A total of 265 patients who had undergone resection of a primary gastric tumor at our institute were enrolled in this study. Tumor specimens were fixed in 10% formaldehyde solution and embedded in paraffin. Sections (4-μm) were cut and mounted on glass slides. The pathologic diagnoses and classifications were made according to the Japanese Classification of Gastric Carcinoma (13). The median follow-up time for all 265 patients was 58 months (range, 1–177 months). The median follow-up time for the patients that succumbed to the disease was 25 months (n=88) compared with 75 months for surviving patients (n=177). Thirty-one patients were lost during more than 60 months of follow-up. Kaplan-Meier overall survival curves were calculated from the date of surgery.

A mouse monoclonal antibody which recognizes α-SMA (clone 1A4) and a mouse monoclonal antibody which recognizes vimentin (clone Vim 3B4) were purchased from DakoCytomation (Cambridge, UK). Normal rabbit serum, normal mouse immunoglobulin G, biotinylated rabbit anti-mouse immunoglobulin G, streptavidin-peroxidase reagent and diaminobenzidine were purchased from Nichirei Corp. (Tokyo, Japan).

Since there is no myofibroblast-specific immunocytochemical marker, characterization of human tumor-associated myofibroblasts is based on a combination of positive markers such as vimentin and α-SMA. The methods for the immunohistochemical determination of α-SMA and vimentin are described in detail in the manufacturer's instructions. Briefly, the slides were deparaffinized in xylene and hydrated in decreasing concentrations of ethyl alcohol. The tissues were heated for 20 min at 105°C and at 0.4 kg/cm² by autoclave in Target Retrieval Solution (Dako Co., Carpinteria, CA). The sections were then dewaxed and incubated with 3% hydrogen peroxide v/v in methanol for 15 min to block endogenous peroxidase activity. Next, the sections were washed in phosphate-buffered saline (PBS) and incubated in 10% normal rabbit serum v/v for 10 min to reduce non-specific antibody binding. The specimens were incubated with α-SMA antibodies (1:200) or vimentin antibodies (1:200) for 1 h at room temperature followed by three washes with PBS. Sections were incubated with biotinylated rabbit anti-mouse immunoglobulin G for 30 min, followed by three washes with PBS. Slides were treated with streptavidin-peroxidase reagent for 15 min and washed with PBS three times. Finally, the slides were incubated in PBS diaminobenzidine and 1% hydrogen peroxide v/v for 20 sec, counterstained with Mayer’s hematoxylin and mounted.

The tumor specimens showed various staining patterns against the anti-α-SMA and anti-vimentin antibodies. Vimentin-positive stromal cells were considered to be fibroblasts. Myofibroblasts were defined as fibroblasts which were positive for α-SMA staining. Smooth muscle was defined as being α-SMA-positive and vimentin-negative. The myofibroblast expression level was semi-quantitatively analyzed according to the percentage of fibroblasts showing α-SMA positivity: 0, 0%; 1+, 1–24%; 2+, 25–49%; 3+, ≥50%. Myofibroblast expression was considered positive when scores were ≥2+, and negative when scores were ≤1+ (Fig. 1). The slides were interpreted by two investigators without knowledge of the corresponding clinicopathological data.

The χ² test was used to determine the significance of the differences between the covariates. Survival durations were calculated using the Kaplan-Meier method and were analyzed by the log-rank test to compare the cumulative survival durations in the patient groups. The Cox proportional hazards model was used to compute univariate and multivariate hazards ratios for the study parameters. For all tests, a p-value <0.05 was defined as statistically significant. The SPSS software program (SPSS Japan, Tokyo, Japan) was used for the analyses.

Myofibroblast expression was positive in 92 (35%) of the 265 gastric carcinoma specimens, in 16 (13%) of the 119 early stage gastric carcinoma specimens, and in 76 (52%) of the 146 advanced gastric carcinoma specimens. The relationships between myofibroblast positivity and clinicopathological features of the tumors are shown in Table I. A significantly (p<0.001) high frequency of myofibroblast positivity was observed in the advanced gastric cancers (76 of 146) in comparison to the early stage cancers (16 of 119). Therefore, a statistically significant correlation was found between myofibroblast expression and scirrhous type gastric cancer (p<0.001). Myofibroblast positivity was also significantly present in patients with lymph node metastasis (p<0.001), positive cytology (p=0.005) and lymphatic invasion (p<0.001). ‘Cytology’ is defined as peritoneal lavage cytology at laparotomy as a standard method for the detection of free tumor cells. There was no statistically significant association between myofibroblast positivity and histological type, peritoneal dissemination and venous invasion.

The prognosis of patients with tumors positive for myofibroblast expression was significantly (p<0.001) worse than that of patients with tumors negative for myofibroblast expression (Fig. 2A). The 5-year survival of the patients with myofibroblast-positive tumors was 57% in comparison to 79% for those patients with negative tumors. The prognosis of patients who underwent a curative resection (R0) with myofibroblast-positive tumors was significantly (p<0.0210) worse than the prognosis of patients with tumors negative for myofibloblast expression (Fig. 2B). Univariate analysis revealed that myofibroblast expression (p=0.003), advanced cancer (p<0.001), peritoneal dissemination (p<0.001), venous invasion (p=0.042) and lymph node metastasis (p<0.001) were significantly correlated with patient survival (Table II). Multivariate analysis indicated that the macroscopic type and peritoneal dissemination were independent prognostic factors (Table III), while myofibroblast expression (p=0.290) was not an independent prognostic factor.