Perineural Invasion in Pancreatic Ductal Adenocarcinoma (PDAC): A Saboteur of Curative Intended Therapies?

(1) Background: Perineural invasion (PNI) is a common characteristic of pancreatic ductal adenocarcinoma (PDAC) and is present in most resection margins. We hypothesized that curative pancreatic tumor resection with long-term survival could only be achieved in PNI-negative patients. (2) Material and Methods: A retrospective investigation of PDAC patients who underwent curative-intended surgery during the period 2008 to 2019 was performed at our institution. (3) Results: We identified 571 of 660 (86.5%) resected patients with well-annotated reports and complete datasets. Of those, 531 patients (93%) exhibited tumors with perineural invasion (Pn1), while 40 (7%) were negative for PNI (Pn0). The majority of patients in the Pn1 group presented advanced tumor stage and positive lymph node infiltration. Patients in the Pn0 group showed an improved disease-free and long-term survival compared to the Pn1 group (p < 0.001). Subgroup analysis of all R0-resected patients indicated improved long-term survival and disease-free survival of R0 Pn0 patients when compared to R0 Pn1 patients (p < 0.001). (4) Conclusion: Our study confirmed that Pn0 improves the long-term survival of PDAC-resected cancer patients. Furthermore, PNI significantly challenges the long-term survival of formally curative (R0) resected patients. We provide new insights into the dynamics of PNI in pancreatic cancer patients which are needed to define subgroups of patients for risk stratification and multimodal treatment strategies.


Introduction
Across stages, pancreatic duct adenocarcinoma (PDAC) has overall poor prognosis, with 5-year survival rates of approximately 10% [1]. The incidence of PDAC is rising worldwide, making it the second predicted leading cause of cancer-related deaths until 2 of 13 2030 [2]. When patients present with symptoms at the time of diagnosis, the infiltration of surrounding tissues or metastasis to distant organs has likely already occurred. Only a small number of patients are amenable to curative treatment, which currently requires a combination of neoadjuvant-/adjuvant chemotherapy and surgical resection [3]. The key for appropriate treatment allocation prior to the operation is the clinical assessment of local and systemic tumor extension by cross-sectional imaging in accordance with current consensus guidelines [3][4][5]. Following curative resection, the standard multimodal treatment protocol depends on the performance status and tumor biology but is commonly based on six months of adjuvant chemotherapy with either modified FOLFIRINOX, gemcitabine/capecitabine combination or gemcitabine alone [6][7][8][9]. Eligible patients are increasingly receiving preoperative treatment with the idea of early control of occult dissemination [10]. Major benefits possibly lie in the higher completion rates of systemic therapy, in downstaging of borderline-resectable tumors and higher rates of negative margins [11][12][13]. Still, the majority of patients eventually relapse and larger, long-term interventional studies are needed to ultimately discriminate a significant effect of neoadjuvant treatment on overall survival and recurrence rates.
Reasons for treatment failure and high recurrence rates are manifold. Recurrence patterns in patients most often show systemic dissemination (>75%), suggesting micrometastatic disease at time of operation [14,15]. However, up to 24% have primarily local recurrence, which often correlates with margin status or perivascular invasion. We understood that accurate histopathological assessment has a major impact on overall clinical outcome. Tumor size is consistently relevant, as well as nodal status (including number of nodes) and resection margin [16][17][18][19]. Additional information on margin status, such as the microscopic assessment of perivascular, lymphatic and perineural invasion, is regularly included, while their exact prognostic relevance remains controversial [20][21][22][23].
Perineural invasion (PNI) is a characteristic feature of PDAC and is present in 70% to 95% of resected pancreatic carcinomas [20]. Compared to other solid malignancies, PNI has the highest prevalence in PDAC [24]. PNI can also be detected in the early stages of pancreatic carcinoma, suggesting that it may be an early event in cancer progression [25,26]. Nerval invasion is a complex process driven by reciprocal interaction between tumor cells and nerves, resulting in axonal growth and tumor cell invasion [27][28][29]. The process is orchestrated by several signaling molecules secreted by neural cells and tumor cells (reviewed in Ref. [30]). With accumulated knowledge of the clinical impact on patient outcomes, perineural invasion has been included in the seventh edition of the TNM classification, defined as either present or absent [31].
At this point, many clinical studies have examined the prognostic value of the PNI of resected PDAC, both indicating that PNI is associated with early tumor recurrence and decreased overall survival [20,26]. To what extent PNI diminishes the long-term survival of curative (R0) resected patients in combination with current (neo-) adjuvant treatment protocols has only been partially examined. We hypothesized that, in the setting of complete cancer eradication through surgery, long-term survival can only be achieved in PNI-negative patients. As a high-volume pancreatic cancer center, we aim to explore the impact of PNI in a large cohort of PDAC patients who have undergone curative intended surgery.

Study Design
Patients undergoing curative-intended surgical resection for PDAC at the Department of Surgery, Campus Charité Mitte|Campus Virchow, Charité-Universitätsmedizin, Berlin, Germany, between January 2008 and December 2019 were included. Ethical approval was obtained from the institutional ethics committee (EA1/208/12; EA1/188/17). Data was retrospectively analyzed using a prospectively updated database. Prior to surgery, all patients received radiographic staging (pancreas protocol CT-scan) for evaluation of resectability according to current consensus guidelines [4,5,32]. Selected patients with borderline or locally advanced tumors received preoperative chemotherapy. Subsequent surgical procedures included the Kausch-Whipple operation, pylorus-preserving pancreaticoduodenectomy (PPPD), distal pancreatectomy (DP) or total pancreatectomy. Patients with incomplete medical history documentation, R-status, PNI status, tumor stage as well as patients with in-hospital mortality (<30 days survival) were excluded.

Statistical Analysis
For statistical analysis, the statistical software R (The R Foundation, version 4.0.0, Vienna, Austria) was used. Continuous variables are presented as means (range, standard deviation). For categorical variables, simple comparisons between groups were performed based on Fisher's exact test. Kaplan-Meier curves were estimated for analyses of timeto-event outcomes, such as overall survival and disease-free survival, and the log-rank test was used for subgroup comparison. To control for confounders, Cox proportional hazards regression was applied. Multivariable analysis with PNI as outcome variable was performed via multiple logistic regression. The confidence level was set to 95% for all analyses.

Pn0 Status Is Associated with Improved Disease-Free and Long-Term Survival
Studying long-term survivors (LTS with survival > 5 years) in our study cohorts, we discriminated a significantly increased number of LTS in the Pn0 group compared to Pn1 patients (p = 0.003). The Kaplan-Meier analysis confirmed that patients in the Pn0 group show dramatically improved overall survival compared to patients in the Pn1 group. Median overall survival was 64.9 months in the Pn0 group compared to 18.1 months in the Pn1 group (p < 0.001) ( Figure 1A). Patients in the Pn0 group also presented with improved disease-free survival compared to patients in the Pn1 group ( Figure 1B). Median disease-free survival was 26 months in the Pn0 group compared to 12.9 months in the Pn1 group (p < 0.001). Well established variables impacting the patients' survival, such as resection margin and N-stage, also negatively correlated with overall-and diseasefree survival in our cohort (Supplementary Figures S1 and S2). We therefore performed Cox proportional-hazards regression model for analysis of PDAC patients' characteristics to identify independent variables impacting the DFS and OS (Table 4) Figure 2B). R0 Pn0 patients also presented with improved overall survival compared to R0 Pn1 patients (Figure 2A). Median overall survival was 20.95 months in the R0 Pn1 group, whereas median survival in the R0 Pn0 group was not reached. We calculated a significantly increased risk for R0 Pn1 patients (HR: 3.21; p < 0.001) compared to R0 Pn0 patients.    Supplementary Table S1.

Postoperative Chemotherapy Improves OS of Pn1 Patients
According to the Kaplan-Meier analysis, patients with Pn1 tumors receiving postoperative chemotherapy showed significantly improved overall survival. Median overall survival was 19.4 months for patients with Pn1 tumors with postoperative chemotherapy compared to 14.1 months for patients with Pn1 tumors without adjuvant chemotherapy (HR: 0.74; p = 0.009, Figure 3A). Median disease-free survival was 14.2 months for patients with Pn1 tumors with postoperative chemotherapy compared to 12.9 months for patients   Supplementary Table S1.

Postoperative Chemotherapy Improves OS of Pn1 Patients
According to the Kaplan-Meier analysis, patients with Pn1 tumors receiving postoperative chemotherapy showed significantly improved overall survival. Median overall survival was 19.4 months for patients with Pn1 tumors with postoperative chemotherapy compared to 14.1 months for patients with Pn1 tumors without adjuvant chemotherapy (HR: 0.74; p = 0.009, Figure 3A). Median disease-free survival was 14.2 months for patients with Pn1 tumors with postoperative chemotherapy compared to 12.9 months for patients with Pn1 tumors without postoperative chemotherapy, not reaching significance ( Figure 3B). Within the first 24 months after curative surgery, however, improved DFS was apparent for patients receiving postoperative chemotherapy. There were no significant differences in OS or DFS for patients with Pn0 tumors receiving postoperative chemotherapy (Supplementary Figure S3).
According to the Kaplan-Meier analysis, patients with Pn1 tumors receiving postoperative chemotherapy showed significantly improved overall survival. Median overall survival was 19.4 months for patients with Pn1 tumors with postoperative chemotherapy compared to 14.1 months for patients with Pn1 tumors without adjuvant chemotherapy (HR: 0.74; p = 0.009, Figure 3A). Median disease-free survival was 14.2 months for patients with Pn1 tumors with postoperative chemotherapy compared to 12.9 months for patients with Pn1 tumors without postoperative chemotherapy, not reaching significance ( Figure  3B). Within the first 24 months after curative surgery, however, improved DFS was apparent for patients receiving postoperative chemotherapy. There were no significant differences in OS or DFS for patients with Pn0 tumors receiving postoperative chemotherapy (Supplementary Figure S3).  Subgroup analysis of patients with R0 Pn1 tumors indicated that postoperative chemotherapy improved OS. Median overall survival was 21.8 months for patients with R0 Pn1 tumors with adjuvant chemotherapy compared to 16.6 months for patients with R0 Pn1 tumors without adjuvant chemotherapy, not reaching statistical significance (HR: 0.79; p = 0.11, Figure 4A). Median disease-free survival was 16.6 months for patients with R0 Pn1 tumors with postoperative chemotherapy compared to 14.1 months for patients with R0 Pn1 tumors without postoperative chemotherapy ( Figure 4B). Again, focusing on first 24 months, improved DFS for R0 Pn1 patients receiving chemotherapy was detected without statistical significance.
tumors with adjuvant chemotherapy compared to 16.6 months for patients with R0 Pn1 tumors without adjuvant chemotherapy, not reaching statistical significance (HR: 0.79; p = 0.11, Figure 4A). Median disease-free survival was 16.6 months for patients with R0 Pn1 tumors with postoperative chemotherapy compared to 14.1 months for patients with R0 Pn1 tumors without postoperative chemotherapy ( Figure 4B). Again, focusing on first 24 months, improved DFS for R0 Pn1 patients receiving chemotherapy was detected without statistical significance.

Discussion
The central aim of this study was to investigate the impact of perineural invasion in curatively treated PDAC patients at our institution. PDAC has the highest rates of PNI compared to other solid malignancies and varies across studies from 60 to 95% [20,24]. In our cohort, 93% of all resected PDAC patients showed histologically confirmed PNI.
Our data indicates that PNI generally correlates with advanced tumor stage. Most patients with Pn1 harbored pT2 and pT3 tumors, while 30% of Pn0 tumors were small (pT1). Patients with Pn1 tumors had significantly higher rates of lymph node metastasis and lymphatic invasion than those with Pn0 tumors. Moreover, as underlined by previous reports, the perineural sheath likely serves as a consecutive route for tumor spread to surrounding tissues and lymph nodes [33,34].
In agreement with the findings of the multicenter study of Crippa S et al. characterizing the clinical impact of PNI in a large cohort, PDAC patients with Pn1 display impaired DFS and OS compared to patients with Pn0 PDAC [26]. This appears particularly relevant in the setting of tumor-free resection margins (R0 Pn1) postsurgery, where Hazard increases up to 3.2 compared to R0 Pn0 tumors. R0 resection was achieved in 62.9% of the entire study cohort, while PNI was detected in the majority of R0 tumors (91%). Multivariate Cox regression analyses confirmed that PNI was an independent predictor of worse DFS and OS, to similar extent as with positive margin (R1) and adjuvant chemotherapy. In consideration of recent studies and meta-analyses about the clinical impact of PNI, this subcohort requests particular attention for improving adjuvant treatment algorithms of affected individuals [20,26,35]. We therefore sought to investigate how PNI

Discussion
The central aim of this study was to investigate the impact of perineural invasion in curatively treated PDAC patients at our institution. PDAC has the highest rates of PNI compared to other solid malignancies and varies across studies from 60 to 95% [20,24]. In our cohort, 93% of all resected PDAC patients showed histologically confirmed PNI.
Our data indicates that PNI generally correlates with advanced tumor stage. Most patients with Pn1 harbored pT2 and pT3 tumors, while 30% of Pn0 tumors were small (pT1). Patients with Pn1 tumors had significantly higher rates of lymph node metastasis and lymphatic invasion than those with Pn0 tumors. Moreover, as underlined by previous reports, the perineural sheath likely serves as a consecutive route for tumor spread to surrounding tissues and lymph nodes [33,34].
In agreement with the findings of the multicenter study of Crippa S et al. characterizing the clinical impact of PNI in a large cohort, PDAC patients with Pn1 display impaired DFS and OS compared to patients with Pn0 PDAC [26]. This appears particularly relevant in the setting of tumor-free resection margins (R0 Pn1) postsurgery, where Hazard increases up to 3.2 compared to R0 Pn0 tumors. R0 resection was achieved in 62.9% of the entire study cohort, while PNI was detected in the majority of R0 tumors (91%). Multivariate Cox regression analyses confirmed that PNI was an independent predictor of worse DFS and OS, to similar extent as with positive margin (R1) and adjuvant chemotherapy. In consideration of recent studies and meta-analyses about the clinical impact of PNI, this subcohort requests particular attention for improving adjuvant treatment algorithms of affected individuals [20,26,35]. We therefore sought to investigate how PNI impacted the patients' survival in correlation with current (neo-)adjuvant treatment protocols.
Current standard treatment for resectable PDACs usually includes surgery followed by adjuvant chemotherapy [3]. In our institutional study cohort, low adherence of patients to adjuvant treatment protocols (61.2%) is likely due to decentralized oncologic patient registries (Germany), which challenges the collection of complete information on administered systemic therapies after surgery. Still, we discriminated that adjuvant chemotherapy improved the clinical outcome of PDAC patients with Pn1, significantly impacting their OS and DFS. Subgroup analyses confirmed effects on patients with negative resection margins (R0 Pn1), whereas patients without perineural invasion (Pn0) did not significantly benefit from adjuvant treatments. Most critically, the prevalence of long-term survivors was significantly increased in patients with Pn0 compared to Pn1 (10% vs. 2,8%). Altogether, our study not only adds to the existing evidence of PNI as an independent negative prognosticator on patients' survival after surgery, but demonstrates that long-term survival is mainly achieved in patients without histological signs of perineural invasion [20,26]. It still remains to be evaluated whether second-line systemic treatments (nonresponders) have an impact on clinical outcome with respect to perineural invasion status [36].
Considering current trends in the oncologic treatment of PDAC patients, we also investigated the effects of preoperative treatment on survival regarding PNI status. Currently, clinical decision making is based on preoperative imaging and multidisciplinary assessment, assigning patients to upfront surgery or systemic therapies [37]. During the entire study period, 11.6% of the patients received neoadjuvant treatment. There were no significant differences in the prevalence of PNI between preoperative treatment compared to surgery upfront. Similar results were also reported by recent studies and have been explained by an observational bias, which included patient cohorts of the "pre-neoadjuvant-" and "pre-FOLFIRINOX" era, a bias we have to consider in our study as well [26,38]. However, some studies were able to discriminate lower rates of PNI in patients with neoadjuvant chemotherapy, underscoring the potential of neoadjuvant protocols in current treatment strategies [39,40].
Nerval invasion is a complex process and the responsible molecular mechanisms have not been sufficiently elucidated. A specific treatment strategy to directly alter the perineural invasion is currently not available. We believe that further research in this field is needed so that PNI can be addressed by targeted treatment modalities which could additionally impact the survival of patients.
We are aware that our study has limitations. As an experienced high-volume pancreatic cancer center, our study cohort includes a large enough dataset to present patients' demographics in line with reports from other centers. However, our retrospective, singlecenter study misses out on other reference centers that could offer locoregional variations in patient cohorts and possible differences of pathologic reporting for perineural invasion. In addition, due to trends with changes in the treatment protocols for PDAC over the past decade, our data likely underlies an observation bias, while total numbers were needed to power our study. We therefore focused on valid and robust statements at the cost of detailed subgroup analyses in an effort to minimize type-two errors.
In conclusion, our study confirms that PNI is a common feature of PDAC and correlates with advanced tumor stage and lymphogenic tumor spread of PDAC. PNI occurs in most curative (R0)-resected pancreatic tumors and diminished long-term survival of formally curative (R0)-resected patients. In turn, Pn0 is associated with improved long-term survival of patients after resection. Based on our study, PNI presents a robust and independent predictor of aggressive tumor biology of curative (R0)-resected PDAC patients. Further studies exploring the dynamics of PNI in pancreatic cancer patients are needed to define subgroups of patients for risk stratification and alternative multimodal treatment strategies.   Institutional Review Board Statement: Ethical approval was obtained from the Charité ethics committee (EA1/208/12; EA1/188/17).
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: Not applicable.