Main

Recent studies have demonstrated that the systemic inflammatory response (SIR) (McMillan et al, 2003) is associated with postoperative survival in patients with several types of cancer (Forrest et al, 2004; Ishizuka et al, 2007; Ramsey et al, 2007). Although there are a number of inflammation-based prognostic systems and clinical measures based on SIR, such as the Glasgow Prognostic Score (GPS) (Forrest et al, 2004; Ramsey et al, 2007; Ishizuka et al, 2012), neutrophil to lymphocyte ratio (NLR) (Bhatti et al, 2010; Kim et al, 2010; Shimada et al, 2010; Chua et al, 2011), and reactive thrombocytosis (Silvis et al, 1970; Monreal et al, 1998; Shimada et al, 2004), their systems of evaluation differ. Although the GPS is based on estimation of two types of protein – the serum levels of C-reactive protein (CRP) and albumin, which are regulated by inflammatory cytokines – the NLR and thrombocytosis are based on cellular components that are also regulated by such cytokines, especially interleukin-6 (IL-6) (Ohsugi, 2007).

Among the inflammatory cytokines, IL-6 is known to be a multifunctional cytokine that acts on a variety of cells, including immune-competent cells and hematopoietic cells, to trigger proliferation and differentiation. From a clinical viewpoint, it is also important to consider that IL-6 stimulates hepatocytes to induce acute-phase proteins, including CRP, and decrease the serum albumin level (Ramadori et al, 1998; Ohsugi, 2007). This phenomenon is merely part of the mechanism reflected in the GPS. Similarly, IL-6 elicits not only neutrophil proliferation but also differentiation of megakaryocytes to platelets (Imai et al, 1991; Ruscetti, 1994), and these phenomena are also related to the mechanisms underlying SIR.

There is an increasing evidence that the GPS, reactive thrombocytosis, and the NLR can be used for prognostication in patients with several types of cancer (Shimada et al, 2004, 2010). Because various combinations of acute-phase proteins can be employed for this purpose, it would be expected that estimation of reactive thrombocytosis (Shimada et al, 2004) and NLR (Shimada et al, 2010) would also be potentially applicable for prognostication.

In the present study, therefore, we evaluated the prognostic utility of a novel inflammation-based prognostic system, named the COP-NLR (COmbination of Platelet count and Neutrophil to Lymphocyte Ratio), in patients undergoing surgery for colorectal cancer (CRC).

Materials and methods

We conducted a retrospective review of a database comprising 490 patients who had undergone elective surgery for CRC. All procedures had been performed by the same surgical team at the Department of Gastroenterological Surgery, Dokkyo Medical University Hospital, between January 2000 and August 2009. Among these patients, 480 were enrolled in this study on the basis of specific criteria. Patients for whom insufficient laboratory data were available in the record, or those who had idiopathic thrombocytopenic purpura (n=2), were excluded.

Routine laboratory measurements including the serum levels of CRP, albumin, and tumour markers such as carcinoembryonic antigen (CEA; upper physiological value 5 ng ml−1) and carbohydrate antigen 19-9 (upper physiological value 37 U ml−1) were carried out on the day of admission in order to exclude any effects attributable to inflammation associated with sequential preoperative examinations. None of the patients had clinical evidence of infection or other inflammatory conditions, and none had received preoperative chemotherapy or irradiation.

The recommended cut-off values of the preoperative NLR and platelet count were decided using receiver operating characteristic (ROC) curve analyses. The recommended cut-off value of the NLR was based on the most prominent point on the ROC curve for sensitivity (0.492) and specificity (0.742), respectively. Because these two parameters indicated a cut-off value of 2.902, the recommended NLR cut-off value was defined as 3.0. The area under the ROC curve was 0.619. Similarly, the recommended cut-off value of the preoperative platelet count was defined as 30 × 104 mm−3, because the most prominent point on the ROC curve indicated a cut-off value of 28.85 × 104 mm−3 for sensitivity (0.413) and specificity (0.742). The area under the ROC curve was 0.589.

The COP-NLR was calculated on the basis of data obtained on the day of admission as follows: patients with both an elevated platelet count (>30 × 104 mm−3) and an elevated NLR (>3) were allocated a score of 2, and patients showing one or neither were allocated a score of 1 or 0, respectively.

Univariate analysis was performed to evaluate clinical characteristics including age (year), sex (male/female), number of tumours (⩾2/1), maximum tumour diameter (>40/⩽40 mm), tumour location (rectum/colon), tumour type (3, 4, and 5/0, 1, and 2), pathology (others/tub1, 2), lymphatic invasion (presence/absence), venous invasion (presence/absence), lymph node metastasis (presence/absence), white blood cell (WBC) count ( × 103 mm−3), platelet count ( × 104 mm−3), neutrophil ratio (%), lymphocyte ratio (%), NLR, the serum levels of CRP (mg dl−1), albumin (g dl−1), CEA (ng ml−1), and CA19-9 (U ml−1), body mass index (BMI) (kg m−2), the GPS (2/0, 1), and the COP-NLR (1, 2/0) relative to cancer-specific survival.

Multivariate analysis was then performed using clinical characteristics with a P-value <0.05 selected in the univariate analysis to assess those that were predictive of cancer-specific survival.

Kaplan–Meier analysis and log-rank test were used to compare the survival curves of the three groups divided according to the COP-NLR scores of 0, 1, and 2.

Estimation of the GPS

The GPS was estimated as described previously. Briefly, patients with both an elevated CRP level (>1.0 mg dl−1) and hypoalbuminemia (<3.5 g dl−1) were allocated a score of 2. Patients in whom only one of these biochemical abnormalities was present were allocated a score of 1, and those in whom neither of these abnormalities was present were allocated a score of 0 (Forrest et al, 2004).

Definition of operative curability

Based on the Japanese Classification of Colorectal Carcinoma (Japanese Society for Cancer of the Colon and Rectum, Second English Edition), residual tumour is diagnosed as follows: R0, no residual tumour; R1, no residual tumour, but tumour suspected at the resection margin; and R2, macroscopically evident residual tumour. Based on this definition, operative curability is defined as follows: curability A, R0 in TNM stage I, II, or III; curability B, R0 in TNM stage IV or R1 in any TNM stage; curability C, R2 in any TNM stage.

Definition of macroscopic tumour types and histological findings

Similarly, macroscopic tumour types are classified as follows: type 0, superficial; type 1, polypoid; type 2, ulcerated with clear margin; type 3, ulcerated with infiltration; type 4, diffusely infiltrating; and type 5, unclassified. On the basis of these definitions, we classified the patients into two groups: those with noninvasive tumours (types 0, 1, and 2) and those with invasive tumours (types 3, 4, and 5).

The pathological types of tumours are defined as follows: tub1, well-differentiated adenocarcinoma; tub2, moderately differentiated adenocarcinoma; por, poorly differentiated adenocarcinoma; muc, mucinous adenocarcinoma; and sig, signet ring cell carcinoma. On the basis of these definitions, the patients were divided into two groups: those with differentiated adenocarcinoma (tub1 and tub2) and those with other types of carcinoma (por, muc, and sig).

Invasion of vessels – that is, lymphatic invasion (ly) and venous invasion (v) – was diagnosed as follows: ly0 (v0), no invasion; ly1 (v1), minimal invasion; ly2 (v2), moderate invasion; and ly3 (v3), severe invasion. On the basis of these definitions, the patients were divided into two groups: those without (ly0, v0) and those with (ly1–3, v1–3) vessel invasion.

Administration of chemotherapy

Most of the stage IV patients undergoing surgery were considered for postoperative chemotherapy. Because, in our department, recently developed chemotherapy regimens such as FOLFIRI (FOLinic acid (leucovorin)/Fluorouracil (5-FU)/IRInotecan (Camptosar)) (Tournigand et al, 2004) and FOLFOX (FOLinic acid (leucovorin)/Fluorouracil (5-FU)/OXaliplatin (Eloxatin)) (de Gramont et al, 2000) were introduced in January 2005, patients who had undergone surgery before January 2005 had been administered oral anticancer drugs based on 5-FU as postoperative chemotherapy (Hernandez et al, 1992).

Similarly, most of the stage III patients undergoing surgery were considered for postoperative chemotherapy. Because, in our department, recently developed chemotherapy regimens such as capecitabine (Twelves et al, 2005) were introduced in October 2008, patients who had undergone surgery before October 2008 were administered the same drugs as those for stage IV patients (Kaser et al, 2001). In addition, most of the stage II patients undergoing surgery were not considered for postoperative chemotherapy.

Statistical analysis

Data are presented as mean±s.d. Differences among the three groups were analysed using the χ2- test and the Kruskal–Wallis test. Odds ratios (OR) with 95% confidence interval (CI) were calculated by uni- or multivariate Cox proportional hazards model analyses.

Kaplan–Meier analysis and log-rank test were used to compare survival curves. Deaths before 31 March 2009 were included in this analysis.

Statistical analyses were performed using the SPSS statistical software package, version 16.0 (SPSS Inc., Chicago, IL, USA) at a significance level of P<0.05.

Results

A total of 480 patients were enrolled (male : female=309 (64.4%) : 171 (35.6%)). There were 356 (74.2%) patients with a platelet count of ⩽30 × 104 mm−3 and 124 (25.8%) patients with a platelet count of >30 × 104 mm−3. Among these patients, 338 (70.4%) patients had a NLR of ⩽3 and 142 (29.6%) patients had a NLR of >3.

Table 1 shows the distribution of the clinical background characteristics of the studied patients in the three groups divided according to the COP-NLR. There were no significant differences among the groups, except for maximum tumour diameter (⩽40/>40, mm) (P<0.001), tumour type (0, 1, 2/3, 4, and 5) (P=0.016), lymphatic invasion (absence/presence) (P<0.001), venous invasion (absence/presence) (P=0.004), operative curability (A/B/C) (P<0.001), GPS (0/1/2) (P<0.001), and TNM stage (0/I/II/III/IV) (P<0.001) (χ2-test).

Table 1 Relationships between clinical characteristics and the COP-NLR in patients with CRC
Full size table

Table 2 shows the relationships between clinicolaboratory characteristics and the three groups of patients. There were no significant inter-group differences, except for maximum tumour diameter (mm; P<0.001), WBC count (P<0.001), platelet count (P<0.001), neutrophil ratio (P<0.001), lymphocyte ratio (P<0.001), NLR (P<0.001), the serum levels of CRP (mg dl−1; P<0.001), albumin (g dl−1; P<0.001) and CEA (ng ml−1; P<0.001), BMI (kg m−2; P=0.002), and survival period (P=0.001) (Kruskal–Wallis test).

Table 2 Relationships between clinicolaboratory characteristics and the COP-NLR in patients with CRC
Full size table

During the observation period, 150 (31.3%) patients died, among whom 30 (20.0%) died of intercurrent disease. Univariate and multivariate analyses were performed to evaluate the relationship between clinical characteristics and cancer-specific survival.

The results of univariate analyses demonstrated that maximum tumour diameter (⩽40/>40 mm) (OR, 0.517; 95% CI, 0.349–0.767; P=0.001), pathology (others/tub1, 2) (OR, 0.338; 95% CI, 0.205–0.558; P<0.001), lymphatic invasion (presence/absence) (OR, 0.167; 95% CI, 0.068–0.409; P<0.001), venous invasion (presence/absence) (OR, 0.376; 95% CI, 0.207–0.683; P=0.001), lymph node metastasis (presence/absence) (OR, 0.299; 95% CI, 0.201–0.444; P<0.001), WBC count ( × 103 mm−3) (OR, 1.108; 95% CI, 1.060–1.159; P<0.001), platelet count ( × 104 mm−3) (OR, 1.027; 95% CI, 1.011–1.043; P=0.001), neutrophil ratio (%) (OR, 1.041; 95% CI, 1.024–1.059; P<0.001), lymphocyte ratio (%) (OR, 0.951; 95% CI, 0.933–0.970; P<0.001), NLR (OR, 1.126; 95% CI, 1.074–1.179; P<0.001), the serum levels of CRP (mg dl−1) (OR, 1.224; 95% CI, 1.159–1.292; P<0.001), albumin (g dl−1) (OR, 0.530; 95% CI, 0.394–0.714; P<0.001), CEA (ng ml−1) (OR, 1.001; 95% CI, 1.001–1.002; P<0.001), and CA19-9 (U ml−1) (OR, 1.000; 95% CI, 1.000–1.000; P<0.001), the GPS (2/0, 1) (OR, 0.321; 95% CI, 0.208–0.497; P<0.001), and the COP-NLR (1, 2/0) (OR, 0.349; 95% CI, 0.240–0.507; P<0.001) were associated with cancer-specific survival (Table 3).

Table 3 Univariate analyses in relation to cancer-specific survival
Full size table

Multivariate analysis was performed using the characteristics shown to have statistical significance (P<0.05) by univariate analysis. This indicated that the COP-NLR (1, 2/0) was associated with cancer-specific survival (OR, 0.464; 95% CI, 0.267–0.807; P=0.007), along with pathology (others/tub1, 2) (OR, 0.377; 95% CI, 0.217–0.655; P<0.001), lymph node metastasis (presence/absence) (OR, 0.377; 95% CI, 0.241–0.591; P<0.001), the serum levels of CRP (mg dl−1) (OR, 1.189; 95% CI, 1.081–1.308; P<0.001), albumin (g dl−1) (OR, 0.547; 95% CI, 0.376–0.794; P=0.002), and CEA (OR, 1.001; 95% CI, 1.001–1.001; P<0.001), and the GPS (2/0, 1) (OR, 2.604; 95% CI, 1.242–5.456; P=0.011) (Table 4).

Table 4 Multivariate analyses in relation to cancer-specific survival
Full size table

The median and maximum follow-up periods for survivors were 1169 and 3936 days, respectively, and the mean follow-up period was 1326±943 days (mean±s.d.). There were significant differences in the periods after surgery among the COP-NLR 0 (1414±841 days), 1 (1253±1058 days) and 2 (1104±1031 days) groups (mean±s.d., P=0.001, Kruskal–Wallis test).

Kaplan–Meier analysis and log-rank test demonstrated that there were significant differences in cancer-specific survival among the three groups (P<0.001) (Figure 1). Thus, the COP-NLR was able to clearly classify such patients into three independent groups.

Figure 1
figure 1

Relationship between the COP-NLR (COP-NLR 0, 1, and 2 from top to bottom) and cancer-specific survival in patients with CRC undergoing surgery.

Full size image

Discussion

Because the COP-NLR consists of two SIR-related characteristics – NLR and platelet count – it is reasonable that a higher proportion of patients with COP-NLR 2 or 1 would have large tumours, invasive-type tumours, presence of lymphatic invasion, presence of venous invasion, operative curability C, GPS 2, and TNM stage IV than patients with COP-NLR 0. Similarly, the COP-NLR showed close relationships with not only tumour-related characteristics, such as maximum tumour diameter and the serum level of CEA, but also SIR-related characteristics such as WBC count, platelet count, neutrophil ratio, lymphocyte ratio, NLR, the serum levels of CRP and albumin, and BMI. The latter characteristics are related to nutritional status, because an elevated CRP level, hypoalbuminemia, and low BMI reflect cachexia due to hypercytokinemia resulting from tumour progression. These characteristics were retained after uni- and multivariate analyses. Univariate analysis also selected tumour-related characteristics including maximum tumour diameter, pathology, lymphatic invasion, venous invasion, lymph node metastasis, CEA, and CA19-9, and SIR-related characteristics including WBC count, platelet count, neutrophil ratio, lymphocyte ratio, NLR, the serum levels of CRP and albumin, the GPS, and the COP-NLR. Among these clinical characteristics, multivariate analysis disclosed that the COP-NLR was associated with cancer-specific survival, along with not only tumour-related characteristics, such as pathology, lymph node metastasis, and CEA, but also SIR-related characteristics such as CRP, albumin, and the GPS. Therefore, the COP-NLR appears to have potential utility for predicting the postoperative survival of patients with CRC, as well as the GPS.

Recent studies have demonstrated that reactive thrombocytosis is associated with survival after surgery for several types of cancer (Silvis et al, 1970; Monreal et al, 1998; Takahashi et al, 1998; Browder et al, 2000; Shimada et al, 2010). There are two explanations for this phenomenon.

First, reactive thrombocytosis is induced in a background of hypercytokinemia through tumour vs host interaction. Among several inflammatory cytokines, IL-6 has an important role in reactive thrombocytosis (Sierko and Wojtukiewicz, 2004), as it is a multifunctional cytokine with a number of physiological actions, stimulating not only CRP upregulation but also albumin downregulation in the liver, as well as protein synthesis (Ramadori et al, 1998). Similarly, IL-6 has a cell-proliferative effect, triggering the differentiation of megakaryocytes to platelets in the bone marrow (Imai et al, 1991; Ruscetti, 1994). Therefore, it is reasonable that reactive thrombocytosis would be associated with survival after surgery, because this phenomenon is based on the same mechanism as that reflected in the GPS, which is determined from the serum levels of CRP and albumin.

Second, thrombocytosis is also induced from the tumour itself (Mohle et al, 1997; Kohrana and Fine, 2004). Generally, thrombocytosis is a feature in 10–57% of patients with malignancy, as a variety of neoplastic cells can stimulate platelet activation (George et al, 2000; Gunsilius et al, 2000). Several studies have revealed that cancer cells secrete vascular endothelial growth factor (VEGF), which also stimulates megakaryocyte differentiation (Troxler et al, 2007). Because VEGF induction promotes tumour growth (Mohle et al, 1997; Khorana and Fine, 2004; Gorelick et al, 2009), thrombocytosis indirectly reflects tumour progression. In fact, a high level of VEGF is found in serum, platelets, and leukocytes of patients with malignant disease (Lavie et al, 1999), and platelet interactions with malignant cells promote metastasis (Gislason and Nou, 1985).

With regard to the platelet count, most previous studies have used a cut-off value of 30–40 × 104 mm−3 (Symbas et al, 2000; Ishizuka et al, 2012). Although the normal platelet count is 15–30 × 104 mm−3, the cut-off value for reactive thrombocytosis is not clearly defined. Therefore, the cut-off value of the preoperative platelet count used in this study, 30 × 104 mm−3, based on ROC curve analysis, was acceptable.

Previous studies have also demonstrated a relationship between the NLR and postoperative survival in patients with several types of cancer (Shimada et al, 2003; Walsh et al, 2005; Malik et al, 2007; Halazun et al, 2008, 2009; Shimada et al, 2010; Chiang et al, 2012). Because the NLR is based on neutrophils and lymphocytes, it would also be affected by SIR. Although most studies have recommended a NLR cut-off value of 5 (Shimada et al, 2003; Szczepanik et al, 2011), a recent report has recommended a value of 3 (Chiang et al, 2012). Therefore, the NLR cut-off value of 3.0 that we adopted in the present study was acceptable and in line with a previous study that derived a similar value based on ROC curve analysis (Cho et al, 2008).

Shimada et al (2010) have reported that a high NLR was associated with a high platelet count, and selected it as an independent indicator of reduced postoperative survival in patients with gastric cancer. They also reported that not only an elevated CRP level (Shimada et al, 2003) but also reactive thrombocytosis (Shimada et al, 2004) was associated with postoperative survival in patients with oesophageal cancer. These results lend strong support to the use of a combination of reactive thrombocytosis and the NLR for prediction of postoperative survival, along with the GPS.

There are now a number of well-established systemic inflammation-based prognostic scores for patients with CRC. In particular, the GPS and the NLR have been well validated (McMillan, 2013). However, the possibility that additional measures of the SIR might add further significance to these scores has been suggested, as the parameters comprising these scores, that is, CRP, albumin, neutrophils, and lymphocytes, are all correlated with each other (Table 2). To investigate this issue, we examined the ability of neutrophil, lymphocyte, and platelet counts to further discriminate patients in individual groups divided according to whether they had a GPS of 0 or 1, or a GPS of 2. The results obtained from Cox proportional hazards model analyses using the recommended cut-off values from ROC curve analyses revealed that only the platelet count was able to divide both patients with a GPS of 0 or 1, and those with a GPS of 2, into two independent groups (data not shown). These results demonstrated that the platelet count was able to add additional discriminatory ability to not only the GPS but also the NLR; combination of the NLR and the platelet count created a novel inflammation-based prognostic score, the COP-NLR. This is considered to be a meaningful advance in the development of an optimal, routinely available, systemic inflammation-based prognostic score that is able to complement tumour staging.

Theoretically, direct measurement of the serum IL-6 level is the best way to estimate SIR based on tumour vs host interaction. However, there are many unsolved problems associated with the routine measurement of IL-6 in cancer patients. Although a recent study has revealed that the serum level of IL-6 is associated with the postoperative survival of patients with gastric cancer (Szczepanik et al, 2011), routine measurement of IL-6 is difficult in a clinical setting because of its high cost and inconvenience.

On the other hand, the COP-NLR is easy to measure routinely because of its low cost and convenience. In addition, in comparison with the GPS, the COP-NLR may have higher applicability for estimation of SIR, because proliferation and differentiation of cellular components occurs much faster than protein synthesis when inflammatory cytokines are released. Moreover, because repeat measurements of the COP-NLR can be performed more easily, not only before but also after surgery, than those of the GPS and tumour markers, it would deliver solid data based on well-known tumour markers.

Thus, the COP-NLR not only appears capable of classifying patients with CRC into three independent groups before surgery but also has potential as a novel predictor of postoperative survival in such patients.