Med Oncol (2015) 32:144 DOI 10.1007/s12032-015-0590-2
ORIGINAL PAPER
Prognostic performance of inflammation-based prognostic indices in patients with resectable colorectal liver metastases Christopher P. Neal1 • Vaux Cairns1 • Michael J. Jones1 • Muhammad M. Masood1 Gael R. Nana1 • Christopher D. Mann1 • Giuseppe Garcea1 • Ashley R. Dennison1
•
Received: 5 January 2015 / Accepted: 14 March 2015 Ó Springer Science+Business Media New York 2015
Abstract A range of prognostic cellular indices of the systemic inflammatory response, namely the neutrophil– lymphocyte ratio (NLR), derived NLR (dNLR), platelet– lymphocyte ratio (PLR), lymphocyte–monocyte ratio (LMR), combination of platelet count and neutrophil–lymphocyte ratio (COP–NLR) and prognostic nutritional index (PNI), have been developed and found to have prognostic utility across varied malignancies. The current study is the first to examine the prognostic value of these six inflammatory scores in patients with resectable colorectal liver metastases (CRLM). Data from 302 consecutive patients undergoing surgery for resectable CRLM were evaluated. The prognostic influence of clinicopathological variables and the inflammatory scores NLR, dNLR, PLR, LMR, COP– NLR and PNI upon overall survival (OS) and cancer-specific survival (CSS) were determined by log-rank analysis and univariate and multivariate Cox regression analyses. High preoperative NLR was the only inflammatory variable independently associated with shortened OS (HR 1.769, 95 % CI 1.302–2.403, P \ 0.001) or CSS (HR 1.927, 95 % CI 1.398–2.655, P \ 0.001) following metastasectomy. When NLR was replaced by dNLR in analyses, high dNLR was independently associated with shortened OS (HR 1.932, 95 % CI 1.356–2.754, P \ 0.001) and CSS (HR 1.807, 95 % CI 1.209–2.702, P = 0.004). The inflammatory scores PLR, LMR, COP–NLR and PNI demonstrated no independent association with either overall or cancer-specific survival in the study population. Our findings support high preoperative
& Christopher P. Neal [email protected] 1
Department of Hepatobiliary and Pancreatic Surgery, Leicester General Hospital, Gwendolen Road, Leicester LE5 4PW, UK
NLR and dNLR as independent prognostic factors for poor outcome in patients undergoing CRLM resection, with prognostic value superior to other cellular-based systemic inflammatory scores. Keywords Colorectal cancer Liver metastasis Hepatectomy Neutrophil-to-lymphocyte ratio Biomarker Prognosis
Introduction According to the American Cancer Society, colorectal cancer (CRC) was the third most common malignancy and the second most frequent cause of cancer death in the USA in 2013 [1]. Worldwide more than 1.2 million new CRC cases are diagnosed each year [2]. Of these, 20–25 % of patients will have colorectal liver metastases (CRLM) at presentation and a further 40–50 % will develop metachronous CRLM following primary tumour removal. Hepatic resection is now accepted as the optimal treatment modality for patients with CRLM, providing 5-year survival rates of 30–50 % depending upon selection criteria [3–5]. Disease recurrence in the remnant liver and/or extrahepatic sites is common following metastasectomy, however, affecting up to three-quarters of patients [5]. The ability to identify patients more likely to develop recurrence and poor outcome following surgery would be of immense benefit, enabling surgical and chemotherapeutic treatment tailored to individual recurrence risk. This is particularly so given the increasing rates of hepatectomy for the treatment of CRLM. Whilst clinicopathological factors and scoring systems have been shown to predict recurrence risk following metastasectomy [3, 6], they lack the sensitivity to enable individual prognostication. Similarly, studies have
123
144
Page 2 of 11
assessed the influence of molecular tumour markers upon recurrence risk, but further validation is required [7]. Whilst it is clear that a genetic basis underlies cancer development, recent work has identified the host inflammatory response to the tumour as an additional ‘hallmark’ of carcinogenesis, playing a key role in cancer development, progression and metastasis [8, 9]. Local tumour-related inflammation is reflected in a systemic inflammatory response (SIR) that may be easily measured preoperatively. Inflammatory cytokines, such as interleukin-1b, tumour necrosis factor-a and the multifunctional cytokine IL-6, are released and induce expression of acute phase proteins, such as CRP, and a decrease in the level of circulating albumin. Elevation of CRP has been shown to independently predict shortened survival in numerous cancer types [10, 11], including resectable CRC [11, 12] and CRLM [13, 14]. The pro-inflammatory cytokine milieu of the SIR triggers changes in immune and haematopoietic cell populations, including myeloid cell proliferation and the differentiation of megakaryocytes to platelets. The theory of ‘immune enhancement’ suggests that neutrophils and monocytes are then recruited from the circulation to play a direct role in promoting aggressive tumour behaviour [15, 16], through the production of pro-proliferative and proinvasive ligands such as transforming growth factor-b (TGF-b) [17] and matrix metalloproteinases [18], the release of pro-angiogenic vascular endothelial growth factor [19] and the direct promotion of cancer cell intravasation and extravasation [20]. Similarly, activated platelets may secrete factors that contribute to tumour growth, invasion and angiogenesis, release particles that increase tumour cell invasive potential and interfere with natural killer (NK) cell tumour recognition [21]. Conversely, a relative lymphopenia is frequently associated with the SIR and is thought to aid cancer progression through effects upon cellmediated immunity [22]. In keeping with these effects upon carcinogenesis, measurable changes to these cellular markers of the SIR, including leucocytosis, neutrophilia, lymphopenia and thrombocytosis, have been shown to have prognostic significance in malignancy, including in the context of resectable CRLM [11]. In an aim to further refine prognostication, haematological indices have been variably combined to generate a number of prognostic inflammatory scores. Most notably, an elevated neutrophil–lymphocyte ratio (NLR) has repeatedly been shown to independently predict disease recurrence and shortened survival in varied cancer types [11, 23], including CRC [24, 25] and resectable CRLM [4, 26, 27]. An elevated platelet–lymphocyte ratio (PLR) has also been shown to be associated with shortened overall survival in varied solid tumours [28]. More recently, the combination of platelet count and neutrophil lymphocyte
123
Med Oncol (2015) 32:144
ratio (COP–NLR) and the lymphocyte–monocyte ratio (LMR) were developed and shown to be independently associated with shortened survival in several tumour types [29–31], including primary CRC [32, 33]. Finally, a low prognostic nutritional index (PNI, derived from lymphocyte count and serum albumin) was shown to exert prognostic significance in several tumour types, including primary CRC [34]. Of note, however, the majority of these studies have studied the prognostic value of a single inflammatory score, without comparison with other inflammatory variables, including the established NLR. In the context of resectable CRLM, no study to date has examined the prognostic value of the COP–NLR, LMR or PNI, with only one study assessing the PLR [35]. Furthermore, whilst data pertaining to NLR and prognosis are convincing, Proctor et al. recently highlighted the issue that the further widespread validation of the NLR using existing clinical trials databases may be limited by the fact that these resources commonly only enter information pertaining to white cell count and neutrophil count, with information on lymphocyte count lacking. To obviate this problem and allow the widespread utilisation of the NLR, Proctor et al. [36] implemented a derived score, the derived neutrophil–lymphocyte ratio (dNLR), composed of neutrophil count divided by white cell count minus neutrophil count, and were able to demonstrate similar prognostic value to the established NLR across a range of malignancies. Subsequent studies have confirmed its prognostic significance in several other cancer types [37– 41], but it has yet to be validated in the context of resectable CRLM. In the present study, therefore, we set out to validate the prognostic significance of a range of pre-treatment cellular inflammatory variables (including the PLR, COP–NLR, LMR and PNI) in a large cohort of patients undergoing resection of CRLM, comparing these indices with the established NLR in order to determine the cellular index of the SIR with optimum prognostic value. In addition, we sought to validate, for the first time, the prognostic value of the dNLR in this context.
Patients and methods Patients undergoing hepatectomy for CRLM between January 2006 and December 2010 at a single hepatobiliary tertiary referral institution (Leicester General Hospital, UK) were identified on a prospectively maintained database. During the study period, clinical data had been recorded prospectively in the database. These data were analysed retrospectively, with any additional information gathered from medical records. The study was approved by our institutional review board.
Med Oncol (2015) 32:144
Clinicopathological variables, including the Clinical Risk Score (CRS), devised by the Memorial Sloan Kettering Cancer Centre [3], were recorded for each patient. This score is composed of the following five independent prognostic variables (one point is assigned to each to yield a score of 0–5): positive nodal status of the primary tumour, disease-free interval from diagnosis of primary tumour to discovery of liver metastases of 12 months or less, number of liver metastases greater than one, diameter of largest liver metastasis [5 cm and preoperative carcinoembryonic antigen (CEA) level [200 ng/mL. Other clinicopathological variables recorded, previously shown to have prognostic significance [5], included age, gender, primary tumour site and the use of adjuvant or neoadjuvant chemotherapy. Haematological variables were obtained from blood tests performed routinely before surgery. At the time of sampling, no patient had clinical evidence of infection or other inflammatory conditions. Preoperative staging comprised computed tomography (CT) of the chest, abdomen and pelvis using an iodinated contrast agent and magnetic resonance imaging of the liver. The follow-up surveillance protocol included a clinical review at 6 weeks post-surgery, followed by clinical evaluation, routine serum investigations (including liver function tests and serum CEA level) and CT scan of the chest, abdomen and pelvis, all performed every 6 months for 2 years and then annually to 5 years post-hepatectomy. Abnormal results during routine post-operative surveillance triggered further investigation. Development of symptoms prompted a review earlier than scheduled. The primary outcome measure was overall survival (as of January 2014). Cancer-specific survival was considered as the secondary end point. The cause of death was determined from case notes, computerised records and death certificates. Patients who died within 90 days of hepatectomy were excluded from long-term survival analysis. For staged procedures, variables were recorded prior to the first-stage procedure and patients who were inoperable at the attempted second stage were excluded from analysis. In the case of patients who underwent a repeat liver resection following the diagnosis of operable recurrence, only the index resection was included in the analysis. Statistics
Page 3 of 11
144
literature. For the inflammatory indices dNLR and LMR, established thresholds do not exist, and hence, the optimal cut-off levels for these two variables were determined by applying receiver operating characteristic (ROC) curve analysis (on the basis of joint sensitivity and specificity), as described previously [44]. For dNLR, 3.0 was identified as the optimal cut-point for both overall survival and cancerspecific survival in our study population. Of note, this cutpoint was also recently found to be the optimal cut-point in predicting disease-free survival in patients with primary colon cancer by Absenger et al. [37]. For LMR, ROC analysis confirmed 2.35 to be the optimal cut-point for both overall survival and cancer-specific survival in the current study population. The Chi-squared and Fisher’s exact tests were used to analyse for significant associations and differences between subgroups within the cohort. Where variables did not follow a normal distribution, the Mann–Whitney U test was applied. Univariate prognostic significance of variables was determined by means of univariate Cox regression analysis, Kaplan–Meier analysis and application of the logrank test. Multivariate analysis was performed, using all variables with P \ 0.10 on univariate analysis, through their entry into Cox proportional hazard regression analysis using a stepwise backward procedure. Statistical significance was defined as P \ 0.05. Statistical analyses were performed using Statistical Package for the Social Sciences 20.0Ò (SPSS, Chicago, IL, USA). Table 1 Inflammation-based prognostic scores Scoring system
Score
Neutrophil–lymphocyte ratio (NLR) Neutrophil count: lymphocyte count C5
1
Neutrophil count: lymphocyte count \5
0
Derived neutrophil–lymphocyte ratio (dNLR) Neutrophil count: (leucocyte count–neutrophil count) C3
1
Neutrophil count: (leucocyte count–neutrophil count) \3
0
Lymphocyte–monocyte ratio (LMR) Lymphocyte count: monocyte count \2.35 Lymphocyte count: monocyte count C2.35
1 0
Platelet–lymphocyte ratio (PLR) Platelet count: lymphocyte count [300
2
Platelet count: lymphocyte count 150–300
1
Platelet count: lymphocyte count \150
0
Combination of platelet count and NLR (COP–NLR)
As reported previously [42], grouping of the continuous variables CRS, leucocyte count, neutrophil count, monocyte count, lymphocyte count, platelet count and serum albumin were carried out using standard thresholds. The inflammatory scores NLR, PLR, COP–NLR and PNI were constructed as setout in Table 1, using widely accepted thresholds [43], allowing comparison with the available
NLR [ 3 and platelet count [ 300 9 109/L
2
NLR [ 3 or platelet count [ 300 9 109/L
1
Neither NLR [ 3 nor platelet count [ 300 9 109/L
0
Prognostic nutritional index (PNI) Albumin (g/L) ? 59 total lymphocyte count (9109/L) \ 45
1
Albumin (g/L) ? 59 total lymphocyte count (9109/L) C 45
0
123
144
Page 4 of 11
Med Oncol (2015) 32:144
Results A total of 311 patients underwent hepatectomy for CRLM over the study period. All resections were performed with curative intent. Ninety-day mortality rate following hepatectomy was 1.9 % (n = 6). Three patients were deemed inoperable at the second stage of an attempted twostage procedure. Three hundred and two patients were, therefore, eligible for inclusion in the study. Baseline patient characteristics and tumour biological factors are shown in Table 2. There were 192 (63.6 %) men and 110 (36.4 %) women. The median age at time of surgery was 66 years (mean age 64.8 years, range 26–85 years). The median number of tumours on preoperative imaging was 2 (range 1–15). Tumour size ranged from 6 to 150 mm, with
a median size of 30 mm. A poor preoperative CRS (score 3–5) was recorded in 39.1 % of patients (n = 118). One hundred and thirty-two patients (43.7 %) had systemic chemotherapy in the 6 months prior to their liver resection, and 126 patients (41.7 %) received systemic chemotherapy following metastasectomy. In total, 187 patients (61.9 %) had a ‘major’ resection (resection of more than three Couinaud segments). Inflammatory scores and their relationship to clinicopathological characteristics The spread of inflammatory scores amongst the study population is shown in Table 2. The relationships between inflammatory variables and clinicopathological features
Table 2 Univariate Cox regression survival analyses for all patients undergoing metastasectomy (n = 302) Variables
Patients (n = 302)
Overall survival
Age: \65/C65 years
134/168
1.207 (0.907–1.605)
0.197
1.158 (0.856–1.566)
0.342
Gender: male/female
192/110
0.947 (0.704–1.273)
0.718
0.856 (0.622–1.179)
0.341
Site of primary tumour: rectum/colon
149/153
0.951 (0.716–1.262)
0.726
0.889 (0.658–1.200)
0.442
Stage of primary tumour: node positive/node negative
183/119
1.128 (0.844–1.509)
0.415
1.183 (0.869–1.610)
0.286
Number of liver metastases: 1/[1
141/161
1.421 (1.067–1.894)
0.016
1.354 (1.000–1.833)
0.050
Hazard ratio (95 % CI)
Cancer-specific survival P value
Hazard ratio (95 % CI)
P value
(a) Clinicopathological variables
1–3/[3
254/48
1.285 (0.879–1.880)
0.196
1.362 (0.918–2.021)
0.125
Diameter of liver metastases: B50/[50 mm
238/64
1.344 (0.961–1.879)
0.084
1.421 (1.002–2.015)
0.049
Temporal presentation of metastases: DFI B 12/[12 months Preoperative CEA: B200/[200 ng/mL
214/88 286/16
1.487 (1.072–2.063) 2.309 (1.377–3.871)
0.017 0.002
1.593 (1.117–2.270) 2.217 (1.277–3.848)
0.010 0.005
Clinical Risk Score: 0–2/3–5
184/118
1.303 (1.128–1.506)
ORIGINAL PAPER
Prognostic performance of inflammation-based prognostic indices in patients with resectable colorectal liver metastases Christopher P. Neal1 • Vaux Cairns1 • Michael J. Jones1 • Muhammad M. Masood1 Gael R. Nana1 • Christopher D. Mann1 • Giuseppe Garcea1 • Ashley R. Dennison1
•
Received: 5 January 2015 / Accepted: 14 March 2015 Ó Springer Science+Business Media New York 2015
Abstract A range of prognostic cellular indices of the systemic inflammatory response, namely the neutrophil– lymphocyte ratio (NLR), derived NLR (dNLR), platelet– lymphocyte ratio (PLR), lymphocyte–monocyte ratio (LMR), combination of platelet count and neutrophil–lymphocyte ratio (COP–NLR) and prognostic nutritional index (PNI), have been developed and found to have prognostic utility across varied malignancies. The current study is the first to examine the prognostic value of these six inflammatory scores in patients with resectable colorectal liver metastases (CRLM). Data from 302 consecutive patients undergoing surgery for resectable CRLM were evaluated. The prognostic influence of clinicopathological variables and the inflammatory scores NLR, dNLR, PLR, LMR, COP– NLR and PNI upon overall survival (OS) and cancer-specific survival (CSS) were determined by log-rank analysis and univariate and multivariate Cox regression analyses. High preoperative NLR was the only inflammatory variable independently associated with shortened OS (HR 1.769, 95 % CI 1.302–2.403, P \ 0.001) or CSS (HR 1.927, 95 % CI 1.398–2.655, P \ 0.001) following metastasectomy. When NLR was replaced by dNLR in analyses, high dNLR was independently associated with shortened OS (HR 1.932, 95 % CI 1.356–2.754, P \ 0.001) and CSS (HR 1.807, 95 % CI 1.209–2.702, P = 0.004). The inflammatory scores PLR, LMR, COP–NLR and PNI demonstrated no independent association with either overall or cancer-specific survival in the study population. Our findings support high preoperative
& Christopher P. Neal [email protected] 1
Department of Hepatobiliary and Pancreatic Surgery, Leicester General Hospital, Gwendolen Road, Leicester LE5 4PW, UK
NLR and dNLR as independent prognostic factors for poor outcome in patients undergoing CRLM resection, with prognostic value superior to other cellular-based systemic inflammatory scores. Keywords Colorectal cancer Liver metastasis Hepatectomy Neutrophil-to-lymphocyte ratio Biomarker Prognosis
Introduction According to the American Cancer Society, colorectal cancer (CRC) was the third most common malignancy and the second most frequent cause of cancer death in the USA in 2013 [1]. Worldwide more than 1.2 million new CRC cases are diagnosed each year [2]. Of these, 20–25 % of patients will have colorectal liver metastases (CRLM) at presentation and a further 40–50 % will develop metachronous CRLM following primary tumour removal. Hepatic resection is now accepted as the optimal treatment modality for patients with CRLM, providing 5-year survival rates of 30–50 % depending upon selection criteria [3–5]. Disease recurrence in the remnant liver and/or extrahepatic sites is common following metastasectomy, however, affecting up to three-quarters of patients [5]. The ability to identify patients more likely to develop recurrence and poor outcome following surgery would be of immense benefit, enabling surgical and chemotherapeutic treatment tailored to individual recurrence risk. This is particularly so given the increasing rates of hepatectomy for the treatment of CRLM. Whilst clinicopathological factors and scoring systems have been shown to predict recurrence risk following metastasectomy [3, 6], they lack the sensitivity to enable individual prognostication. Similarly, studies have
123
144
Page 2 of 11
assessed the influence of molecular tumour markers upon recurrence risk, but further validation is required [7]. Whilst it is clear that a genetic basis underlies cancer development, recent work has identified the host inflammatory response to the tumour as an additional ‘hallmark’ of carcinogenesis, playing a key role in cancer development, progression and metastasis [8, 9]. Local tumour-related inflammation is reflected in a systemic inflammatory response (SIR) that may be easily measured preoperatively. Inflammatory cytokines, such as interleukin-1b, tumour necrosis factor-a and the multifunctional cytokine IL-6, are released and induce expression of acute phase proteins, such as CRP, and a decrease in the level of circulating albumin. Elevation of CRP has been shown to independently predict shortened survival in numerous cancer types [10, 11], including resectable CRC [11, 12] and CRLM [13, 14]. The pro-inflammatory cytokine milieu of the SIR triggers changes in immune and haematopoietic cell populations, including myeloid cell proliferation and the differentiation of megakaryocytes to platelets. The theory of ‘immune enhancement’ suggests that neutrophils and monocytes are then recruited from the circulation to play a direct role in promoting aggressive tumour behaviour [15, 16], through the production of pro-proliferative and proinvasive ligands such as transforming growth factor-b (TGF-b) [17] and matrix metalloproteinases [18], the release of pro-angiogenic vascular endothelial growth factor [19] and the direct promotion of cancer cell intravasation and extravasation [20]. Similarly, activated platelets may secrete factors that contribute to tumour growth, invasion and angiogenesis, release particles that increase tumour cell invasive potential and interfere with natural killer (NK) cell tumour recognition [21]. Conversely, a relative lymphopenia is frequently associated with the SIR and is thought to aid cancer progression through effects upon cellmediated immunity [22]. In keeping with these effects upon carcinogenesis, measurable changes to these cellular markers of the SIR, including leucocytosis, neutrophilia, lymphopenia and thrombocytosis, have been shown to have prognostic significance in malignancy, including in the context of resectable CRLM [11]. In an aim to further refine prognostication, haematological indices have been variably combined to generate a number of prognostic inflammatory scores. Most notably, an elevated neutrophil–lymphocyte ratio (NLR) has repeatedly been shown to independently predict disease recurrence and shortened survival in varied cancer types [11, 23], including CRC [24, 25] and resectable CRLM [4, 26, 27]. An elevated platelet–lymphocyte ratio (PLR) has also been shown to be associated with shortened overall survival in varied solid tumours [28]. More recently, the combination of platelet count and neutrophil lymphocyte
123
Med Oncol (2015) 32:144
ratio (COP–NLR) and the lymphocyte–monocyte ratio (LMR) were developed and shown to be independently associated with shortened survival in several tumour types [29–31], including primary CRC [32, 33]. Finally, a low prognostic nutritional index (PNI, derived from lymphocyte count and serum albumin) was shown to exert prognostic significance in several tumour types, including primary CRC [34]. Of note, however, the majority of these studies have studied the prognostic value of a single inflammatory score, without comparison with other inflammatory variables, including the established NLR. In the context of resectable CRLM, no study to date has examined the prognostic value of the COP–NLR, LMR or PNI, with only one study assessing the PLR [35]. Furthermore, whilst data pertaining to NLR and prognosis are convincing, Proctor et al. recently highlighted the issue that the further widespread validation of the NLR using existing clinical trials databases may be limited by the fact that these resources commonly only enter information pertaining to white cell count and neutrophil count, with information on lymphocyte count lacking. To obviate this problem and allow the widespread utilisation of the NLR, Proctor et al. [36] implemented a derived score, the derived neutrophil–lymphocyte ratio (dNLR), composed of neutrophil count divided by white cell count minus neutrophil count, and were able to demonstrate similar prognostic value to the established NLR across a range of malignancies. Subsequent studies have confirmed its prognostic significance in several other cancer types [37– 41], but it has yet to be validated in the context of resectable CRLM. In the present study, therefore, we set out to validate the prognostic significance of a range of pre-treatment cellular inflammatory variables (including the PLR, COP–NLR, LMR and PNI) in a large cohort of patients undergoing resection of CRLM, comparing these indices with the established NLR in order to determine the cellular index of the SIR with optimum prognostic value. In addition, we sought to validate, for the first time, the prognostic value of the dNLR in this context.
Patients and methods Patients undergoing hepatectomy for CRLM between January 2006 and December 2010 at a single hepatobiliary tertiary referral institution (Leicester General Hospital, UK) were identified on a prospectively maintained database. During the study period, clinical data had been recorded prospectively in the database. These data were analysed retrospectively, with any additional information gathered from medical records. The study was approved by our institutional review board.
Med Oncol (2015) 32:144
Clinicopathological variables, including the Clinical Risk Score (CRS), devised by the Memorial Sloan Kettering Cancer Centre [3], were recorded for each patient. This score is composed of the following five independent prognostic variables (one point is assigned to each to yield a score of 0–5): positive nodal status of the primary tumour, disease-free interval from diagnosis of primary tumour to discovery of liver metastases of 12 months or less, number of liver metastases greater than one, diameter of largest liver metastasis [5 cm and preoperative carcinoembryonic antigen (CEA) level [200 ng/mL. Other clinicopathological variables recorded, previously shown to have prognostic significance [5], included age, gender, primary tumour site and the use of adjuvant or neoadjuvant chemotherapy. Haematological variables were obtained from blood tests performed routinely before surgery. At the time of sampling, no patient had clinical evidence of infection or other inflammatory conditions. Preoperative staging comprised computed tomography (CT) of the chest, abdomen and pelvis using an iodinated contrast agent and magnetic resonance imaging of the liver. The follow-up surveillance protocol included a clinical review at 6 weeks post-surgery, followed by clinical evaluation, routine serum investigations (including liver function tests and serum CEA level) and CT scan of the chest, abdomen and pelvis, all performed every 6 months for 2 years and then annually to 5 years post-hepatectomy. Abnormal results during routine post-operative surveillance triggered further investigation. Development of symptoms prompted a review earlier than scheduled. The primary outcome measure was overall survival (as of January 2014). Cancer-specific survival was considered as the secondary end point. The cause of death was determined from case notes, computerised records and death certificates. Patients who died within 90 days of hepatectomy were excluded from long-term survival analysis. For staged procedures, variables were recorded prior to the first-stage procedure and patients who were inoperable at the attempted second stage were excluded from analysis. In the case of patients who underwent a repeat liver resection following the diagnosis of operable recurrence, only the index resection was included in the analysis. Statistics
Page 3 of 11
144
literature. For the inflammatory indices dNLR and LMR, established thresholds do not exist, and hence, the optimal cut-off levels for these two variables were determined by applying receiver operating characteristic (ROC) curve analysis (on the basis of joint sensitivity and specificity), as described previously [44]. For dNLR, 3.0 was identified as the optimal cut-point for both overall survival and cancerspecific survival in our study population. Of note, this cutpoint was also recently found to be the optimal cut-point in predicting disease-free survival in patients with primary colon cancer by Absenger et al. [37]. For LMR, ROC analysis confirmed 2.35 to be the optimal cut-point for both overall survival and cancer-specific survival in the current study population. The Chi-squared and Fisher’s exact tests were used to analyse for significant associations and differences between subgroups within the cohort. Where variables did not follow a normal distribution, the Mann–Whitney U test was applied. Univariate prognostic significance of variables was determined by means of univariate Cox regression analysis, Kaplan–Meier analysis and application of the logrank test. Multivariate analysis was performed, using all variables with P \ 0.10 on univariate analysis, through their entry into Cox proportional hazard regression analysis using a stepwise backward procedure. Statistical significance was defined as P \ 0.05. Statistical analyses were performed using Statistical Package for the Social Sciences 20.0Ò (SPSS, Chicago, IL, USA). Table 1 Inflammation-based prognostic scores Scoring system
Score
Neutrophil–lymphocyte ratio (NLR) Neutrophil count: lymphocyte count C5
1
Neutrophil count: lymphocyte count \5
0
Derived neutrophil–lymphocyte ratio (dNLR) Neutrophil count: (leucocyte count–neutrophil count) C3
1
Neutrophil count: (leucocyte count–neutrophil count) \3
0
Lymphocyte–monocyte ratio (LMR) Lymphocyte count: monocyte count \2.35 Lymphocyte count: monocyte count C2.35
1 0
Platelet–lymphocyte ratio (PLR) Platelet count: lymphocyte count [300
2
Platelet count: lymphocyte count 150–300
1
Platelet count: lymphocyte count \150
0
Combination of platelet count and NLR (COP–NLR)
As reported previously [42], grouping of the continuous variables CRS, leucocyte count, neutrophil count, monocyte count, lymphocyte count, platelet count and serum albumin were carried out using standard thresholds. The inflammatory scores NLR, PLR, COP–NLR and PNI were constructed as setout in Table 1, using widely accepted thresholds [43], allowing comparison with the available
NLR [ 3 and platelet count [ 300 9 109/L
2
NLR [ 3 or platelet count [ 300 9 109/L
1
Neither NLR [ 3 nor platelet count [ 300 9 109/L
0
Prognostic nutritional index (PNI) Albumin (g/L) ? 59 total lymphocyte count (9109/L) \ 45
1
Albumin (g/L) ? 59 total lymphocyte count (9109/L) C 45
0
123
144
Page 4 of 11
Med Oncol (2015) 32:144
Results A total of 311 patients underwent hepatectomy for CRLM over the study period. All resections were performed with curative intent. Ninety-day mortality rate following hepatectomy was 1.9 % (n = 6). Three patients were deemed inoperable at the second stage of an attempted twostage procedure. Three hundred and two patients were, therefore, eligible for inclusion in the study. Baseline patient characteristics and tumour biological factors are shown in Table 2. There were 192 (63.6 %) men and 110 (36.4 %) women. The median age at time of surgery was 66 years (mean age 64.8 years, range 26–85 years). The median number of tumours on preoperative imaging was 2 (range 1–15). Tumour size ranged from 6 to 150 mm, with
a median size of 30 mm. A poor preoperative CRS (score 3–5) was recorded in 39.1 % of patients (n = 118). One hundred and thirty-two patients (43.7 %) had systemic chemotherapy in the 6 months prior to their liver resection, and 126 patients (41.7 %) received systemic chemotherapy following metastasectomy. In total, 187 patients (61.9 %) had a ‘major’ resection (resection of more than three Couinaud segments). Inflammatory scores and their relationship to clinicopathological characteristics The spread of inflammatory scores amongst the study population is shown in Table 2. The relationships between inflammatory variables and clinicopathological features
Table 2 Univariate Cox regression survival analyses for all patients undergoing metastasectomy (n = 302) Variables
Patients (n = 302)
Overall survival
Age: \65/C65 years
134/168
1.207 (0.907–1.605)
0.197
1.158 (0.856–1.566)
0.342
Gender: male/female
192/110
0.947 (0.704–1.273)
0.718
0.856 (0.622–1.179)
0.341
Site of primary tumour: rectum/colon
149/153
0.951 (0.716–1.262)
0.726
0.889 (0.658–1.200)
0.442
Stage of primary tumour: node positive/node negative
183/119
1.128 (0.844–1.509)
0.415
1.183 (0.869–1.610)
0.286
Number of liver metastases: 1/[1
141/161
1.421 (1.067–1.894)
0.016
1.354 (1.000–1.833)
0.050
Hazard ratio (95 % CI)
Cancer-specific survival P value
Hazard ratio (95 % CI)
P value
(a) Clinicopathological variables
1–3/[3
254/48
1.285 (0.879–1.880)
0.196
1.362 (0.918–2.021)
0.125
Diameter of liver metastases: B50/[50 mm
238/64
1.344 (0.961–1.879)
0.084
1.421 (1.002–2.015)
0.049
Temporal presentation of metastases: DFI B 12/[12 months Preoperative CEA: B200/[200 ng/mL
214/88 286/16
1.487 (1.072–2.063) 2.309 (1.377–3.871)
0.017 0.002
1.593 (1.117–2.270) 2.217 (1.277–3.848)
0.010 0.005
Clinical Risk Score: 0–2/3–5
184/118
1.303 (1.128–1.506)
