EHR-based phenome wide association study in pancreatic cancer

Area of Science:

• Computational biology
• Medical informatics
• Oncology

Background:

• Pancreatic cancer is a leading cause of cancer mortality in the U.S. with poor prognosis.
• Early detection remains a significant challenge in managing pancreatic cancer.
• Novel methods are needed for effective clinical hypothesis generation.

Purpose of the Study:

• To present a novel method for large-scale clinical hypothesis generation.
• To utilize phenome-wide association studies (PheWAS) on Electronic Health Records (EHR) data.
• To identify clinical concepts associated with mortality in a pancreatic cancer cohort.

Main Methods:

• A cohort of 1,154 pancreatic cancer patients from 2004-2013 was analyzed.
• Over 2.5 million structured clinical observations from EHRs formed the phenome.
• Pairwise associations between 10,164 clinical concepts and patient death were tested.

Main Results:

• After correcting for multiple testing, 231 clinical concepts showed significant association with death.
• This suggests a link between various clinical observations and pancreatic cancer mortality.
• The findings highlight potential areas for further investigation.

Conclusions:

• Phenome-wide association studies (PheWAS) are a viable method for rapid hypothesis generation.
• Leveraging structured EHR data combined with knowledge engineering is effective.
• This approach can accelerate discoveries in pancreatic cancer research.