the negative predictive value of TTF-1 increased to 96.2%. Therefore, when the status of EGFR mutations cannot be tested in a timely manner because of inadequate tumor tissue or when treatment for NSCLC is urgent for clinical reasons, the IHC staining result of TTF-1 may be used to guide systemic anticancer therapy. In particular, negative TTF-1 expression can be a surrogate marker to recommend conventional chemotherapy.
This study has several limitations. First, the included studies showed considerable diversity in the methods used for TTF-1 staining and EGFR mutation detection. In addition, the cut-off criteria for positive TTF-1 staining also varied among studies. Second, the rate of TTF-1 overexpression and EGFR mutations are significantly higher with NSCLC patients with no smoking history. However, we could not perform a subgroup analysis according to the smoking status because of the limited data. Third, the substantial heterogeneity observed among studies could not be interpreted thoroughly, even though the random-effects model was selected. Finally, about half of the included studies were retrospective and therefore may carry the biases inherent to the study design.

Conclusions
This meta-analysis indicates a significant correlation between TTF-1 overexpression and EGFR mutations status in patients with NSCLC. In clinical practice, the status of TTF-1 expression may be a biomarker to guide anticancer therapy in advanced NSCLC patients with unknown EGFR mutation status. Especially, negative TTF-1 expression has such a high negative predictive value for EGFR mutations that it might be a surrogate marker to recommend conventional chemotherapy first when the status of EGFR mutations cannot be tested in a timely manner.