Abbreviated Name

Colon Ref Set App: Disis (2008)

Lead Investigator

Disis, Mary L — University of Washington

Coordinating Investigator

Disis, Mary L — University of Washington

Involved Investigators

• Disis, Mary L — University of Washington
• Zheng, Yingye — Fred Hutchinson Cancer Center

Abstract

No abstract availalbe.

Aims

To validate the diagnostic value of a panel of serum antibodies (cathepsin D, p53, cyclin B1, topoisomerase IIalpha, IGFBP2, CEA) for the early detection of colorectal cancer.

Analytic Method

We will run ELISA blinded to case status, data along with our prediction of disease status for each sample, will be sent to EDRN-DMCC at Fred Hutchinson Cancer Research Center for statistical analysis. DMCC already has the combination role from our preliminary study. Our preliminary data suggested that at high specificity 90%, sensitivity is 57%. With n=50 in each group, we have more than 80% power to confirm this level of sensitivity against an inferior sensitivity of 37% at 90% specificity comparing CRC versus normal controls. We understand that the reference set is not powered for definitive validation but to help prioritization. We did not have the data on patients with adenoma but the performance on early-stage CRC suggests that it is possible our marker panel could detection early lesions like adenoma,

Comments

Goodell V, Lu H, Delong J, Disis ML. Antibody Immunity to a Panel of Oncogenic Proteins May Predict Presence of Colorectal Cancer and Stage of Disease. Manuscript in preparation.

Outcome

1. Assess the Classification Error of their prediction: The data set they used for prediction of case/control contains 130 subjects and 30 of them are cases. According to their predicted probability, we classify a subject as a case if his/her predicted risk ≥ 0.5 and a control vice versa. Classification Error= # of false prediction / total # of subjects=0.12. 2. When assessing individual marker’s performance, among 6 tumor markers, only catD showed some potential of discriminating cancers from polyps, and cancer from controls. None of the rest seems to be good. 3. When using their algorithm (2 methods), the single score calculated based on the Ag weights is significant at discriminating polyps from cancers and polyps from control (but in the reverse direction with polyps has lower score than controls). 4. We also did logistic regression with all 6 tumor markers. The risk models are doing well in terms of classification accuracy (cancer vs. control, polyps vs. control, and polyps vs. cancer). But after correcting for model Overfitting, none of them is significant. The results are not as promising as that observed in your preliminary study but catD and p53 do identify a small subset of patients with high specificity. However this set is quite small, about 20-30% of cancer cases. Your prediction model picked about a bit more than 50% of cancers with about 70-76% specificity (see the tables in the file if we combined polyps and controls as non-cancer controls to calculate specificity). Your markers seem to be cancer specific and are not related to polyps. The marker panel is still premature for a large validation study. catD and p53 definitely are elevated for some cancer cases but their sensitivites are not high enough

Secure Outcome

No secure outcome available.

Publications

• No publications available at this time for this protocol.

Biomarkers

• CCNB1
• CEACAM5
• CTAG1B
• CTSD
• ERBB2
• IGFBP2
• TOP2A
• TP53

Data Collections

• No data collections available at this time for this protocol.

 Team Project

Start Date

Oct 6 2008

Estimated Finish Date

Feb 6 2009

Finish Date

Aug 28 2008

Protocol ID

249

Protocol Type

Reference Set

Fields of Research

• Proteomics

Collaborative Group

G.I. and Other Associated Cancers Research Group

Cancer Types

• Malignant neoplasm of colon