To Revise or Not Revise? Isolated Margin Positivity in Localized Pancreatic Ductal Adenocarcinoma

Background The study determined the proportion of patients with pancreatic adenocarcinoma (PDAC) who had margin-positive disease and no other adverse pathologic findings (APF) using institutional and administrative datasets. Methods Patients with clinical stage I or II PDAC in the National Cancer Database (NCDB 2010–2020) and those who underwent pancreatectomy at the authors’ institution (2010–2021) were identified. Isolated margin positivity (IMP) was defined as a positive surgical margin with no APF (negative nodes, no lymphovascular/perineural invasion). Results The study included 225 patients from the authors’ institution and 23,598 patients from the NCDB. The margin-positive rates were 21.8% and 20.3%, and the IMP rates were 0.4% and 0.5%, respectively. In the institutional cohort, 68.4% of the patients had recurrence, and most of the patients (65.6%) had distant recurrences. The median recurrence-free survival (RFS) was 63.3 months for no APF, not reached for IMP, 14.8 months for negative margins & 1 APF, 20.3 months for positive margins & 2 APFs, and 12.9 months with all APF positive. The patients in the NCDB with IMP had a lower median OS than the patients with no APF (20.5 vs 390 months), but a higher median OS than those with margin positivity plus 1 APF (20.5 vs 18.0 months) or all those with APF positivity (20.5 vs 15.4 months). Based on institutional rates of IMP, any margin positivity, neck margin positivity (NMP), and no APF, the fraction of patients who might benefit from neck margin revision was 1 in 100,000, and those likely to benefit from any margin revision was 1 in 18,500. In the NCDB, those estimated to derive potential benefit from margin revision was 1 in 25,000. Conclusions Isolated margin positivity in resected PDAC is rare, and most patients experience distant recurrence. Revision of IMP appears unlikely to confer benefit to most patients. Supplementary Information The online version contains supplementary material available at 10.1245/s10434-024-15616-y.

3][4][5] Despite major improvements in surgical management for PDAC and a postoperative mortality rate lower than 5% in specialized centers, patients still encounter a disproportionately high risk of postoperative morbidity and disease recurrence. 62][13][14][15][16][17][18] Several factors can contribute to a positive surgical margin, including definitions based on margin distance, size, and location To Revise or Not Revise?… of the tumor; the presence of lymph node metastasis; and surgical technique. 7,9,10,18,19etrospective studies of intraoperative margin revision in patients with PDAC have reported heterogeneous outcomes.][27] Recently, Malleo et al. 28 reported on 671 patients who underwent pancreatoduodenectomy after neoadjuvant therapy for PDAC with frozen section analysis of neck margin.The authors demonstrated that conversion of an initially positive pancreatic neck margin by additional resection was not associated with oncologic benefits.However, that study included only patients who received neoadjuvant therapy followed by pancreatoduodenectomy and did not quantitatively contribute to the argument concerning the futility of neck margin revision.
In this study, we performed a hybrid analysis using institutional and administrative databases to investigate the frequency of isolated margin positivity (IMP) among patients with localized PDAC who underwent pancreatectomy (pancreatoduodenectomy, distal pancreatectomy, or total pancreatectomy).Our primary aim was to determine the proportion of patients with IMP.Our secondary aims were to determine the fraction of patients who would potentially benefit from revising surgical margins and to examine the association of IMP with OS and RFS.

Institutional Assurances
The Institutional Review Board (IRB) approved this retrospective analysis (STUDY20180966).Data were collected using REDCap, made available by the Clinical and Translational Science Award grant UL1TR002548.Given the retrospective nature of the analysis and minimal harm to the study patients, the need for informed consent was waived.Retrospective analyses of public, deidentified datasets are exempt from IRB review.

Patient Cohort
We identified patients who underwent a pancreatectomy at our institution between 2010 and 2021.The study enrolled patients with a histologically confirmed diagnosis of PDAC.
In the National Cancer Database (NCDB), patients with PDAC were identified using International Classification of Diseases for Oncology (third edition) histology codes 8140 and 8500.The study included only patients who underwent pancreatectomy (pancreatoduodenectomy, distal pancreatectomy, total pancreatectomy) between 2010 and 2021.The study excluded patients with clinical stage III or IV PDAC at presentation and those with unknown clinical stage.Furthermore, patients were excluded if they had unknown pathologic findings, including Patients with unknown margin status were excluded, nodal positivity, and lymphovascular invasion.Patients also were excluded if they had unknown status for receipt of systemic chemotherapy, radiation therapy, immunotherapy, or palliative care.Finally, patients were excluded if they had unknown sociodemographic characteristics, including race, median income, education, health insurance, facility type, and distance to treating facility.Figure S1 shows the flowchart for selection of the NCDB cohort.
The NCDB contains de-identified patient data and is maintained by the American College of Surgeons and the American Cancer Society.These societies are not responsible for the accuracy of the data presented or the statistical analyses performed.The conclusions reported in this study represent the authors' interpretation.

Variables of Interest
The clinicodemographic characteristics included in the institutional dataset were age at surgery, sex, race, clinical stage, pancreatectomy type (pancreatoduodenectomy, distal pancreatectomy, total pancreatectomy), and recurrence patterns (none, local, distant).The treatment-related characteristics included receipt of preoperative therapy, type of preoperative therapy if received (systemic chemotherapy, radiation therapy, or both), chemotherapy regimen if received, and receipt of adjuvant chemotherapy.The pathologic data included surgical margin status, number of positive margins, location of positive margins, lymph node positivity, lymphovascular invasion, and perineural invasion.
The clinicodemographic characteristics in the NCDB were age, sex, race, clinical stage, and pancreatectomy type.The treatment-related characteristics included receipt of systemic chemotherapy (if received, single-agent vs multi-agent and neoadjuvant vs adjuvant chemotherapy) and administration of radiation therapy (if received, neoadjuvant vs adjuvant radiation therapy).The pathologic data included surgical margin status, lymph node positivity, and lymphovascular invasion.The NCDB does not capture perineural invasion or number and location of positive margins.

Outcome Measures and Definitions
The primary outcome of interest was the proportion of patients with IMP in the institutional and NCDB datasets.Isolated margin positivity was defined as margin positivity (MP) and absence of other adverse pathologic findings, including lymph node metastases, perineural invasion, and lymphovascular invasion (institutional data) or MP and absence of lymph node metastases or lymphovascular invasion (NCDB data).0][31][32] The average rate of perineural invasion was calculated using the total number of patients with perineural invasion divided by the total number of patients reported in all studies.The secondary outcomes included associations between adverse pathologic findings (all negative, IMP, MP with additional adverse pathologic features, all adverse pathologic features positive), OS, and RFS.

Clinicopathologic Biostatistical Modeling
In consistency with previous reporting, 33 a pathologic biostatistical model was devised to estimate the fraction of patients undergoing pancreatectomy who might derive a survival benefit (FDPB) from revising a positive neck margin to negative, thus converting a patient with IMP to having no adverse pathologic findings (NAPF).We estimated FDPB by the following equation: where IMP is the fraction of patients with IMP, NMP is the fraction of patients with neck margin positivity (NMP), and NAPF is the fraction of patients with no adverse pathologic findings.
The FDPB # in the institutional cohort was calculated using MP rather than NMP, where MP is the fraction of patients with positive surgical margins.Therefore, FDPB # = IMP × MP × NAPF.
Similarly, the FDPB* in the NCDB cohort (location of positive margins is not captured) was calculated using MP.Because the NCDB does not also capture perineural invasion, we corrected the rates of IMP and NAPF by multiplying them by the average rate of perineural invasion obtained from the current and previous studies 27,[29][30][31][32] to calculate the corrected IMP (i.e., IMP*) and NAPF (i.e., NAPF*).Thus, FDPB* = MP* × MP × NAPF*.

Statistical Analysis
Patient characteristics were described as median (interquartile range [IQR]) and compared using the Wilcoxon rank-sum test if they were continuous and non-normally distributed.Categorical variables were described as frequencies and percentages and compared using Pearson's chi-square test.
In both the institutional and NCDB cohorts, OS was defined as the time from diagnosis to death or last followup visit, and RFS was defined as the time from surgery to recurrence in the institutional cohort only because the NCDB does not provide data on RFS.The median OS and RFS were calculated using the Kaplan-Meier method, and comparisons among pathologic categories were performed using the log-rank test.
Multivariable Cox proportional hazards regression analyses were used to examine the association between IMP and OS in the NCDB cohort.The factors included in all multivariable analyses were age, sex, race, education, median income, primary insurance payer, Charlson-Deyo Comorbidity Index, clinical stage, pancreatectomy type, facility type, facility volume (determined based on the total annual hepatopancreatobiliary malignancy case volume, operative and nonoperative, as previously described [34][35][36] ), distance to treating facility, receipt of preoperative therapy, and year of diagnosis.This model was determined a priori based on clinical relevance.All multivariable models were clustered on facility identifiers to account for intra-class correlation for patients nested within the same hospital.Stata 17.0 (Stata-Corp, College Station, TX, USA) was used for all statistical analyses.
Only 1 patient (0.4%) in the institutional cohort had IMP compared with 744 patients (3.2%) in the NCDB cohort (Tables 1 and 2).After integration of the average perineural invasion rate (83.7%) into the calculation (Table S3), the IMP* rate in the NCDB cohort was 0.5%.Stratification of the patients by treatment sequence showed that 171 patients (3.8%) in the NCDB cohort who received neoadjuvant therapy had IMP compared with 573 patients (3.0%) who underwent upfront surgery (Table 3).The only patient in the institutional cohort who had IMP received neoadjuvant therapy.
In the institutional cohort, the median OS was not reached for IMP or NAPF, but was 32.6 months for negative margins and one adverse pathologic finding,
We then used our biostatistical models to estimate the fraction of patients who might derive benefit from margin revision.In the institutional cohort, the FDPB was calculated using the following values: IMP, 0.4%; NMP, 4.0%; and NAPF, 6.2%.Thus, FDPB was 0.4% × 4.0% × 6.2%.Therefore, 1 in 100,000 patients might derive a potential survival benefit from revision of a positive pancreatic neck margin (Table 4A).If MP (rather than NMP) were to be included in the equation, FDPB # would equal 0.4% × 21.8% × 6.2%.Therefore, approximately 1 in 18,500 patients might benefit from revision of any positive margin.
Among the patients who received neoadjuvant therapy, the fraction of patients who potentially would benefit from revision of a positive neck margin was 1.8% × 7.1% × 8.9%, which is approximately 1 in 9000 patients.
In the NCDB cohort, the FDPB* was calculated using the following values: IMP*, 0.5%; MP, 20.3%; and NAPF*, 3.9%.Thus, FDPB* was 0.5% × 20.3% × 3.9%, which is approximately 1 in 25,000 patients (Table 4B), similar to our findings on the institutional level.In addition, the fraction of patients who potentially would benefit from revision of surgical margins was approximately 1 in 18,000 patients receiving neoadjuvant therapy and 1 in 28,000 patients undergoing upfront surgery.

DISCUSSION
This study demonstrated that IMP is rare, experienced by less than 1% of our institutional cohort and the estimated NCDB cohort.The fractions of patients in the institutional cohort who would potentially benefit from revision of all margins or neck margins were approximately 1 in 18,500 and 1 in 100,000 patients, respectively.Similarly, in the NCDB cohort, the fraction of patients potentially benefiting from revision of surgical margins was 1 in 25,000 patients.Finally, the patients with IMP had an associated decrease in OS versus the patients with NAPF, but had an associated increase in OS versus the patients with MP and positive lymph nodes or lymphovascular invasion or those with all pathologic findings positive.
Prior studies have shown a discrepancy between the results of intraoperative frozen section analysis and the final histologic report. 22,37Patients who undergo operative intervention to revise positive resection margins have higher associated complication rates. 37,38The pancreatic neck margin is the most straightforward margin to remediate intraoperatively. 39However, the optimal management of a microscopically positive pancreatic neck margin on intraoperative frozen-section pathologic examination is a matter of debate.2][43][44] Our study adds quantitatively to this philosophical argument because we demonstrated that IMP is an extremely rare event.Our biostatistical model predicts that 100,000 patients with a positive pancreatic neck margin would need to be revised to potentially benefit one patient.This seems to be impractical.
Although much of the debate centers on the pancreatic neck margin, other margins include the proximal (duodenal/gastric) margin, the distal jejunal margin, the bile duct margin, and the SMA/uncinate margin.These margins are infrequently assessed on intraoperative frozen section and rarely revised.It seems to us that the amount of general attention given to the pancreatic neck margin while ignoring the other potentially revisable margins and non-modifiable pathologic factors is unwarranted.Ghaneh et al. 13 performed a post hoc analysis of the European Study Group for Pancreatic Cancer (ESPAC)-3 randomized controlled trial data and included 1151 patients with resected PDAC.The authors found that R1 neck and posterior/retroperitoneal and medial/ uncinate margins were separately and independently associated with worse OS and RFS.Furthermore, the presence of more than one positive margin was associated with worse OS and RFS, with an increasing number of margin-positive sites associated with progressively worse OS.
The findings of our study must be interpreted with consideration of some limitations.We did not distinguish between R1 less than 1 mm and R1-direct because this information was not available in all pathologic reports.Nonetheless, this classification has previously been accepted by the American Joint Committee on Cancer and is becoming more relevant in terms of patient outcomes. 45However, this classification system is not routinely used in intraoperative frozen section analysis.
Another possible limitation was the very small number of patients with IMP in the institutional cohort, which precluded a more formal survival analysis.We attempted to address this by using the NCDB for this purpose.However, the NCDB does not capture perineural invasion or number and site of surgical margins involved.
A further limitation of this study could have been that its definition of IMP was not validated.It would be ideal to validate this definition in an institutional database with recurrence data to identify variables that associate the most with local-only recurrence.We could not perform such validation given the small number of patients with local-only recurrence in our institutional cohort and that the NCDB does not capture local or distant recurrence data.External validation was beyond the scope of the current study, but we believe our current investigation sets the groundwork for future collaborative multi-institution or consortium-based studies to further this question in greater detail.We also tried to account for the absence of perineural invasion in the NCDB by incorporating the average rate of perineural invasion from the current study and other large published cohorts.However, this may not have been a perfect alternative.Despite these limitations, our findings quantitatively illustrate the rarity of IMP among patients with resected PDAC and the likely futility of intraoperative revision of a positive neck margin.

CONCLUSIONS
Developing IMP in patients with resected PDAC is rare.Patients with IMP were more likely to have a better OS than patients with all pathologic findings positive or at least one adverse pathologic finding but worse OS than patients with no adverse pathologic findings.The fraction of patients who would potentially benefit from revision of positive neck margins is very small, suggesting that routine practice likely is futile.

TABLE 1
Clinicodemographic characteristics of the patients in the institutional cohort IMP, Isolated margin positivity; IQR, Interquartile range; FU, Fluorouracil a The denominator denotes patients who had cancer recurrence b The denominator denotes patients who received neoadjuvant chemotherapy c The denominator denotes patients who received adjuvant chemotherapy 31.1 months for MP and one or more adverse pathologic findings, and 15.6 months for all positive pathologic findings (Fig.1A).The median RFS was not reached for IMP, 63.3 months for NAPF, 14.8 months for negative margins and one adverse pathologic finding, 20.3 months for MP and one or more adverse pathologic findings,

TABLE 2
Clinicodemographic characteristics of the patients within the NCDB cohort

TABLE 4
Summary of the calculated number of patients who would potentially benefit from revising surgical margins IMP, Isolated margin positivity; NMP, Neck margin positivity; MP, Margin positivity; NAPF, Rate of no adverse pathologic findings; IMP*, Isolated margin positivity rate multiplied by average rate of perineural invasion; NAPF*, No adverse pathologic findings multiplied by average rate of perineural invasion; FDPB, Fraction of patients who might derive a potential benefit after revising neck margin calculated as FDPB = IMP × NMP × NAPF; FDPB # , Fraction of patients who may derive a potential benefit after revising all surgical margins calculated as FDPB # = IMP × MP × NAPF; FDPB*, Fraction of patients who may derive a potential benefit after revising all surgical margins calculated as FDPB* = IMP* × MP × NAPF*; NCDB, National Cancer Database