Lack of Prognostic Value of Pretreatment Neutrophil-to-Lymphocyte Ratio in Early Breast Cancer

Abstract

Background

Breast cancer is a highly heterogeneous disease with large differences in the risk of recurrence. An elevated neutrophil-to-lymphocyte ratio (NLR) is correlated with a poor prognosis in a variety of tumors, and although it is still controversial in breast cancer, there are multiple studies, including meta-analysis, suggesting this. The purpose of this study was to analyze the prognostic value of preoperative NLR in an Argentine population of patients with nonmetastatic breast cancer, not exposed to neoadjuvant treatment.

Methods

Retrospective multicenter cohort study that includes patients over 18 years of age from three centers in the city and province of Buenos Aires who have had surgery for early breast cancer between January 1, 1999, and December 31, 2014. Based on the previous literature, a cutoff value of 2.0 was defined.

Results

A total of 791 patients were eligible for the analysis. Median age was 55 years (IQR 45–65). Median NLR was 1.92 (IQR 1.50–2.56). The distribution of groups according to the 8th edition of the AJCC was 54.1% for stage I, 35.6% stage II, and 10.4% stage III. Among the different tumor phenotypes, 79.0% were HR+/HER2−, 11.4% were HR+ or−/HER2+, and 9.2% were HR−/HER2−. With a median follow-up of 5.3 years, 112 patients (14.2%) had disease recurrence. Stage III patients had a higher NLR than stage I and stage II patients (p = 0.002). The rest of the clinical and pathological characteristics did not show differences in the groups according to NLR. There were no differences in relapse-free survival according to the NLR (p = 0.37), and it did not change after adjusting for other prognostic variables.

Conclusion

We consider it is important to determine the efficacy of prognostic markers that are easily accessible and of simple, systematic application. However, NLR does not appear to be an independent prognostic factor for recurrence in our population. In this sense, we consider it is important to publish negative results in order to avoid publication bias.

Introduction

Breast cancer is the most common tumor in women worldwide and it continues to be the leading cause of cancer death in developing countries [1]. On molecular level, it is a highly heterogeneous disease, and this is reflected in the varied clinical presentation and associated prognosis. The molecular subtype is currently considered one of the main determinants of the risk of recurrence [2]. Several multigene platforms have been incorporated into clinical practice to guide treatment in early stages, particularly in hormone-dependent disease. Although they have shown to be an important prognostic and predictive factor of response to chemotherapy in early disease [3], this is not always accessible to the entire population, especially due to a problem of access.

On the other hand, tumor-infiltrating lymphocytes and tumor-associated neutrophils (TANs) have been recognized as prognostic variables in patients with solid tumors. Attention has been paid to the “neutrophil-to-lymphocyte ratio” (NLR) as a parameter reflecting systemic inflammation and host immune response [4]. In breast cancer setting, the association between elevated NLR and worse course of disease has been a subject of controversy in several studies. The purpose of this study was to analyze the prognostic value of preoperative NLR in an Argentine population of patients with nonmetastatic breast cancer, not exposed to neoadjuvant treatment in terms of recurrence-free survival (RFS).

Materials and Methods

We carried out a multicenter retrospective cohort study in which three hospitals in the city and province of Buenos Aires, Argentina, participated (Austral University Hospital [HUA], Alexander Fleming Institute [IAF], and Center for Medical Education and Clinical Research [CEMIC]).

Patient Eligibility

Patients were eligible if they were 18 years old or older, had underwent surgery between January 1999 and December 2014, and had a confirmed diagnosis of invasive breast carcinoma in early stage of the disease. They also had to have a laboratory performed by an automated counter at the participating institutions within 30 days prior to surgery. Patients receiving neoadjuvant treatment were excluded, as well as patients with in situ carcinoma, synchronous bilateral breast carcinoma, and inflammatory breast carcinoma.

The primary outcome of the study was disease RFS. RFS was defined as a composite end-point including the following events: ipsilateral invasive breast tumor recurrence (invasive breast cancer involving the same breast parenchyma as the original primary), regional invasive breast cancer recurrence (invasive breast cancer in the axilla, regional lymph nodes, chest wall, and skin of the ipsilateral breast), distant recurrence (metastatic disease), and contralateral invasive breast cancer. RFS analysis time was censored at the last date the patient was known to be alive and recurrence free.

All patients included in the retrospective database signed informed consent forms allowing to share their data. The majority of the Argentinian population is white. The information was entered independently by each participating center, according to pathological and medical records. Patients were considered to have a positive family history if they described breast cancer events in first- and second-degree relatives. Tumors were classified as hormone receptor-positive (HR+)/HER2−, HR+ or −/HER2+, or HR−/HER2− according to hormone receptors and HER2 expression. If any of these variables were missing, the patient was excluded from phenotype classification. HR was considered positive if immunohistochemical analysis (IHC) was greater than 1% for estrogen and/or progesterone receptors. HER2 was considered positive if it presented overexpression (3+) by IHC. If IHC was 2+, tumors were analyzed by fluorescence in situ hybridization, and they were considered positive if HER2 gene copy was ≥6.0 or HER2/CEP17 ratio ≥2.0 and HER2 gene copy was ≥4.0. Neither patient with surgical procedures before 2005 nor patients with microinvasive breast carcinoma, defined as invasive carcinoma of the breast with no invasive focus measuring more than 1 mm, had routine HER2 assessment in our country. The NLR cutoff value used was the cutoff point in an Italian group cohort (equal to or greater than 2) [4].

Statistical Analysis

Categorical variables were expressed as absolute numbers and percentages. Continuous variables were described in terms of means and standard deviations if normally distributed or medians and interquartile ranges otherwise. Comparisons of continuous variables between clinical groups were performed using the Wilcoxon rank sum test or Student's t test, depending on sample distribution, and the χ² test was used to estimate the association between categorical variables. The NLR value was dichotomized, defining a cutoff value equal to or greater than 2. Survival distributions were estimated with the Kaplan-Meier method, and the significance of differences between survival rates was ascertained with the log-rank test. Cox proportional hazard models were used to assess associations between NLR and RFS. Hazard ratios (HRs) and 95% CIs were estimated from univariable and multivariable models. Median follow-up time was calculated using the reverse Kaplan-Meier method. For estimating the predicting ability of NLR, we used receiver operator characteristics (ROC) analysis. The area under the curve (AUC) was estimated for NLR and its 95% confidence interval. A p value <0.05 was considered statistically significant. Statistical analysis was performed with R (RStudio, Version 1.3.1093).

Results

Patient and Tumoral Characteristics

A total of 791 patients were included. Clinical and pathological characteristics of the population are shown in Table 1. NLR was determined in a blood sample obtained in a median of 9 days before surgery (IQR 6–15). Median NLR for the whole population was 1.92 (IQR 1.50–2.56). With the predefined cutoff value, 364 (46.0%) patients were considered to have a high NLR.

Table 1.

Demographic, tumor, and treatment characteristics

Characteristic	N (%)*
Institution
HUA	322 (40.7)
CEMIC	176 (22.3)
IAF	293 (37.0)
Age	55.31 (13.26)
Menopause status
Premenopausal	329 (41.6)
Postmenopausal	462 (58.4)
Tumoral size, cm+	1.60 [1.00–2.50]
Node-positive disease	114 (32.2)
Disease stage, %
I	427 (54.1)
II	281 (35.6)
III	82 (10.4)
Histological subtype
NST	661 (83.6)
Lobular	69 (8.7)
Mixed	30 (3.8)
Others	13 (1.6)
Unknown	18 (2.3)
Histological grade
1	94 (11.9)
2	422 (53.3)
3	229 (29.0)
Not specified	46 (5.8)
Ki67+	10 [5–25]
LVI	262 (35.3)
Phenotype
HR+/HER2−	625 (79.0)
ER%+	88 [70–95]
PR%+	70 [36.5–90]
HR+ or−/HER2+	90 (11.4)
HR−/HER2−	73 (9.2)
Unknown	5 (0.6)
Breast surgery
Mastectomy	220 (27.8)
BCS	571 (72.2)
Axillary procedure
SLNB	515 (65.1)
ALND	261 (34.9)
Adjuvant treatment
CT	400 (50.6)
Hormone therapy	654 (82.7)

^*Continuous variables are described as mean and SD or median and IQR depending on their distribution.

Association of NLR with Other Clinical and Pathological Variables

A difference in median NLR was observed according to the stage. Stage III patients had a higher NLR (median 2.22) than stage I and II patients (1.90 and 1.84, respectively, p = 0.002) (shown in Fig. 1). NLR was not associated with other clinical or pathological characteristics.

Fig. 1.

Box plot for NLR distribution by stages. Stage III patients had a higher NLR than stage I and II patients (median 2.22).

RFS by NLR Status

With a median follow-up of 5.2 years (IQR 3.2–6.7), 112 patients (14.2%) had disease recurrence, with a 5-year RFS of 87.3% (95% CI: 84.7–90.0). There was no difference in RFS at 5 years between the groups according to NLR, being 86.6% (95% CI: 83.0–90.4) for NLR <2 and 87.1% (95% CI: 83.2–91.2) for NLR ≥2 (p = 0.37) (shown in Fig. 2a). Sub-analysis of NLR prognostic role were carried out in the 3 main breast cancer subtypes, showing no statistical significant association (HR+/HER2−, p = 0.51, HR + or−/HER2+ p = 0.50 and HR−/HER2−, p = 0.78). Kaplan-Meier curve for triple negative breast cancer patients are shown in online supplementary Figure 1 (see www.karger.com/doi/10.1159/000525287 for all online suppl. material).

Fig. 2.

a RFS according to the NLR value. NLR variable was dichotomized as high (NLR ≥2) or low (NLR <2). No statistical association was observed. b RFS according to AJCC stages. Kaplan-Meier curve shows statistical differences, p < 0.0001. c RFS according to tumor phenotype.

Analysis of Other Variables Associated with RFS

Other clinicopathological variables already known to be prognostic factors for recurrence were assessed. A significant association was observed with the tumor stage (p < 0.0001) (shown in Fig. 2b). No statistical significance was observed with the following variables analyzed, axillary lymph node involvement (p = 0.586), LVI (p = 0.17), and tumor phenotype, although in this case a trend toward statistical significance was observed (p = 0.095) (shown in Fig. 2c).

In multivariate analysis, the observed results were maintained, particularly significant differences were observed in patients with stage 2 (HR = 1.76, 95% CI: 1.14–2.72, p = 0.01) and stage 3 (HR = 4.27, 95% CI: 2.52–7.26, p < 0.001) (shown in Fig. 3). NLR and phenotype did not show association with RFS when adjusted by stage.

Fig. 3.

Multivariate analysis. Worst relapse-free survival was observed in patients with stage 2 (HR = 1.76, 95% CI: 1.14–2.72, p = 0.01) and stage 3 (HR = 4.27, 95% CI: 2.52–7.26, p < 0.001), but NLR and phenotype did not show association with RFS when adjusted by stage.

ROC Analysis

Figure 4 shows the ROC curves for ability of NLR in predicting breast cancer relapse. The diagnostic ability of NLR as continuous measures for breast cancer relapse is estimated by area under ROC curve, indicating a very poor capability with an AUC of 0.55 (95% CI: 0.49–0.60). The best cutoff value for this population is 1.68 with a sensibility of 0.65 and specificity of 0.45.

Fig. 4.

ROC analysis for NLR. AUC and 95% CI are presented.

Discussion/Conclusion

Host cell-mediated immunity plays an important role in the elimination of any residual tumor cells and micro-metastases. NLR is an easy and accessible immunological parameter that has been widely studied as a marker of the host systemic inflammatory response during cancer development and progression. In this study, we assessed the correlation between preoperative NLR and prognosis in a cohort of early breast cancer patients treated in three Argentinehospitals, showing that NLR has no prognostic role in this population.

Early breast cancer is considered a curable disease in 70–80% of patients. The risk of recurrence varies between 10% and 40% in patients who complete locoregional treatment along with chemotherapy [5]. Classical parameters used to decide the use of adjuvant chemotherapy include lymph node involvement, histological grade, tumor size, lymphatic vascular invasion, expression of estrogen and progesterone receptors, overexpression of HER2 and Ki67 [6]. Molecular, multigene platforms such as Oncotype DX® and MammaPrint® are considered to be major determinants of the risk of recurrence and have been incorporated into clinical practice to guide treatment in the early stages and HR-positive disease [2, 7]. Although they have facilitated clinical decision-making, they are not easily accessible on a massive scale, which makes it necessary to use indicators that are not only accurate, standardized, and reproducible but also inexpensive and easy to perform. The NLR can be calculated by dividing the neutrophil count and the peripheral blood lymphocyte count. Therefore, the NLR is of interest to be properly analyzed.

The association between an elevated NLR and poor prognosis is complex and largely unclear, but with several possible explanations. Breast cancer subtypes greatly differ not only by ER, PR, or HER2 expression but also by tumor microenvironment (TME). TME is now recognized as a pivotal regulator of the immune response against cancer and is clearly influenced by the specific molecular subtype of breast cancer [8]. An elevated NLR may reflect a key role of systemic inflammation in the stimulation of angiogenesis, tumor growth, and the development of metastases. Associations between NLR and markers of systemic inflammation have been described in patients with operable cancer [9].

NLR may reflect the immune cell infiltrate in tumor stroma and inversely correlate with tumor-infiltrating lymphocytes (TILs) [10]. Cytotoxic T-lymphocytes are mainly responsible for executing the antitumor response by recognizing the target cell through its membrane receptor, and inducing its death by cytolysis mediated by enzymes and/or by activation of receptors of the TNF family such as TNFR, Fas, or TRAIL. A low lymphocyte count has been associated with worse outcomes in some patients [11]. Tumor infiltration by lymphocytes has been associated with the generation of an effective specific antitumor cellular immune response, and in fact, increased lymphocyte infiltration is correlated with a better prognosis [12]. We could speculate that tumors with higher numbers of TILs are associated with lower NLR, and this condition affects immune response and patients' prognoses.

While a different function of TILs across the different breast cancer subtypes has been demonstrated, very little is known about the role and activation of neutrophils and in general, neutrophilia has been described as a negative prognostic factor [13]. Neutrophils are emerging as major players in defining the fate of cancer development, promoting tumor growth and progression toward a metastatic disease, or favoring killing of tumor cells and cancer regression [14]. Neutrophils can be polarized by the TME toward an antitumor (N1) or a pro-tumor (N2) phenotype, depending on the media [15]. Some subtypes could have a pro-tumor effect by inhibiting the immune system by, for example, suppressing the cytolytic activity of lymphocytes, natural killer cells, and activated T cells.

Numerous studies have supported the value of NLR as a negative prognostic factor for patients with solid tumors [16]. Templeton et al. [16] conducted a systematic review to evaluate its prognostic value with overall survival (OS) as primary outcome. 100 studies with a total of 40,559 patients were included, of which only 1,195 (2.9%) were patients with breast cancer. Median cutoff for NLR was 4. Overall, an NLR greater than the cutoff value was associated with a hazard ratio for OS of 1.81 (95% CI = 1.67–1.97; p < 0.001). This effect was observed across all disease subgroups, sites, and stages. This analysis showed a clear evidence of publication bias with fewer small studies reporting negative results than would be expected [16].

In breast cancer setting, the published results are controversial. Studies are highly heterogeneous including patients in different stages of disease and treatment. Chen et al. [17] reported a small meta-analysis including a total of 4,293 patients in 8 studies; 4 conducted in Asian and 4 in Caucasian population. Patients with early and metastatic stage were included. Although the cutoff values applied in the studies were not consistent, a high level of NLR significantly affected OS and DFS in breast cancer in overall population [17]. In 2017, Liu et al. [18] published a meta-analysis with a larger sample that demonstrated that an elevated NLR is associated with worse DFS and OS. It included 18 studies with the majority of patients of Asian origin. The cutoff value used varies in the different studies, making it difficult to generalize the conclusions. Most of these studies have been carried out in Asia with little information on European or American populations.

Orditura et al. [4], an Italian group, conducted a study with 300 Caucasian patients to assess the correlation between presurgery NLR and distant metastasis-free survival in patients with early breast cancer. The NLR cutoff value used in this work was 1.97. The results showed that elevated NLR was associated with a worse prognosis. We have chosen this cutoff value. Our group published a work with a small cohort with 85 patients, in which the NLR was analyzed, observing the same prognostic value as this last study [19].

Although numerous studies have succeeded in demonstrating the negative prognostic role of NLR, we did not find statistical significance in our population that allows us to use NLR as an independent prognostic factor for recurrence. This was not even detected in triple negative cases where the role of inflammation and immune response is most important, although these analyses should be considered exploratory because of the multiple testing. ROC analysis shows a poor discriminant capability with AUC marginally higher than 0.5. Adjusting this type of analysis to best cutoff values derived from the same population tend to generate even more confusion in discovering the real impact of NLR in prognosis, as every publication uses the one that fits best for its own numbers.

Negative results, which are an important building block for science, are commonly banned from publication. Scientists choose not to proceed with their negative findings that yield less scientific interest and fewer impacts. Consequently, the amount of nonsignificant data reported is progressively declining. Data from a 2012 study of more than 4,000 published papers show that the scientific literature as a whole is trending toward more positivity [20]. The frequency at which papers testing a hypothesis returned a positive conclusion increased by more than 22% from 1990 to 2007, and by 2007, more than 85% of published studies claimed to have produced positive results. When negative results are not published in high-impact journals, other scientists cannot learn from them and end up repeating failed experiences or believing certain asseverations, leading to a delay in genuine progress. We need reviewers and publishers to commit to publishing negative results in their journals. Only by reshaping the scientific culture will negative results be revered for their true value. Negative findings are still a low priority for publication, so we need to find ways to make publishing them more attractive in order to prevent publication bias as we understand is happening in NLR prognostic role.

There are several caveats in our analysis. First, we observed a high percentage of recurrences in patients with luminal phenotypes compared to other subtypes, which is probably reflecting the population included in the study, given that patients with more aggressive tumor phenotypes probably received neoadjuvant chemotherapy and were excluded from our study. Second, retrospective analysis from databases may have errors and follow-up may be compromised.

Despite not having obtained positive results for our final objective, it seems to us of great importance to publish them in order to try to balance previous results regarding NLR. Most of the previously published studies include highly heterogeneous populations, and there is no consensus on the cutoff value which also makes it difficult to generalize the results. We believe more studies are needed before even thinking to use NLR systematically in clinical practice.

Statement of Ethics

This research complied with the guidelines for human studies and was ethically conducted in accordance with the World Medical Association Declaration of Helsinki. The inclusion of patients in the database was approved by the local institutional review boards of all participating institutions. All patients included in the retrospective database signed informed consent forms allowing sharing their data. This study protocol was reviewed and approved by Comité de Ética e Investigación of CEMIC, approval number 1330.

Conflict of Interest Statement

The authors declare that they have no competing interests.

Funding Sources

No funding was received for the present research project.

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: Maria del Rosario Sifon, Pablo Mandó, Manglio Rizzo, Florencia Perazzo, and Estrella Mariel Levy; data collection: Maria del Rosario Sifon, Pablo Mandó, Sergio Rivero, Nicolas Marcolini, Maria Julia Barber, Ignacio Mclean, Adrian Nervo, Maria Victoria Costanzo, and Gabriel Crimi; analysis and interpretation of results: Maria del Rosario Sifon, Pablo Mandó, Manglio Rizzo, Florencia Perazzo, and Estrella Mariel Levy; draft manuscript preparation: Maria del Rosario Sifon and Pablo Mandó. Maria del Rosario Sifón, Nicolas Marcolini, Maria Julia Barber, Ignacio Mclean, Manglio Rizzo, Sergio Riveros, Maria Victoria Costanzo, Adrian Nervo, Gabriel Crimi, Florencia Perazzo, Estrella Mariel Levy, and Pablo Mandó reviewed the results and approved the final version of the manuscript.

Data Availability Statement

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.

Supplementary Material

Supplementary data

Funding Statement

No funding was received for the present research project.