Article Text
Abstract
Aims The prognostic significance of perineural invasion (PNI) for gastric cancer (GC) patients was under debate. This study aimed to review relevant studies and evaluate the impact of PNI on the survival outcome of GC patients.
Methods Systematic literature search was performed using PubMed and Embase databases. The relevant data were extracted, and the association between PNI and clinicopathological characteristics or survival outcome in GC patients were evaluated using a fixed-effect model or random-effect model.
Results A total 13 studies involving 7004 GC patients were included in this meta-analysis. The positive rate of PNI was 35.9% (2512/7004) in GC patients, ranging from 6.9% to 75.6%. There were significant relationships between PNI and a series of unfavourable clinicopathological factors including undifferentiated histology type (OR: 1.78, 95% CI 1.37 to 2.33, p<0.001; I2=75.3%), diffuse type (OR: 1.96, 95% CI 1.07 to 3.60, p=0.029; I2=79.5%), lymphatic invasion (OR: 7.00, 95% CI 3.76 to 13.03, p<0.001; I2=83.6%), vascular invasion (OR: 5.79, 95% CI 1.59 to 21.13, p=0.008; I2=95.8%), deeper tumour invasion (OR: 4.79, 95% CI 3.65 to 6.28, p<0.001; I2=65.0%) and lymph node metastasis (OR: 3.60, 95% CI 2.37 to 5.47, p<0.001; I2=89.6%). In addition, PNI was significantly associated with worse survival outcome in GC patients (HR: 1.69, 95% CI 1.38 to 2.06, p<0.001; I2=71.0%).
Conclusion PNI was frequently detected in surgically resected specimens of GC patients, and it was a predictive factor for survival outcomes in these patients.
- gastric cancer
- perineural invasion
- survival
- prognostic factor
Statistics from Altmetric.com
Introduction
Although the overall incidence of GC has declined in the past few years, it remains the common cause of cancer-related deaths in the world according to Global Cancer Statistics 2018.1 The optimal treatment option for resectable GC patients remains curative resection with negative margins and adequate lymphadenectomy.2 However, a large proportion of GC patients experienced tumour recurrence and had a poor survival outcome even after curative resection. The depth of tumour invasion (T stage) and lymph node metastasis (N stage) were important prognostic factors for GC patients.3 4 To better select patients and determine appropriate therapeutic strategies, identifying as many biological or pathological indicators for overall survival (OS) and recurrence as possible still be necessary.
Perineural invasion (PNI), which also has been called neurotropic carcinomatous spread or perineural spread, was a pathological feature characterised by the infiltration of tumour cells along the perineurium or the neural fascicles. This is an important pathway for the local spread of cancer cells, which represented an aggressive biological behaviour of tumours.5 According to the previous reports, the prevalence of PNI in GC patients varied from 31.7% to 65.0%.6–9 Positive PNI was not infrequent in surgically resected specimens of GC patients with the improvement of detection technology and pathologists’ experience. However, we still lack sufficient understanding of PNI pathogenesis for cancer patients up to now. Besides, the definition of PNI was under debate. Some researchers hold that it should be defined as the presence of tumour cells in any three layers of the nerve sheath (epineurium, perineurium and endoneurium) or in foci outside the nerve sheath with the involvement of 33% the nerve circumference.10
It has been reported that PNI was significantly associated with recurrence and poor survival outcome in pancreatic cancer, prostate cancer and colorectal cancer.11–14 Furthermore, there have been several studies evaluating the prognostic significance of PNI for GC patients, but the results were conflicting. Some studies regarded PNI as a useful prognostic factor for GC patients who underwent curative resection,7 8 15 but others reported that it could not provide more additional prognostic information than well-known clinicopathological factors such as tumour, node, metastases (TNM) stage and differentiation type.9 16 In view of no consensus on this topic, more research evidence need to be provided to help us better understand the prognostic significance of PNI and guide individual treatment for GC patients. In the present study, we reviewed relevant studies and performed a meta-analysis to determine the relationship between PNI and clinicopathological characteristics and evaluate the impact of PNI on survival outcome in GC patients.
Methods
Literature search strategy
The systematic search for relevant studies was independently conducted by two investigators using PubMed, Embase and Cochrane Library databases up to April 2019. The following keywords and search terms were used for this search strategy: “gastric cancer” or “gastric adenocarcinoma” and “perineural invasion”. Through these combined keywords, the primary collection of studies was defined. Besides, the reference lists of relevant articles were manually searched to identify additional relevant studies. The search was limited to published full articles in English and Chinese language. The titles and abstracts of each retrieved studies were scanned by two investigators for evaluating the topic relevance. The full text of potentially relevant studies was obtained and further assessed.
Eligibility criteria
Studies included in this systematic review and meta-analysis were required to meet the following criteria: (1) positive PNI was defined as the presence of tumour cells along the perineurium or the neural fascicles in surgically resected specimens. All GC patients were divided into two groups based on the PNI status (positive PNI and negative PNI group). (2) Included studies investigated the relationship between PNI and clinicopathological characteristics or evaluated the prognostic significance of PNI for GC patients. (3) The survival outcome was presented as OS. HR with 95% CIs should be reported, and these data could be extracted directly from multivariate Cox regression analysis or calculated indirectly from original studies. (4) In case of the same study cohorts or the same populations, only the most informative or highest quality or the latest study was included in this meta-analysis. Studies were excluded if they were letters, comments, conference abstracts, review articles or case reports.
Data extraction and quality assessment
The data from included studies were independently extracted by two investigators after reviewing the abstracts and full text. Any discrepancies were resolved by discussion among all investigators until consensus was reached. The following information was collected from included studies: first author, publication year, country of study population, study design, study period and sample size. In addition, the following clinicopathological characteristics were extracted: age (≥60 vs <60 years), gender (male vs female), tumour size (≥3 cm vs <3 cm or ≥5 cm vs <5 cm), tumour location (upper and middle 1/3 vs low 1/3), differentiation type (undifferentiated vs differentiated type), Lauren classification (diffuse type and intestinal type), Borrmann type (III–IV vs I–II), T stage (T3-T4 vs T1-T2), lymph node metastasis (yes vs no), lymphatic invasion (yes vs no), vascular invasion (yes vs no), pathological TNM stage (III–IV vs I–II) and HR with 95% CIs for OS. All predefined outcomes were summarised in a 2*2 contingency table for further analysis.
The methodological quality of individual retrospective studies was independently evaluated by two investigators using the Newcastle-Ottawa Quality Assessment Scale.17 All included studies were scored according to the scoring items. Studies with ≥7 points were recognised as high-quality in methodology for observational cohort studies. Our study was designed, performed and reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.18
Statistical analysis
ORs and 95% CIs were used as measures to assess the relationship between PNI and clinicopathological characteristics. The HR and its corresponding 95% CIs for OS were pooled and calculated to evaluate the prognostic significance of PNI for GC patients. The pooled HR greater than one indicated a poorer survival outcome in the research group (positive PNI) as opposed to the control group (negative PNI). The degrees of heterogeneity between different studies were quantified and evaluated using the Cochran Q test and I2 statistic. The results were expressed as I2 value and Q test p value, where I2>50% and/or p<0.01 was considered as a high heterogeneity. If a significant statistical heterogeneity was observed, a random effect model (inverse variance method) should be adopted; otherwise, a fixed-effect model should be used.
To explore potential source of heterogeneity, subgroup analysis was performed according to the population geography (eastern vs western), tumour stage (I–III vs I–IV stage), sample size (≥350 vs <350), publication year (after 2010 vs before 2010), resection type (R0 vs R+), the incidence of positive PNI (≥40% vs <40%) and the quality of included studies (≥7 vs <7). Also, the funnel plot was constructed and potential publication bias was detected through visual inspection of its symmetry. All statistical analyses for evaluating the association between PNI and clinicopathological characteristics or its prognostic value were performed by the Stata V.13.0 software. The p value <0.05 was considered to be of statistical significance.
Results
Search results and study characteristics
The flow diagram of study selection was shown in figure 1. From an electronic database search, a total of 254 potentially relevant studies were initially identified with the predefined search strategy. After scanning the titles and abstracts, 232 irrelevant studies were further excluded. Full text of the remaining 22 studies were obtained for review and assessment. Among these studies, three were review articles, four studies did not report related outcomes or data for pooled analysis and two studies did not meet the inclusion criteria. These studies were not included in this systematic review and meta-analysis. Ultimately, we identified 13 relevant studies for evaluating the clinicopathological characteristics and prognostic significance of PNI in GC patients.6–9 15 16 19–25
The baseline characteristics of 13 included studies were summarised in table 1. No randomised controlled trial was eligible for inclusion. All studies were retrospective in nature and were published between 1994 and 2018, and the research period ranged from 1976 to 2014. Five studies were from China, three studies were from Italy, three studies were from Turkey, one study was from Japan and one study was from Finland.
In this meta-analysis, a total of 7004 GC patients underwent surgical treatment, including 2512 patients with positive PNI and 4492 patients without PNI, with a range of sample size from 103 to 1801. According to the included studies, the positive rate of PNI was 35.9% in GC patients, ranging from 6.9% to 75.6%.
Relationship between PNI and clinicopathological characteristics
The results of meta-analysis indicated that positive PNI was significantly associated with undifferentiated histology type (OR: 1.78, 95% CI 1.37 to 2.33, p<0.001; I2=75.3%), diffuse type (OR: 1.96, 95% CI 1.07 to 3.60, p=0.029; I2=79.5%), upper or middle 1/3 tumour (OR: 1.26, 95% CI 1.12 to 1.42, p<0.001; I2=40.5%), deeper tumour invasion (T3-T4 vs T1-T2: OR: 4.79, 95% CI 3.65 to 6.28, p<0.001; I2=65.0%), lymph node metastasis (OR: 3.60, 95% CI 2.37 to 5.47, p<0.001; I2=89.6%) and more advanced tumour stage (stage III–IV vs stage I–II: OR: 4.43, 95% CI 3.40 to 5.75, p<0.001; I2=61.5%) (figures 2–4). On the other hand, lymphatic invasion (OR: 7.00, 95% CI 3.76 to 13.03, p<0.001; I2=83.6%) and vascular invasion (OR: 5.79, 95% CI 1.59 to 21.13, p=0.008; I2=95.8%) were more frequent in GC patients with PNI (table 2 and online supplementary figure 1). The cases with PNI had a larger tumour size than those without PNI (≥3 cm vs <3 cm: OR: 2.92, 95% CI 1.98 to 4.28, p<0.001; I2=0%; ≥5 cm vs <5 cm: OR: 1.83, 95% CI 1.59 to 2.11, p<0.001; I2=0%) (table 2 and online supplementary figure 2). However, there was no significant difference between positive PNI and negative PNI in terms of patients age, gender and Borrmann type (table 2 and online supplementary figure 3).
Supplemental material
Supplemental material
Supplemental material
Prognostic significance of PNI in GC patients
A total of 11 studies involving 5906 GC patients evaluated the impact of positive PNI on survival outcome, and nine of these studies reported a significant result. The meta-analysis of 11 studies revealed a pooled HR of 1.69 (95% CI 1.38 to 2.06, p<0.001), with a significant heterogeneity (I2=71.0%, p<0.001) (figure 5). This result indicated that positive PNI was a predictive factor for survival outcome in GC patients.
Table 3 showed the results of subgroup analysis for OS in these patients. We found that geographic area, publication year, sample size, tumour stage, resection type, the prevalence of PNI and the quality of included studies had no significant impact on the pooled results for OS, which further supported a strong consistent association between PNI and poor survival outcome in GC patients. Besides, the sensitivity analysis suggested that the pooled results remain significant when those studies with the largest weight were in turn removed.
Three studies provided available data on the number of recurrence cases. The pooled result showed that the patients with positive PNI had a higher risk of recurrence than those with PNI negative despite a significant heterogeneity (OR: 3.60, 95% CI 1.67 to 7.74, p<0.001; I2=76.5%).
Publication bias
Based on the visual assessment of funnel plots, the asymmetric distribution of main studies was not observed in the pooled analysis of HR for OS (figure 6).
Discussion
In the present study, we performed a systematic review and meta-analysis to determine the relationship between PNI and other clinicopathological characteristics and evaluate the prognostic significance of PNI for GC patients. The results demonstrated that PNI was significantly associated with a series of unfavourable clinicopathological factors including undifferentiated histology type, diffuse type, tumour size, lymphatic invasion, vascular invasion, lymph node metastasis and tumour stage. The patients with positive PNI were more likely to have more aggressive oncological features than those without PNI.
Previous studies reported a strong association between lymphovascular invasion and PNI.7 15 26 The abundant lymphatic network around the nerves and direct infiltration of the vasa nervorum by cancer cells may partly explain why PNI and lymphovascular invasion were more likely to be simultaneously detected in resected specimens.15 Hwang et al retrospectively reviewed histopathological information of 206 consecutive stage II–III patients who underwent curative resection for GC, the results demonstrated that the proportion of concomitant existence of PNI and lymphovascular invasion accounted for 42.7% of all patients.26 In terms of survival outcome, the median disease-free survival (DFS) and OS time for the patients with both positive PNI and positive lymphovascular invasion were the poorest among all GC patients. These data suggested a high propensity between PNI and lymphovascular invasion. In addition, positive PNI seem to be more frequent in the upper and middle 1/3 tumours. The distribution of large autonomic nerves was relatively broader in the upper and middle 1/3 tumours, so cancer cells could easily spread through the gap between the nerves and tissue.7 Also, tumour size was significantly larger, tumour invasion was significantly deeper and tumour stage was more advanced in patients with positive PNI than those with negative PNI. These findings seem to further support a fact that the patients with positive PNI had a poorer survival outcome than those without PNI. According to the study of Aurello et al,8 the 5-year DFS rate of positive PNI and negative PNI patients was 22% and 69%, respectively. Similarly, Bilici et al reported that the median OS of positive PNI patients was 28.1 months, which was significantly worse than that of negative PNI patients (64.9 months, p=0.001).15
Although PNI was significantly associated with poorer survival outcome in GC patients, some studies reported that it could not provide more additional prognostic information than other well-known clinicopathological factors.16 20 Zhou et al found that PNI was not an independent prognostic factor after classical and well-established clinicopathological variables were adjusted in Cox multivariate regression analysis.9 Similar results were reported by De Franco et al in the multivariate analysis. However, further subgroup analysis based on the Lauren histology type indicated that PNI was a valuable prognostic factor for GC patients with intestinal type (HR: 1.99, 95% CI 1.24 to 3.19, p=0.005), but not for those with diffuse type.20 In view of current conflicting results, we extracted adjusted HR and its corresponding 95% CIs for OS and performed a pooled analysis to determine the prognostic significance of PNI. The results indicated that PNI was a predictive factor for survival outcome in GC patients. It has been proved that the depth of tumour invasion (T category) and the number of positive lymph nodes (N category) were two important prognostic factors for GC patients.3 4 The TNM classification system based on these prognostic parameters has been widely used to predict survival outcome of GC patients and guide postoperative treatment strategy. Recently, Jiang et al incorporated PNI into the 7th edition of TNM staging system for GC, and the results indicated that the novel staging system could provide a better prognostic stratification for stage III patients than single TNM staging.7 Given its prognostic significance in GC patients, a thorough follow-up plan and intensive adjuvant chemotherapy should be considered for those patients with positive PNI.
The exact pathological mechanism of PNI in a different type of cancers remains unclear. Some researchers attributed PNI pathogenesis to the close anatomical relationship between the tumour and neural plexus.16 However, the incidence of PNI was not high in rectal cancer patients despite the proximity to the presacral autonomic nerve plexus.27 28 Xia et al found that GC cells not only could promote the proliferation of neural progenitor cells but also enhance neurite elongation and branching of postmitotic neural cells in cancer-neural cell coculture condition in vitro.29 A large amount of cell adhesion molecule and chemokines secreted from tumour cells may recruit and facilitate migration and invasion of cancer cells toward surrounding neuronal axons and neural tissue.30 Lv et al reported that the expression level of CX3CL1 chemokines and its receptor CX3CR1 was significantly higher in GC patients with positive PNI than in those with negative PNI, showing a significant association between CX3CL1/CX3CR1 expression and PNI in gastric carcinoma patients.31
PNI was more frequent in pancreatic cancer and biliary tract carcinoma than other malignant tumours (approximately 80%–88% in pancreatic cancer and biliary tract carcinoma).32 33 According to the previous reports, the incidence of PNI in surgically resected specimens of GC patients varied from 6.9% to 75.6%.8 15 20 22 The significant variation among different studies for the incidence of PNI may be partially attributed to detection technology and the experience of pathologists in addition to patient selection. In general, most of research institutions evaluated the presence of PNI using H&E staining. However, nerve fibres in histological sections were not always clearly identified by pathologists due to the invasion and severe damage by tumour cells.9 Previous studies have demonstrated that the detection rate of PNI could be significantly improved by immunohistochemical staining using S-100 or laminin.9 34 Zhou et al reported that the positive rate of PNI was 56.9% (91/160) in GC patients using S-100 stain, but only 38.1% of patients (61/160) were identified as positive PNI when the tissue sections were evaluated by HE staining.9 Despite improved visual detection of nerves, the identification of PNI was still not an easy task for pathologists. It is very difficult to distinguish tumour cells from inflammatory cells even in S-100 stained slides, especially when these cells were small and distributed in a diffuse way. Recently, Zhou et al used double immunohistochemical staining to label nerve fibres and epithelial-derived tumour cells for evaluating the PNI status in GC patients. The results indicated that the positive rate of PNI was 65.0% (104/160) in the double staining group and 56.9% (91/160) in single S-100 staining group, respectively.9 Compared with double staining, 19 false-negative and six false-positive cases were observed in single S-100 stained slides, especially for diffuse histology type.9 In the future, detection methods for PNI need to be further improved for better diagnostic accuracy and efficiency.
In addition, the prognostic value of PNI may be partly depended on how it is evaluated by pathologists. Most of the current studies evaluated the PNI status in submucosa or muscularis propria (mural perineural invasion, mPNI), but few studies focus their attention on extramural perineural invasion (ePNI, invasion of the nerve plexus beyond muscularis propria). Recently, España-Ferrufino et al subdivided 73 pT3-T4a stage GC patients with positive PNI into mPNI and ePNI group, and they found that the patients with ePNI had a poorer disease-specific survival than those with mPNI.35 However, these findings need to be confirmed by more studies.
The present study had some limitations. First, all included studies were retrospective in nature and selection bias should be considered. Second, different detection methods and diagnostic criteria for PNI as well as heterogeneous patient populations in included studies could result in a significant variation for the positive rate of PNI, which could further affect the prognostic evaluation.
In summary, the positive rate of PNI was high in surgically resected specimens of GC patients. Our results demonstrated that PNI was significantly associated with a series of unfavourable clinicopathological factors including undifferentiated histology type, diffuse type, larger tumour size, lymphovascular invasion, lymph node metastasis and advanced tumour stage. PNI was a predictive factor for survival outcome in GC patients, and it may provide additional prognostic information for these patients.
Take home messages
The positive rate of perineural invasion (PNI) was very high in surgically resected specimens of gastric cancer (GC) patients.
The patients with positive PNI were more likely to have more aggressive oncological features than those without PNI.
PNI was a predictive factor for survival outcome in GC patients, and it may provide additional prognostic information for these patients.
In view of its prognostic significance in gastric cancer patients, a thorough follow-up plan and intensive adjuvant chemotherapy should be considered for those patients with positive PNI.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Handling editor Runjan Chetty.
Contributors BZ contributed to study design, data analysis, data interpretation and drafted this manuscript; WL and DM contributed to data analysis, data interpretation and performed statistical analysis; RL contributed to the preparation of this manuscript; SB contributed to the acquisition of data and clinicopathologic information; BH and JL critically reviewed this manuscript and provided important guidance.
Funding This work was supported by the Natural Science Foundation of Liaoning Province (No. 20180530026).
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; internally peer reviewed.
Data availability statement Data are available upon reasonable request.