1 Introduction
With 350,000 deaths each year, head and neck cancer accounts for a significant part of global cancer mortality and is the sixth most common cancer worldwide, affecting more than 650,000 people each year [
1]. The most common types of head and neck cancer are laryngeal, oral, and oropharyngeal cancer [
2]. Although more than 95 % of these cancers are squamous cell carcinoma, clinical and molecular characteristics of these tumors are heterogeneous [
3‐
5]. Frequently, head and neck squamous cell carcinoma (HNSCC) is detected at an advanced stage implying that the primary tumor has already metastasized to the neck. Advanced stage HNSCC is nowadays treated with varying combinations of radiation therapy, chemotherapy and surgery.
Current chemoradiation therapy regimens cause severe short- and long-term side-effects in more than 80 % of HNSCC patients [
6]. Additionally, 5-year relative survival rates have slightly improved over the past three decades, but remain low at 65 %. Persisting poor survival rates of HNSCC with current treatment regimens have led to a search for novel therapeutic targets and prognostic biomarkers [
6,
7]. The effort to resolve these problems has led to a quest for novel predictive and prognostic biomarkers with the intention to individualize treatment and reserve aggressive therapy for biologically aggressive tumors. As a result, molecular carcinogenesis has become a major focus of cancer research. Previous research endeavors in the pursuit of novel therapeutic targets have identified potential predictive and prognostic molecular biomarkers in HNSCC [
3,
8]. One of them is the fibroblast growth factor receptor (FGFR) family [
9].
FGFRs are upcoming promising therapeutic targets and possible prognostic biomarkers in multiple types of cancer, including HNSCC [
9,
10]. The FGFR family comprises five (FGFR1–5) cell membrane-bound tyrosine kinase receptors linked to multiple intracellular downstream signaling pathways. FGFRs regulate tissue homeostasis in normal human tissues [
11,
12]. In cancer cells, oncogenic aberrations in FGFR pathway-related genes dysregulate and constitutively activate the FGFR pathway, resulting in particular hallmark capabilities in cancer cells: to sustain proliferative signaling, resist cell death, induce angiogenesis, and activate invasion by cell migration [
13‐
15]. These genomic aberrations include gene fusion, translocation, amplification, and somatic DNA mutations [
9]. Because of their major role in cancer cell biology, FGFR family members provide promising opportunities for targeted therapies in a wide spectrum of solid tumors [
9,
16,
17]. In addition, previous studies have identified a possible role of FGFRs as prognostic biomarkers, by which they could select patients for adjuvant systemic therapy [
18,
19]. However, the prognostic value of FGFRs in HNSCC remains a subject of debate and has not yet been reviewed in this type of cancer. Therefore, the aim of this study was to systematically review current evidence on the prognostic value of FGFR1–5 in HNSCC and analyze it in a clinically relevant meta-analysis.
2 Materials and Methods
2.1 Eligibility Criteria and Information Sources
Studies were eligible for inclusion if they were English original articles and addressed the prognostic value of FGFRs in any type of HNSCC (i.e. overall survival, disease-specific survival, disease-free survival, recurrence-free survival, or progression-free survival). FGFR was required to be investigated as a molecular biomarker with laboratory techniques (protein expression, gene amplification, mutation, translation, polymorphism, or mRNA). Original articles were defined as primary research studies with new, unpublished results and written by the researchers who performed the study. Studies were excluded if they investigated fibroblast growth factor (FGF) instead of FGFR, had no prognostic study design, were repetitive studies on same samples, or were non-English. Also, animal studies, case reports, reviews, meta-analysis, and commentaries were excluded. A systematic search of PubMed, Embase, and the Cochrane Library was performed for publications up to 14 May 2014.
2.2 Search Strategy and Study Selection
The search terms ‘FGFR1–5’ and ‘head and neck cancer’ with all relevant synonyms were used (Online resource
1). Using predefined inclusion and exclusion criteria, two authors (NI and KL) independently screened all retrieved records on title and abstract and excluded duplicate titles to select potentially eligible articles. Subsequently, relevant articles were screened on full text, and a further selection of eligible articles was made on relevance of full text. Finally, review articles on this topic and references of selected articles were manually screened for titles not identified by the initial search. Selection was based on consensus.
2.3 Data Collection and Data Items
Two authors independently (NI and KL) extracted all data of the selected studies using a standardized data extraction form. Discrepancies were resolved by discussion. The following information was extracted from each study: first author’s name, year of publication, sample size, head and neck site, treatment, FGFR aberration investigated, type of survival outcome, patient material, laboratory techniques, statistical analysis, cut-off value (if applicable), prevalence of FGFR aberration, correlations with survival outcome, and length of follow-up.
2.4 Risk of Bias in Individual Studies
The methodological quality of the remaining eligible articles was independently assessed by two reviewers (NI and KL), using criteria of the Quality In Prognosis Studies (QUIPS) tool [
20]. This tool has been evaluated in 43 groups reviewing prognostic studies and has been identified as a reliable and useful tool for systematic reviewers. Among nine review groups, inter-rater agreement statistic (kappa) varied between 0.56–0.82 [
21]. According to the QUIPS tool, risk of bias was scored as low, moderate or high for the following six items; study participation, study attrition, prognostic factor measurement, outcome measurement, study confounding and statistical analysis, and reporting (Online resource
2). Studies scoring low risk of bias on three or more items were considered to be of ‘high’ methodological quality, while studies scoring high risk of bias on three or more items were considered to be of ‘low’ methodological quality. All other studies were of moderate quality. The use of proper positive and negative controls for laboratory techniques on FGFR and the use of well-defined scoring criteria for FGFR aberrations were also considered as methodological quality criteria. When there was disagreement on (certain items of) risk of bias of a study, two authors (NI and KL) discussed reasons for disagreement in order to mutually agree on final risk of bias score using the QUIPS tool (Online resource
2).
2.5 Synthesis of Results
Because of clinical and methodological heterogeneity of the included studies (study population, type of survival outcome, material, techniques, cut-off values, and applied statistical models), results were not quantitatively pooled. Therefore, a clinically relevant meta-analysis could not be performed. A forest plot was produced using Comprehensive Meta-Analysis Version 2.2.064 software (Biostat, Englewood, NJ, USA).
4 Discussion
During the past decade, multiple molecular biomarkers which play a role in tumor growth and metastatic spread have been identified. Molecular carcinogenesis has become a major field of cancer research and is driven by the need for novel targetable and prognostic molecular biomarkers. Among these are the FGFR family members [
9]. There is increasing evidence for the therapeutic role of FGFRs in HNSCC, but their prognostic role remained unclear.
Herein, we present the first systematic review of currently available evidence on the prognostic value of FGFRs in HNSCC. All included articles were critically appraised using the QUIPS tool items [
20]. Studies addressed the prognostic value of FGFR1 gene amplification, FGFR1 protein expression, FGFR4 genotype, and FGFR4 protein expression in 1870 HNSCC cases of various sites, including oral, oropharyngeal, hypopharyngeal, and laryngeal SCC. To date, 12 studies focused on the prognostic value of FGFRs in HNSCC [
18,
19,
23‐
32]. While our extensive search retrieved multiple articles reporting on FGFR2 and FGFR3 in HNSCC, none of these studies assessed their prognostic value [
33‐
39]. Therefore, only FGFR1 and FGFR4 could be assessed as prognostic biomarkers in HNSCC.
Current evidence suggests that
FGFR1 gene amplification and protein overexpression are not prognostic biomarkers of value in HNSCC. However, in one study, FGFR1 protein expression in cancer-associated fibroblasts of HNSCC was a prognostic biomarker which should be validated in a larger cohort [
27]. Interestingly, evidence on the prognostic value of the
FGFR4 Gly388Arg polymorphism was conflicting [
18,
19,
23,
28‐
32]. This could possibly be explained by differences in duration of follow-up time among studies with positive and negative results. In one study, differences in overall survival only occurred after 24 months of follow-up, while two out of three studies with negative results had a follow-up period of less than three years [
23,
31,
32]. This in contrast to positive results in the other four studies with a follow-up time of more than 3 years [
18,
19,
28,
30]. Remarkably, studies with statistically significant survival results used FFPE tissues, while studies without significant results used DNA from either fresh frozen tumor tissue, peripheral blood, or cell lines [
18,
19,
23,
28‐
32]. Perhaps the type of study material may influence the PCR techniques used to detect the
FGFR4 Gly388Arg polymorphism. Differences in tumor localization of HNSCC might explain the conflicting results on DFS related to the
FGFR4 Gly388Arg polymorphism found by Azad et al and Dutra et al [
18,
29]. Azad et al also included hypopharyngeal and laryngeal tumors besides oral and oropharyngeal tumors. HNSCC from different anatomical locations show different clinical and molecular characteristics and are acknowledged as different entities. As such, tumor localization might affect patient outcome. Other differences in study populations or patient treatment could not explain the conflicting results. However, none of the studies were non-randomized controlled trials.
The prognostic value of FGFR has been investigated in numerous other types of cancer. Our findings are in agreement with studies on other types of cancer that also showed conflicting results. For example, studies on
FGFR1 gene amplification and
FGFR4 Gly388Arg polymorphism in lung SCC [
40‐
46] and breast cancer [
47‐
49] reported conflicting results on overall survival and disease-free survival. Only studies on
FGFR3 mutations in bladder cancer are in agreement with each other; nearly all studies found correlations with better progression-free survival and disease-specific survival [
50]. For all other FGFRs in cancer, evidence on their prognostic value is limited or absent and therefore inconclusive, which is similar to HNSCC.
This review focusses on the prognostic value of FGFR aberrations in HNSCC. These prognostic associations do not necessarily predict the response to FGFR-inhibitor therapy as the latter is mainly, yet exclusively, predicted by the underlying aberration itself. For example, previous clinical studies in breast and lung cancer showed that FGFR-inhibitor therapies were only effective in FGFR-amplified or FGFR-mutated tumors [
51,
52]. The
FGFR genes are frequently aberrated in HNSCC,
FGFR1 is amplified in 10 % of HPV-negative HNSCC and
FGFR3 is aberrated in 11 % of HPV-positive HNSCC [
53]. As such, HNSCC patients with
FGFR-aberrated tumors may benefit from FGFR-inhibitor therapies as these tumors may be sensitive. In addition, targeting FGFR family members has shown to enhance radiotherapy and chemotherapy sensitivity of cancer cells. Radiotherapy resistant cancer cells upregulate FGFR3 protein and chemoradiotherapy resistant cancer cells upregulate FGFR4 protein. Targeting FGFR3 in resistant HNSCC cells restored sensitivity to radiotherapy and targeting FGFR4 restored sensitivity to chemoradiotherapy [
54‐
56].
The probative value of the present review depends on methodological quality of included studies to some extent. All included studies have their specific strengths but the methodological quality is questionable in many. Poor description of study populations and incomplete data in two studies might have introduced selection bias [
24,
25]. Six studies had a mean/median follow-up time less than three years [
23,
24,
26‐
28,
31]. Risk of information bias is present in all included studies as none of them provided information on blinding of investigators for the current vital status of patients. Survival rate was only standardized in one study [
25]. In all other studies, the definition and assessment of the specific survival outcomes were unclear. Some articles have relatively low sample sizes and did not use multivariable statistical methods, such as multivariable Cox proportional hazard models. Because of risk of bias among included studies, cautious interpretation of results is recommended.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.