Risk Factors and Clinical Presentation of Pancreatic Cancer
In the United States, there are about 65000 cases of pancreatic cancer per year (1). Risk factors for pancreatic cancer include age older than 55, male sex, obesity, smoking, and genetics (ten percent of patients with pancreatic cancer have an inherited mutation). Patients with pancreatic cancer usually present with weight loss (about 90% of patients) and abdominal pain (about 75% of patients). Jaundice due to blockage of the common bile duct by a tumor in the head of the pancreas is seen in 70% of patients. Other common findings include weakness, loss of appetite, and itchiness due to elevated levels of bilirubin. Lab findings include elevated liver function tests, elevated bilirubin, and elevated amylase and lipase. Some patients may present with new-onset diabetes despite stable or declining weight (3).
Incidence and Mortality Of Pancreatic Cancer
The average lifetime risk of pancreatic cancer is about 1 in 64. Pancreatic cancer accounts for 3% of all cancers in the US and 7% of cancer-related deaths. The 5-year survival rate is about 12% and the disease is the fourth leading cause of cancer death in men and women (1). Of note, the dismal 5-year survival rates for pancreatic cancer have only improved marginally over the past four decades due to small advances in chemotherapeutic options. About 80% of patients with pancreatic cancer have unresectable disease at the time of diagnosis (2).
Promise of AI To Improve The Early Diagnosis Of Pancreatic Cancer
Pancreatic cancers are often asymptomatic in their early stages and correspondingly diagnosed when they are advanced and no longer amenable to surgical intervention and potential cure. Improving the sensitivity of Computed Tomography (CT) and Magnetic Resonance imaging for early pancreatic cancer is a major clinical need. About 40% of pancreatic cancers smaller than 2cm are missed on Abdominal CT (3).
A new artificial intelligence model provides an approach for early detection. The AI model was trained on 1080 control CT scans and 696 diagnostic CT scans with pancreatic cancer. The model was tested on public datasets of CT scans in patients with new onset diabetes, some of whom had pancreatic cancer, and pre-diagnostic CT scans performed up to 3 years before diagnosis of pancreatic cancer. This model had an 80% sensitivity for stage 1 pancreatic cancers and detected pancreatic cancer on pre-diagnostic CT scans at a median of 475 days before clinical diagnosis.
This AI based model shows the ability to decrease diagnostic errors in the reading of CT scans and identify pancreatic cancer at an earlier stage when there is increased potential for cure. Targeted screening of higher risk groups in combination with emerging blood-based biomarkers could markedly impact the mortality of pancreatic cancer (4).
Bradford Rabin MD aims to provide care by the best concierge doctors in the Bay Area for patients in Palo Alto, Menlo Park, Atherton, Los Altos, Portola Valley, Hillsborough, and Woodside.
1 Pancreatic Cancer Statistics
2 Pancreatic Cancer
3 Pancreatic Cancer On CT Scans With Deep Learning: A Nationwide Population-Based Study
4 Automated Artificial Intelligence Model Trained On A Large Data Set Can Detect Pancreatic Cancer On Diagnostic Computed Tomography Scans As Well As Visually Occult Preinvasive Cancer On Prediagnostic Computed Tomography Scans