The Neutrophil-to-Lymphocyte Ratio as a Prognostic Indicator in Head and Neck Cancer: A Systematic Review and Meta-Analysis

Background The aim of this systematic review and meta-analysis was to investigate the relationship between the Neutrophil-to-Lymphocyte Ratio (NLR) and prognosis in HNC. Methods Studies were identified from Pubmed, Embase, Scopus, and the Cochrane Library. A systematic review and meta-analysis were performed to generate the pooled hazard ratios (HR) for overall survival (OS), disease free survival (DFS), and progression free survival (PFS). Results Our analysis combined the results of over 6770 patients in 26 cohorts (25 studies). The pooled data demonstrated that an elevated NLR significantly predicted poorer OS, DFS, and PFS. Heterogeneity was found for OS, PFS, and marginally for DFS. Subgroup analysis in OS demonstrated that elevated NLR remained an indicator of poor prognosis. Conclusions Elevated pretreatment NLR is a prognostic marker for HNC. It represents a simple and easily obtained marker that could be used to stratify groups of high-risk patients that might benefit from adjuvant therapy.

Background: The aim of this systematic review and meta-analysis was to investigate the relationship between the Neutrophil-to-Lymphocyte Ratio (NLR) and prognosis in HNC.
Methods: Studies were identified from Pubmed, Embase, Scopus, and the Cochrane Library. A systematic review and meta-analysis were performed to generate the pooled hazard ratios (HR) for overall survival (OS), disease free survival (DFS), and progression free survival (PFS). weighted and pooled using the generic inverse-variance. 11 Because of anticipated heterogeneity, a more conservative approach applying the random effects model (DerSimonian and Laird method) was chosen for all analyses. Forest plots were constructed for all outcomes displaying the random-effects model of the summary effect measure and 95% CI. Heterogeneity was assessed using Cochran's Q and Higgins's I 2 .

Results
Cochrane's Q p-value of <0.1 and I 2 > 50% were considered as markers of significant heterogeneity. To assess publication bias, Begg's Funnel Plot and Egger's bias indicator test were used. If publication bias was detected, the influence of bias on the overall effect was assessed by Duval's "Trim and fill" method. 26 A Failsafe N measure was also calculated with the methods described by Rosenthal. 27 All analyses was done using the RevMan 5.3 analysis software (Cochrane Collaboration, Copenhagen, Denmark). 28 Tests for publication bias were performed by Meta-Essentials (ERASMUS Research Institute, Rotterdam, Netherlands). 29 All statistical tests were two-sided, and a p-value of less than 0.05 was considered statistically significant. No correction was made for multiple testing. The PRISMA flow chart of the systematic review can be found in Figure 1. An initial search done using the search strategy (Table 1) obtained an initial 900 results.

Results
De-duplication was then performed, which reduced the number of results to 500. The first phase of screening was performed next on titles and abstracts, which reduced the number of results to 65. The Thus, 25 studies published between 2011 and 2017 were included in our meta-analysis, with sample sizes ranging from 59 to 1410 patients. 20-25, 36, 67-84 As the study by Charles et al 68 had split their data into two cohorts with separately reported HR and 95% CI, we have designated both groups of data as Charles1 and Charles2 respectively. The characteristics of the included studies are summarized in Table 2. Eleven studies were from China, three from Japan, two from Korea, two from the USA, one from Australia (two cohorts), one from Austria, one from India, one from Italy, one from Singapore and one from the UK. Out of 25 studies, one was a prospective cohort study. The rest of the studies were based on retrospectively collected data.
NLR was calculated from laboratory data in all of the studies. NLR cutoffs ranged from 1.92 -5.56 (Median 2.895), with cutoffs unavailable from two studies. (Table 3)   The included studies were at low to moderate risk of bias with regards to study participation and study attrition. As a large majority of the studies were retrospective cohort studies, we found that there was an inherent risk of bias in patient selection. There were also studies that did not adequately report the attrition rate, or numbers of patients who were excluded because of unavailable data. Prognostic factor measurement (method of obtaining the NLR cutoff) for most of the studies was at a moderate risk of bias.
A number of the selected studies had derived their NLR cutoff values from previous literature, from percentile values of NLR, or was not even reported. There were some high quality studies that derived the NLR cutoff from Receiver Operator Characteristic (ROC) curve analysis alone or with a 'training cohort' data set. Outcome measurement was mostly at a low risk of bias for the selected studies, with many studies having clear definitions and descriptions of their endpoints. Study confounding was at a low to moderate risk of bias, with most studies having appropriate and sufficient covariates. Statistical analysis and reporting was mostly at a low risk of bias, with most studies having the appropriate statistical designs and data reporting. There was only a minority of studies that only reported univariate/unadjusted data. Figure 2 shows the overall summary of the quality assessment grading, and Table 4 Figure 3.
Data from 11 studies (12 cohorts) were synthesized in the meta-analysis for NLR and DFS/RFS in HNC.
An elevated NLR value above the cutoff was found to be significantly associated with poorer DFS/RFS Higgins's I 2 indicated a significant interaction existed between the subtotal estimates for the subgroups.
Thus, it can be concluded that these subgroup stratifications estimated different population parameters.
The low within-group heterogeneity in the tumor site and ethnicity subgroups may point to different NLR cutoffs existing for each patient population. This is clinically consistent as we would expect tumors of different subsites and from different patient populations to have different characteristics that would affect survival. The high between-group heterogeneity and low within-group heterogeneity in subgroups analyzing study design (sample size, NLR cutoff method, NLR cutoff) could also point to the conclusion that higher quality studies had different results from lower quality studies. These differences were shown to be quantitative in nature, as the differences of effect were still in the same direction. It was expected that studies using more robust statistical techniques would come to a similar conclusion, as demonstrated by the low within group heterogeneity of studies using ROC curve analysis (I 2 =0%). These findings suggest that the effect of a dichotomized cutoff for NLR may have utility in different populations, and could be used to guide clinical stratification and decision making with regard to outcomes for HNC patients. Notwithstanding, due to the intrinsic limitations to meta-analyses, we recommend prudence to avoid over interpreting the results of the subgroup analysis. To the best of our knowledge, this is the first meta-analysis reporting the relationship between elevated pretreatment NLR and outcomes in all sites of the head and neck.
The paradigm of local and systemic inflammatory states interacting with the local tumor microenvironment is based on strong evidence. 2,86 However, the mechanism behind the association between a high NLR and poor cancer prognosis remain poorly understood. 5 A high NLR indicates a relative neutrophilia and lymphopenia, and neutrophilia has been known to inhibit the cytolytic activity of T cells and NK cells. 87, 88 On the other hand, the significance of lymphocyte infiltration of tumors has been shown to improve prognosis and response to treatment. 89 Perhaps the prognostic ability of NLR lies in its measurement of the pro-tumor versus anti-tumor dynamic in the host immune system. 5 93 have also received interest as prognostic indicators in HNC. It remains to be seen which of these markers, or combination of markers, is the superior option for clinical use as a prognostic biomarker.
In spite of our findings, there are several weaknesses of our study that we acknowledge. Heterogeneity was found in the pooled results for OS, PFS, and marginally so for DFS. Subgroup analysis on OS showed that tumor site, ethnicity and study design factors could account for the heterogeneity. It is very likely that the heterogeneity is secondary to the above factors, together with the unreported genetic diversity of head and neck cancers as well as other confounders (such as HPV status). Most of the studies included were also retrospective in nature, with only one study collecting data prospectively.
Furthermore, because of a lack of individual patient data in many of the studies, we were unable to perform meta-analyses of individual patient data (MAIPD). We were also unable to include all cohorts in the subgroup analysis due to the diverse patient populations represented in the included studies. Another limitation of this paper is the publication bias detected for OS, as there were significantly more papers published that reported a poorer OS for higher NLR. However, the adjusted trim and fill analysis did not change the original conclusion. Lastly, the primary endpoint chosen for inclusion of studies was OS, therefore DFS and PFS data were drawn from studies that reported OS as an endpoint.
The advantages of our study were the relatively high amount of studies included, agreement of our results with the existing literature in other cancers, and significance using the random effects model. The effect of NLR on OS was also stable after performing subgroup analysis, sensitivity analysis, and "trim and fill" publication bias adjustment. The quality of studies that were included also showed low-moderate bias as assessed via QUIPS, with only few studies having a high amount of bias.
To conclude, the results of our meta-analysis suggest that an elevated pretreatment NLR is a negative prognostic factor in patients with HNC. The NLR value could have utility in stratifying patients and determining patient-specific treatment plans, particularly identifying high-risk patients that might benefit from adjuvant therapy. Our results should be interpreted with some degree of caution in view of the limitations described above. Therefore, further research with high-quality prospective studies is needed to fully validate the prognostic utility of NLR in HNC.       H  d  e  s  c  r  i  p  t  o  r  :  [  P  h  a  r  y  n  x  ]  e  x  p  l  o  d  e  a  l  l  t  r  e  e  s  #  2  0  M  e  S  H  d  e  s  c  r  i  p  t  o  r  :  [  O  r  o  p  h  a  r  y  n  x  ]  e  x  p  l  o  d  e  a  l  l  t  r  e  e  s  #  2  1  M  e  S  H  d  e  s  c  r  i  p  t  o  r  :  [  L  a  r  y  n  x  ]  e  x  p  l  o  d  e  a  l  l  t  r  e  e  s  #  2  2  M  e  S  H  d  e  s  c  r  i  p  t  o  r  :  [  N  a  s  o  p  h  a  r  y  n  x  ]  e  x  p  l  o  d  e  a  l  l  t  r  e  e  s  #  2  3  M  e  S  H  d  e  s  c  r  i  p  t  o  r  :  [  E  s  o  p  h  a  g  u  s  ]  e  x  p  l  o  d  e  a  l  l  t  r  e  e  s  #  2 T  a  b  l  e  4  .  S  t  u  d  y  -L  e  v  e  l  Q  u  a  l  i  t  y  A  s  s  e  s  s  m  e  n  t  U  s  i  n  g  t  h  e  Q  u  a  l  i  t  y  I  n  P  r  o  g  n  o  s  i  s  S  t  u  d  i  e  s  T  o  o  l  (  Q  U  I Table represents bar charts on the summary of risk of bias assessment with the Quality In Prognosis Studies tool (QUIPS). 19 The x-axis represents the percentage of studies graded to a specific risk of bias: low, moderate, or high risk of bias. The y-axis represents the 6 domains that were graded: study participation, study attrition, prognostic factor measurement, outcome assessment, confounding factors, and statistical analysis and reporting. A detailed scoring criterion for each domain is available in the Supplementary Materials.