Abstract 56P
Background
Biliary tract cancer (BTC) has a very poor prognosis. The only potentially curative treatment is surgery, but most patients are ineligible for this treatment due to advanced disease. Thus, biomarkers that can identify BTC at an early stage are needed. We aimed to identify protein signatures that could discriminate patients with BTC from non-cancer participants.
Methods
The study included 191 patients with all stages of BTC (gall bladder cancer (n=37), intrahepatic cholangiocarcinoma (CC) (n=92), and extrahepatic CC (n=62), and 250 controls (healthy (n=90), benign biliary tract disease (n=25), and other benign diseases (n=135)). We analyzed serum levels of immuno-oncology (I-O) related proteins using the Olink I-O assay (Olink Proteomics, Sweden) as well as CA19-9. To identify protein signatures and test their performance, the cohort was split randomly in a training (2/3) and replication set (1/3). Signatures were identified in the training set using logistic elastic-net (Lasso and Ridge) regressions. We generated signatures with decreasing number of proteins based on the proteins’ stability as predictors in Lasso regression. Signature performance was evaluated in both datasets using receiver operating characteristics (ROC) and area under the ROC curve (AUC). Sensitivity and specificity were calculated using best point.
Results
Sixteen unique protein signatures including 2 to 84 proteins were generated. All signatures included CA19-9 and chemokine (C-C motif) ligand 20 (CCL20). Overall, all signatures showed promising performance in both the training and replication set with AUC ranging from 0.95 – 0.99 for BTC vs. all controls. The lowest AUC was observed for signatures with less than 6 proteins. The best point sensitivity ranged from 91% -100% and specificity from 85% - 96% in the replication set. The best performing signatures achieved an AUC of 0.99 with a sensitivity of 96% and a specificity of 96%. All signatures identified patients with BTC independent of primary tumor location and stage (AUC ≥ 0.95).
Conclusions
We identified several protein signatures that could discriminate BTC patients from non-cancer participants with high sensitivity and specificity. A validation study in an independent cohort is ongoing.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
The Danish Cancer Society, Beckett Fonden, Fonden til fremme af klinisk cancerforskning, Tømremester Holms Mindelegat, Fonden til Lægevidenskabens Fremme.
Disclosure
All authors have declared no conflicts of interest.