Identification of High-Risk Cohorts using PLCO data for Targeted Gastric Cancer Prevention and Control Strategies
A potential biomarker that is currently being investigated for gastric cancer is Pepsinogen (PG). Serum pepsinogen (PG) testing is considered the best available non-invasive option to identify individuals with precancerous lesions, in particular, atrophic gastritis. Given PG identifies atrophic gastritis, a precursor of gastric cancer, PG has great potential to be used as a screening tool for gastric cancer risk.
The goal of this proposal is to develop risk prediction models that identify high-risk subgroups of individuals who might benefit from antibiotic treatment of H. pylori infection or undergo UGI endoscopy for diagnosis and treatment of gastric pre-cancer and early GC. That is, we want to develop a risk-based GC prevention and control algorithm for use in the US. We propose to use the PLCO data on serum Pepsinogen I, Pepsinogen II and H pylori along with the baseline information on race/ethnicity and demography and dietary history of gastric cancer cases.
PLCO is one of the few large scale prospective cohort studies where data has been collected on many known and potential GC risk factors. PLCO also serves as a rich repository of data for biomarkers related to GC. This makes it particularly important for GC study in the US since most studies examining H pylori infection and Pepsinogen tests have been conducted in other countries and studies in the US are limited. The availability of such data in the PLCO study gives us a unique opportunity to estimate the risk of GC risk factors and develop a risk-prediction model using a US population sample.
We will use the data collected on GC and non-GC controls available in the PLCO data to produce risk estimates of specific items such as diet, diabetes, smoking, alcohol intake, BMI, family history of cancer, place of birth, race/ethnicity and immigration, socio-economic status, lab results on H pylori and Pepsinogen for gastric cancer using logistic regression and Cox proportional hazards models. Risk prediction modeling techniques will be applied to PLCO data to identify a shortlist of the most predictive items that can be used as a pre-screening tool. The results of this proposal will establish supportive evidence for our efforts to develop a risk prediction tool to be used in targeted gastric cancer prevention and control strategies.
GC mortality can be reduced by two, evidence-based approaches: 1) prevention through detection of H. pylori (HP) which causes 90% of all gastric cancers, followed by treatment with antibiotics and 2) early detection by upper gastrointestinal (UGI) endoscopic screening followed by endoscopic or surgical treatment of pre-cancer and early cancers. H. pylori test and treat strategy has been shown to reduce incidence of gastric cancer by 35-50% in high-risk groups. The implementation of endoscopic screening and treatment brought about stage shift from late- to early-stage GC and better overall 5-year survival resulting in 30-60% decrease in mortality rates in high-risk counties.
The goal of this proposal is to develop risk prediction models that identify high-risk sub-groups of individuals who might benefit from antibiotic treatment of HP for primary prevention or undergo UGI endoscopy for diagnosis and treatment of gastric pre-cancer and early GC for secondary prevention. That is, we want to develop a risk-based GC prevention and control algorithm for use in the US. There are several promising markers that may be applicable but need further investigation with respect to the aforementioned goal. Serologic testing for HP and pepsinogen have been proposed as markers to detect persons at elevated risk of HP and GC. We will utilize the PLCO data to obtain information on H pylori testing and Pepsinogen I and II levels in GC cases and controls. The survey data will provide us information on other specific risk factors associated with gastric cancer such as diet, diabetes, smoking, alcohol intake, BMI, family history of cancer, place of birth, race/ethnicity and immigration, socio-economic status.
We propose the following Specific Aims to:
1. Develop a risk prediction model for H. pylori
2. Develop a risk prediction model for GC
All predictor variables will be considered for the prediction model. Analyses will first focus on establishing effect estimates for HP and GC risk factors in the US. We will consider nonlinear and threshold effects through polynomial, and change point models, and categorical predictors.
The above risk prediction modeling will be accompanied by the development of parsimonious models that can serve as efficient pre-screening tools to be used in the community as well as clinical settings.
Dr. Philip E Castle Albert Einstein College of Medicine