Background: Gastric cancer (GC) lacks specific symptoms, resulting in diagnosis at later stages and high mortality. Serum pepsinogen is a biomarker for atrophic gastritis, which is an intermittent phase in the development of GC. A serum based biomarker that indicates increased risk of GC would be useful for targeted prevention strategies. Methods: The Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial is a large randomized trial conducted over 10 centers in the US between 1993 and 2001. Serum samples were collected at baseline and participants were followed for development of incident GC. ELISA-based pepsinogen tests were conducted on pre-diagnostic serum samples of patients who developed GC and age, sex, and race-matched controls. Pepsinogen negative (PG-) and positive (PG+) status was determined using pepsinogen I (PGI) and pepsinogen I to II ratio (PGR). Those with PGR ≥ 3 and PGI > 70µg/L were considered PG- while those with lower levels on either or both were considered PG+. Cox proportional hazard models were used to determine the hazard ratio (HR) and 95% confidence intervals (95%CI) of PG+ for GC. To examine the association of PG with non-cardia and cardia GC separately, propensity matching was used to create control subsets matched on age, sex, and race. Results: Pepsinogen test results were available on 105 GC cases (70 non-cardia and 35 cardia) and 220 matched controls. Median and range of time from pepsinogen measurement to incident GC diagnosis was 78.6 and 3.0-152.2 months, respectively. GC patients were more likely to be PG+ (31.4% vs 5.5%, p < 0.001) and current smokers (20.0% vs 7.7%, p = 0.004) at baseline than controls. Compared to PG-, PG+ were at an increased risk of any GC (HR = 3.77; 95%CI = 2.50-5.71). After adjusting for family history of GC, smoking, and BMI, PG+ were still at a significantly increased risk of GC compared to PG- (adjusted HR = 4.42; 95%CI = 3.14-6.21). For 138 controls matched to 70 non-cardia cancers, PG+ were at an increased risk of non-cardia GC compared to PG- (HR = 5.65, 95%CI = 3.67-8.70; adjusted HR = 7.26, 95%CI = 4.84-10.90). For 70 controls matched to 35 cardia cancers, PG+ were not at greater risk of cardia GC compared to PG- (HR = 1.79, 95%CI = 0.72-4.44; adjusted HR = 1.95, 95%CI = 0.81-5.37). Conclusions: Pre-diagnostic serum pepsinogen levels from a large prospective cohort study predicted development of non-cardia GC but not cardia GC. PG could be considered as a potential risk biomarker to identify individuals at higher risk of non-cardia GC for targeted screening or interventions.