Relation of baseline neutrophil-to-lymphocyte ratio to survival and toxicity in head and neck cancer patients treated with (chemo-) radiation

Background A high neutrophil-to-lymphocyte ratio (NLR) is a marker of systemic inflammation and together with the platelet-to-lymphocyte ratio (PLR) is associated with worse outcomes in several solid tumors. We investigated the prognostic value of NLR and PLR in patients with head and neck squamous cell carcinoma (HNSCC) treated with primary or adjuvant (chemo)radiotherapy ((C)RT). Methods A retrospective chart review of consecutive patients with HNSCC was performed. Neutrophil-to-lymphocyte ratio and PLR were computed using complete blood counts (CBCs) performed within 10 days before treatment start. The prognostic role of NLR and PLR was evaluated with univariable and multivariable Cox regression analyses adjusting for disease-specific prognostic factors. NLR and PLR were assessed as log-transformed continuous variables (log NLR and log PLR). Endpoints of interest were overall survival (OS), locoregional recurrence-free survival (LRFS), distant recurrence-free survival (DRFS), and acute toxicity. Results We analyzed 186 patients treated from 2007 to 2010. Primary sites were oropharynx (45%), oral cavity (28%), hypopharynx (14%), and larynx (13%). Median follow-up was 49 months. Higher NLR was associated with OS (adjusted HR per 1 unit higher log NLR = 1.81 (1.16–2.81), p = 0.012), whereas no association could be shown with LRFS (HR = 1.49 (0,83-2,68), p = 0.182), DRFS (HR = 1.38 (0.65–3.22), p = 0.4), or acute toxicity grade ≥ 2. PLR was not associated with outcome, nor with toxicity. Conclusion Our data suggest that in HNSCC patients treated with primary or adjuvant (C)RT, NLR is an independent predictor of mortality, but not disease-specific outcomes or toxicity. Neutrophil-to-lymphocyte ratio is a readily available biomarker that could improve pre-treatment prognostication and may be used for risk-stratification.


Background
Risk stratification of patients diagnosed with head and neck squamous cell carcinoma (HNSCC) poses an important challenge [1]. Currently, some of the widely used factors are smoking and human papillomavirus (HPV) status, age, performance status, and tumor stage. Nomograms based on baseline characteristics can enhance prognostic prediction [2].
Inflammation is a hallmark of cancer [3], which is shown to play an important role in tumor development and progression [4][5][6]. An elevation of circulating neutrophil count is thought to be the result of tumor cells releasing cytokines, which stimulate the bone marrow to produce neutrophils [7][8][9]. Cytokines released by neutrophils also promote angiogenesis leading to tumor growth and metastasis [10][11][12][13][14][15]. There is an increasing interest in the use of hematological parameters as prognostic factors in malignancies. Neutrophil, lymphocyte, and platelet counts, either as individual values or in relation to each other, could be associated with the cancer prognosis [16,17]. The neutrophil-to-lymphocyte ratio (NLR) is an emerging marker of host inflammation, which reflects the relation between circulating neutrophil and lymphocyte counts. It can be easily calculated from routine complete blood counts (CBCs) with differentiation. The independent prognostic value of NLR has been shown for a variety of solid malignancies [17][18][19][20]. In addition to NLR, the platelet-to-lymphocyte ratio (PLR) has also been shown to be a potential prognostic factor [2,19]. Several studies involving HNSCC have shown an association between inflammation and worse prognosis [21][22][23][24][25][26][27]. However, information about the possible value of pretreatment NLR or PLR on toxicity is limited [18][19][20][21][22][23][24][25][26][27][28][29].
In this study, we retrospectively evaluated the prognostic impact of pretreatment NLR and PLR on oncological outcomes and toxicity in HNSCC patients treated with primary or adjuvant curative-intended (chemo-) radiotherapy ((C)RT). We hypothesized that elevated NLR and/or PLR are associated with detrimental survival; we also explored NRL and PLR associations with acute treatment-related toxicity since it has prognostic value in primary and adjuvant (C)RT for HNSCC [30,31].

Patient selection
Medical records of HNSCC patients consecutively treated with primary or adjuvant curative-intent intensity-modulated radiation therapy with or without concomitant systemic therapy between January 2007 and December 2010 at the Department of Radiation Oncology, Inselspital, Bern University Hospital were retrospectively analyzed. Patients diagnosed with oral cavity (OCC), oropharynx (OC), hypopharynx (HC) and laryngeal cancers (LC) were included in the analysis. History of another malignancy within 5 years of diagnosis, prior radiation to the head and neck, non-squamous cell carcinoma histology, distant metastases, lack of differentiated CBC within 10 days before oncologic surgery or RT start, and early abortion of RT were defined as exclusion criteria. This study was approved by the local ethics committee (289/2014).

Treatment and follow-up
The standard treatment was based on institutional policies following the multidisciplinary tumor board decision as previously published [32,33]. All cases were presented at the weekly institutional interdisciplinary head-and-neck tumor board. After completion of staging examinations and final TNM staging (AJCC), selection of treatment modalities and treatment sequencing were defined. The standard treatment in OCC was to perform surgery followed by adjuvant radiotherapy (RT) [30,32], while in OC, HC and LC the joint recommendations of the multidisciplinary meeting was primary RT [31,33]. Case-based decisions were made concerning the use of concomitant systemic therapy and up-front neck dissection. The delivery of radiotherapy, the definition of clinical target volume (CTV) and planned target volume (PTV) followed departmental guidelines [32,33] based on international recommendations [34][35][36]. All treatment plans were contoured and calculated using Eclipse treatment planning system (Varian Medical Systems, Palo Alto, CA). The standard concomitant therapy consisted of cisplatin 100 mg/m2 day 1 in three-week intervals for all patients. In few cases of induction chemotherapy, cisplatin, docetaxel, and 5-fluorouracil were used. Patients not deemed medically fit for cisplatin chemotherapy because of pre-existing co-morbidities were evaluated for weekly treatment with monoclonal antibody cetuximab [37] or carboplatin three weekly. Pre-treatment CBC with differential values was used to calculate NLR and PLR.

Statistical analysis
NLR was calculated by dividing absolute neutrophil count by absolute lymphocyte count measured in peripheral blood. PLR was calculated by dividing absolute thrombocyte count by absolute lymphocyte count. Due to its non-normal distribution, NLR and PLR were log e -transformed to obtain symmetric distributions and then analyzed as continuous variables. Frequencies and percentages are reported for categorical variables, medians with range or interquartile range for continuous variables. The primary endpoint of the study was overall survival (OS), and the secondary endpoints were locoregional relapse-free survival (LRFS) and distant recurrence-free survival (DRFS). Time-to-event was calculated for OS, LRFS, and DRFS from the start of RT to death (OS), locoregional relapse (LRFS), and distant recurrence (DRFS), respectively, with censoring of patients without such events at last follow up. Median times to event were estimated using the Kaplan Meier method. The prognostic value of NLR and PLR, and other variables (i.e. age, gender, smoking status, Karnofsky Performance Status (KPS), UICC stage, tumor grade, hemoglobin level) were assessed by univariable Cox regression analysis. Subsequently, multivariable analysis with forward elimination was planned with inclusion of all variables with a p-value < 0.05 in the univariable analysis. The association of NLR and PLR with acute and late toxicities (i.e. pain, dermatitis, mucositis, dysphagia, xerostomia) was examined using logistic regression. Analyses were carried out using SPSS version 23 (IBM Corp., Chicago, IL). The threshold for statistical significance was set at p < 0.05, and no correction for multiple testing was performed.

Patients
One hundred and eighty-six patients were included in the study. Patients' and disease characteristics are presented in Table 1. The majority of patients were male and in good performance status (KPS ≥ 70). The primary

Overall survival
At a median follow-up time of 40 months, 60 patients (32%) died; median OS was not reached. Higher NLR was associated with lower OS (Table 2). When dividing the population into two groups according to the median NLR, there was a significant OS difference between the groups (Fig. 1). For PLR there was a non-significant association between higher PLR and lower OS (Fig. 2). On univariable analysis log e NLR was associated with OS. Also, older age, worse Karnofsky Performance Status (KPS ≤ 70), and UICC stage IV were associated with lower OS. Performance status, UICC stage IV and log e NLR remained of prognostic value in multivariable analysis ( Table 2).

Recurrence
Of the variables tested, only UICC stage IV was associated with increased loco-regional, distant, and any recurrence rate, whereas no association was found for all other variables tested (Table 3). Consequently, no multivariable analyses were conducted. In patients with high NLR, recurrences occurred earlier, but the correlation was not statistically significant (Fig. 3).

Toxicity
Rates and grades of the most common acute toxicities are summarized in Table 4. There was no correlation between baseline NLR or PLR and the grade of toxicity (data not shown).

Discussion
NLR is the object of numerous previously published studies. Not only in oncology but also in other disciplines, blood counts reflecting the complexity of the immune system can be easily obtained at low costs, which may impact daily clinical practice. About 15-20% of all cancer deaths worldwide seem to be associated with underlying infections and inflammatory reactions [38]. Many triggers of chronic inflammation increase the risk of developing cancer. These triggers, for example, include microbial infections such as Helicobacter pylori (associated with stomach cancer), inflammatory bowel disease (associated with bowel cancer) and prostatitis (associated with prostate cancer) [38]. Despite conflicting studies, treatment with non-steroidal anti-inflammatory agents has been associated with reduced cancer incidence and mortality [38][39][40][41].
Increased NLR is associated with poorer outcomes in many solid tumors, be it early or advanced stage cancer [17]. An early decrease in NLR may be associated with more favorable outcomes and higher response rates [42], whereas an increase in NLR in the first weeks of treatment had the opposite effect [42].
In this study with a relatively large cohort of HNSCC patients treated with (C)RT with curative intention, an elevated NLR at baseline was associated with a shorter OS but not with disease recurrence or toxicities. Our findings of a negative prognostic role of NLR are in accordance with other studies [26,43] that have investigated NLR in HNSCC. In contrast to our results, Rassouli et al. [44] have demonstrated a statistically significant impact of PLR on OS. Worth to note, such associations were observed at various cut-offs in different studies. They have also shown that an increased NLR was  Fig. 1 Overall survival of NLR higher than median vs. equal or lower than median Fig. 2 Overall survival of PLR higher than median vs. equal or lower than median not only associated with decreased OS but with higher recurrence rates too [44]; which was not shown in our cohort and another study from the United Kingdom [45]. Along with the increased NLR in malignant disease, a possible explanation for a lower OS could also be a cause of death not attributable to cancer, but other co-morbidities such as a cardiac cause where it could also be shown that an increased NLR is predictive for cardiac mortality [46]. It is also known that smokers have a "smoker's leukocytosis" [37,38,47,48]. In our cohort, most patients are at least ex-smokers (80%), and at least one third continued smoking during and after radiation. Therefore, it is possible that the patients with a smoker's leukocytosis have died earlier from smoking-related comorbidities [49].
Several limitations to our study should be considered. First, this was a retrospective analysis with possible selection bias and confounding variables. We included 16 patients (9%) with early-stage disease and 15 (8%) patients who had neoadjuvant chemotherapy, which may have introduced some heterogeneity to our cohort. Second, we were unable to capture data on HPV status systematically. Studies have shown an important interaction between HPV status, immunomodulation and  [50]. Therefore, there might be different results in HPV-associated and unassociated tumors [51]. Since this is a retrospective study, there might be unknown causes of CBC changes that have not been identified. Beside patient and tumor-specific factors which may influence the complex cascades of the immune system, it must also be noted that despite clinical benefit, the dichotomization or grouping of continuous variables in statistical analysis is accompanied by a loss of the statistical power. To account for this, NLR and PLR were analyzed as (log-transformed) continuous variables. Lastly, an overestimation of statistical significance due to multiple testing is possible. Although these results should be validated in other cohorts, we reproduced some of the previously reported studies [26,43] on the interface of systemic inflammatory pathways and OS. Therefore, we provide data on surrogate values for inflammation as predictors of clinical outcomes; however, a causal relationship and its impact on tumor aggressiveness or tumor microenvironment warrants further investigation.

Conclusion
Our data suggest that in HNSCC patients treated with primary or adjuvant (C)RT, NLR is an independent predictor of OS. NLR is a readily available biomarker that could improve pre-treatment risk stratification.