Johannes Fahrmann

Ph.D.

MD Anderson Cancer Center

Instructor

PLCO (Learn more about this study)

2016-0018

Jun 2, 2016

Validation of serum autoantibody for pancreatic cancer early detection

Pancreatic cancer (PDAC) is the 4th leading cause of cancer death in the United States and one of the most lethal types of cancer with only about 5% of overall 5 year relative survival rate. Although resectable PDACs are associated with better survival, only about 10% of PDAC patients present with localized disease. Given the low incidence of PDAC, imaging modalities are not suitable for screening in terms of throughput, cost effectiveness, or invasiveness. Therefore, development of minimally invasive, sensitive and specific blood-based early detection biomarker assays will have a very significant impact in reducing mortality and increasing overall survival rates of individuals predisposed to and diagnosed with PDAC. However, there are no blood-based tests to detect asymptomatic PDAC at early stages due to the lack of adequate sensitivity and specificity of biomarkers that have been tested.
MD Anderson Cancer Center (MDACC) has launched two major initiatives that address PDAC early detection. One is a new Center for Global Early detection (CGED), led by two co-investigators on this proposal Drs Sam Hanash and Ziding Feng. The other initiative led by Dr Anirban Maitra, also co-investigator on this project, aims directly at the development of innovative strategies aimed at reducing the mortality and morbidity associated with PDAC, taking advantage of multi-disciplinary teams that have been assembled and taking advantage of patient cohorts and biospecimen resources at MDACC.
Increased levels of circulating autoantibodies have been known to precede the development of cancer-associated symptoms. Therefore autoantibody levels have been proposed as potential diagnostic biomarkers for early detection of cancer. On this basis, we have identified a set of autoantibody biomarkers that can discriminate early stage PDAC subjects (n = 60) from healthy control subjects (n = 73), with utilizing the focused protein array consisting of 186 recombinant proteins for which there is evidence of their association with autoantibodies in solid tumors. The autoantibody panel consisting five autoantibodies yielded AUC of 0.890 (standard deviation = 0.040) after cross validation and significantly improved the performance compared to CA19-9 alone in discriminating early stage PDAC from healthy controls. Of note, the panel yielded 59% and 26% of sensitivity at 95% and 99% specificity, suggesting the potential usefulness of the panel in the screening setting.
Our next critical step is to validate our results in another independent pre-diagnostic cohort, so that we can move forward to retrospective and prospective screening studies. In addition, there is a need to assess and compare the relative contribution of different types of biomarkers side by side in the same sample sets, enabling us to make the best choice and to develop an optimal biomarker model. Therefore we will also aim to integrate findings from autoantibodies with findings from previously validated biomarkers such as CA19-9. Thus we view the PLCO samples as a most suitable sample set to validate the potential utility of the autoantibody panel we have assembled.

We have applied focused protein array technologies and identified a set of potential autoantibody biomarkers. The autoantibody panel significantly improved the performance in discriminating early stage PDAC from healthy controls compared with the performance of CA19-9 alone. Therefore the main goal of our proposed study is to determine if our autoantibody panel also can detect PDAC before clinical diagnosis among healthy subjects in an independent, pre-diagnostic sample cohort and to determine marker performance in relation to time to diagnosis. In addition, we will integrate protein biomarkers that were previously validated in the PLCO samples and refine the model to further improve the performance of the model.

Specific Aim #1: Validation of autoantibody panel for PDAC early detection using an established combination rule.
The primary objective of this study is to validate an autoantibody biomarker panel for PDAC early detection. Autoantibodies will be assayed with using the same focused protein array platform. A logistic regression model-based combination rule has already been established. We will apply this established combination rule in an independent PLCO serum sample set, consisting of about 120 pre-diagnostic PDAC cases collected within 5 years before diagnosis and 480 matched healthy controls (1:4 case: control ratio). We will also examine the association of performance of the model with time to diagnosis. A panel that has 25% sensitivity at the threshold corresponding to 98% specificity was set as the null hypothesis. Therefore our autoantibody panel will be validated if its sensitivity is better than 25% at 98% specificity.

Specific Aim #2: Integration and refinement of the panel for PDAC early detection.
Protein biomarker candidates have been extensively assayed by Lokshin and colleagues in PLCO samples, resulted in identification of CA 19-9, CEA, NSE, bHCG, CEACAM1 and PRL as potential biomarkers for PDAC early detection. Therefore we will integrate these protein biomarker candidates with autoantibodies and refine our biomarker panel for further improvement of the performance. We will consider multivariate logistic regression models with using penalized likelihood based model estimation methods such as LASSO or smoothly clipped absolute deviation. Clinical factors will be also integrated in the model as appropriate. We view a panel that has 50% sensitivity at the threshold corresponding to 98% specificity after cross validation as useful for stratifying risk and prompting subjects with a positive test to undergo a further work-up.