Society: SSAT

LUNCH AND TRAINEE JEOPARDY! 12:00 PM - 1:00 PM LEADERSHIP FORUM, NETWORKING SESSION AND ROUNDTABLES 4:00 PM - 5:30 PM KEYNOTE ADDRESS: FUTURE SCENARIO PLANNING TO PREDICT THE FUTURE OF SURGERY 4:00 PM - 4:30 PM

Introduction
The role of lymph node (LN) parameters in pancreatoduodenectomy (PD) for cancer has been mainly investigated in the upfront surgery setting. Yet, due to the impact of neoadjuvant therapy (NAT) on nodal status, these results cannot be directly translated to post-NAT PD. This study aimed to examine LN yields and metastases per anatomical stations and how the extension of LN dissection affects nodal staging in post-NAT PD. Lastly, the prognostic role of LN parameters was investigated.
Methods
An institutional lymphadenectomy protocol was prospectively applied to all post-NAT PDs from June 2013. Lymphadenectomy included stations 5/6/8a-p/12a-b-c-p/13/14a-b/17 and jejunal mesentery LNs. Stations embedded in the PD specimen (13/14/17/jejunal) were defined as first-echelon, those sampled separately (5/6/8/12) as second-echelon. The prognostic impact of LN parameters in N+ patients was evaluated using uni- and multivariable Cox regression. To avoid collinearity, separate multivariable models were designed for each nodal parameter.
Results
Among 288 patients 61% received FOLFIRINOX, 30% Gem-Abraxane. The median number of examined (ELN) and positive LNs (PLN) were 43 and 1, and 185 patients were N+ (64%). The commonest metastatic sites were stations 13 (51%), 14 (34%) and 17 (32%). The overall rates of first and second echelon involvement were 60% and 20%. The median number of ELN and PLN in the first echelon were 29 and 1. The addition of second echelon LNs increased nodal counts by 9 ELN and 0 PLN, resulting in only minor changes in staging.
The median follow-up was 25.1 months, 35.8 in censored cases. At multivariable analysis, second echelon involvement, ≥4 metastatic stations, metastases to station 8 and jejunal mesentery LNs, but not N2 status, were independently associated with survival of N+ patients, along with adjuvant treatment.
The median recurrence-free survival (RFS) was 14.8 months and 176 patients experienced recurrence (71%), among which 41 were local relapses (23%). In N+ patients, nodal echelons, ≥4 metastatic stations and tumor involvement of station 8,14 and jejunal mesentery LNs were independent predictors of RFS, along with Ca 19.9 response, T- and R-status and adjuvant treatment. Distant recurrences incrementally increased with nodal involvement (Figure).
Conclusion
LN metastases most commonly occur in first-echelon LNs, and first-echelon dissection provides an adequate number of ELN for optimal staging. Examining second-echelon LNs does not improve the staging process substantially. Yet, second-echelon involvement is prognostically relevant, as well as metastases to station 8 and jejunal mesentery LNs. These data have potential implications when assessing surgical indication after NAT. Moreover, intraoperative frozen section of station 8 might help decision-making, especially in technically demanding cases or fragile patients.

Type of recurrence stratified by nodal parameters in node-positive patients

Type of recurrence stratified by nodal parameters in node-positive patients

Background:
The concept of “experience” in surgery remains nebulous and multifactorial, encompassing both the surgeon and the institution as pivotal variables. While a surgeon’s career volume seems to be a determinant in improving outcomes for pancreatoduodenectomy (PD), the influence of individual surgeon experience within high-volume institutional settings remains undefined. Within such a framework, the present investigation analyzes the association of cumulative surgeon volume experience with risk-adjusted postoperative outcomes after PDs.

Methods:
A total of 8,189 PDs performed by 82 surgeons at 18 international institutions (median:140 PD/year) were accrued from 2003 to 2020. Surgeon’s cumulative PD volume was categorized in 4 quartiles (≤150, 151 to 285, 286 to 525 and ≥526 PDs). Associations of categorical and continuous variables were analyzed with appropriate univariate tests. Fistula Risk Score (FRS)-stratified performance comparisons of postoperative outcomes across each volume quartile were quantified through multivariable analyses. Next, the same methodology was implemented when considering the ten most impactful scenarios (previously defined as a combination of occurrence and severity) for the development of clinically relevant pancreatic fistula (CR-POPF; n = 2,830 patients).

Results:
Within the overall cohort, 18.7% patients suffered severe complications (Accordion≥3), 14% developed CR-POPF, 4.8% were reoperated upon, and 2.2% expired. Surgeons performed a median of 68.5 career PDs (IQR 21-136), with a median FRS of 4 (IQR 3-5). When compared with those with less experience, the top-quartile surgeons more often operated on intermediate/high FRS cases (73% vs 61%, p <0.001); yet, their performance was associated with significant declines in CR-POPF, severe complications, reoperations, and length of stay (8 vs 9 d), whereas mortality and failure-to-rescue were not affected (Figure). This same outcome profile was accentuated even more when considering the most frequent and impactful FRS scenarios that surgeon encounter. In the overall cohort, risk-adjusted models indicate male gender, increasing age, ASA class and FRS, but not surgeon experience, as predictors for severe complications, failure-to-rescue and mortality. Instead, in advanced fistula risk circumstances, upper-echelon experience demonstrates significant reductions in CR-POPF, reoperations and LOS (Table).

Conclusion:
At specialty institutions, mortality and failure-to-rescue depend primarily on baseline patient and systemic characteristics, while cumulative surgical experience independently impacts pancreatic fistula occurrence and its attendant effects - even more so for riskier PDs. These data suggest an extended learning curve exists for this operation and reinforce the notion that surgeon experience is a key contributor for outcome improvement.

Outcomes of pancreatoduodenectomy based on individual surgeon’s experience.

Outcomes of pancreatoduodenectomy based on individual surgeon’s experience.

Surgical outcomes of pancreatoduodenectomy for the top-quartile experienced surgeons (n = 12).

Surgical outcomes of pancreatoduodenectomy for the top-quartile experienced surgeons (n = 12).

Importance: Guidelines recommend surgical exploration in selected patients with locally advanced pancreatic cancer (LAPC) following induction chemotherapy. However, surgical exploration, has potential drawbacks related to surgical risks and treatment breaks, which apply in particular to patients undergoing exploration without resection (i.e. non-therapeutic laparotomy). Data regarding the impact of non-therapeutic laparotomy for LAPC treated with (m)FOLFIRINOX induction chemotherapy could guide aggresiveness of surgeons for this patient population.
Objective: To assess the incidence and oncologic impact of a non-therapeutic laparotomy for LAPC treated with (m)FOLFIRINOX induction chemotherapy.
Design: Retrospective cohort study
Setting: International multicenter study including patients from 5 referral centers in the USA and The Netherlands (2012-2019).
Participants: Patients diagnosed with pathology-proven LAPC treated with ≥1 cycle (m)FOLFIRINOX (± radiotherapy). Patients with metastatic disease on radiologic (re)staging or clinical deterioration during induction therapy were excluded. Patients undergoing non-therapeutic laparotomy (group A) were compared to those not explored (group B). Patients undergoing resection were assigned to group C.
Main outcomes and measures: 90-day mortality, palliative systemic treatment, and median OS from date of pathology-proven diagnosis.
Results: Overall, 663 patients with LAPC were included, of whom 78 (11.8%) subsequently received a second-line induction chemotherapy after (m)FOLFIRINOX and 413 (66.8%) received radiotherapy. In total, 67 patients (10.1%) were included in group A, 425 patients (64.1%) in group B, and 171 patients (25.8%) in group C. Resection was aborted in 28.2% (n=67/238) of all surgical explorations, commonly due to occult metastases (n=30/238, 12.6%). The 90-day mortality in group A was 3.0% (n=2/67). The proportion of patients receiving palliative therapy did not differ between groups A and B (65.9% vs. 73.1%; P=0.307). Median OS for groups A and B were 20.4 (95%CI; 15.9-27.3) and 20.2 (95%CI; 19.1-22.7) months respectively (P=0.752). Median OS in group C was 36.1 (95%CI; 30.5-41.2) months. Corresponding 3-year survival rates for all groups were 25.0%, 21.4% and 51.1%, respectively. Compared to unexplored patients, non-therapeutic laparotomy was not associated with reduced OS (HR=0.88 [95%CI 0.61-1.27]) in Cox regression analysis.
Conclusion and relevance: Even in experienced hands, about ¼ of surgically explored LAPC patient will remain unresectable. However, non-therapeutic laparotomy does not appear to substantially reduce short- and long-term outcomes compared to similar patients who are not explored.

Background

Long-term survival in patients with localized pancreatic adenocarcinoma (PDAC) or ampullary adenocarcinoma (AA) who undergo resection is rare, even in lymph node (LN)-negative disease. We aimed to assess the frequency of occult metastases (OM) in patients with resected PDAC or AA discovered with a detailed pathologic examination technique on LNs previously considered negative with conventional analysis. We also examined the association between OM and overall survival (OS).

Methods

Patients with LN-negative disease on conventional pathologic analysis following resection of PDAC or AA from 2010 to 2020 were identified from our institutional database, and those with available tissue for re-analysis were included. LNs were selected for re-examination based on proximity to the tumor and size. Original hematoxylin & eosin slides, three 4-micron-thick sections from deeper levels, and one pan-cytokeratin (AE1/AE3/PCK26) immunostain were examined for each block. The primary outcome was the frequency of OM. The secondary outcome was OS.

Results

A total of 598 LNs from 74 LN-negative patients (PDAC=71; AA=3) were re-examined in detail. A total of 49 patients (66.2%) underwent pancreatoduodenectomy, 17 (23.0%) underwent distal pancreatectomy/splenectomy, and 7 (10.8%) underwent total pancreatectomy. The median LN yield was 19. Sixteen patients (21.6%) had positive surgical margins, 18 (24.3%) had lymphovascular invasion, and 47 (63.5%) had perineural invasion. Twenty-six patients (35.1%) received neoadjuvant therapy and 35 (47.3%) received adjuvant chemotherapy.
On detailed LN analysis, 19 patients (25.7%) had OM. Of these, 9 OM (47.4%) were found only with immunohistochemistry but not on hematoxylin & eosin staining. The number of positive lymph nodes ranged from 1-3. On multivariable analysis, no clinicodemographic or pathologic factors were associated with OM.
The proportion of OM was 10.5% for patients with operative LN yields of <10 LNs, 42.0% for 10-19 LNs, 37.0% for 20-29 LNs, and 10.5% for ≥30 LNs. On conventional pathologic analysis, 3 patients (15.8%) had stage IA disease, 9 patients (26.5%) had stage IB disease, and 7 patients (36.8%) had stage IIA disease, all upstaged to stage IIB on detailed LN analysis.
On survival analysis, patients with OM had an associated decrease in OS as compared to those without OM (median OS: 22.3 vs. 50.5 months; HR=3.86, 95% CI: 1.53-9.78; Figure).

Conclusions

There is a high discordance rate between conventional and detailed LN pathologic analysis in resected PDAC and AA. The presence of OM is associated with worse OS. The high rate of occult nodal disease may in part explain poor survival outcomes in patients with node-negative disease.

Background. A wrong diagnosis of nature is common in pancreatic cystic neoplasms (PCNs). The aim of the current study is to reappraise the diagnostic errors for presumed PCNs undergoing surgery.
Methods. All pancreatic resections performed for presumed PCNs at the Verona Pancreas Institute between 2011 and 2020 were analyzed. “Misdiagnosis” was defined as the discrepancy between preoperative diagnosis of nature and final pathology. “Mismatch” was defined as the discrepancy between the preoperative suspect of malignancy (or its absence) and final pathology. Features considered suggestive for malignancy at preoperative work-up and at final pathology are described in Figure 1. Diagnostic errors considered “clinically relevant” implied a potential over- or under-treatment for the patient.
Results. A total of 601 patients were included. Endoscopic Ultrasound (EUS) was performed in 301 (50%) patients. Overall misdiagnosis and mismatch were 19% and 34%, respectively, with no significant benefit for those patients who underwent EUS. The highest rate of misdiagnosis was reached for cystic neuroendocrine tumors (61%) and the lowest for solid pseudopapillary tumors (6%). Several diagnostic errors had clinical relevance, including 7 (13%) presumed serous cystic neoplasms eventually found to be other malignant entities, 50 (24%) intraductal papillary mucinous neoplasms (IPMN) with high-risk stigmata (HRS) revealed to be non-malignant, and 38 (33%) IPMN without HRS revealed to be malignant at final pathology. A preoperative presumption of malignant mucinous cystic neoplasm was correct in only 20 (16%) patients (Table 1).
Conclusions. Despite not always clinically relevant, diagnostic errors are still common among resected PCNs when applying International Guidelines. New diagnostic tools beyond EUS are needed to refine diagnosis of those lesions at higher risk for unnecessary surgery or accidentally observed nevertheless being malignant.

<b><i>Figure 1. Features of malignancy at preoperative work-up and at final pathology.</i></b><br /> <i>HRS, High Risk Stigmata; IAP, International Association of Pancreatology; IPMN, Intraductal Papillary Mucinous Neoplasms; MCN, Mucinous Cystic Neoplasms; MPD, Main Pancreatic Duct; NET, Neuroendocrine Tumor; PDAC, Pancreatic Ductal Adenocarcinoma; SCN, Serous Cystic Neoplasms; SPT, Solid Pseudopapillary Tumor. </i>

Figure 1. Features of malignancy at preoperative work-up and at final pathology.
HRS, High Risk Stigmata; IAP, International Association of Pancreatology; IPMN, Intraductal Papillary Mucinous Neoplasms; MCN, Mucinous Cystic Neoplasms; MPD, Main Pancreatic Duct; NET, Neuroendocrine Tumor; PDAC, Pancreatic Ductal Adenocarcinoma; SCN, Serous Cystic Neoplasms; SPT, Solid Pseudopapillary Tumor.

<b><i>Table 1. Correct diagnosis and misdiagnosis rate between preoperative diagnosis and final pathology.</i></b><br /> <i>HRS, High Risk Stigmata; IPMN, Intraductal Papillary Mucinous Neoplasms; MCN, Mucinous Cystic Neoplasms; NET, Neuroendocrine Tumor; PDAC, Pancreatic Ductal Adenocarcinoma; SCN, Serous Cystic Neoplasms; SPT, Solid Pseudopapillary Tumor. Correct diagnoses are showed inside black-contoured squares. Clinically relevant errors are showed inside grey squares. </i>

Table 1. Correct diagnosis and misdiagnosis rate between preoperative diagnosis and final pathology.
HRS, High Risk Stigmata; IPMN, Intraductal Papillary Mucinous Neoplasms; MCN, Mucinous Cystic Neoplasms; NET, Neuroendocrine Tumor; PDAC, Pancreatic Ductal Adenocarcinoma; SCN, Serous Cystic Neoplasms; SPT, Solid Pseudopapillary Tumor. Correct diagnoses are showed inside black-contoured squares. Clinically relevant errors are showed inside grey squares.

Background: A standardized future liver remnant (sFLR) <30% and a kinetic growth rate (KGR) <2% are associated with increased risk of hepatic insufficiency and death from liver failure after hepatectomy. Here, we sought to identify clinicopathologic factors associated with inadequate sFLR and KGR to help predict which patients may not achieve sufficient hypertrophy with portal vein embolization (PVE) alone and inform selection for liver venous deprivation (LVD).

Methods: A prospectively maintained single institution database was evaluated for patients undergoing PVE between 1998 and 2020. Clinicopathologic variables, including age, sex, BMI, a known diagnosis of liver disease, diabetes, cycles of neoadjuvant chemotherapy, liver function tests, baseline sFLR, and extended PVE (segment 4 embolization) were evaluated for associations with sFLR and KGR.

Results: A total of 474 patients were identified who underwent right PVE and had both pre- and post-PVE volumetric assessments. Median patient age was 58 years (interquartile range [IQR] 49-66) and median BMI was 27kg/m² (IQR 25-30). The most common histology was metastatic colorectal cancer (66%) followed by hepatocellular carcinoma (12%) extrahepatic cholangiocarcinoma (7%), and intrahepatic cholangiocarcinoma (4%). Most patients (77%) received neoadjuvant chemotherapy prior to PVE (median 6 cycles [IQR 4-11]). Median baseline sFLR was 22% (IQR 16-29%). A sFLR >30% was achieved in 60% of patients following PVE, while KGR >2% was noted in 58%. 71% of patients ultimately underwent surgery, which involved right hepatectomy in 58% and extended right hepatectomy in 41%. Multiple logistic regression revealed that baseline sFLR (OR 1.39 [95% CI 1.30-1.50]) was predictive of post-PVE sFLR >30%. Extended PVE (OR 0.44 [95%CI 0.25-0.77]) and planned two-stage hepatectomy (OR 0.51 [95%CI 0.32-0.82]) were predictive of KGR <2%. ROC analysis revealed that a baseline sFLR ≥19 is 90% sensitive and 78% specific for sFLR >30% (AUC 0.92) (Figure).

Conclusions: Patients with a baseline sFLR <19% or those requiring extended hepatectomy may not achieve adequate hypertrophy with PVE alone. In this subset of patients, LVD should be considered to optimize hepatic regeneration.

Introduction. Microvascular invasion (MVI) is the main risk factor for overall mortality and recurrence after surgery for hepatocellular carcinoma (HCC). Its diagnosis can be made only postoperatively on the histological specimen. The aim of this preliminary study was to train machine-learning models to predict MVI on preoperative CT scan (fig.1).
Methods. Clinical data and 3-phases CT scans were retrospectively collected among 4 Italian centres. DICOM files were manually segmented to detect the liver and the tumor(s). An already available segmentation algorithm (Nnunet) was retrained to obtain automatic detection focused on HCC. An implementation was added to automatically extract radiomics features from the tumoral, peritumoral (among 5mm from the tumor margin) and healthy liver areas in each phase. Performance comparison between manual and algorithm segmentations was measured by intersection over union (Jaccard Index). Data obtained were explored and principal component analysis (PCA) was performed to reduce the dimensions of the dataset, keeping only the PC’s explaining 95% of the variability. After normalization, data were divided between training (70%) and test (30%) sets. Random-Forest (RF), fully connected MLP Artificial neural network (neuralnet) and extreme gradient boosting (XGB) models were fitted to predict MVI. Hyperparameters tuning was made per each model to reduce the out-of-bag error. Prediction accuracy was estimated in the test set and employed as the study end-point.
Results. Between 2008 and 2022, 218 consecutive preoperative CT scans of patients affected by HCC and submitted to surgery were collected with the relative clinical data. At the histological specimen, 72 (33.02%) patients had MVI. The Jaccard index between manual and algorithm segmentations was overall 90%. First and second order radiomics features were extracted, obtaining 672 variables per patient. After data exploration, PCA selected 58 dimensions explaining >95% of the variance. After standardization and normalization, RF, neuralnet and XGB were fitted to predict the presence of MVI. Tuning parameters were: 1) RF: n.tree=500, mtry=30; 2) Neuralnet: 2 hidden layer with 40 and 20 neurons, learning rate= 0.001, threshold for termination= 1%, activation function= sigmoid; 3) XGB: nrounds = 100, max_depth = 3, eta = 0.3. The models were then fitted in the testset to estimate prediction accuracy by confusion-matrix. RF was the best performer (Acc=98.4%, 95%CI: 0.91-0.99, Sens: 95.2%, Spec: 100%, PPV: 100% and NPV: 97.7%, fig.2).
Conclusion. Our model allowed an impressive prediction accuracy of the presence of MVI at the time of HCC diagnosis, never reached until now. This could lead to change the treatment allocation, the surgical extension and the follow-up strategy for those patients. The algorithm will be freely distributed online for medical purpose.

Fig.1 The steps of the study. From 3-phases CT scans, automatic segmentation of the liver and the tumor were obtained by a modified Nnunet. Automatic radiomics features extraction were obtained, and 3 models (RF, ANN and XGB) were fitted to predict MVI.

Fig.1 The steps of the study. From 3-phases CT scans, automatic segmentation of the liver and the tumor were obtained by a modified Nnunet. Automatic radiomics features extraction were obtained, and 3 models (RF, ANN and XGB) were fitted to predict MVI.

Fig.2 confusion matrices per each model investigated, reporting the accuracy and the number of true and false positive and negative cases identified.

Fig.2 confusion matrices per each model investigated, reporting the accuracy and the number of true and false positive and negative cases identified.

Fabrizio Romano

Appearances