Ann Killary

M.S, Ph.D.

The University of Texas M. D. Anderson Cancer Center

Professor

PLCO (Learn more about this study)

2023-0056

Jun 12, 2023

Validation of Pathways-Based Biomarkers for the Early Detection of Prediagnostic Pancreatic Adenocarcinoma in the Blood

Early detection biomarkers for pancreatic adenocarcinoma (PDAC) are urgently needed. The vast majority of PDAC patients are detected at late stage when the cancer is symptomatic and the prognosis extremely poor. Individual biomarkers with the sensitivity and specificity needed for population-based screening have not been discovered. CA-19-9 has been studied extensively and yet has failed to demonstrate the predictive value necessary for early detection and diagnosis. As MPIs of an NCI UO1 grant associated with the NCI Pancreatic Cancer Detection Consortium (PCDC), Drs.Killary and Sen have validated biomarker panels that improve CA19-9 to detect early stage PDAC in the blood . They pioneered a genomic pathways approach to biomarker discovery targeting the earliest genomic intervals initiating loss or gain identified in pancreatic cancer and common to smoking related cancers. Dr. Killary’s laboratory identified a novel “migration signature” that detected pancreatic cancer in the blood (Balasenthil, et al. Cancer Prev Res. 2011) and subsequently validated this signature in multiple blinded validation cohorts. Results, published in JNCI, (Balasenthil, et al, 2017), demonstrate that this biomarker signature significantly improves CA19-9 for detection of early stage pancreatic cancer in the blood. Using the signature in a large NCI EDRN blinded reference set validation, a significant improvement in AUC was observed that distinguished all early stage cancer from both healthy controls and chronic pancreatitis. In particular, among the subcohort free of diabetes and chronic pancreatitis, the biomarker panel with optimal cut offs (risk score) achieved an AUC of 0.90, 0.93, and 0.90 respectively for discriminating stage IA/IB/IIA, stage IIB, or all early stage cancer from healthy controls. Furthermore, for all cases of early PDAC in the EDRN cohort, results were strengthened demonstrating statistical significance for multiple combinations of cases and controls. Dr. Sen’s lab utilized a cancer-related pathways approach for discovery of novel microRNAs in pancreatic cyst fluid samples to better distinguish low grade from high grade lesions (Wang, et al, 2015) and validated microRNAs for early PDAC detection in the blood using small volume assays. Results demonstrated that a large number of candidate microRNA biomarkers improved CA19-9 performance and among those, three were selected based on their association with cancer hallmark pathways. Using a blinded validation set of EDRN samples comprising of 44 stage II cases and 18 healthy controls, microRNA markers were found that were significantly higher values for cases than controls (P=0.0470). When combined with CA19-9, microRNA yielded an improved AUC of 0.87( Dittmar, R. et al, Cancer Prev Res, 2021). Having completed validation studies in Phase 2 ProBe design studies for detection of early stage PDAC in the plasma, we hypothesize that our marker panels could also improve detection at a prediagnostic stage. The overarching goal of this application for acquisition of PLCO PDAC specimens is to determine whether our markers are elevated prior to PDAC diagnosis compared to controls. We propose to validate integrated blood-based biomarker signatures for early detection at a prediagnostic stage when outcomes could be significantly improved.

AIM 1: Validation of pathway-associated, blood-based biomarker panels from protein and microRNA in the PLCO cohort of PDAC using timed interval detection prior to diagnosis of cancer. For this aim, we will utilize the migration signature protein biomarkers TFPI, TNC FNIII-C, and CA19-9 using ELISA assays in blinded validation of PLCO cases and controls. A second panel of microRNA markers will be validated using HTG technology and added to the gold standard CA19-9. CA19-9 has recently been shown to be elevated starting at 2 years prior to diagnosis in the PLCO cohort, so testing our biomarkers could provide additional important utility to increase significance of early pre-diagnostic detection. PLCO EDTA plasma specimens requested include approximately 175 PDAC cases and a 2:1 ratio of healthy controls (compared to cases) (350 total). Controls will be matched to cases based on age, sex, race, and collection dates. Plasma samples obtained from these cases and controls will be from blood collection times equal to 4 years, 3 years, 2 years, and 1 year or less from diagnosis of PDAC. The goal of the analysis is to determine if the anchor panel of protein or microRNAs has sufficient performance for detecting prediagnostic PDAC. Two performance measures will be used: (1) the area under the ROC (AUC), and (2) ROC (0.05), the sensitivity of the biomarker panel corresponding to 95% specificity on the receiver operating characteristic (ROC) curve. The anchor panel performance will be considered acceptable if AUC is at least 90% or the ROC (0.05) is significantly greater than 50. We can also use leave-m out cross validation to re-estimate a combination rule for the anchor panel and validate its performance.

AIM 2: Statistical analysis and selection of optimal panel for detection of prediagnostic PDAC in the blood. Biomarkers validated in Aim 1 from the protein and microRNA platforms will be considered to refine the anchor panels of markers. We will refine the optimal biomarker panel using either a backward elimination procedure or a lasso penalty applied to a logistic regression model with the validated anchor markers, such as validated candidate markers CA19-9, TFPI, and TNC, and patient characteristics such as age, gender and smoking. Non-linear trends and logic rules (e.g. OR rule: classify positive if either marker A or B is elevated) will be examined as well. The optimal panel will determine the risk score for each patient, and its performance of the cutoff value corresponding to 95% specificity will be tested. Power Analysis: A sample of 150 cases and 300 controls will achieve at least 80% power to detect at least 5% improvement in the area under the ROC curve (AUC) from 90% using a two-sided significance level of 0.05. The study has >90% power to reject the null hypothesis with a one-sided p-value = 0.05 if the true sensitivity is at least 74% at 95% specificity.