Comparison of the prognostic value of a panel of tissue tumor markers and established clinicopathological factors in patients with gastric cancer.

J. Wiksten,J. Lundin,S. Nordling,A. Kokkola,C. Haglund
2008-07-01
Anticancer Research
Abstract:AIM The purpose of this study was to assess if the addition of a panel of tumor markers to established clinicopathological factors could improve the accuracy of 5-year outcome prediction in gastric cancer. The studied markers were chosen to represent different aspects of tumor biology. PATIENTS AND METHODS The expression of syndecan-1, tenascin-C, tumor-associated trypsin inhibitor (TATI), p53, p21 and bcl-2 was analyzed by immunohistochemistry informalin-fixed, paraffin-embedded specimens from 337 patients with gastric cancer. In addition, the DNA ploidy and S-phase fraction (SPF) were assessed by flow cytometry. RESULTS The loss of epithelial syndecan-1, strong stromal syndecan-1, the loss of stromal tenascin-C, the loss of tumor tissue TATI, high p53 and high p21 expression, aneuploidy and high (> or =7.6%) SPF were all associated with an unfavorable prognosis in univariate survival analysis. In multivariate survival analysis, p53 (hazard ratio (HR) 1.58; confidence interval (CI) 95% 1.16-2.16), p21 (HR 1.67; CI 95% 1.09-2.56) and DNA ploidy (HR 1.50; CI 95% 1.10-2.05) were independent prognostic factors, in addition to penetration depth (pT), lymph node status (pN), age at diagnosis and estimated radicality of surgery. The difference in prognostic accuracy between a base model with pT, pN, age and radicality of surgery (area under the curve (AUC) 0.898; CI 95% 0.86-0.94) and an extended model including p53, p21 and DNA ploidy (AUC 0.900; CI 95% 0.86-0.94) was not statistically significant (p=0.85). CONCLUSION In gastric cancer, p53 and p21 expression, as well as DNA ploidy, are independent prognostic factors in addition to standard clinicopathological factors. However, the established indicators of the extent of disease show an impressively high accuracy in 5-year outcome prediction and adding the examined tumor markers to the base model does not significantly improve the prognostic accuracy.
What problem does this paper attempt to address?