Background
Head and neck squamous cell carcinoma (HNSCC) is an aggressive disease and is the sixth most common cancer worldwide, with approximately 650,000 cases diagnosed worldwide annually and nearly 400, 000 deaths [
1,
2]. HNSCC encompasses a wide variety of malignancies deriving from the mucosal epithelium of the upper aerodigestive tract, including lip, oral cavity, paranasal sinuses, nasal cavity, pharynx and larynx [
3]. Data from the USA indicates over two-thirds of patients present with advanced-stage disease with either locoregional spread to the lymph nodes or distant metastasis [
4]. Historically, up to 50 % of patients will experience locoregional recurrence within 2 years of treatment with limited options for salvage surgery or reirradiation [
4,
5]. To date, there is limited molecular characterisation of the driver mutations of the various subtypes of HNSCC, with human papilloma virus (HPV), smoking and alcohol the only identified causative agents. Therefore, understanding the biological mechanisms that lead to cancer progression and identification of prognostic factors are essential to improve the clinical management of HNSCC.
A hallmark of many cancers, including HNSCC, is the presence of a tumour promoting phenotype of chronic, low-grade cancer-related inflammation [
6‐
8]. Recent studies have demonstrated that cancer-related inflammation derives from communication between the host and tumour cells to develop a reciprocal interplay that often results in systemic alterations, immune suppression and evasion and malignant progression [
6]. In HNSCC, cancer-related inflammation is characterised by increased circulating concentrations of pro-inflammatory cytokines and acute phase reactant proteins (C-reactive protein, serum amyloid A protein) that enhance the recruitment of circulating neutrophils, monocytes [
9], myeloid derived suppressor cells (MDSC) [
10,
11], and thus total leucocyte numbers, whilst also inhibiting the recruitment of lymphocytes to the circulation. These changes lead to the development of cancer-related syndromes, including fever, night sweats, fatigue, cachexia and bone and muscle pain [
12].
Over the last few years, there has been a proliferation in clinical studies measuring the systemic inflammatory response in cancer patients to identify patients with poor prognosis (reviewed in [
7,
13]). One of the key biomarkers of systemic inflammation is the neutrophil-to-lymphocyte ratio (NLR). An NLR score is obtained from a patients full blood count by dividing the absolute neutrophil count by the absolute lymphocyte count. An elevated NLR is strongly related to other inflammatory markers, including the Glasgow Prognostic Score, platelet-lymphocyte ratio and elevated C-reactive protein levels, which have been associated with increased tumour burden and spread of disease. NLR is elevated in patients with laryngeal squamous cell carcinoma compared to patients with benign and precancerous lesions [
14]. NLR is also an independent prognostic marker of reduced overall survival (OS) in most epithelial cancers [
6,
15].
There have been numerous studies of the prognostic role of NLR in various selected populations of HNSCC. Small studies conducted in site-specific populations of nasopharyngeal, oropharyngeal and oral cavity cancers, showed elevated NLR was predictive of local and regional recurrence or reduced progression free survival and/ or poorer OS [
16‐
20]. Investigations in small cohorts of unselected HNSCC patients have shown that HNSCC patients have an elevated NLR compared to healthy controls and univariate analyses have associated elevated NLR to recurrence, tumour and nodal stage [
21‐
23]. A pilot study in 46 unselected HNSCC patients was conducted by our group and univariate analysis found that NLR was predictive of shorter overall survival [
24]. However, in these investigations of heterogeneous populations of HNSCC, multivariable analysis of NLR as prognostic of recurrence free survival (RFS) or OS was not undertaken.
Additionally, literature shows that HPV mediated overexpression of p16 is an important marker of reduced risk for recurrence and survival in HNSCC [
25,
26]. Recent in vitro and animal studies of cervical cancer have shown that HPV positive (HPV+) cells are more efficient at producing a pro-inflammatory tumour microenvironment [
27] leading to enhanced myeloid cell proliferation in the bone marrow and spleen and increased recruitment of leucocytes to the tumour [
28]. Thus, the p16 status of a patient may also alter the inflammatory response and contribute both directly and indirectly to cancer outcomes. Huang et al. [
9] identified that p16 positive oropharyngeal cancer patients with high circulating neutrophil levels have a reduced OS and RFS. Interestingly, this association was not seen in the p16 negative oropharyngeal patients. Furthermore, higher levels of circulating lymphocytes were predictive of improved RFS and marginally improved OS in the p16 positive population but not in the p16 negative patients. Additionally, in a study by Ward et al. [
29], HPV+ oropharyngeal cancer patients with high or moderate tumour infiltrating lymphocyte expression had significantly improved survival compared to HPV+ low tumour infiltrating lymphocytes and HPV negative (HPV-) patients regardless of lymphocyte expression. This would suggest within the HPV+ oropharyngeal cancer population the systemic and local inflammatory environment may be important for determination of clinical outcomes. In both studies there is a significant minority of HPV+ patients (20 %) that have poor OS. Identification of this high risk group is important in an era of potential treatment de-escalation and introduction of molecularly targeted therapies. In addition, systemic inflammation has not been well investigated as predictive biomarker for all clinical outcomes in the non-oropharyngeal cancer population and identification of the high risk group of patients is also essential.
In this retrospective analysis, we sought to investigate whether NLR was an independent prognostic factor of RFS and OS in a prospectively collected, non-selected HNSCC population from one treatment centre. In addition we investigated whether elevated NLR was associated with clinicopathological features, including p16 status, which may aid in treatment decisions.
Discussion
NLR is an easily obtainable, inexpensive marker of systemic inflammation that may assist in clinical decisions regarding recurrence and survival in a heterogeneous HNSCC population. This study aimed to investigate the predictive role of NLR in an unselected population of HNSCC patient, but we found that oropharyngeal patients had significant differences in baseline characteristics compared to non-oropharyngeal patients. As expected from the growing literature, a very high percentage of oropharyngeal patients were p16 positive (84 %). Thus, we conducted total and sub-site analyses due to the differences reflecting the potential diverging molecular etiology of oropharyngeal and non-oropharyngeal disease. In univariate analysis, NLR status did not associate with any other clinicopathological variables other than neutrophil and lymphocyte levels in either subgroup as expected. Patients with an elevated NLR were associated with shorter RFS and OS in both oropharyngeal and non-oropharyngeal populations. Univariate survival analysis showed ECOG PS, smoking status, age and NLR associated with RFS and OS to varying degrees in both populations. Multivariable analysis confirmed NLR significantly predicted RFS in oropharyngeal patients only, while NLR strongly predicted OS in both sub-populations. Additionally, ECOG PS significantly showed associations to OS in oropharyngeal patients and non-oropharyngeal patients. Interestingly, smoking status remained predictive of RFS and OS only in non-oropharyngeal patients. This may not be unexpected considering the low numbers of p16 negative patients in the oropharyngeal cancer cohort, which may represent the contribution of smoking habits in the causation of disease in these patients.
The majority of oropharyngeal patients were p16 positive in this study and oropharyngeal patients were younger, had better ECOG PS but had increased nodal spread compared to non-oropharyngeal patients. Eighty-four percent of tested oropharyngeal patients had p16 positive tumours. This percentage is comparable to other American, Swedish and British studies (summarised in [
32]) although higher than the average rate (~40 %) in most developed countries. Patients with p16 positive tumours are generally younger [
33] and have been noted to have better ECOG PS and higher nodal stages when compared to p16 negative patients [
34,
35]. The high prevalence of p16 in the oropharyngeal population most likely accounts for the younger age and better ECOG PS compared to non-oropharyngeal patients seen in this study. The higher nodal stage but improved outcomes in p16 positive patients is the most likely cause of AJCC stage not being significant in our study, similar to other reports [
34]. The 3-year OS of oropharyngeal patients was 86 % and 69 % for p16 positive and negative patients respectively, which is comparable to larger studies [
36‐
38]. In vitro studies with p16 positive HNSCC cells lines have shown that these cells are more radiosensitive [
39]. Oropharyngeal patients in our unit are unlikely to undergo primary surgical intervention due to the perceived high risk of morbidity if extensive surgery is required. Our excellent rates of locoregional control in this population further support this recommendation. However with availability of transoral robotic assisted surgery [
40], a biomarker to predict a poor performing oropharyngeal subgroup may aid selection of patients for surgery in the future.
With decreasing smoking rates due to extensive anti-smoking campaigns, as seen in countries such as Australia, HPV+ oropharyngeal cancer is increasingly becoming the prominent subtype. Therefore, additional predictive biomarkers of clinical outcomes are needed within the HPV+ or p16 positive oropharyngeal cancer population. Additionally, classification of patients as HPV+ is not without difficultly as the various techniques of assessment produce variable results and there is no universally agreed classification system. The results of this study show that on a background of high p16 positive status, elevated NLR was associated with recurrence and survival outcomes under univariate analysis and many of the recurrences and deaths occurred within the first year following radiotherapy. Multivariable analysis showed that NLR remained a predictor of OS independent of AJCC stage, tumour site, treatment modality and sex in oropharyngeal and non-oropharyngeal sub-populations. Additionally, NLR also predicted RFS in oropharyngeal patients. The results of this study identified NLR as a prognostic marker of OS in an unselected HNSCC cohort, supporting previous findings from other studies in nasopharyngeal, oral squamous cell carcinoma and preliminary investigations in unselected HNSCC cohorts [
14,
16‐
22,
41]. These findings are also consistent with other cancer types including other head and neck associated cancers, such as thyroid cancer [
42,
43].
The association between NLR and poor OS and recurrence is not well understood. However, it is hypothesised that elevated NLR reflects a more aggressive tumour phenotype that is immune evasive and/or suppressive. Elevated NLR is more often seen in patients with advanced disease, as denoted by increased AJCC stage, tumour depth of invasion or metastatic spread [
7]. In our study, we did not find evidence to confirm NLR was associated with higher AJCC staging and thus may represent aspects reflecting immune suppression. Recent analysis conducted by The Cancer Genome Atlas project, shows that within the HPV+ population of HNSCC there is an increase in loss of
TNF receptor-associated factor 3 gene and presence of activating mutations in
PIK3CA gene, which enhance NF-κB signalling and promote a pro-inflammatory microenvironment [
44]. This data supports the role of cancer-related inflammation in determining the outcomes of HPV+ HNSCC patients.
In the tumour microenvironment, innate immune cells, such as neutrophils, macrophages and myeloid derived suppressor cells, regulate both immune surveillance and suppression [
45]. Increased abundance of these cells is observed in more advanced stages of HNSCC and is associated with poorer survival [
9‐
11,
46]. Mechanistic studies conducted in animal models and ex vivo cultures of immune cells from HNSCC patients have demonstrated that myeloid derived suppressor cells are critical for regulating the immunosuppressive phenotype and function of co-operating lymphoid-derived cells in the tumour and circulation [
10,
11,
47,
48]. In terms of adaptive immune cells, the low infiltration of T cells, particularly T regulatory cells, combined with functional deficits in T helper cells, cytotoxic T cells and natural killer cells leads to the highly immune suppressive tumour microenvironment that allows for unrestrained tumour growth [
29,
49‐
52].
Improved understanding of the various interactions of the tumour and immune system suggest that the ideal biomarker would measure both the innate and adaptive immune response, such as the NLR, as this may provide a better indication of the impact of tumour growth on both arms of the host immune response. In a mixed cancer population (not including HNSCC patients) elevated NLR was found to positively correlate with circulating MDSC levels and suppression of lymphocyte function [
53]. However, there is no evidence to date that specifically links elevated NLR to immune cell behaviour in HNSCC tumours or circulation. Unfortunately, we do not have blood samples from our patient cohort, but it would be interesting to investigate circulating NLR values in studies that have measured peripheral blood and tumoural MDSC or T cell populations and overall survival to clarify the biological relationships between NLR and immune suppression in cancer.
More recently, the NLR has been suggested as a Phase I clinical trial patient selection tool by the Royal Marsden Hospital, UK [
54]. Pharmacological inhibitors of key immunosuppressive mediators (anti-PD1 or PD ligand 1 antibodies, STAT3 and PDE5 inhibitors) have been shown to reduce the number and function of MDSC, Tregs and/or immune T cell-mediated anti-tumour responses in mice and are increasingly being investigated in clinical trials [
11,
55,
56] . New data from the The Cancer Genome Atlas [
44] has also suggested novel pathways for intervention, such as the PIK3 pathway due to activating mutations in
PI3KCA for HPV+ cancers. Thus, NLR could be useful as inclusion criteria for clinical trial participation investigating these molecular targeted and immune modulating therapies.
In addition to NLR predicting survival outcomes, other exploratory variables including smoking status and ECOG PS were predictive of RFS and OS. Smoking status was a significant predictor of RFS and OS but only in the non-oropharyngeal population. Smoking is not only a risk factor for the development of head and neck cancer but patients who maintain smoking during treatment are also at increased hazard of worse clinical outcomes [
57]. Patients with poorer ECOG PS had an increased hazard of death in both sub-populations which has been suggested previously in HNSCC [
58] and observed in other advanced cancers [
59].
This study is limited by inherent selection bias due to the retrospective analysis of this study and being conducted in one metropolitan area hospital. However, this cohort reflects the heterogeneous nature of HNSCC in the community. Our population has a large proportion of p16 positive oropharyngeal tumours with comparable clinical outcomes to other international sites. Similar to other cancers (breast, colon, lung), the management of this patient group will change in the future from one single treatment to individualised treatments based on patient and tumour characteristics. One of the main limitations of the study was the incomplete analysis of p16 status in all patients. It would be of interest to investigate if the p16 positive oropharyngeal patients alone mimic the results of the total oropharyngeal population. Due to the high rates of positive patients it is probable that the results would be similar, unfortunately, our study did not have large enough numbers of tested patients to confirm this assumption. Although all patients had radiotherapy and chemotherapy at the one site, surgery was conducted over multiple hospital sites. Unfortunately, we were unable to collect diagnostic blocks from some surgical sites and private pathology laboratories. As such, we assumed based on the lack of significant differences in major covariates in the tested and untested populations and consistency with overall rates of p16 positive oropharyngeal cancer patients in our local area health service, that p16 status was not a statistically relevant covariate in our patient population. Using this assumption we may have missed an important interaction between p16 and NLR. In addition, we used p16 immunohistochemistry as the method for HPV positivity. There is a known discordance between DNA and protein detection methods [
60,
61]. The p16 positive immunohistochemistry staining method, as performed in this paper, assumes that the overexpression of p16 is predominantly due to HPV infections, however HPV-independent mechanisms such as alterations in the retinoblastoma pathway may also drive p16 expression [
44]. A variety of DNA-based and immunohistochemical methods have been used in various studies and consensus methods are being developed.
Competing interest
The authors declare that they have no competing interests.
Authors’ contributions
KC, SC and TE conceived the design of the study. CH, SC, AG, MS, MB, DV and TE were involved in data collection, data entry and clinical management of patients. TD and AJG interpreted immunohistochemistry staining. KC, BH, CH carried out statistical analysis. All authors contributed to data interpretation. All authors contributed to drafting and revising the final manuscript.