Methods: Oral cancer and matched controlled saliva samples were obtained at UCLA under approved IRB. Three independent validations were done at UCLA followed by three consecutive independent validations by an NCI Biomarker Reference Lab (BRL). Sample size in each validation study ranged from 50 to 100 subjects totaling 324 cases and 281 controls. Seven mRNA biomarkers (IL8, SAT, IL1B, H3F3A, S100P, OAZ1 and DUSP1) were evaluated using qPCR. Three protein biomarkers (IL8, IL1B and M2BP) were evaluated using ELISA or immuno-blotting. Statistical analysis was performed using the R software for Wilcoxon rank sum test and computing ROC value. Meta-analysis was performed on individual markers as well as the original 4-mRNA marker prediction model (IL8, SAT, IL1B, OAZ1) across six different datasets.
Results: Three salivary biomarkers (IL8 mRNA, IL8 protein and SAT mRNA) were statistically significant across all six datasets. All others were statistically significant in most datasets. AUCs ranges from 0.66 to 0.83 while sensitivity ranges from 0.61 to 0.82 and specificity ranges from 0.63 to 0.84 for three markers across all dataset. Meta-analysis shows that sensitivity and specificity are homogenous across datasets for most markers. The prediction model based on original 4 mRNA marker panel (IL8, SAT, IL1B, OAZ1) were validated in all six datasets and comparable to the performance of the best panel of markers in each individual dataset.
Conclusions: Salivary biomarkers for oral cancer detection have been validated. Salivary mRNA and protein biomarkers are robust translational and clinical tools for highly specific and sensitive detection of OC. Supported by PHS Grant R01 DE017170.