The study included 145 patients who underwent curative surgery for gastric cancer at Chonnam National University Hospital between January 1992 and December 1993. Formalin-fixed and paraffin-embedded tissue blocks were selected by viewing original pathologic slides and choosing blocks that show the junction between carcinoma and benign tissue. This allowed for direct comparison of carcinoma and benign tissue side by side after immunohistochemistry. Patient characteristics, including sex, age, histologic grade, stage and survival data, were obtained by medical records and pathologist and physician contact when necessary. No patient had received anticancer therapy prior to the operation. The histologic grade was classified according to the criteria of Lauren and the World Health Organization^{21, 22)}. The tumors were staged at the time of surgery by the standard criteria for TNM staging using the American Joint Committee on Cancer²³⁾. This study group comprised 99 males and 46 females. The mean age was 59.2±10.3 (mean±standard deviation) with a range from 28 to 79 years. The mean size of the tumor was 5.1±2.8 cm (mean±standard deviation) with a range from 0.5 to 15.0 cm.

All procedures for immunohistochemical staining were done by the Micro-Probe staining system (Fisher Scientific, Pittsburgh, PA) based on capillary action²⁴⁾. Paraffin sections, of 4 μm in thickness with mounted probe on slides, were immunostained with anti-mouse monoclonal antibodies by the avidin-biotin peroxidase complex method²⁴⁾. Sections were deparaffinized and rehydrated. They were immersed in 0.6% hydrogen peroxide for 5 minutes to block the endogenous peroxidase activity. A polyclonal antibody against VEGF (A-20; diluted 1:50; Santa Cruz Biotechnology, Santa Cruz, Calf, USA), a monoclonal antibody against CD34 (QB-END/10; diluted 1:25; Novocstra Lab., Newcastle, UK) and a monoclonal mouse antihuman p53 antibody (DO-7, diluted 1:100; Dakopatts, Glostrup, Denmark) were used as primary antibodies. The primary antibodies, in the aforementioned concentrations, were diluted in phosphate- buffered saline supplemented with 5% normal horse serum and 1% bovine serum albumin and then incubated with tissues for 15 minutes at 45°C. Anti-mouse immunoglobulin G (Sigma, St. Louis, MO) labeled with biotin was added as a secondary antibody for the detection of primary antibodies and the samples were incubated for 7 minutes at 45°C. After multiple rinses with universal buffer, streptavidin-alkaline phosphatase detection system (Biomeda, Foster, CA) was applied for 7 minutes. As the final step, the slides were developed for 7 minutes with the enzyme substrate 3 amino-9-ethyl carbazole (AEC, Sigma, St. Louis, MO). The slides were counterstained with hematoxylin solution for 1 minute (Research Genetics, Huntsville, AL). After dehydration, the tissue was sealed with a universal mount (Research Genetics, Huntsville, AL). For negative controls, the primary antibody was omitted and replaced with phosphate-buffered saline.

Staining intensity was classified from zero (no staining) to 3 (strong staining) and the percentage of staining area was classified as 0 for no positive staining of tumor cells, 1 for positive staining in <10% of the tumor cells, 2 for positive staining in 10% to 50% of the tumor cells, or 3 for positive staining in >50% of the tumor cells. A staining index was calculated as the product of staining intensity and staining area¹⁸⁾. Assessment of the staining was evaluated by two independent observers without knowledge of the clinical outcomes, such as tumor stage, grade and survival. Consensus scores were assigned for each case by reviewing the slides with discrepancies in scoring. All sections on which the two observers disagreed were re-evaluated and, after discussion, there was total agreement on the classification. The tumors were categorized as positive expression (staining index>4) or negative expression (staining index≤4).

Microvessels were highlighted with a monoclonal antibody against CD34 (QB-END/10; diluted 1:25; Novocstra Lab., Newcastle, UK) using the Micro-Probe staining system (Fisher Scientific, Pittsburgh, PA) based on capillary action. Microvessel quantification was performed according to an international consensus²⁵⁾. Any single brown-stained cell or cluster of endothelial cells that was clearly separate from adjacent microvessels was regarded as a microvessel (Figure 1). Neither vessel lumens nor the presence of red blood cells were used to define a micovessel. Large vessels with thick muscular walls were excluded from the counts. The stained sections were screened at ×40 magnification to identify the areas of the highest vascular density within the tumor. Vessels were counted in the 5 areas of highest vascular density at ×200 magnification. MVD was expressed as the mean number of vessels in these areas.

The χ²-test and Fisher’s exact test, where appropriate, were used to compare expression of the VEGF and p53 with various clinicopathological parameters. The relationship between VEGF or p53 expression and MVD was evaluated by the Mann-Whitney U test. Actuarial survival rates of patients were evaluated according to the Kaplan-Meier method and the differences were tested with a log-rank test. The Cox regression model was used to determine the prognostic significance of each parameter by a multivariate analysis. The statistical software program used was Statistical Package for the Social Sciences (SPSS/PC+ 10.0, Chicago, IL). A p value of less than 0.05 was accepted as statistically significant.

Normal gastric mucosa was not immunoreactive with an anti-VEGF antibody. VEGF was mainly localized to the cytoplasm or the membrane of the tumor cells (Figure 2). Tumor cells that stained strongly for VEGF were observed more often in the invasive front than in the tumor center. In cancerous tissues, positive expression of VEGF was 31.0% (45/145) (Table 1). Abnormal accumulation of the p53 protein was evident in the nuclei of tumor cells (Figure 3), and heterogeneously distributed. Based on our criteria, the positive expression of p53 in cancerous tissues was 35.9% (52/145) (Table 1).

The expression of p53 did not correlate with VEGF expression (p=0.959, Table 1). The correlation between VEGF or p53 expression and clinicopathological parameters is summarized in Table 2, 3. Positive expression of VEGF correlated significantly with lymph node and distant metastasis (p=0.040, 0.048, respectively, Table 2). There was a trend towards an association between the positive expression of VEGF and poor survival (p=0.087, Figure 4). However, no significant correlation was found between p53 expression and various clinicopathological parameters, including survival (Table 3, Figure 5).

The MVD for 145 tumors ranged from 23.0 to 182.0 with a mean MVD of 74.0±31.1. When a mean MVD value of 74.0 was chosen as the cut-off point for discrimination of the 145 patients into two subgroups, 74 patients were categorized as high MVD and 71 as low MVD. High MVD was significantly associated with the depth of tumor invasion (p=0.004) and also did correlate with distant metastasis (p=0.021) (Table 4). Moreover, the overall survival for patients with high MVD was significantly lower than that of low MVD (p=0.048) (Figure 6).

The mean MVD value of VEGF positive tumors was 78.5±31.3 and was a significantly higher MVD than that of VEGF negative tumors (p=0.028). However, the relationship between the status of p53 expression and MVD was not statistically significant (p=0.525). The mean MVD value of VEGF and p53 positive tumors was 75.6±30.3 and was higher than that of VEGF and p53 negative tumors, but the mean MVD value of both groups did not show statistically significant difference (p=0.147) (Table 5).

When the status of VEGF and p53, MVD and conventional clinicopathological parameters were analyzed by the Cox regression model, the stage and status of metastasis were found to be significant, independent, prognostic factors, while the status of VEGF, p53 expression and MVD were not significant (data not shown).