||5R21CA220073-02 Interpret this number
||Cedars-Sinai Medical Center
||Predicting the Diagnosis of Pancreatic Cancer By Leveraging Big Data
Only 10-20% of pancreatic ductal adenocarcinoma (PDAC) cases in the U.S. are diagnosed at a resectable
stage. Existing biomarkers such as CA19-9 and CA-125 have not translated to real gains in early detection due
to low diagnostic accuracy. Increasing the pre-test probability of PDAC by incorporating patient-centered data
can improve existing and future biomarker performance. Leveraging comorbidities that commonly develop in
parallel with PDAC pathogenesis is an efficient way to identify persons at high risk for PDAC. Recent initiation
of insulin treatment is a notable example, as it is associated with >5 times greater risk of PDAC compared to
persons without diabetes. Our own analysis of population-based sample of elderly patients in the U.S. shows
that healthcare claims for poorly controlled diabetes increases prior to PDAC, suggesting that quantifying
healthcare utilization for diabetes and related conditions may help to identify persons who may be developing
PDAC. The overarching aim of this proposal is to develop a systematic data-driven model that relies on
cumulative medical histories to identify individuals with undiagnosed PDAC for early detection. Using 16 years'
worth of data from the Veterans Affairs Clinical Data Warehouse, we will retrospectively identify persons with
progressing diabetes, in whom we will estimate the 6-, 12-, 24-, 36- and 60-month incidence of PDAC. In this
population we will extract and harmonize national level data on risk factors and clinical indicators, diagnoses,
prescription drug fills. We will then develop a multivariable model for PDAC to estimate the strength of the
relationship between covariates and PDAC. Finally, we will validate the model within the Veteran Affairs
database to evaluate the predictive performance of the model. The aims will be carried out by an interdisciplinary
team with expertise in epidemiology, statistics, gastroenterology, and oncology with strong shared interests in
PDAC epidemiology, prediction modeling and value-centered health care. Ultimately our project will result in a
quantitative tool that can estimate the future probability of PDAC given information readily available in medical
records. With such a tool we will be able to enhance the performance of existing and future biomarkers and
ultimately intervene on a large population of PDAC patients earlier with treatment modalities that have better
prognosis for the patients.
Association of Pancreatic Steatosis With Chronic Pancreatitis, Obesity, and Type 2 Diabetes Mellitus.
, Jeon C.Y.
, Li L.
, Joon A.Y.
, Seltman T.A.
, Sankar M.
, Persohn S.A.
, Territo P.R.
Pancreas, 2019 03; 48(3), p. 420-426.
Do changes in health reveal the possibility of undiagnosed pancreatic cancer? Development of a risk-prediction model based on healthcare claims data.
, Kim S.
, Risch H.A.
, Nuckols T.K.
, Wu B.U.
, Hendifar A.E.
, Pandol S.J.
, Pisegna J.R.
, Jeon C.Y.
PloS one, 2019; 14(6), p. e0218580.