Pancreatic head malignancies are aggressive cancers that are often inoperable when they are diagnosed. In the ~20% of patients who are diagnosed when the disease is still operable, surgery is the only treatment that can provide a chance of cure. Unfortunately, up to 75% of patients undergoing surgery will have the cancer come back (recur). One of the reasons for this is the challenge of removing the whole tumour with some surrounding non-cancerous tissue to ensure that every tumour cell has been removed. This is difficult because there are many structures very close to the pancreas (such as the blood vessels that supply the intestines) that cannot be removed. A recent review study of >1700 patients who had a Whipple's operation (the cancer operation that is performed to remove the head of pancreas) and found that whilst the majority of patients had cancer recurrence in distant sites (like the liver) that would not be affected by how the operation was performed, 12% of patients had the cancer recur just at the site of where the operation had been; this is known as 'local' recurrence. This suggests that a small amount of cancer was not removed at the time of surgery in these patients. Very few studies have looked at the relationship between the Computerised Tomography (CT) scan before surgery and the histology results (information about the tumour after it has been examined under the microscope) and whether this can predict exactly where the tumour recurs. If investigators can find factors that predict which patients get local only recurrence, investigators may be able to offer improved surgical techniques or other therapies during or immediately after the operation to these patients, hopefully leading to improved cure rates.

This retrospective international study will look at these factors in patients who underwent a Whipple's operation for pancreatic cancer, bile duct cancer or ampullary cancer over a three year period between 2012 and 2015. Participating centres will provide data on pre-operative scans, complications around the time of surgery, any therapies (e.g. chemotherapy) that the patients had and if and where the cancer recurred. With this information, investigators hope to find ways to predict which patients will get local-only recurrence, so researchers can select them for future studies to see if additional treatments can improve the chance of cure from surgery for these patients.

Research window:

This retrospective observational cohort study will investigate the outcomes of patients who underwent pancreaticoduodenectomy at 30 hepatopancreaticobiliary (HPB) surgery centres in the UK and abroad between 01/06/2012 and 31/05/2015. These dates were chosen because:

• Most HPB units would expect to see 30-50 patients undergoing pancreaticoduodenectomy per annum. Based on the chief investigator's unit's prospectively maintained database of consecutive patients, 310 patients underwent pancreaticoduodenectomy during this window (benign and malignant). A five year research window would provide a large number of patients that is appropriate for subset analysis. This is especially important in distal cholangiocarcinoma, which is less commonly treated with pancreaticoduodenectomy compared to pancreatic and ampullary cancer. Even if all centres only provided 100 patients (20 per annum), this would provide a cohort of ~3000 patients. This number of cases will have greater resolution to detect small but significant statistical differences in studied variables.
• The amount of data available on patient records reduces with chronicity, in part due to a heavier reliance on paper rather than digital records. In addition, other reporting standards (e.g. TNM staging and histology reporting) change over time. 2010 was selected as a starting year as the 7th edition of the UICC TNM staging came into effect that year.
• An end date of June 2015 allows the study to collect full five-year follow up data on all patients to June 2020, thus providing a complete data set.

Data collection:

Data will be collected by each participating centre on a purpose-built REDCap database. REDCap is a well-established secure web-based data collection tool that is frequently used in medical research involving several centres. Advantages include:

• Real-time collection of data from all centres visible to the research team.
• Guaranteed anonymity of patients as only anonymised data will be collected.
• In-built data validation (e.g. set limits on expected ranges of values) to reduce poor quality/erroneous data collection.
• Easy download of data into a format suitable for processing on an appropriate statistical software package (e.g. SPSS).
• The ability to display or hide questions based on previous responses to tailor data collection to each record.

REDCap access is provided through University Hospitals Plymouth NHS Trust (UHPNT). Data is stored on the Microsoft Azure web-based cloud service. Servers are based in the EU and are GDPR compliant. REDCap access will be provided to all participating centres (one user log in per centre) for data collection.

Data collected falls into the following categories (examples given after each category are not exhaustive):

• Participant ID number (anonymised).
• Demographics: Age, sex, body mass index.
• Comorbidities: Diabetes, cardiovascular disease, respiratory disease, previous history of cancer.
• Pre-operative imaging: Dates, modalities, maximum tumour size, radiological TNM stage.
• Pre-operative biliary drainage: Approach, stent type.
• Neoadjuvant therapies: Type, duration.
• Pre-operative bilirubin.
• Surgery: Date, type, ASA grade, intraoperative procedures and findings.
• Post-operative complications: Types, date of occurrence, Clavien-Dindo grade, treatment, 30-day readmission, 90-day mortality and cause of death.
• Histology: Cancer type, differentiation, tumour size, pathological TNM stage, R status, involved margins and distance, total and involved number of resected lymph nodes, perineural, microvascular and named vessel invasion.
• Adjuvant therapies: Type, duration.
• Recurrence: Date of recurrence, site(s) of recurrence.
• Palliative therapies: Type, duration.
• Survival: Disease free survival (DFS) and overall survival (OS).

Screening of eligible patients:

Patients will be screened to ensure that they meet the inclusion and exclusion criteria (see below). Each unit will be responsible for screening patients for eligibility. All participating units have confirmed that they already have an existing list of consecutive patients who underwent PD during the research window. The clinical team at each participating unit will be responsible for maintaining a password-protected participant look up database that links the local patient hospital number to the anonymised participant ID number on REDCap.

Plymouth sub-study on sarcopenia and myosteatosis:

There is evidence that sarcopenia and myosteatosis may indicate a higher risk of postoperative complications, although its impact on DFS and OS is not yet established. Patients entered into the study from Plymouth will also have their pre-operative CT scan reviewed by a member of the research team trained in sarcopenia and myosteatosis estimation (an academic radiology trainee). The clinicians taking the measurements will be unaware of any outcomes or complications.

This is not being extended to other centres for the following reasons:

• Normal values of tissue attenuation are specific to local populations due to the variation in morphology of patients and the HU cut-offs in one country are not applicable to other countries. It would therefore not be possible to pool data on myosteatosis and sarcopenia from multiple countries.
• As the measurement relies on pre-operative CT imaging, and CT scanners and protocols vary between hospitals and change over time, it is difficult to retrospectively pool such data from multiple sites and maintain meaningful data. Using a single centre provides more robust data for analysis.
• The investigators will be using an inbuilt software package that is part of Plymouth's InSightWeb. This package may not be available in other centres.

As this reduces the number of patients in the cohort, Plymouth will extend its research window to 01/05/2006. This date was chosen as it is the beginning of their prospectively maintained database. This provides a pre-screened population of 365 patients who underwent PD at our unit, with a predicted 276 patients for analysis.

Inclusion Criteria:

• Patients who underwent pancreaticoduodenectomy for pancreatic head malignancy.
• Date of surgery from 01/06/2010* to 31/05/2015 inclusive (*01/05/2006 for Plymouth sub-study).
• Post-operative surgical histology confirmed pancreatic ductal adenocarcinoma (PDAC), ampullary adenocarcinoma (AA) or distal bile duct cholangiocarcinoma (DBCC).

Exclusion Criteria:

• Postoperative surgical histology confirmed benign pathology, non-invasive neoplasia or malignant tumours other than adenocarcinoma of pancreatic, ampullary or biliary origin.
• Patients who underwent distal pancreatectomy or total pancreatectomy as their primary procedure.
• Patients in whom five-year follow up data is not available.

Patients who underwent pancreatioduodenectomy for pancreatic head malignancy (PDAC, ampullary cancer or distal extrahepatic cholangiocarcinoma) between 01/06/2010* and 31/05/2015 inclusive in participating centres (*01/05/2006 for Plymouth sub-study).