Neutrophil to Lymphocyte Ratio as Prognostic and Predictive Factor in Breast Cancer Patients: A Systematic Review.

Inflammatory blood markers (IBM), such as the neutrophil to lymphocyte ratio (NLR), have emerged as potential prognostic factors in various cancers, including breast cancer (BC), potentially allowing an easy, minimally invasive evaluation of a given cancer‘s prognosis and treatment outcome. We report here a systematic overview of the published data evaluating NLR as a prognostic factor or predictive factor for pathological complete response (PCR) and toxicity in early and advanced BC. A total of 45 articles were identified. NLR was found to be an independent prognostic factor for survival in most of the adjuvant treatment studies. However, no significant correlation was found between survival and NLR for early BC patients receiving neo-adjuvant chemotherapy (NACT) and advanced BC patients. Most studies failed to find a significant correlation between NLR and PCR after NACT. Finally, some data showed that IBM could be predictive of chemotherapy-related toxicity.


Introduction
Breast cancer (BC) prognosis depends not only on the tumor stage (localized versus metastatic disease), but also on the molecular subtype (luminal, HER2+, or triple-negative BC). Currently, BC management includes multidisciplinary and multimodal treatments: surgery, radiation therapy, chemotherapy, endocrine therapy and/or targeted therapies [1,2]. Although some predictive and/or prognostic factors are available (for example, hormone receptor status, HER2 overexpression/amplification, histological grade or stage), additional predictive and prognostic biomarkers are needed to better adapt the treatment to each individual patient.
In the last few years, inflammatory blood markers have emerged as predictive and prognostic factors, particularly the neutrophil to lymphocyte ratio (NLR); that is, the ratio between the absolute neutrophil count and the absolute lymphocyte count. Lymphopenia and a high NLR before chemotherapy initiation have been inconsistently associated with worse responses to neo-adjuvant chemotherapy and with poor prognoses of different cancer types, including BC [3,4].
The role of inflammation in cancer is now well established [5], and has been described at different stages of cancer development (initiation, promotion, invasion, and metastasis). Activated inflammatory cells are sources of reactive oxygen species and reactive nitrogen intermediates that can induce DNA damage and genome instability, thus promoting cancer initiation [6,7], or interfere with the DNA repair systems [8]. Inflammation increases the production of growth factors and cytokines that can confer a stem cell-like phenotype to tumor progenitors. Inflammation also promotes upregulation of angiogenic factors (known as angiogenetic switch) that favor tumor progression.
Neutrophils have been the focus of much research, and there is now evidence that they can promote tumor growth and play a role in metastasis development [9,10]. Their ability to secrete proteases, particularly matrix metalloproteases, contributes to favor a tumor's invasion. Neutrophils are also involved in tumor progression through their capacity to activate signal transducers and activators of transcription 3 (STAT3) and to promote neo-angiogenesis [11].
In recent years, the role of tumor infiltrating lymphocytes (TILs), especially in BC, has also been studied [12]. TILs are a selected population of T cells that show a high specific immunological reactivity against tumor cells. These lymphocytes, which are part of the innate immune system, can detect cancer cells and alert the immune system that will destroy them. Therefore, a low TIL count could be predictive of a lower response to neo-adjuvant chemotherapy [13], and might be associated with poor prognosis [14,15].
Peripheral inflammatory blood markers could be helpful for predicting patients' prognoses and also their response to neo-adjuvant chemotherapy in BC. Many studies evaluated the NLR, with conflicting results about its value as a predictive and/or prognostic factor [11,[16][17][18][19][20][21]. Studies on the total white blood cell count, neutrophil count, lymphocyte count and other blood cell ratios, such as the platelet to lymphocyte ratio (PLR), also gave inconsistent results.
Here, we systematically collected published data on the predictive or prognostic role of NLR in patients with BC.
We first summarized the published data on NLR and treatment efficacy, evaluated in terms of disease-free survival (DFS), breast cancer-specific survival (BCSS) and overall survival (OS), in patients with early BC (treated with neo-adjuvant or adjuvant chemotherapy) and with advanced BC. We also summarized the data on the predictive value of NLR for pathological complete response (PCR).
Then, we evaluated the correlation between chemotherapy-related toxicity and NLR and/or lymphopenia.

Search Strategy
We performed a systematic search of the PubMed database using the following search terms: "neutrophil to lymphocytes ratio" or "lymphopenia" AND "breast cancer". We also looked for articles using the search terms "toxicity" AND "neutrophil to lymphocyte ratio" AND "breast cancer" and "toxicity" AND "lymphopenia" AND "breast cancer". We updated the literature search in February 2019. We also screened the references of the selected articles found in PubMed in order to ensure exhaustivity. We followed the PRISMA guidelines throughout the process.

Study Selection
For the first part on NLR and treatment efficacy, we included all articles with only BC cohorts and PCR and/or survival analysis as the primary objectives. We excluded articles that concomitantly assessed different conditions (e.g., neo-adjuvant, adjuvant and metastatic settings). For the second part on toxicity, we included all articles reporting data on BC, inflammatory blood markers and chemotherapy-related toxicity. For both parts, we excluded all articles that were not in English ( Figure 1).

Data Extraction
Two reviewers independently extracted the following data from the selected studies: name of the first author and year of publication, population of interest, description of endpoints based on BC molecular subtypes, number of enrolled patients, ethnicity, treatment received, chosen NLR cut-off, primary objective and results (univariate analysis), results of subgroup analyses (if applicable), secondary objectives and their results, and multivariate analysis of NLR results (multivariate models and covariates used for adjustment). The two reviewers distinguished three populations: patients receiving neo-adjuvant chemotherapy, patients receiving adjuvant chemotherapy and patients with advanced BC.
For the studies on toxicity, the two reviewers selected articles on chemotherapy-related toxicity in the function of the lymphocyte count or NLR, and collected data on the population of interest, received treatments and toxicity.
If the selected articles also reported data on other inflammatory blood markers (e.g., lymphocyte count or PLR) as predictive and/or prognostic factors, we included this information in our tables.

Definitions
DFS was defined as the time from diagnosis (or the date of surgery for patients receiving adjuvant chemotherapy) to the date of relapse (local recurrence or metastases to distant sites) and/or death from any cause. In some papers, recurrence-free survival (RFS) was the primary objective. As RFS has the same definition as DFS, we used DFS for both DFS and RFS in this work.
BCSS was calculated from the date of diagnosis to the date of death by cancer or of the last follow-up visit. Some articles reported data on disease-specific survival (DSS), which has the same definition as BCSS. Therefore, we used the term BCSS for both BCSS and DSS in this work.
OS was defined as the time from the date of diagnosis (or the date of surgery) to the date of death due to any reason, or the date of the last follow-up (for some papers).
Progression-free survival (PFS) was defined as the time from treatment initiation to the date of disease progression, or death from any cause.

Data Extraction
Two reviewers independently extracted the following data from the selected studies: name of the first author and year of publication, population of interest, description of endpoints based on BC molecular subtypes, number of enrolled patients, ethnicity, treatment received, chosen NLR cut-off, primary objective and results (univariate analysis), results of subgroup analyses (if applicable), secondary objectives and their results, and multivariate analysis of NLR results (multivariate models and covariates used for adjustment). The two reviewers distinguished three populations: patients receiving neo-adjuvant chemotherapy, patients receiving adjuvant chemotherapy and patients with advanced BC.
For the studies on toxicity, the two reviewers selected articles on chemotherapy-related toxicity in the function of the lymphocyte count or NLR, and collected data on the population of interest, received treatments and toxicity.
If the selected articles also reported data on other inflammatory blood markers (e.g., lymphocyte count or PLR) as predictive and/or prognostic factors, we included this information in our tables.

Definitions
DFS was defined as the time from diagnosis (or the date of surgery for patients receiving adjuvant chemotherapy) to the date of relapse (local recurrence or metastases to distant sites) and/or death from any cause. In some papers, recurrence-free survival (RFS) was the primary objective. As RFS has the same definition as DFS, we used DFS for both DFS and RFS in this work.
BCSS was calculated from the date of diagnosis to the date of death by cancer or of the last follow-up visit. Some articles reported data on disease-specific survival (DSS), which has the same definition as BCSS. Therefore, we used the term BCSS for both BCSS and DSS in this work.
OS was defined as the time from the date of diagnosis (or the date of surgery) to the date of death due to any reason, or the date of the last follow-up (for some papers).
Progression-free survival (PFS) was defined as the time from treatment initiation to the date of disease progression, or death from any cause.

Results on NLR and PCR
Studies including All BC Molecular Subtypes: the PCR rate in the function of NLR was the primary objective in seven studies [22][23][24][25][27][28][29]. In all studies (except for two without definition), PCR was defined as the complete disappearance of invasive tumors in breast and lymph nodes (patients with residual ductal carcinoma in situ were also considered to have achieved PCR: ypT0/is pN0). Among these seven studies, only three (43%) found a significant correlation between NLR and PCR in the univariate analysis [23,25,28]. Only Qian et al. analyzed data using a multivariate model and found that NLR was not an independent prognostic factor for PCR (p = 0.254) [28].
Studies including Only Specific BC Molecular Subtypes: Chae et al. did not find any relation between NLR and PCR in the univariate and multivariate models (odds ratio (OR) = 4.274; 95%CI 1.451−12.658; p = 0.008) [31] in 87 patients with TNBC.
Subgroup Analyses: Suppan et al. [24] did not find any correlation between NLR and PCR in patients who received both anthracyclines and taxanes and in patients who received only anthracyclines or taxanes. They also compared the different BC molecular subtypes, and did not identify any correlation between NLR and PCR in any subtype.
Sub Group Analyses: Asano et al. [23] showed that in the subgroup of patients with TNBC (n = 61), NLR was correlated with DFS in the univariate analysis, but was not an independent prognostic factor of DFS.

Results on NLR and OS and BCSS
Studies including All BC Molecular Subtypes: four studies evaluated NLR and OS by a univariate analysis [23,26,28,29], and only one (25%) [26] found a significant association between NLR and OS, but NLR was not an independent prognostic factor for OS (p = 0.543). On the other hand, Chen et al. [25] showed that in patients with stage II and III BC (n = 215), NLR was an independent prognostic factor for BCSS (HR = 2.21; 95%CI 1.01-4.39; p < 0.05).
Sub Group Analyses: Koh et al. [30] also found that the correlation between NLR and OS was stronger in patients with stage III BC than in patients with stage II BC.

Conclusion on NLR as Prognostic and Predictive Factor in Patients with Early BC Receiving Neo-Adjuvant Chemotherapy
Six articles reported the results of the multivariate analyses of the data on NLR and PCR, DFS or OS in patients receiving neo-adjuvant chemotherapy [24][25][26]28,30,31]. NLR was correlated with PCR in one article [31] (50% of n = 2), with DFS in two studies (50% of n = 4) [25,30], with OS in two articles (50% of n = 4) [30,31] and with BCSS in one article (100%) [25]. These analyses included a total of 1558 patients. The conclusion should be taken with caution because two of these six selected studies included only one molecular subtype (TNBC or hormone receptor-positive/HER2 negative BC). Table 2 summarizes the results of these multivariate analyses and the adjustment factors.

Results on other Inflammatory Blood Markers
Among these studies, four [26][27][28][29] also evaluated the total lymphocyte count (n = 2) [26,28] or total neutrophil count before treatment (n = 2) [26,28], and other ratios, such as the lymphocyte to monocyte ratio (LMR) (n = 2) [26,29], neutrophil to monocyte ratio (NMR) (n = 2) [26,29] and PLR (n = 2) [27,29], as predictive and prognostic factors of survival. Lymphocyte count was not correlated with DFS and OS in the study by Marin Hernandez et al. [26], whereas it was an independent predictive factor for PCR in the multivariate analysis in the work by Qian et al. (OR = 4.37; 95%CI 1.43-13.39; p = 0.01) [28]. Neutrophil count was not correlated with PCR and OS in the studies by Qian et al. [28] and by Marin Hernandez et al. [26], respectively. Conversely, it was an independent prognostic factor for DFS (multivariate analysis) in the study by Qian et al. [28]. PLR was not correlated with PCR [27], DFS or OS [29] (univariate analyses). Similarly, NMR and LMR were not correlated with survival (DFS and OS) in the study by Marin Hernandez et al. (univariate and multivariate analyses) [26] and also in the study by Losada et al. (univariate analysis) [29].
Subgroup Analyses: many studies reported subgroup analyses based on the BC stage, molecular or pathological subtypes. Five studies tried to determine whether the results varied in the function of the molecular subtypes [32,38,39,44,46]. In two studies, NLR was correlated with DFS (and to DSS) only in patients with luminal cancer [44] and luminal A cancer [32]. Conversely, in two other studies, this correlation was observed only in patients with TNBC [38,39]. Another article reported that NRL was correlated with DFS in all molecular subtypes [46].
Moreover, in patients with lymph node invasion, NLR was correlated with DFS, but only in the univariate analysis [44].
Finally, two studies [45,46] showed that NLR was correlated with DFS in early stage BC (I and/or II according to the AJCC staging system), but not in stage III BC.
Subgroup Analyses: NLR was correlated with OS in the luminal A and TNBC subgroups in the study by Yao et al. [36], but only in the TNBC subgroup in the work by Jia et al. [39].
Two studies found that a derived NLR (dNLR, calculated as the ratio of neutrophils over white blood cells minus neutrophils) was correlated with DFS, OS and DSS in the univariate analyses [37,44], but not in the multivariate analyses [44].
LMR was correlated with DFS, OS and DSS (univariate analysis) in two [39,44] of the three studies that assessed this biomarker [39,43,44]. In the multivariate analyses, no correlation was found between LMR and survival.
Among the four studies on lymphocyte count and survival [33,44,47,52], only one (25%) found that lymphocyte count was correlated with DFS and DSS [44], but never with OS. No multivariate analysis was performed.
Finally, two studies [33,44] showed that neutrophil count, platelet count, and monocyte count were not correlated with survival. Cihan et al. [33] also reported no correlation between white blood cell count, eosinophil cell count, basophil cell count and survival (univariate analyses).
(2) Results on NLR and OS Studies including All Molecular Subtypes: among the three studies with data on NLR and OS [54,56,58], two (66%) [54] found that NLR was an independent prognostic factor for OS in patients with metastatic BC.
Studies including Only Specific BC Molecular Subtypes: NRL was correlated with OS in patients with HER2-positive BC, in the univariate and multivariate analyses (HR = 0.35; 95%CI 0.13-0.84; p = 0.018) [61], and also in patients with hormone receptor-positive in the univariate analysis (multivariate results not available) [59].
(3) Conclusion on NLR as Prognostic Factor in Patients with Advanced BC Six articles reported the results of the multivariate analyses on NLR and survival in patients with advanced BC [54,56,[58][59][60][61]. All found a significant correlation between NLR and survival. However, such results should be considered with caution because the multivariate analyses did not concern all types of survival outcomes and the studied populations were very heterogeneous. Table 6 summarizes the results of the multivariate analyses and the adjustment factors used.

(4) Results for Other Inflammatory Blood Markers
Three articles had data on other inflammatory blood markers as prognostic factors for survival in advanced BC [55,60,61]. In the univariate analysis, PFS was correlated with the absolute lymphocyte count, PLR and LMR [55,60]; however, one study found no correlation between PLR and PFS [61]. Vernieri et al. [60] found that in the multivariate analysis, PLR was an independent prognostic factor for PFS. One analysis showed that PLR was not a prognostic factor for OS [61].

Toxicity
We identified four articles that reported data on inflammatory blood markers and treatment-related toxicity [62][63][64][65] as the primary objective. Only one study analyzed the correlation between NLR and toxicity in a cohort of patients with BC (Yamanouchi et al. [65]). The three other studies focused on lymphopenia (defined as lymphocytes value < 0.7 G/L) in a cohort of patients with different cancer types (including BC). Two studies evaluated a French population [62,63], and the other two concerned Asian patients [64,65]. (Table 7).
Ray Coquard et al. [62] evaluated the predictive factors for early death after chemotherapy (defined as death within one month after the administration of cancer treatment) in a prospective study. They included 1051 patients among whom 756 (33%) had BC. They found that lymphopenia was a predictive factor of worse survival, in the univariate and multivariate models (OR = 3.1; 95%CI 1.8-5.8; p < 0.001). In a successive article, the same authors [63] determined whether a lymphocyte count <0.7 G/L at day 1 was predictive of febrile neutropenia in three cohorts of 950, 321 and 329 patients (including the cohort used in the previous work) among whom 24%, 33% and 42% had BC, respectively. They found that in the largest cohort, a lymphocyte count <0.7 G/L at day 1 was predictive of febrile neutropenia in the univariate and multivariate analyses (OR = 1.75; 95%CI 1.49-4.8; p = 0.02). Table 7. Articles with data on inflammatory blood markers as predictive factor of toxicity in patients with breast cancer. A Korean study [64] on 82 patients (including 11 (13%) patients with BC) did not find any correlation between lymphopenia at day 1 of the first chemotherapy course and risk of febrile neutropenia. However, due to the small number of patients with BC, no definitive conclusion could be drawn for this population.
Finally, Yamanouchi et al. [65] did not find any correlation between peripheral neuropathy occurrence and NLR, PLR or MLR in a cohort of 67 patients with BC who received at least four cycles of docetaxel (75 mg/m 2 ).

Conclusions
We report here a comprehensive and exhaustive overview of the published literature on NLR as a predictive and/or prognostic factor in patients with BC.
Results are less clear-cut for patients with a localized disease receiving neo-adjuvant chemotherapy because only two of four studies (50%) found a correlation between NLR and DFS [25,30], and one of two studies (50%) found a correlation between NLR and OS [30]. The only study that analyzed NLR and BCSS found a significant correlation [25]. Conclusions on NLR and PCR in this population cannot be drawn because only one of two studies (50%) found a correlation between these factors [31].
Some heterogeneity in the NLR results in the neo-adjuvant chemotherapy setting could be linked to the different results observed in the different molecular subtypes (positive correlation only in patients with TNBC) and also to the small patient samples. Indeed, the total number of patients enrolled in the selected studies in this setting was 1558, compared with 18,153 patients in the adjuvant chemotherapy setting.
Fewer data have been published on patients with metastatic BC. NLR appears to be a good prognostic factor for PFS and OS in this population, although studies are hard to compare due to the population heterogeneity [54,56,[58][59][60][61] (i.e., BC molecular features, treatment type, number of previous chemotherapy courses). Therefore, no firm conclusion can be made on inflammatory blood markers as prognostic factors in patients with metastatic BC.
Only very few studies focused on NLR or lymphopenia as predictive factors of chemotherapy-related toxicity [62][63][64][65]. Lymphopenia at day 1 of treatment was correlated with early death after chemotherapy [62]. However, no definitive conclusion could be made for patients with BC because the studied population included patients with various cancer types, and no specific subgroup analysis was performed in the BC subgroup. Lymphopenia at day 1 was inconsistently correlated with febrile neutropenia (two studies with opposite results [63,64]), and NLR was not correlated with peripheral neuropathy in patients receiving docetaxel [65]. Additional studies with larger populations should be performed to bring more data on NLR/lymphopenia as predictive factors of treatment toxicity.
Due to its worse prognosis and higher chemosensitivity, TNBC was the most extensively evaluated molecular subgroup. Indeed, the results of the subgroup analyses on TNBC were reported in 10 studies [23,24,32,36,38,39,46,53,56]. Most of them (80%) found a correlation between NLR and survival in this subgroup [23,36,38,39,46,53,56], a much higher proportion of positive results than for the other BC subgroups, possibly related to TNBC-specific clinical history.
Another limitation concerns the drugs used in the neo-adjuvant and adjuvant settings. Although most studies reported the use of anthracyclines and/or taxanes, only 47% of them (16/34) described the sequential or concomitant use of anthracycline-and taxane-based regimens, as recommended by the international learned societies. Data on chemotherapy regimens were missing in 53% of the studies (18/34).
Finally, all selected studies had a retrospective design. Although in most of them (>75%) at least one multivariate analysis was carried out, these results need to be validated in prospective studies or by retrospective evaluation in a prospective clinical trial.
In conclusion, NLR appears to be a prognostic factor for DFS and OS in patients with early BC receiving adjuvant chemotherapy. Additional investigations in prospective studies would strengthen these results. On the other hand, the correlation between NLR and survival in neo-adjuvant chemotherapy settings remains unclear because most studies failed to show evidence of an independent correlation. Finally, most of the available data show that NLR is not a predictive factor for PCR in patients treated with neo-adjuvant chemotherapy; however, due to the population heterogeneity and/or small sample size of the published studies on this question, dedicated clinical trials are needed. Funding: There has been no financial support for the conduct of the research and the preparation of the article.

Conflicts of Interest:
The authors have no conflict of interest.