FGFR4 Gly388Arg Polymorphism Reveals a Poor Prognosis, Especially in Asian Cancer Patients: A Meta-Analysis

The fibroblast growth factor-4 receptor (FGFR4) is a member of receptor tyrosine kinase. The FGFR4 Gly388Arg polymorphism in the transmembrane domain of the receptor has been shown to increase genetic susceptibility to cancers. However, its prognostic impact in cancer patients still remains controversial. Herein, we performed this meta-analysis to evaluate the clinicopathological and prognostic impacts of the FGFR4 Gly388Arg polymorphism in patients with cancer. We carried out a computerized extensive search using PubMed, Medline, and Ovid Medline databases up to July 2021. From 44 studies, 11,574 patients were included in the current meta-analysis. Regardless of the genetic models, there was no significant correlation of the FGFR4 Gly388Arg polymorphism with disease stage 3/4. In the homozygous model (Arg/Arg vs. Gly/Gly), the Arg/Arg genotype tended to show higher rate of lymph node metastasis compared with the Gly/Gly genotype (odds ratio = 1.21, 95% confidence interval (CI): 0.99-1.49, p = 0.06). Compared to patients with the Arg/Gly or Arg/Arg genotype, those with the Gly/Gly genotype had significantly better overall survival (hazard ratios (HR) = 1.19, 95% CI: 1.05-1.35, p = 0.006) and disease-free survival (HR = 1.25, 95% CI: 1.03-1.53, p = 0.02). In conclusion, this meta-analysis showed that the FGFR4 Gly388Arg polymorphism was significantly associated with worse prognosis in cancer patients. Our results suggest that this polymorphism may be a valuable genetic marker to identify patients at higher risk of recurrence or mortality.

In 2010, Frullanti et al. conducted a meta-analysis of 21 studies to evaluate the role of the FGFR4 Gly388Arg polymorphism as a prognostic factor in cancers. They found a statistically significant association of the FGFR4 Arg388 allele and overall survival (hazard ratio (HR) = 1.21, 95% confidence interval (CI) 1.05 -1.40) and LN involvement (odds ratio (OR) = 1.33, 95% CI 1.01-1.74) (66). Because of a limited number of eligible articles, however, they only included three or less studies in the subgroup analysis according to the primary tumor type. Given the amount of accumulated data thereafter, an updated quantitative synthesis has been deemed worthy. Herein, we performed this meta-analysis to evaluate the clinicopathological and prognostic impacts of the FGFR4 Gly388Arg polymorphism in patients with cancer.

Publication Search Strategy
This meta-analysis was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (67). We considered all studies that examined the clincopathological or prognostic value of the FGFR4 Gly388Arg polymorphism in any types of cancer. We carried out a computerized extensive search using PubMed, Medline, and Ovid Medline databases up to July 2021. The search used the following keywords variably combined: 'fibroblast growth factor receptor 4 or FGFR4', 'polymorphism or SNP', 'Gly388Arg or rs351855' 'prognosis or survival' and 'cancer or carcinoma or tumor.' The 'snowball' method was adopted to identify additional relevant articles and the reference lists of identified articles were hand-searched (68). In case of duplicate publication, the recent paper was selected.

Eligible Criteria
Studies should meet the following eligible criteria: (i) prospective or retrospective cohort study; (ii) study investigating the association of the FGFR4 Gly388Arg polymorphism with clinicopathological features (LN metastasis or disease stage) or survival outcomes (disease-free survival or overall survival); (iii) the use of adequate method to assess the FGFR4 Gly388Arg SNP; (iv) adequate data to estimate OR with 95% confidence interval (CI) for pathological parameters and/or HR with 95% CIs for survival; (v) study published only in peer-reviewed journal; and (vi) article written in English.

Data Extraction
Two investigators (STP and SYJ), working independently and in parallel, screened the literature and extracted from the eligible articles according to the inclusion criteria. The following data were collected from each article and recorded using the predesigned data-collection form: first author, year of publication, country, ethnicity, inclusion period, sample size, cancer type, genotyping method, genotype counts for the FGFR4 Gly388Arg polymorphism, LN status, disease stage, survival outcomes (disease-free survival or overall survival), and ORs with 95% CIs for pathological parameters and HRs with their 95% CIs for survival outcomes. When both univariate and multivariate analysis were performed for survival times, the HR with 95% CI from multivariate analysis was selected. Any conflicts were resolved by discussion, with input from the other investigators (JHK and HSK).

Statistical Analyses
The strength of the association between the FGFR4 Gly388Arg polymorphism and pathological findings was estimated by the ORs with their 95% CIs in the four genetic models: homozygous [Arg/Arg (AA) vs. Gly/Gly (GG)]; heterozygous (Arg/Gly (AG) vs. GG)); recessive (AA vs. AG+GG); and dominant (AG+AA vs. GG). For the survival analyses, HRs with 95% CIs according to the FGFR4 Gly388Arg polymorphism status were combined. Statistical values were directly obtained from the original articles. If HRs with their 95% CIs were not reported, the Engauge Digitizer software was utilized to calculate them from the corresponding data and Kaplan-Meier curves. The RevMan version 5.4 software was used to combine the data. The heterogeneity across studies was assessed by the Q statistic and I 2 inconsistency test. If significant heterogeneity was detected (p < 0.1 or I2 > 50%), the random-effects model was selected. Otherwise (p ≥ 0.1 and I2 ≤ 50%), the fixed-effects model was used. Statistical significance of the pooled HR or OR was determined by Z test. The combined OR or HR > 1.0 implies that cancers harboring the FGFR4 Gly388Arg polymorphism had worse clinicopathological features or survival, respectively.
Publication biases were evaluated by the Begg's funnel plot (69) and the Egger's linear regression test (70). All the statistics were two-sided, with p-value < 0.05 considered significant.

Results of Search
The flow diagram of search process is shown in Figure 1. Except for duplicates, a total of 190 relevant articles were initially retrieved, but 103 of them were excluded after careful screening of the titles and abstracts. Of the remaining 87 potentially eligible studies, 43 which did not meet the eligible criteria were further excluded. Finally, 44 studies were selected for the qualitative synthesis (9, 12-18, 21, 23-27, 34-45, 47-53, 55-65).

Characteristics of the Included Studies
The main characteristics and clinicopathological findings of the included studies are summarized in Table 1 and its supplement. In total, data were obtained on 11,574 subjects from the 44 included studies. Studies were more commonly conducted in Western countries (26 studies). Polymerase chain reactionrestriction fragment length polymorphism (PCR-RFLP) method was most frequently used to assess the FGFR4 Gly388Arg polymorphism status. Most studies used formalin-fixed and paraffin-embedded tissue to extract DNA, except for some utilizing fresh frozen tumor tissue or peripheral blood. The impact of the FGFR4 Gly388Arg SNP was most often studied in HNSCC (9 studies), followed by LC (8 studies), BC (7 studies), and CRC (6 studies). Other investigated tumor types were GC, PC, HCC, retinoblastoma, sarcoma, melanoma, lymphoma, bladder cancer, ovarian cancer, cervical cancer, and panNET. Twenty-four studies analyzed the pathological or prognostic parameters according to the four genetic models. Seven studies had a small sample size with less than 100 subjects in total (36,42,45,51,59,60,64), and most studies used univariate statistical method to compare survival times, except for seven studies (37,39,40,44,56,58,61).

Clinicopathological Impact of the FGFR4 Gly388Arg Polymorphism
From 24 studies, 6,157 patients were included in the metaanalysis of ORs with 95% CIs for LN metastasis. The odds of LN metastasis at the time of diagnosis were not different between patients with the GG genotype and those with the AG or AA genotype (OR = 1.08, 95% CI: 0.91-1.29, p = 0.39, randomeffects, Table 2, forest plot not shown). In the homozygous model (amino acid: AA vs. GG), the AA genotype tended to show higher rate of LN metastasis compared with the GG genotype. (OR = 1.21, 95% CI: 0.99-1.49, p = 0.06, fixedeffects) ( Figure 2 and Table 2).
From 21 studies, 5,585 patients were pooled to assess the effect of the FGFR4 Gly388Arg SNP on disease stage. There was no significant difference of advanced disease (stage 3/4) between patients with the GG genotype and those with the AG or AA genetic type (OR = 1.05, 95% CI: 0.89-1.25, p = 0.56, randomeffects, forest plot not shown). The association of the FGFR4 Gly388Arg SNP with disease stage was not significant in the other genetic models, either ( Table 2).
From 10 studies, 2,803 patients were included in the pooled analysis to evaluate the correlation between the FGFR4 Gly388Arg SNP and disease-free survival. Patients with the GG genotype showed significantly longer disease-free survival than those with the AG or AA genotype (HR = 1.25, 95% CI: 1.03-1.53, p = 0.02, random-effects) ( Figure 3B).  (Figure 4). We did not perform the subgroup analyses for other types of cancer in which only two or less studies were included.

DISCUSSIONS
The FGFR4 Gly388Arg polymorphism results in the replacement of the glycine residue with a charged arginine residue in the transmembrane domain of the receptor, which consequently exposes proximal STAT3 binding site and then enhances STAT3 signal to stimulate cell proliferation (71). Thus, the FGFR4 Arg388 variants may promote tumorigenesis by enhancing cell motility, invasiveness, and proliferation. The FGFR4 gene rs351855 G>A polymorphism has been known to confer increased genetic susceptibility to cancers (29)(30)(31)(32)(33). Many researchers also examined the relationship between the FGFR4 gene SNP and its pathological or prognostic roles among diverse cancer types. However, the results were inconsistent. In the current meta-analysis, we evaluated the clinicopathological and prognostic significance of the FGFR4 Gly388Arg polymorphism in cancers. The study included 11,574 patients with various types of cancer from the 44 eligible articles.
In the current study, the FGFR4 Gly388Arg SNP failed to show a significant correlation with disease stage 3/4 at the time of diagnosis, regardless of the genetic models. In the homozygous model (AA vs. GG), patients with the AA genotype showed a tendency of higher rate of LN metastases than those with the GG genotype (OR = 1.21, 95% CI: 0.99-1.49, p = 0.06). In terms of   (66). These results indicate that the FGFR4 Gly388Arg polymorphism is a potential genetic marker associated with worse prognosis in cancer patients. The association of the FGFR4 Arg388 polymorphism with the susceptibility to cancer has mainly been described in PC and BC (15,16,(29)(30)(31)33). In the meta-analysis of 27 studies comprising 8,682 cases by Xiong et al., interestingly, the FGFR4 rs351855 G>A polymorphism increased the risk of PC and BC, but decreased the susceptibility of LC (33). This finding indicates that the FGFR4 Arg388 SNP might has opposite effects on different types of cancer, suggesting that this polymorphism may modify cancer susceptibility in a tissue specific manner. Therefore, the prognostic role of the FGFR4 Arg388 polymorphism may also be different among different types of cancer. In our subgroup analyses using two most common primary sites, however, there was no significant overall survival difference between the GG and AG/ AA genotypes in both HNSCC (HR = 1.10, 95% CI: 0.78-1.54, p=0.65) and LC (HR = 1.16, 95% CI: 0.96-1.42, p = 0.13). In HNSCC, various tumor locations among studies may explain the negative result on the prognostic value of the FGFR4 Arg388 SNP, since different anatomical locations show different clinical and molecular characteristics (72). In terms of LC, the prognostic impact of the FGFR4 Arg388 variant may be different among histologic subtypes. Indeed, the prognostic role of this   SNP was more frequently observed in patients with ADC (9,18).
In the current study, unfortunately, we could not perform the subgroup analysis according to the tumor location in HNSCC and histologic subtype in LC due to a limited number of relevant articles. The previous studies have reported that there is significantly different in the prevalence of the FGFR4 Arg388 allele between Asians (37.2-40.1%) and Caucasians (29.5-30.4%) (29,33). The higher frequency of the Arg388 allele in Asian populations might lead to a higher statistical power of studies in Asians than in non-Asians. However, the meta-analysis by Xu at al. reported that the association between Arg variant genotype and increase risk of cancer was significant only in Asians, not in Caucacians or Africans (33). In another meta-analysis by Liwei et al., the significant association between Arg variant genotype and susceptibility of PC was observed among European ethic descents, not among African-Americans (31). When we stratified by the ethnicity in the current study, the FGFR4 Arg388 allele was associated with worse overall survival in Asian population (HR = 1.37, 95% CI: 1.19-1.57, p < 0.00001), but not in non-Asians (HR = 1.07, 95% CI: 0.89-1.27, p = 0.47). These findings suggest that the genetic effect of the FGFR4 Gly388Arg SNP may be different among the ethnicities.
Beside the FGFR4 rs351855 G>A polymorphism, there are also other SNPs or mutations of FGFR4 which may affect the risk of cancer development or prognosis of cancer patients. The recent meta-analysis by Moazeni-Roodi et al. revealed that the FGFR4 rs1966265 C>T polymorphism significantly reduced the risk of cancer in the recessive model (TT vs CT+CC) and the rs7708357 G>A variant was significantly associated with increased cancer development in the dominant model (AG +AA vs GG) (32). The Y367C FGFR4 mutation in the extracellular juxtamembrane domain promotes the FGFR4 dimerization on the cell surface and thereby leads to ligandindependent activation of downstream signaling pathways (73). In addition, the mutations in FGFR4 kinase domain such as N535K and V550E cause receptor autophospholyation and then activate the STAT3 signal pathway (73). Some FGFR4 mutations (N535K, V548M and V550L) were reported to be relatively resistant to tyrosine kinase inhibitors (74). However, the clinical and pathological significance of these genetic variations involving FGFR4 are still needed to be investigated in further studies.
There were some inherent limitations of this meta-analysis. First, there might be a selection bias since we were only able to acquire data from published articles written in English. Second, most studies were performed retrospectively and therefore, may carry the biases of the retrospective design. Third, considerable number of included studies had a relatively small sample size, and most studies utilized univariate statistical method to compare survival outcomes. Forth, because of a paucity of relevant articles, we could not perform the subgroup analyses in other types of cancers than HNSCC and LC. Finally, there was a substantial heterogeneity in the pooled outcomes, which might weaken reliability of the meta-analysis although the randomeffects model was adopted.
In conclusion, this meta-analysis elucidated that FGFR4 Gly388Arg polymorphism was associated with worse prognosis in cancer patients. Our results suggest that this SNP may be a valuable genetic marker to identify patients at higher risk of recurrence or mortality. Considering the limitations of the current study, however, large prospective researches with genotyping of the whole FGFR4 locus are warranted to reveal the clinicopathological and prognostic roles of the FGFR4  Gly388Arg SNP among various types of cancer, histology, and ethnicity.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

AUTHOR CONTRIBUTIONS
JHK and HSK conceived and designed the study. STP and SYJ searched the literatures and extracted the data. HSK and STP carried out the statistical analyses and data interpretation. JHK and HJJ wrote the manuscript. All authors contributed to the article and approved the submitted version.