Abstract
G A A b st ra ct s were 0.674, 0.727, 0.680, 0.739, respectively. Two peaks were found in the autofluorescence spectrum of gastric juice from all patients. The first peak (P1FI) was distinctively higher in gastric cancer than that in benign lesion (P<0.01). The AUC from ROC for P1FI was 0.872, the sensitivity, specificity, and accuracy of P1FI were 83.8%, 75.2%, and 78.8%, respectively, while the false positive rate was 24.8% (41/165). Furthermore, all of the four candidate genes and P1FI were evaluated in univariate and multivariate logistic regression (LR) models. The multivariate model included logarithmic transformations of E-cadherin, DAPK and P1FI, and the AUC from ROC was 0.933, which was statistically superior to either parameter alone (P<0.01), the sensitivity, specificity and accuracy were 89.7%, 86.7% and 87.9%, respectively, the false positive rate was 13.3% (22/165). Additionally, CART (Classification and Regression Trees) (Salford Systems) was carried out to set up the diagnostic model for gastric cancer and similar to the results of LR, P1FI, E-cadherin and DAPK were identified as tree nodes(Fig.1). The sensitivity, specificity, and accuracy were 81.2 %, 91.5%, and 87.2%, respectively. It is noticeable that the false positive rate was 8.5% (14/165). Conclusions The complementary use of both LR and CART models could be a promising approach to analyze multiple markers. The model, which combining the autofluorescene spectrum with the DNA promoter hypermethylation of E-cadherin and DAPK of gastric juice, could be a potential promising diagnostic model for better decision making for gastric cancer screening.
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call