Exploring the prognostic value of the neutrophil-to-lymphocyte ratio in cancer

Abstract

In cancer patients, a high pre-treatment neutrophil-to-lymphocyte ratio (NLR) is associated with poorer survival outcomes. Significant variation in the magnitude of this association has been observed between studies, but sources of this variation are poorly understood. Here, we explore differences in the prognostic potential of NLR between patient subgroups stratified by demographic and clinical characteristics using a retrospective cohort of 5,363 patients treated at Moffitt Cancer Center (Tampa, FL). We identify patients for whom NLR has maximum prognostic potential via adjusted hazard ratios (HRs) calculated using multivariable Cox proportional hazards models and area under the curve analysis. NLR demonstrates stronger associations (HRs > 2) with survival among African-American patients, patients receiving radiation therapy, stage IV patients, and melanoma patients when compared with the overall study population (HR = 1.58). Sensitivity and specificity of NLR as a prognostic marker are also higher in these patient subgroups, and increase further with combinations of multiple "high-risk" demographic or clinical characteristics. In summary, NLR may have greater prognostic value in patients with certain demographic and clinical features. Future prospective studies could validate this hypothesis, after further characterization of populations in which NLR has maximum prognostic potential and the identification of meaningful thresholds for risk stratification.

Figure 1

Variation in pre-treatment NLR. Box plots demonstrating baseline NLR in patients according to age, race, disease stage, cancer type and sex. Each point on the scatter plot represents an individual patient within the subgroup specified. The overlaid box plot presents the median and interquartile range of NLR for those patients.

Figure 2

Survival Analysis. Kaplan Meier survival curves demonstrating survival probability with increasing time post diagnosis in patients with above-median NLR (blue) as compared to below-median NLR (red) in the cohort as a whole (A) and for each studied cancer type (B–I). The shaded areas reflect 95% confidence intervals. The hatched lines show median survival time for patients above or below the median NLR. Group-specific medians for determining high NLR cutoff are calculated within each cancer subtype.

Figure 3

Prognostic potential of NLR. ROC curves demonstrating the diagnostic potential (5-year overall survival) of NLR as a binary classifier. Each colored curve represents a different patient subgroup. The AUC is increased in patients with stage IV disease (A) and melanoma (B) as compared to those with lower stage disease and/or other cancer types. In the example, the accuracy of NLR as a prognostic marker is increased significantly in patients exhibiting multiple “high risk” characteristics.