Study
PLCO
(Learn more about this study)
Project ID
2020-1013
Initial CDAS Request Approval
Sep 23, 2021
Title
Proteomic Analyses of Serial Prediagnostic PLCO Plasma in Cases and Controls to Identify Early Detection Ovarian Cancer Biomarkers Rising in a Substantial Fraction of Cases and Stable in Most Controls
Summary
This project aims to discover and validate plasma biomarkers for the early detection of ovarian cancer. A hallmark of cancer is uncontrolled cell division, leading to a doubling time of the tumor. This exponential growth stands in stark contrast to the stable or slowly changing profile of plasma proteins in almost all other diseases or in healthy subjects. We will discover early detection (ED) plasma protein biomarkers by identifying the proteins that significantly rise over time in an exponential fashion in a substantial fraction of cases and yet remain relatively stable over time in most controls. This requires plasma assays over a large suite of proteins with CVs lower than the protein's biological variation over time which can be as low as a CV of 10%. Furthermore, a low volume requirement is essential for access to precious biospecimens formed from long-term large early detection trials. Olink AB has developed proximity extension assays (PEAs) for a suite of ~1,500 proteins with CVs ranging from 6-12% and with a minimal volume requirement of 3 μL. Applying the Olink proteomic assays to serial pre-diagnostic plasma from subjects in the PLCO who were diagnosed with ovarian cancer during the study (cases n=50) and to serial plasma samples from a 4:1 matched control (n=200) : case (n=50) cohort will provide longitudinal data on ~1,500 plasma proteins from cases and controls by which to identify ED candidate biomarkers. Prior to cancer developing in each case, a biomarker will be stable over time, while after cancer inception the biomarker will rise exponentially reflecting tumor doubling. This behavior is represented by a change-point model in cases while the same biomarker in women without ovarian cancer (controls) will have a flat profile. ED biomarkers will be the proteins which have a change-point in a substantial fraction of cases while remaining stable in most (98%) controls. We will identify the top 20 ED biomarkers where the criteria for inclusion is a combination of fraction of cases, complementarity to proteins already selected, and time of rise with earlier risers having priority. After identification of the 20 ED biomarkers, Olink will develop a custom panel of 20 ED markers with absolute quantification. The custom panel will assay the same PLCO plasma samples as used in discovery. These data will be analyzed with a multivariate longitudinal change-point model to form a multiple marker longitudinal algorithm for ED. This classifier will be locked down and then assessed for validation by assaying the custom panel of 20 ED biomarkers on an independent PLCO serial plasma sample set, from cases (n=50) and 10:1 matched controls (n=500). From these data the classifier will be assessed for two dimensions of sensitivity for ED: (i) number of months prior to detection in PLCO, and (ii) proportion of cases detected, while (iii) maintaining a high specificity goal of 98% - a false positive rate of 2%. This low false positive rate requires a large number of controls (n=500) for accurate assessment.
Aims
This project aims to perform:
• Proteomic analyses of longitudinal pre-diagnostic PLCO plasma samples from a training cohort of ovarian cancer cases (n=50) and 4:1 matched controls for ~1,500 proteins using minimal volume (3 mL) and with accurate assays (low CVs 6 – 12%).
• Hierarchical longitudinal statistical modeling of each plasma protein with change-point models for cases and flat models for controls to identify proteins which rise above each woman’s baseline for a substantial fraction of cases while remaining stable over time in controls.
• Identification of the top 21 proteins with criteria of sensitivity in earlier in time and coverage of the spectrum of ovarian cancer while maintaining high fixed specificity from which to form a custom panel of proteins measured with absolute quantification.
• Remeasurement of the training cohort with the custom panel from which development of an optimal longitudinal multi-marker classifier with the same criteria as SA3.
• Assessment of the performance characteristics (sensitivity, specificity, PPV) of the classifier in longitudinal plasma samples from an independent PLCO validation cohort of ovarian cancer cases (n=50) and 10:1 controls.
