Background: Pancreatic cancer is often diagnosed at an incurable stage when it has already spread outside of the pancreas. This is because until the cancer is quite advanced, it does not cause specific symptoms. However, patients often present to their GP several times before diagnosis with symptoms that could be caused by early pancreatic cancer. Identifying these patients earlier could lead to a diagnosis when the cancer is less advanced, so more patients could undergo surgery with potential cure. Purpose of the study: Using the Clinical Practice Research Datalink (CPRD) of over ten million patients, we will compare the symptoms of patients later diagnosed with pancreatic cancer to patients with the same symptoms who did not have pancreatic cancer. We will also look at their medical history and medications to identify other risk factors for pancreatic cancer. Potential importance of the findings: We aim to create an electronic risk calculator (known as a 'clinical decision support tool'). After reviewing a patient with suspicious symptoms, doctors could enter in the patient's risk factors and other relevant information and the tool would estimate the risk that that patient's symptoms are caused by pancreatic cancer, prompting referral for investigations or a specialist opinion.
Objective: To create a clinical decision support tool to help doctors determine the risk that a patient with a particular set of symptoms and risk factors has a cancer within the pancreas, to assist in deciding whether or not to refer a patient for further investigation. Methods: The anonymised records of ten million patients in the Clinical Practice Research Datalink (CPRD) dataset from 2000 - 2016 will be analysed. Patients with a cancer within the pancreas (pancreatic ductal adenocarcinoma, pancreatic neuroendocrine tumours, or biliary tract cancers) diagnosed between 01/01/2001 and 31/12/2015 will be compared to patients without cancer to determine the likelihood that a patient presenting with a particular group of symptoms and risk factors has one of these cancers. Data analysis: Descriptive statistics will be used to determine the changing incidences of the three cancer types over the study period. Multivariable regression analysis will examine the associations between the three cancer types and age, gender and level of deprivation, as well as determining whether survival from these cancers has changed in the last 15 years. Regression models, cluster and factor analyses will be used to develop an algorithm to estimate the risk of pancreatic cancer based on an individual patient's symptoms and risk factors.
Primary health outcomes
- Diagnosis of a pancreatic ductal adenocarcinoma, pancreatic neuroendocrine tumour or biliary tract cancer
Secondary health outcomes
- Trends in incidences of the three different cancer types over the study period
- Overall survival of patients after they have been diagnosed with a pancreatic or biliary tract cancer
- How patients use health care resources (e.g. GP or hospital attendances) before they are diagnosed with a cancer arising from within the pancreas
Stephen Pereira - Chief Investigator - University College London ( UCL )
Stephen Pereira - Corresponding Applicant - University College London ( UCL )
Arturo Gonzalez-Izquierdo - Collaborator - University College London ( UCL )
Georgios Lyratzopoulos - Collaborator - University College London ( UCL )
Harry Martin - Collaborator - University College London ( UCL )
Margaret (Geri) Keane - Collaborator - University College London ( UCL )
Peter Labib - Collaborator - University College London ( UCL )
Rebecca White - Collaborator - University College London ( UCL )
Shahida Islam - Collaborator - University College London ( UCL )
HES Accident and Emergency;HES Admitted Patient Care;HES Diagnostic Imaging Dataset;HES Outpatient;NCRAS Cancer Registration Data;ONS Death Registration Data;Patient Level Townsend Score;Practice Level Index of Multiple Deprivation