Surgery and pathology
Anatomic resectability definitions were applied from the 2009 expert consensus statement except in the upfront resection group prior to 2009 when resectable was defined by the surgeon (4). Patients with pancreatic head tumors underwent pancreaticoduodenectomy with pylorus sparing based on evolving surgeon preference over the study period. Those with pancreatic body or tail tumors underwent distal pancreatectomy and splenectomy. A small percentage of patients required total pancreatectomy due to concerns of diffuse tumor infiltration or numerous intraductal mucinous papillary neoplasms. A vascular surgeon was available in case of need for repair or resection of the SMV or PV.
College of American Pathology protocols were used for pathologic examination, including tumor response grade to neoadjuvant therapy. Grading was performed by pathologists with expertise in gastrointestinal pathology. Margin negative was defined as no tumor cells at the specimen edge.
Follow-up and analysis
All charts were reviewed to identify post-operative morbidity, mortality, adjuvant treatment, and toxicities. Side effects were graded according to version 4.0 of the National Cancer Institute Common Terminology Criteria for Adverse Events. Patients were followed every 3–6 months for the first 5 years and yearly thereafter.
Statistical testing was performed using IBM SPSS (Windows Version 20.0; IBM Corp, Armonk, NY, USA). Comparisons between groups were performed by 2-sided exact Mann-Whitney test for continuous variables or 2-sided Fisher’s exact test for nominal variables. Survival was estimated by Kaplan-Meier method and compared by log-rank (Mantel-Cox) test, with time starting at the first cancer directed treatment (chemotherapy or surgery).
Patient characteristics and neoadjuvant treatments for the surgery analysis group (see Figure 1) are summarized in Table 1. The neoadjuvant therapy group had higher T classification and N classification compared to upfront resection (P<0.001 and P=0.005, respectively; unknowns excluded).
Within the surgery analysis group, grade ≥3 toxicities occurred in 3 patients (4.9%) that were potentially attributable to SBRT. One patient had grade 3 duodenal bleed associated with tumor invasion and ulceration 3 weeks after SBRT. The patient underwent endoscopic cauterization and pancreaticoduodenectomy soon afterwards with uneventful post-operative course. One patient suffered biliary ductal stenosis 7 months after pancreaticoduodenectomy and underwent repeat pancreaticojejunostomy. Another patient experienced portal hypertension and malabsorption resulting from portal venous stricture 18 months after pancreaticoduodenectomy. Grade ≥3 toxicities occurred in 26 patients (42.6%) that were potentially attributable to neoadjuvant chemotherapy. Most toxicity was hematologic as neutropenia (n=15), anemia (n=4), or thrombocytopenia (n=2). Twelve (19.7%) patients experienced at least one grade 3+ symptomatic toxicity due to chemotherapy, including vomiting (n=3), febrile neutropenia (n=2), palmar-plantar erythrodysesthesia (n=2), oral mucositis (n=2), diarrhea (n=2), toxic epidermal necrolysis (n=1), and hypertension (n=1).
Table 2 demonstrates no differences in the documented post-operative morbidities listed (all P>0.1). The overall rate of a patient experiencing surgical morbidity was 39.3% in the neoadjuvant therapy group and 31.1% in the upfront resection group (P=0.226). There was less estimated blood loss (EBL) and shorter surgery time in the upfront resection group, while neoadjuvant patients had shorter hospitalization times. However, when stratified for vein resection or repair, hospitalization and EBL were not statistically different. Median surgery time for patients not requiring vascular repair remained 1 h longer in the neoadjuvant therapy group.
The positive margin rate was only 3.3% after neoadjuvant therapy compared to 16.2% with upfront resection (P=0.006). Table 3 further reports that the median pathologic size of the primary was smaller with neoadjuvant therapy (2.5 vs. 3.0 cm, P<0.001) and the pathologic T classification was lower (P=0.010) with 4 pathologic complete responses (6.6%). Despite more lymph nodes resected in the neoadjuvant therapy group (20 vs. 12, P<0.001), there was a trend towards fewer positive lymph nodes in the neoadjuvant therapy group (2 vs. 3, P=0.054).
Comparisons of overall survival (OS) are presented in Figure 2. The intention-to-treat survival analysis (Figure 2A) includes all 159 BRPC and LAPC patients treated with neoadjuvant therapy (61 resected, 98 not resected). As expected, these mostly unresected pancreatic adenocarcinoma patients had worse survival than those who underwent upfront resection (median OS 17.0 vs. 22.1 months, P=0.029).
Of the 241 upfront resection patients, 45 (18.7%) received adjuvant chemotherapy alone and 147 (61.0%) received chemoradiation guided by Radiation Therapy Oncology Group 97-04 protocol (22). To compare outcomes for patients who received comparable treatment (surgery and at least neoadjuvant or adjuvant chemotherapy), we excluded from further survival analysis the remaining 49 (20.3%) who received no adjuvant treatment or were lost to follow-up, leaving 192 patients. For the survival analysis, the neoadjuvant therapy group includes the 61 who were resected.
Thus in the survival analysis group (Figure 2B), the estimated median OS for the neoadjuvant therapy group (n=61, median follow-up 18.1 months) was 33.5 vs. 23.1 months in the upfront resection group (n=192, median follow-up 22.2 months) (Figure 2B, P=0.057). Locoregional failures, isolated or concurrent with first distant metastasis, were observed in 5 of 61 (8.2%) neoadjuvant therapy patients and in 22 of 192 (11.5%) upfront resection patients (log rank P=0.786).
To attempt to control for baseline differences between groups, univariate and multivariate analyses for OS are presented in the Supplemental materials. On multivariate analysis, no baseline characteristic was significant for survival, but trends remained for neoadjuvant therapy group (P=0.077) and N classification (P=0.091). To attempt to control for year of therapy, a matched analysis was performed for patients who underwent treatment between 2010 and 2012. Margin positive rates remained significantly different within neoadjuvant therapy (n=41) and upfront resectable (n=59) cohorts (2% vs. 16%, P=0.029). OS was similar (33.5 vs. 22.9 months, P=0.317).