Expression of CD44 according to Clinicopathologic Characteristics of Gastric Cancer

The patients included in this study were selected from the data of a previous study that analyzed the clinicopathologic characteristics and prognosis of signet ring cell carcinoma [15]. The process of selecting patients is shown in Fig. 1. The records of 771 patients who underwent curative or palliative gastrectomy due to gastric cancer at Ewha Womans University Mokdong Hospital from May 1998 to December 2009 were retrospectively reviewed. Forty-eight patients with gastric cancer located in the cardia (n=34), fundus (n=6), or whole stomach (n=11) were excluded from this study to analyze the difference in CD44 expression between the antrum and corpus. Of the remaining 723 patients, the tumor was located in the antrum in 427 patients and in the corpus in 296 patients. Of this group, 157 patients were randomly selected with the same sex ratio as the entire group and with an equal division based on tumor location (antrum vs. corpus) and degree of differentiation (well and moderately differentiated vs. poorly and un-differentiated vs. signet ring cell carcinoma). Fourteen patients were excluded because their specimens were not able to be stained with CD44 antibody due to technical difficulties. Finally, 143 patients were analyzed in this study. Patients were followed up until either death or the cutoff date of December 31, 2012. Additionally, the previously reported pathologic results were retrospectively reviewed.

The clinical data included age, sex, tumor location (antrum or corpus), CEA level (normal range, 0 to 5 ng/mL), CA19-9 level (normal range, 0 to 27 U/mL), date of operation, date of recurrence, date of death, cause of death, and purpose of operation. Recurrence was defined as evidence of either a local recurrence or distant metastasis based on radiologic findings and medical records after curative surgery. The pathologic data included a classification as either early gastric cancer or advanced gastric cancer, World Health Organization (WHO) histological classification, histologic grade of the tumor, Lauren classification, venous invasion, perineural invasion, lymphatic invasion, and TNM stage. Early gastric cancer was defined as carcinoma limited to the mucosa and/or submucosa regardless of lymph node metastasis [16]. Gastric cancer was classified as either tubular adenocarcinoma, papillary adenocarcinoma, mucinous adenocarcinoma, or signet ring cell carcinoma according to the WHO histological classification, and tubular adenocarcinoma was further classified as either well-differentiated, moderately differentiated, or poorly differentiated [17]. The Lauren classification identified the gastric cancer as either intestinal type, diffuse type, or mixed type [18]. The TNM stage was classified according to the 7th edition of the American Joint Committee on Cancer/International Union Against Cancer TNM staging system [19]. The depth of the tumor (T stage) was classified as either Tis, T1, T2, T3, or T4; lymph node metastasis (N stage) as either N0, N1, or N2; and distant metastasis (M stage) as either M0 or M1. This study was approved by the institutional review board of Ewha Womans University Mokdong Hospital (2014-09-010-003).

Tissues were cut in serial sections (3.5-μm thickness), fixed in 10% formalin, embedded in paraffin, and stained using standard histological methods. For each tissue sample, one slide was stained with hematoxylin-eosin, and immunohistochemical staining of CD44 was performed on the remaining sections using standard procedures. An antigen retrieval process was performed with hydrocitrate buffer at pH 6.0, and the sections were incubated with an anti-CD44 monoclonal antibody (1:100 dilution; SC-7297, Santa Cruz Biotechnology, Santa Cruz, CA, USA) at 37°C for 15 minutes.

The stained slides were evaluated by one pathologist who was blinded as to the patients’ information. At ×200 magnification, three hot spots, defined as the most highly stained areas, were identified. Expression of CD44 was measured using intensity and proportion in the semi-quantitative method [20,21]. First, intensity was represented as either 0 (no staining), 1 (weak staining), 2 (moderate staining), or 3 (strong staining). A sample of the scoring intensity is shown in Fig. 2. The overall intensity score was calculated as the mean value of the three measured points. Second, the proportion was determined as the percentage of stained cells in the observed areas and scored from 0% to 100% in 5% increments. The average proportion measured in the three hot spots was calculated and categorized as either 0 (0% positive cells), 1 (1% to 25% positive cells), 2 (26% to 75% positive cells), or 3 (>75% positive cells). Based on a previous study on evaluating immunohistochemistry, each intensity and proportion score was added to obtain the final score [22]. A final score of 2 or less was deemed ‘CD44-negative’ and 3 or more as ‘CD44-positive.’

To analyze the relationships between CD44 expression and clinicopathologic features, Pearson’s chi-square test and Fisher’s exact test were used. The odds ratios (ORs) and corresponding 95% confidence intervals (CIs) of factors affecting CD44 expression were calculated from the coefficients in the logistic regression models. The overall survival was defined from the operation date to the date of death or the last follow-up date. The 5-year survival was evaluated from patients who were alive at 5 years after surgery. Disease-free survival was calculated for patients who either died or experienced a recurrence through the last follow-up date after curative surgery. Patients who died due to other causes were censored. The Kaplan-Meier method was used to analyze the overall, 5-year, and disease-free survival rates. Also, the differences between the curves were measured using the log-rank test. The hazard ratios (HRs) and 95% CIs of the factors affecting overall survival and disease-free survival were computed using the Cox regression model. All statistical procedures were performed using IBM SPSS Statistics ver. 20.0 (IBM Corp., Armonk, NY, USA). P-values <0.05 were considered statistically significant.

The baseline clinicopathologic data of the study subjects are summarized in Table 1. The mean age of the study subjects was 60.7±12.9 years (range, 27 to 84 years), and 91 (63.6%) were male. The primary tumor was located in the antrum in 74 patients (51.7%) and in the corpus in 69 patients (48.3%). The median overall survival was 45 months (range, 0 to 155 months). Among the 66 patients who died during the follow-up period, 47 (71.2%) died of gastric cancer. One hundred seventeen patients (81.8%) received curative surgery, and 26 patients (18.2%) received palliative surgery. Recurrence of gastric cancer occurred in 21 patients (14.7%). The prevalence of advanced gastric cancer was 59.4%. Following WHO classification for histopathology of gastric cancer, tubular adenocarcinoma was observed in 90 (62.9%) patients and signet ring cell carcinoma in 43 (30.1%) patients. According to the Lauren classification, intestinal, diffuse, and mixed type carcinoma was observed in 68 (47.6%), 54 (37.8%), and 21 (14.7%) patients, respectively. Thirty-two patients (22.9%) had vascular invasion, 63 patients (44.1%) had lymphatic invasion, and 41 patients (29.3%) had perineural invasion. Among the patients with intestinal type, 67.6% (46/68) were older than 60 years, and 57.4% of those with diffuse type (31/54) were younger than 60 years (P=0.021).

CD44 expression was analyzed to determine its association with various clinicopathologic factors (Table 2). In the univariate analysis, CD44 expression was significantly higher in patients with age ≥60 years (vs. age <60 years, P=0.003), those with intestinal type (vs. diffuse and mixed type, P=0.038), those with the presence of lymphatic invasion (vs. no lymphatic invasion, P=0.010), and those with lymph node metastasis (P=0.022). CD44 expression was not statistically different between the antrum and corpus (P=0.442). When only patients with intestinal type gastric cancer were analyzed, there was no significant difference in CD44 expression between the antrum and corpus (P=0.468). Additionally, there was no difference in CD44 expression according to sex, pathologic classification, vascular invasion, perineural invasion, T or M stage, CEA level, or CA19-9 level.

In multivariate analysis adjusted for age, Lauren classification, lymphatic invasion, and N stage (which were statistically significant in univariate analysis), older age (≥60 years) was independently associated with CD44 expression (adjusted OR, 2.27; 95% CI, 1.09 to 4.70; P=0.028) (Table 3).

We carried out Kaplan-Meier analysis with log-rank tests to evaluate the association of clinicopathologic factors with prognosis (overall, 5-year, and disease-free survival) in the study subjects. The associations between clinicopathologic factors and overall survival of gastric cancer patients are shown in Fig. 3. High expression of CD44 was significantly correlated with poor overall survival (P=0.006) (Fig. 3H). Also, older age (≥60 years), male gender, vascular invasion, lymphatic invasion, perineural invasion, and advanced TNM stage (TNM 2, 3, or 4) were associated with poor overall survival (P=0.003, P=0.017, P<0.001, P<0.001, P<0.001, and P<0.001, respectively) (Fig. 3A-3G). The Lauren classification was not significantly associated with overall survival (P=0.953). In the multivariate analysis using a Cox regression model, older age (≥60 years) (HR, 2.15; 95% CI, 1.21 to 3.82; P=0.009), male gender (HR, 1.88; 95% CI, 1.04 to 3.41; P=0.038), and lymphatic invasion (HR, 3.11; 95% CI, 1.46 to 6.64; P=0.003) were independent predictors for poor overall survival (Table 4). However, CD44 expression was not associated with poor overall survival in the multivariate analysis (P=0.735).

Kaplan-Meier analysis of 5-year survival revealed that older age (≥60 years), male gender, vascular invasion, lymphatic invasion, perineural invasion, advanced TNM stage, and CD44 expression were associated with poor 5-year survival (P=0.004, P=0.035, P<0.001, P<0.001, P<0.001, P<0.001, and P=0.006, respectively) (Fig. 4). The Lauren classification were not related with 5-year survival (P=0.802). The multivariate analysis using a Cox regression model showed that older age (≥60 years) (HR, 2.06; 95% CI, 1.15 to 3.71; P=0.016) and lymphatic invasion (HR, 3.22; 95% CI, 1.47 to 7.05; P=0.004) were significantly correlated with poor 5-year survival (Table 4).

Older age (≥60 years), vascular invasion, lymphatic invasion, perineural invasion, advanced TNM stage, and CD44 expression were factors associated with poor disease-free survival in the Kaplan-Meier analysis (P=0.019, P=0.002, P<0.001, P<0.001, P<0.001, and P=0.043, respectively) (Fig. 5). Sex and Lauren classification were not related with disease-free survival (P=0.235 and P=0.842, respectively). In the multivariate analysis using a Cox regression model, only lymphatic invasion was identified as an independent factor of poor disease-free survival (HR, 4.60; 95% CI, 1.68 to 12.59; P=0.003) (Table 4).