Cancer patients commonly present with metabolic and pathophysiological changes, as well as an inflammatory environment that leads to malnutrition1,2. In addition, this state of malnutrition is associated with worsened clinical prognosis, immune dysfunction, higher infection susceptibility and increased clinical complications, such as negative treatment effects, longer length of hospitalisation, increased costs, and mortality3.

The host inflammatory response to tumours supports angiogenesis, damage to DNA, and tumour progression and invasion by the increased production of cytokines4,5. There exist different inflammatory markers among malignant tumours and the most common used by oncologists are C-reactive protein (CRP), Glasgow Prognostic Score (GPS) and platelet-to-lymphocyte ratio (PLR). Most recently, the neutrophil-to-lymphocyte ratio (NLR) has been studied5,6. Cancer patients exhibit alterations of the immune cells and as a result, lymphopenia has often been detected, due to damage of cell-mediated immunity, as well as, neutrophilia representing a weak inflammatory response5.

Previous evidence has shown that the NLR is predictive of nutritional status and/or sarcopenia3,7,8. The early and effective screening of nutritional risk is important to alleviate complications in these patients3. The nutritional risk screening 2002 (NRS 2002) is the tool commonly used in hospitalised patients9, and it defines patients’ nutritional risk, assisting as a strategy for nutritional support3,9.

Interestingly, both NLR and NRS 2002 are tools routinely used in the hospital that are of low cost and easily available. To our knowledge, no study has investigated whether systemic inflammatory response, as measured by NLR, is related to NRS 2002, which is an easy screening tool for nutritional risk. We hypothesised, in unselected cancer patients, a relationship between NLR and nutritional risk; specifically, a relationship between increased NLR values and malnutrition.

This study aims to determine, in hospitalised, unselected cancer patients, (i) whether there is an association between blood NLR values and nutritional risk, and (ii) the NLR cut-off value that best predicts malnutrition, as evaluated by the nutritional screening tool NRS 2002.

Materials and methods

Study design and patients

This investigation was a cross-sectional study carried out in a public hospital (Hospital das Clínicas da UFG, Goiânia, Goiás, Brazil) during the years 2018–2019, after approval by the Ethical Committee at the Federal University of Goias (nº 2.916.391/2018). All methods were performed in accordance with relevant guidelines and regulations and informed consent was obtained from all patients.

The patients who agreed to participate in the study signed consent forms. Inclusion criteria were patients of both sexes, ≥ 18 years, diagnosis of haematologic or solid tumour (gastrointestinal tract, accessory organs of digestion, pancreas, hepatic, prostate, gallbladder, or skin), and receiving chemotherapy (fluoropyrimidine drugs, topoisomerase inhibitors, platinum-based antineoplastic drugs, and/or anthracyclines) or undergoing surgical treatment. A total of 198 patients were enrolled and 79 were excluded due to incomplete data of the NRS 2002 questionnaire or biochemical assessments of neutrophils and lymphocytes. Thus, 119 patients were eligible for the study.

Sample size was calculated using the G*Power 3.1 software and a regression model; with an alpha-error 5% and beta-error of 95%, we found a sample size minimum of 74 patients.

Clinical and lifestyle data

Clinical variables tumor e treatment type was acquired by medical records. Lifestyle data, age, sex, current smoker, current alcohol, historic physical activity, and enteral nutrition or oral nutritional supplements were collected by trained nutritionists.

Anthropometric and nutrition risk screening

Body weight (kg) and height (m) were obtained using an electronic weighing scale and stadiometer (Filizola®). Body mass index (BMI, kg/m2) was calculated accordingly. Arm (mm) and calf (cm) circumferences were assessed using a flexible tape measure10,11,12.

Nutrition risk screening (NRS 2002) was applied within 24 h of hospitalization to assess nutritional risk. The final score of ≥ 3 was classified as nutritional risk9,13. The anthropometric variables were collected by the researchers, who were nutritionists and nutrition students, and the NRS tool were applied by the hospital’s own nutritionists.

Performance status

The Performance status was used from the Karnofsky Performance Status scores, where to range of 0 to 100. The higher score the more apt the patient is to perform their daily activities 14.

Systemic inflammation evaluation

Blood collection was performed by a trained professional. After collecting venous blood, it was centrifuged at 4000 rpm for 10 min at 4 °C (Hitachi® CF16RN, Ibaraki, Japan) and stored in a freezer at − 80 °C for future analysis. The concentrations of platelets, lymphocytes and neutrophils in the blood were analysed using an automated system. CRP levels were quantified by immunoturbidimetric assay.

Statistical analyses

The Kolmogorov–Smirnov test was used to verify the normality data. Differences between groups were applied using the Student t-test for parametric variables, Mann–Whitney U test for non-parametric variables, and Chi-Square for categorical variables.

The best cut-off point for NLR in determining nutritional risk was estimated by calculating the sensitivity, specificity, positive and negative predictive values, and area under the receiver operating characteristic (ROC) curve. From the ROC curve, we evaluated the predictive power of the study specific cut-off points, and we found that the prediction increased with the Roc curve. The NRS 2002 was dichotomised at its cut-off points as obtained by the ROC curve and, the analysis of the main component identified the cut-off level of the NLR.

Logistic regression analyses were performed to assess the association between NLR (variable-dependent) and nutritional risk (NRS 2002 values). We used model 1 without adjustment and model 2 adjusted by sex, age, physical activity, alcohol intake, smoking status, BMI, cancer type, treatment type and performance status.

Medcalc® software (version 11.1.1.0, Belgium) was used for all analyses and p < 0.05 was considered as statistically significant.

Results

Utilising the analysis of the area under the curve (AUC), we found that the best cut-off point to identify the nutritional risk of these patients was NLR > 5.0, in an AUC of 0.72 ± 0.04, with sensitivity of 60.9% (47.9–72.9%) and specificity of 76.4% (63.0–86.8%) (Fig. 1).

Figure 1
figure 1

Receiver operating characteristic (ROC) curve, confidence interval, cut-off point, sensitivity, specificity, predictive value for the best cut-off point of the NLR value to discriminate nutrition risk (NRS, 2002) in hospitalized cancer patients. NLR: neutrophil–lymphocyte rate.

Full size image

Of the 119 included patients, we did not observe differences between the groups for age, sex, smoking, alcohol intake, history of physical activity, nutritional supplementation, treatment type, or performance status (Table 1). However, regarding the tumour sites, we observed that gastrointestinal and accessory organs of digestion tumours were more prevalent for both groups. In addition, the NLR group ≥ 5.0 had a higher prevalence for nutritional risk than the NLR < 5.0 group (NLR ≥ 5.0: 73.6% vs. NLR < 5.0: 37.9%, p = 0.001) (Table 1).

Table 1 Socio-demographic and clinical characteristics of the patients according to neutrophils-lymphocytes ratio.
Full size table

The groups are similar for height and arm circumference. Nevertheless, the NLR ≥ 5.0 group had lower body weight, BMI, and calf circumference than the NLR < 5.0 group (Table 2). In addition, blood CRP and NLR values were higher in the NLR ≥ 5.0 group than the NLR < 5.0 group (Table 2). The NLR ≥ 5.0 group also showed the highest nutritional risk, according to the NRS 2002, when compared to the NLR < 5.0 group (NLR ≥ 5.0: 3.0 ± 1.1 vs. NLR < 5.0: 2.3 ± 1.2, p = 0.0004) (Fig. 2).

Table 2 Anthropometric and biochemical characteristics of the patients according to neutrophils-lymphocytes ratio.
Full size table
Figure 2
figure 2

Nutrition risk screening (NRS, 2002) values according to blood neutrophil-lymphocytes ratio (NLR).

Full size image

The logistic regression analyses revealed an association between malnutrition risk (NRS 2002) and NLR values in a model without adjustment and adjusted for age, sex, physical activity, alcohol intake, smoking, BMI, tumour and treatment type, and performance status (Table 3). Therefore, high NLR values indicate an increase in nutritional risk of 1.5–1.7 times.

Table 3 Logistic regression between the nutrition risk screening (NRS, 2002) and neutrophil-lymphocytes ratio (NLR) in hospitalised cancer patients.
Full size table

Discussion

Nutritional risk was found in 73.6% of patients with NLR ≥ 5.0, and in 37.9% of those in the NLR < 5.0 group. In addition, higher NLR values were associated with nutritional risk, independent of confounding variables.

In clinical practice, our results showed the importance of establishing the best cut-off point value for blood NLR in hospitalised cancer patients. We found that the value with a higher sensitivity and specificity for NLR in predicting the risk of malnutrition is ≥ 5.0. A recent study performed by our group (Borges et al. 2020) identified a cut-off point of NLR ≥ 6.5 as predictive for the risk of sarcopenia in cancer patients7.

In a meta-analysis, it was demonstrated that NLR has the ability to assess survival in cancer patients; for example, classifications were defined as normal (NLR > 3), moderate (NLR 3–5) and high (NLR > 5). Those patients with higher values of NLR had lower survival probability15. In another prospective study performed in preoperative esophageal patients, NLR, PLR and NRS were found to be promising as prognostic factors of progression-free survival16.

In the present study, we highlighted that establishing the best NLR value is of significance in oncology since it can facilitate nutritional counselling. In the literature, there exists a divergence in relation to the reference values. In this context, in our study, values of NLR ≥ 5.0 may be justified since this parametre was obtained from hospitalised cancer patients that suffer from increased systemic inflammation, in comparison to outpatients7.

During hospitalisation, cancer patients are susceptible to nutritional risk due to several factors. Among them are reduced appetite, mechanical difficulties in chewing and swallowing food, adverse effects during antitumour therapy, and oncology treatment3,17. Cancer also causes immunometabolic changes that induce increased systemic inflammatory response and energy expenditure. As a result, these patients probably present with nutritional risk that may increase exposure to malnutrition outcomes1,2,18.

Currently, many evaluation methods to screening the nutritional status exist; one is the NRS 2002, a valid method for identifying hospitalised patients at risk and those who will benefit from nutritional treatment16. A cohort study adopted the NRS 2002 as a tool for assessing nutritional risk and found this tool can accurately predict the association between nutritional risk and clinical outcome19. In this scenario, the present study identified a high prevalence of nutritional risk, lower body weight, BMI, CRP, and calf circumference. It also found that high values of NLR, used as a systemic inflammatory marker, are associated with increased nutritional risk in hospitalised cancer patients.

The strengths of this study are the use of practical and easily accessible tools, such as the NRS 2002 and biochemical markers of lymphocyte and neutrophil count. In addition, the NLR cut-off point (5.0) may be used to distinguish nutritional risk during the nutritional and physician consultations.

However, limitations include that our study include a sample with underlying disease of heterogeneous tumour types, since patients had both haematological and solid tumours. In addition, the use of various antitumour drugs was not evaluated during the study and blood NLR values may have been affected during data collection. Another limitation was that we were unable to determine the tumour stage of patients, since some had haematologic tumours. Lastly, a cross-sectional designed study did not allow establishment of a relationship between cause and effect, and further studies are warranted to confirm our findings.

Conclusions

In hospitalised, unselected cancer patients, systemic inflammation, when measured by blood NLR, was associated with nutritional risk. Additionally, our results also exhibited the importance of establishing a cut-off point for NRL during use in clinical practice. According to our findings, the cut-off point for NRL in predicting nutritional risk is ≥ 5.0 and more studies may use this value to predict clinical outcomes during cancer patient hospitalisation.