Development of Novel Ovarian Cancer Biomarkers for Early Detection Algorithms

点击翻译按钮获取中文摘要

ABSTRACT Ovarian cancer (OC) is a deadly but often silent disease, showing no specific signs until it reaches advanced stages. The 5-year survival rate for advanced OC is only 50%, as most tumors ultimately become resistant to treatment.1,2 Advances in cytoreductive surgery and combination chemotherapy have improved 5-year survival in patients with epithelial OC, but the rate of cure has not improved over the last two decades. Computer models suggest that detection of OC in early stages (I-II) could substantially improve cure rates, but the low prevalence of OC in the general postmenopausal population restricts early detection efforts. Definitive diagnosis requires operative intervention, but a consensus is that no more than 10 operations should be performed to diagnose a single OC (>10% positive predictive value, PPV). According to current requirements, a first-line biomarker-based screening test must achieve a sensitivity (SN) of at least 75% and a specificity (SP) of 98%, which can then be further increased to 99.6% by adding a second-line screening modality such as transvaginal sonography (TVS). 1,3-6 Because available screening tests remain inadequate to merit wide implementation, based on our strong preliminary findings the proposed project aims to develop a novel, widely translatable, and economically feasible test that can reduce OC mortality rates. Currently, the only promising strategy developed in the United Kingdom Collaborative Trial for OC screening (UKCTOCS), is sequential analysis of the marker CA125 in serum over time (Risk of OC Algorithm, ROCA), followed by TVS. UKCTOCS yielded only a modest 20% decrease in mortality, insufficient to prompt the US Preventive Services Task Force to change its recommendation against population-based OC screening. 1 The most likely reason for such modest mortality reduction by CA125 measures is their insufficient lead-time (estimated interval for detection prior to symptoms-based diagnosis). Bio- mathematical modeling suggests that OC progresses to late stages more than 1 year before symptoms onset, a time range when CA125 levels offer only limited diagnostic power. Therefore, to improve current clinical practice, novel screening algorithms allowing substantially longer lead-times are needed. Based on our strong preliminary findings, we aim to develop and validate a 2-pronged approach, whereby a first-line multi-biomarker test recognizes OC with high SN (>80%) and modest SP (>80%), followed by a second-line biomarker velocity-based test in women who tested positive in the first test, that then yields a combined SP of 98%. Supporting this approach, we have generated a preliminary classification algorithm (threshold-based algorithm, TBA) based on one-time measurement of multiple biomarker concentrations, that identifies with 80%SN-70%SP women who will develop OC 1-7 years later. We further identified several biomarkers that display robust temporal dynamics (velocity) associated with OC development in the 1-7 YTD interval. We thus hypothesize that we can generate a 2-step algorithm that provides >75%SN at >98%SP, by combining our novel TBA with a velocity-based algorithm (VBA). In this approach, similar to ROCA, the positive results of the TBA would trigger frequent follow- up screening with VBA. The crucial advantage of our proposed algorithm vs. UKCTOCS' ROCA is that our novel combined algorithm will recognize OC more than 1 YTD, increasing the probability of detecting OC at early, treatment-responsive stages. We have discovered, and will prioritize for integration into the tests, several promising candidate pre-diagnostic OC biomarkers, including autoantibodies (AAbs). Our long-term goal is to develop a robust, accurate and widely translatable early-stage screening algorithm for risk of OC. Our immediate objectives are to enhance our biomarker-based classifiers for pre-diagnostic samples, developed in preliminary studies, by adding new promising candidate biomarkers we have identified, and validate them in independent pre-diagnostic samples. The Specific Aims are: 1. Generate and validate an optimized first-line threshold-based classification algorithm with 1.5-7 years lead-time. We will assess whether new candidate biomarkers can further improve the algorithm we developed in preliminary studies, and then validate the optimized algorithms in pre-diagnostic PLCO samples. 2. Generate and validate a biomarker temporal dynamics (velocity)-based algorithm. We will validate the promising candidate velocity-based biomarkers identified in Aim 1 in pre-diagnostic serial samples from UKCTOCS and NROSS prospective studies and generate a velocity-based classification algorithm for detecting OC, to complement and enhance the cut-off- based algorithm(s) developed in Aim 1. 3. Determine the performance of a 2-step (threshold+velocity)– based OC screening algorithm with 1.5-7 years lead-time in serial samples. We will determine the cumulative performance of sequential algorithms including the threshold-based algorithm developed in Aim 1, followed by the velocity-based algorithm developed in Aim 2, for OC screening in the 1.5-7 YTD interval, in serial UKCTOCS samples. In summary, we anticipate our results will yield development and validation of the first blood biomarker-based algorithms with the required >75% SN, >98% SP, for reliably classifying OC in preclinical samples collected 1.5-7 YTD. These algorithms will be ready for validation in prospective screening clinical trials to evaluate the effect of early detection upon OC survival. The proposal is supported by extensive preliminary data and will be carried out by a highly qualified, multi-disciplinary research team.