Clinical usefulness of conversion surgery for unresectable pancreatic cancer diagnosed on multidetector computed tomography imaging: Results from a multicenter observational cohort study by the Hokkaido Pancreatic Cancer Study Group (HOPS UR‐01)

Abstract Background and Aim Effective multidisciplinary approaches for unresectable pancreatic cancer (UR‐PC) that include modern chemotherapeutic regimens and subsequent conversion surgery (CS) are being developed. The aim of this study was to evaluate outcomes of patients clinically diagnosed with UR‐PC, focusing on the efficacy of CS. Methods Patients ineligible for two multicenter phase II studies conducted by the Hokkaido Pancreatic Cancer Study Group (HOPS) were recruited. Sequential treatment regimens, conversion to radical surgery, and overall survival (OS) were analyzed by multidetector computed tomography (MDCT)‐based UR factors. Univariate and multivariate analyses were performed to identify predictors of OS. Results Sixty‐six of 247 intended recruits for HOPS studies from October 2013 to April 2016 were included. Unresectability was due to locally advanced (LA) disease and metastasis (M) in 42 and 24 patients, respectively. Induction therapy began with chemotherapy (CT) and chemoradiotherapy (CRT) in 44 and 17 patients, respectively, of whom 23 received modern CT regimens. Radical surgery was completed in 12 (LA, 10; M, two) with a median treatment interval of 10.3 months (range, 2‐32). Eleven patients (91.6%) achieved pathological R0 resection. Median OS was significantly longer in patients who underwent CS than those who did not (44.1 vs 14.5 months, P < 0.0001). CS was an independent predictor of OS (hazard ratio, 0.078; 95% confident interval, 0.017‐0.348; P = 0.001). Conclusion Conversion surgery after a favorable response to sequential treatment might prolong survival in patients with UR‐PC. Precise diagnosis on MDCT followed by sequential multimodal anticancer treatment is essential.


| INTRODUC TI ON
Pancreatic cancer (PC) is one of the most aggressive malignancies. 1 In 2018, there were 44 330 estimated patients with PC-related death in the USA and 34 990 patients in Japan, suggesting that PC is the fourth leading cause of cancer-related death in both countries. 1,2 In the National Comprehensive Cancer Network (NCCN) guidelines, 3 resectability is categorized as resectable (R), borderline resectable (BR), or unresectable (UR) based on multidetector computed tomography (MDCT) evaluation. Upfront surgery followed by postoperative adjuvant therapy was generally recommended for potentially resectable PC (R-PC) 3,4 as well as neoadjuvant treatment followed by surgery for BR-PC in order to achieve R0 resection. 3 Despite marked improvements in diagnostic modalities, PC often presents as a systemic disease, which precludes early detection.
More than 80% of patients are diagnosed with UR because of its high metastatic (M) potential. 5 Recent advances in anticancer treatment for locally advanced (LA) UR, or M-PC facilitate good disease control; such patients sometimes convert to surgical resection. 6 This surgical strategy is called conversion surgery (CS). 7 Several reports on CS in patients with UR-PC have shown that it has a favorable effect on overall survival (OS). 6,8,9 In recent meta-analyses of reports from 2009 to 2015, the rate of conversion from UR-LA-PC to surgery was 26% and OS ranged from 18.7 to 24.2 months. 10,11 The entire cohort examined in these meta-analyses comprised patients recruited into clinical trials conducted before 2013.
The Hokkaido Pancreatic Cancer Study Group (HOPS) conducted multicenter phase II studies to investigate the efficacy of neoadjuvant treatment for BR-PC and R-PC. To analyze the data from patients with UR-PC whose diagnosis was based on central review of MDCT findings but were ineligible for these two HOPS studies, we conducted a multicenter study. Since those patients were managed at referral hospitals in Hokkaido prefecture thereafter, their survival data were recognized as real-world patient outcomes.

| Study design
In this multicenter, retrospective study by HOPS, we assessed the outcomes of patients clinically diagnosed with UR-PC and treated at tertiary referral hospitals around Hokkaido prefecture. The institutional review boards of Sapporo Medical University Hospital (282-39, University Hospital Medical Information Network Clinical Trials Registry, UMIN000035454) and each participating hospital approved the study protocol. longer in patients who underwent CS than those who did not (44.1 vs 14.5 months, P < 0.0001). CS was an independent predictor of OS (hazard ratio, 0.078; 95% confident interval, 0.017-0.348; P = 0.001).

| Patients
Conclusion: Conversion surgery after a favorable response to sequential treatment might prolong survival in patients with UR-PC. Precise diagnosis on MDCT followed by sequential multimodal anticancer treatment is essential.
locally advanced, metastatic, multidisciplinary treatment, radical surgery, unresectable pancreatic cancer the circumference in other situations, and metastatic disease were categorized as UR. Regarding the common hepatic artery, a diagnosis of LA-UR disease was made when safe, complete resection and reconstruction were very difficult because of tumor extension to the bifurcation of the hepatic artery. 8 All MDCT interpretations were performed by two radiologist (Y.S. and D.A.) and verified by a surgeon (T.N.) and a gastroenterologist (M.K.). Suspicious liver metastasis on MDCT was confirmed with gadoxetic acid-enhanced magnetic resonance imaging (EOB-MRI), contrast-enhanced ultrasonography, or both. Positron emission tomography (PET)-CT was used for confirming other types of metastasis. Peritoneal metastasis was diagnosed when there was an intra-abdominal nodule or mass separate from the pancreas located on the surface of the peritoneum, greater omentum, or intestine. Peritoneal metastasis was suspected if an area with the density of water was found in the abdominal cavity on MDCT, for example, in the pouch of Douglas, paracolic gutter, or around the liver or the surface of the spleen. Patients defined as having UR disease were ineligible for these two studies and based on HOPS central MDCT review were introduced into this retrospective study. In general, the diagnosis of PC was confirmed by endoscopic ultrasound-guided fine-needle aspiration or brush cytology during endoscopic retrograde cholangiopancreatography.
The attending physician suggested CS to patients with UR-PC who met the following conditions: (a) treatment effect was evalu-

| Assessment
Clinical treatment effect was assessed using RECIST version 1.1. 13 The histologic assessment of the extent of preoperative treatment response was evaluated using the Evans grading system. 14 The Clavien-Dindo classification was used to assess postoperative complications. 15 Mortality was defined as death during the hospital stay when surgery was performed. Individual survival was defined as the duration between the date of treatment initiation and death or latest hospital visit. Median follow-up was defined as the duration between the date of MDCT consultation and the latest hospital visit for censored patients.

| Outcome measures and statistical analysis
Variables included individual patient data, imaging findings, diagnostic information including cTNM stage according to the General Rules for the Study of Pancreatic Cancer (July 2016, seventh edition), 16 cytology or histology results, date of MDCT consultation, type of biliary drainage, details on sequential treatment regimens, details about attempted radical surgery, pathological findings, and outcomes including disease recurrence or death.
Outcome measures included sequential treatment regimens, conversion to radical surgery, and OS. Outcome measures were analyzed by MDCT-based UR factors including LA and M disease. Univariate and multivariate analyses were performed to determine predictors of OS.
Comparisons between two groups were performed using the Chi-squared test, Mann-Whitney U test, or Cox proportional hazards regression modeling for nonparametric data. Factors with P < 0.2 on univariate analysis without potential confounding were included in multivariate logistic regression models to calculate adjusted odds ratios. OS was calculated using the Kaplan-Meier method and compared using the log-rank test. All calculations were done with StatMate V (ATMS Co., Ltd., Tokyo, Japan), or SPSS version 16.0 (SPSS Inc., Chicago, IL, USA). All results are expressed as medians (range). P < 0.05 was considered statistically significant.

| Conversion to radical surgery and surgical outcomes
Surgery was not recommended for 51 of 66 patients with UR-PC.
Median progression-free survival in these patients was estimated to be 5.7 months (95% confident interval [CI], 4.7-6.8). Twenty-one patients (12 with UR-LA and nine with UR-M disease) had disease control for more than 6 months. Fifteen patients developed progressive disease and six patients continued to have PR or SD for 11.8-44.8 months until the latest follow-up. The reasons why surgery was not recommended for these six patients included persistent liver metastasis in two patients, persistent ascites in one patient, and plexus involvement in three patients that extended into the superior mesenteric (n = 2) and common hepatic arteries (n = 1).
Radical surgery was performed in the remaining 15 patients (12 patients with UR-LA disease and three with UR-M disease; Figure 1) but completed in 12 (10 patients with UR-LA disease and two patients with UR-M disease; Table 2). Three patients with metastasis at the time of surgery underwent palliative procedures that included probe laparotomy, choledochojejunostomy, and laparoscopic gastrojejunostomy. The median preceding treatment interval was

| Predictors for conversion
Among background patient characteristics, various clinicopathologic parameters, and treatment history, none were identified as significant predictors of conversion (Table S2). Interestingly, patients with CS had neither nodal nor peritoneal metastasis on initial imaging (Table S1). Median duration from initial treatment to initial recurrence was estimated to be 29.0 months (range, 10.3-37.3; Figure 2B, Table 3).
Modern chemotherapeutic regimens such as FOLFIRINOX or gemcitabine plus nab-paclitaxel as second-line treatment had comparable survival (MS, 16.7 vs 22.2 months; log-rank P = 0.9482; Figure 3B).
However, patients with CRT as second-line treatment had significantly better MS than those without (24.2 vs 14.5 months; log-rank P = 0.046; Figure 3C). It is obvious that there was bias in treatment selection because CRT as second-line treatment was used in eight    In the current study cohort, postoperative recurrence occurred in more than half of patients with a median duration from CS to ini-  29 This definition acknowledges that resectability is not just about the anatomic relationship between the tumor and vessels, but that biological and conditional dimensions including the status of the tumor marker CA19-9 (> or ≤500 units/mL) and performance status are also important. In order to improve OS, future studies should assess whether to consider CS after initial induction treatment for patients with initial UR-PC and its timing. In Japan, the PREP-04 trial (UMIN 000017793), a multi-institutional prospective observational study to investigate the effects of CS in patients with initial UR-PC, is already ongoing.

| CON CLUS ION
Conversion surgery following a favorable response to sequential treatment may be a good option to prolong survival in patients with UR-PC. Precise imaging diagnosis based on MDCT followed by sequential multimodal anticancer treatment is essential.

ACK N OWLED G M ENTS
We deeply appreciate the support of Daisuke Abo for MDCT interpretations and the affiliate contributors from HOPS for patient management or data acquisition (Appendix S1).

D I SCLOS U R E
Conflict of interests: authors declare no conflict of interests for this article.
Author Contributions: Kimura and Imamura had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Kimura, Imamura, Yamaguchi, Motoya, and Yoshida. Acquisition or interpretation of data: Kimura, Imamura, Yamaguchi, Motoya, Yoshida, Nagayama, Yamakita, Goto, and Takahashi. All MDCT interpretations were performed by Sakuhara and were verified by Nakamura