Lead time trajectory of blood-based protein biomarkers for detection of pancreatic cancer based on repeat testing

In the current study, we assessed whether repeated measurements of a panel of protein biomarkers with relevance to pancreatic ductal adenocarcinoma (PDAC) improves lead time performance for earlier detection over a single timepoint measurement. Specifically, CA125, CEA, LRG1, REG3A, THBS2, TIMP1, TNRFSF1A as well as CA19-9 were assayed in serially collected pre-diagnostic plasma from 242 PDAC cases and 242 age- and sex-matched non-case control participants in the PLCO cohort. We compared performance estimates of a parametric empirical Bayes (PEB) algorithm, which incorporates participant biomarker history, to that of a single-threshold (ST) method. We demonstrated improvements in AUC estimates (2-13 %) for all biomarkers when considering the PEB approach compared to ST. For CA19-9, the PEB^CA19-9 yielded an AUC of 0.88 when at least one repeat measurement was within 3 years of clinical diagnosis. At a specificity of 98.5 %, the PEB^CA19-9 identified 15 of the 41 PDAC cases and signaled positive at an average lead-time of 1.09 years whereas the ST approach captured 11 of the 41 PDAC cases with an average positive signal at 0.48 years. Among CA19-9 low individuals, a PEB algorithm based on repeat measurements of TIMP1 yielded an additional 14 % sensitivity at 98.5 % specificity. An adaptive algorithm that considers repeated CA19-9 measurements improves sensitivity and lead-time detection of PDAC compared to a single-threshold method. Additional protein biomarkers may improve sensitivity for earlier detection of PDAC among cases with low CA19-9.