Prognostic significance of pre-operative C-reactive protein and the neutrophil–lymphocyte ratio in resectable pancreatic cancer: a systematic review

Abstract

Background

Better pre-operative risk stratification may improve patient selection for pancreatic resection in pancreatic cancer. C-reactive protein (CRP) and the neutrophil–lymphocyte ratio (NLR) have demonstrated prognostic value in some cancers. The role of CRP and NLR in predicting outcome in pancreatic cancer after curative resection is not well established.

Methods

An electronic search of MEDLINE, EMBASE and CINAHL was performed to identify studies assessing survival in patients after pancreatic cancer resection with high or low pre-operative CRP or NLR. Systematic review was undertaken using the PRISMA protocol.

Results

In total, 327 studies were identified with 10 reporting on survival outcomes after a pancreatic resection in patients with high or low CRP, NLR or both. All but one paper showed a trend of lower inflammatory markers in patients with longer survival. Three studies from six showed low CRP to be independently associated with increased survival and two studies of eight showed the same for NLR. All studies were retrospective cohort studies of low to moderate quality.

Discussion

Inflammatory markers might prove useful guides to the management of resectable pancreatic cancer but, given the poor quality of evidence, further longitudinal studies are required before incorporating pre-operative inflammatory markers into clinical decision making.

Introduction

In spite of the relatively low incidence of around 11 per 100 000 person-years, pancreatic cancer is the 5th highest cause of cancer-related mortality with all stage, 5-year survival remaining poor at less than 5%. Advances in staging and adjuvant chemotherapy regimens¹ has resulted in little improvement in the past 30 years.² Surgical resection is considered for all patients with Stage I or II^3,4 disease but these patients comprise only around 20% of all new diagnoses.⁵ Even in this group, survival remains poor with approximately 20% of patients alive at 5 years.^6,7 This poor survival is likely in part as a result of unrecognized stage-specific heterogeneity of tumour characteristics.

Better risk stratification may improve patient selection for pancreatic resection and therefore optimize patient outcomes. The role of inflammation in cancer is widely recognized⁸ with inflammatory and immune mediators known to modulate carcinogenesis, tumour invasion and metastasis.^9–12 A number of markers of inflammation have been investigated in various cancers with a view to use in risk stratification.¹³

Inflammatory burden can be measured using a variety of parameters, with some markers being easily accessible and relatively inexpensive. C-reactive protein (CRP), which is released by the liver as an acute phase protein in response to raised interleukin (IL)-6 released by activated macrophages, is used to define the level of inflammatory response.¹⁴ The neutrophil–lymphocyte ratio (NLR) is readily derived from standard pre-operative blood tests by dividing the absolute circulating neutrophil concentration by the absolute circulating lymphocyte population. NLR is raised in cancer as the disease process produces a cytokine milieu of raised myeloid growth factors, tumour necrosis factor-α, IL-10 and transforming growth factor-β leading to neutrophilia with relative lymphocytopaenia.^15,16

Both CRP^17–19 and NLR^20–23 have been demonstrated to be inversely proportional to survival in a number of malignancies. Indeed, CRP has been incorporated in to clinical scoring systems to predict outcome in both resectable oesophageal²⁴ and colorectal cancer.²⁵

The aim of this study was to systematically review the existing literature and consolidate current knowledge on the prognostic significance of pre-operative CRP and NLR in patients undergoing pancreatic resection for pancreatic cancer.

Methods

Search strategy and study selection

An electronic search of Medline (1946–present), EMBASE (1974–present) and CINAHL (1981–present) was performed independently by L.S. and S.P. Search terms used were ‘Pancreas’, ‘Pancreatic’, ‘Peri-ampullary’, ‘Neoplasia’, ‘Cancer’, ‘Carcinoma’, ‘Adenocarcinoma’, ‘Tumour’, ‘Outcome’, ‘Survival’, ‘Prognosis’, ‘Neutrophil–Lymphocyte Ratio’, ‘NLR’, ‘C-reactive protein’ and ‘CRP’. Bibliographic references of search results were reviewed to identify relevant studies not produced by the electronic search. The final search was undertaken on 24 December 2013.

All citations identified by our search strategy were reviewed independently by L.S. and S.P., by sequential review of title, abstract and finally full text to establish inclusion or exclusion as per Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidance.²⁶

Inclusion criteria

Included studies analysed the effect of raised pre-operative CRP or NLR on survival after pancreatic resection for pancreatic cancer. Given that different centres used various assays and laboratory techniques, coupled with a lack of agreed definition of what constitutes an elevated CRP or NLR, only studies which dichotomized patients to high or low inflammatory markers as determined by the study authors were included. Some studies analysed the effect of the Glasgow Prognostic Score (GPS) or the modified Glasgow Prognostic Score (mGPS) on survival. GPS assigns a score of 0 to patients with CRP < 10 mg/l and albumin ≥35 g/l, a score of 1 to those with CRP >10 mg/l or albumin <35 g/l and a score of 2 for patients with CRP >10 mg/l and albumin <35 g/l. The mGPS score is identical to GPS except that a score of 1 is assigned only when CRP is greater than 10 mg/l whereas an isolated fall in albumin is scored as 0. Only studies that published patient numbers with each GPS/mGPS score were included. Patients with a GPS or mGPS score of 0 were analysed as having low CRP whereas those with a GPS of 2 or a mGPS of 1 or 2 were grouped as having a high CRP. Patients with a GPS of 1 were excluded as it was not possible to ascertain whether the CRP or the albumin level was raised. Only published, English-language studies reporting survival outcomes for resectable pancreatic cancer were included.

Exclusion criteria

Search results were carefully reviewed to exclude duplicate studies or a second study using the same data set. Studies not reporting survival data for high or low inflammatory markers independently were also excluded, as were studies in which data sets included unresectable disease. Epidemiological studies, case reports and conference presentations were excluded.

Quality assessment and data extraction

There is no internationally recognized tool for grading quality of prognostic factor studies. Huguet et al. suggest a modification of the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system to quality assess prognostic factor studies.²⁷ The modified GRADE system was used in conjunction with the Quality in Prognosis Studies (QUIPS) tool²⁸ to assess study quality including risk of bias. A pro forma was used to extract study author and date, number of patients, age, gender, tumour stage, use of adjuvant therapy and primary outcome data, all as defined by the authors of each study.

Results

The search strategy identified 325 articles for initial review with a further 2 identified by review of bibliographies of search results. Ninety-two papers were excluded as duplicates while title review excluded a further 210 studies as either case reports, abstracts, letters, reviews or reporting inappropriate outcome measures. Twenty-five articles underwent abstract review with a further 10 papers excluded as epidemiological studies or not reporting primary outcomes. After full text review of the remaining 15 papers, 5 were excluded owing to inclusion of unresectable disease in the final analysis, failure to dichotomize an inflammatory marker to high or low or inability to extract outcome data from the published text. Ten studies underwent data extraction (Fig. 1).

Figure 1.

PRISMA 2009 flow diagram

Six of the final 10 papers analysed survival related to high or low CRP and 8 studies investigated NLR, whereas 4 studies investigated both CRP and NLR (Table 1).

Table 1.

Demographic data of studies analysing the relationship of CRP or NLR and prognosis after a pancreatic cancer resection

Author	Year of publication	Study period	CRP or NLR	Number	Age (years)	Male : female	Stage	Staging system	Resection performed	Adjuvant therapy	Median follow-up & range (months)
Stotz, M29	2013	2004–2010	Both	110	<65 = 55	51:59	I – 2	UICC 7th Edition3	Undefined	88a	Undefined
					≥65 = 55		II – 99
							III – 9
Hamed, M30	2013	2000–2010	NLR	85	Undefined	Undefined	Undefined	N/A	PD	Undefined	23.5 (IQR = 11.5–43.5)
La Torre, M31	2012	Apr 2003–Nov 2009	Both	101	<65 = 39	53:48	I – 21*	AJCC4 – edition undefined	PD or PPPD	7a	Undefined
					≥65 = 62		II – 72*			7b
							III – 9*			12 c
							IV – 1*
Sanjay, P32	2012	2002–2008	Both	51	70	34:17	Undefined	N/A	PD	18d	Undefined
Jamieson, NB33	2011	Jan 2006–Apr 2009	Both	135	<65 = 73	78:57	Undefined	N/A	PD or PPPD	74d	Undefined
					≥65 = 62
Garcea, G34	2011	2001–2011	NLR	74	66	46:28	Undefined	N/A	Undefined	53a	1–125.8
Bhatti, I35	2010	Jun 1998–Jun 2008	NLR	84	65	48:36	Undefined	N/A	PD	30a	Undefined
Pine, JK36	2009	2001–2006	CRP	58	68	29:29	Undefined	N/A	PD	Undefined	Undefined
Clark, E37	2007	1998–2005	NLR	44	65	Undefined	Undefined	N/A	PD	Undefined	Undefined
Jamieson, NB36	2005	Jan 1993–Jul 2001	CRP	65	<65 = 37	33:32	I – 17	UICC – edition undefined	PD or PPPD	0	20 (Minimum 15)
					>65 = 28		II – 2
							III – 46

NLR, neutrophil–lymphocyte ratio; CRP, C-reactive protein; UICC, International Union against Cancer; AJCC, American Joint Committee on Cancer; PD, pancreaticoduedenectomy/Whipple's procedure; PPPD, pylorus-preserving pancreaticoduodenectomy.

^*Numbers reported at stage breakdown exceeds participants. ^aChemotherapy, ^bradiotherapy, ^cchemoradiotherapy, ^dundefined.

The prognostic significance of CRP was analysed in 485 patients with 233 patients (48%) defined as having high CRP. The cut-off value for a high CRP varied from 3 to 10 mg/l. Four of the 6 studies reported a significantly decreased median survival in patients with a higher pre-operative CRP on univariate analysis, with only 1 study not demonstrating such a correlation. Three studies found CRP to be an independently significant prognosticator on multivariate analysis (Table 2).

Table 2.

Results of analysis of papers studying the prognostic value of CRP in resected pancreatic cancer

Author	Year	CRP	Cut off value (mg)	Number	Median survival & range (months)	Univariate analysis	Multivariate analysis
Stotz, M29	2013	Low	≤10	mGPS of 0 = 73	20	HR 1.095 (P = 0.585)	Not performed
		High	>10	mGPS of 1 = 21	20
				mGPS of 2 = 16	13
La Torre, M31	2012	Low	≤10	GPS of 0 = 32	37.2	P = 0.0001	HR 1.7745 (P = 0.005)
		High	>10	GPS of 2 = 34	7.3
Sanjay, P32	2012	Low	≤3	13	21 (6–36)	P = 0.015	HR 3.969 (P = 0.011)
		High	>3	38	13 (3–70)
Jamieson, NB33	2011	Low	≤10	mGPS of 0 = 74	26.7 (19.1–34.3a)	P < 0.001	HR 2.26 (P = 0.0001)
		High	>10	mGPS of 1 = 31	16.5 (9.9–22.2a)
				mGPS of 2 = 30	13.1 (6.7–19.4a)
Pine, JK36	2009	Low	≤5	28	16.9b	HR 1.14 (P = 0.69)	Not performed
		High	>5	30	16.6b
Jamieson, NB38	2005	Low	≤10	32	18.2 (14.9–21.4)	HR 2.56 (P = <0.001)	P > 0.05
		High	>10	33	8.3 (6.6–10.0)

CRP, C-reactive protein.

^a95% confidence interval,

^bextrapolated from Kaplan–Meier survival curves.

NLR as a prognostic indicator in resectable pancreatic cancer was investigated in 685 patients across 8 studies, with 160 patients (23%) having a high NLR. The cut-off value for a raised NLR was set at 5 in all but one study, which used an NLR cut-off value of 4 (Bhatti et al.²⁸). All but one of the studies showed a trend of increased median survival in patients with a low pre-operative NLR. Three studies demonstrated statistically significant lengthened survival on univariate analysis whereas two studies showed NLR to be an independent prognostic indicator on multivariate analysis (Table 3).

Table 3.

Results of analysis of papers studying the prognostic value of NLR in resected pancreatic cancer

Author	Year	NLR	Cut off value	Number	Median survival & range (months)	Disease free survival (months)	Univariate analysis	Multivariate analysis
Stotz, M29	2013	Low	<5	73	21.25b		HR 1.852 (P = 0.006)	HR 1.611 (P = 0.039)
		High	≥5	37	11.2b
Hamed, M30	2013	Low	<5	67	20.6 (11–36.6)		0.153	Not performed
		High	≥5	18	11.3 (8.1–26.3)
La Torre, M31	2012	Low	<5	81	21		0.13	Not performed
		High	≥5	21	17
Sanjay, P32	2012	Low	≤5	38	16.2 (3–63)		P = 0.272	Not performed
		High	>5	13	9.18 (3–70)
Jamieson, NB33	2011	Low	<5	105	20.9 (15.5–26.3a)		P = 0.51	Not performed
		High	≥5	30	25.7 (16.2–35.1a)
Garcea, G34	2011	Low	<5	50		52	P = 0.0057	Not performed
		High	>5	24		12
Bhatti, I35	2010	Low	≤4	71	14 (13.0–15.8)		HR 1.784 (P = 0.023)	HR 1.21 (P = 0.039)
		High	>4	13	7.1 (1.4–12.9)
Clark, E37	2007	Low	<5	40	10.5 (7.1–21.2)		P = 0.16	Not performed
		High	≥5	4	8.9 (5.3–13.3)

NLR, neutrophil–lymphocyte ratio.

^a95% confidence interval,

^bextrapolated from Kaplan–Meier survival curves.

Study quality

All studies were retrospective, cohort studies of low to moderate quality as per the GRADE classification and are particularly susceptible to study level selection bias and publication bias.²⁸ All studies but one included consecutive patients undergoing a pancreatic resection whereas Hamed et al.³⁰ excluded 22 patients owing to incomplete data. In all studies, only patients with histologically confirmed pancreatic cancer were included. There is significant population heterogeneity with regard to confounding factors such as patient age, nature of pancreatic resection performed, tumour stage and co-morbidities such as active infection, cardiovascular disease, pulmonary disease and obesity which are all known to be associated with raised inflammatory markers.^39–41 The distribution of each potential confounder within high and low inflammatory marker populations is poorly reported across the studies, making external validation difficult. Furthermore there was inconsistent inclusion of patients who had biliary intervention such as ERCP and PTC prior to surgery or measurement of an inflammatory marker. Only in the studies by Pine et al.³⁶ and Sanjay et al.³² were patients who underwent biliary intervention pre-operatively, excluded from data analysis. Jamieson et al.^33,38 recorded blood measurements at least 1 week after biliary decompression and, along with Bhatti et al.,³⁵ excluded patients with concurrent inflammatory or infective processes.

Discussion

The 5-year survival for pancreatic cancer, considered curable at presentation, remains poor in spite of increasingly sophisticated staging techniques.^6,7 Better patient stratification is needed to optimize management planning and patient outcomes. Correlation between raised inflammatory markers, such as CRP and full blood count, and prognosis has been reported in various malignancies.¹³ This probably represents a modulated inflammatory response as a result of tumour burden and the tumour micro-environment. The aim of this study was to review the evidence for CRP and NLR as prognostic indicators in resectable pancreatic cancer to determine their usefulness in refining patient management.

The majority of studies reviewed demonstrated that an elevated CRP or NLR is associated with a decreased median survival be it overall, cancer-specific or disease-free. Two studies demonstrated NLR to be an independent prognostic indicator whereas three studies showed CRP to be an independent prognostic indicator in resectable pancreatic cancer.

The evidence base does have significant limitations with all studies being retrospective and of low quality. Population groups are heterogeneous with regard to cut-off between high or low inflammatory marker, timing of sample, presence of associated inflammatory condition, use of anti-inflammatory medications and tumour stage. There is inconsistent reporting of data with regard to range of survival for patients with high and low inflammatory markers making meaningful pooling of data for meta-analysis impossible.

Accepting these limitations, the available studies do suggest a correlation between pre-operative inflammatory markers and post-operative survival. These easily measured parameters may therefore help decision making in the management of pancreatic cancer. This may include improved staging, selection of patients for neoadjuvant and adjuvant therapies and help in determining follow-up arrangements. Before this is possible, however, high-quality longitudinal studies are required to establish their independent prognostic value.

Most recurrence and cancer-related mortality in patients after resection of pancreatic cancer is observed in later stage disease or with positive resection margins.⁴² Given the significantly better prognosis for patients achieving full tumour clearance with negative margins,⁴³ differentiation between tumour stage IIB (borderline resectable) and stage III (unresectable) is paramount to allow more tailored management plans. Pine et al.²⁸ showed the CRP level to be associated with tumour resectability whereas Ong et al.⁴⁴ found NLR to be significantly higher in patients found to have unresectable disease at laparotomy. Further research is warranted to determine if pre-operative inflammatory markers may help define resectable stage IIB disease and reduce futile surgical intervention.

Neoadjuvant chemotherapy and radiotherapy are not routinely used for pancreatic cancer although recent trials to investigate its usefulness, particularly in borderline resectable disease, have had promising results.⁴⁵ CRP has been shown to predict a response to palliative chemotherapy in pancreatic cancer⁴⁶ whereas NLR has been shown to predict a complete pathological response to neoadjuvant chemoradiotherapy in locally advanced colorectal cancer.⁴⁷ Research into inflammatory markers as predictors of effectiveness of neoadjuvant therapy in pancreatic cancer is therefore justified.

Garcea et al.³⁴ showed that pre-operative NLR value correlates with disease recurrence. Patients without recurrence had a mean NLR of 3.1 compared with 4.7 in those with recurrence (P = 0.02). There may therefore be a role for pre-operative inflammatory markers being used to guide follow-up, with patients defined as higher risk being followed up more closely or with a lower threshold for further cross-sectional imaging post-operatively.

Conclusion

The evidence suggests an association between CRP and NLR and survival after a pancreatic resection. However, there is insufficient high-quality data available to justify the use of either CRP or NLR in clinical decision-making for the management of resectable pancreatic cancer at present. There is a need for high-quality longitudinal studies, which should be prospective in design to allow for adequate data collection, particularly potential confounders such as inflammatory disease and tumour stage. To have a maximal impact, any study should be robustly reported to allow external validation.

Conflict of interest

None declared.