Introduction

The diagnostic criteria for “gastrointestinal cancer” differ between Western countries and Japan. In Western countries, gastric cancer (GC) is diagnosed when there is histological evidence of invasion into the lamina propria or beyond the submucosa [1–4], whereas in Japan and some other countries, diagnosis of GC is based on the degree of structural and cytological abnormality of the tumor glands [4–6]. In such a situation, lesions diagnosed as high-grade dysplasia (HGD) by Western pathologists have been diagnosed as GC by Japanese pathologists [2–4, 7, 8]. The differences in the two sets of diagnostic criteria are internationally well known. In the 1990s, gastrointestinal pathologists from all over the world proposed the Padova and Vienna gastrointestinal epithelial neoplasm classification as a means of interpreting the Western and Japanese diagnostic systems [9, 10]. However, the underlying difference between the two sets of diagnostic criteria has still not been resolved [11, 12], and they are still used independently in the West and East.

The choice of therapy is determined on the basis of pretreatment biopsy diagnosis. The lesion should be treated if it is diagnosed as cancer, but may be followed up if it is diagnosed as dysplasia. With improvements in endoscopy, small lesions can now be detected more easily than before [13, 14]. It has thus become increasingly important to diagnose pretreatment biopsy samples accurately. However, the Western diagnostic criteria, which are based on invasion, have a number of ambiguities, and similarly, the Japanese diagnostic criteria, which are based on structural and cytological abnormality, may have certain shortcomings. Accordingly, the main outcome of the present study was to compare the Western and Japanese diagnostic criteria in terms of accuracy for detection of GC. In addition, interobserver agreement of both diagnostic criteria was evaluated.

Materials and methods

Materials

We enrolled 233 consecutive patients with a postoperative diagnosis of submucosal invasive GC who underwent gastrectomy (142 cases) or endoscopic submucosal dissection (ESD, 91 cases) at Shizuoka Cancer Center Hospital (SCCH) between January 2010 and December 2011. All the patients underwent preoperative biopsy using endoscopic biopsy forceps (Radial Jaw 4, standard capacity; Boston Scientific, USA). The specimens obtained were fixed in 10 % formalin for 24 h at room temperature and then embedded in paraffin. Sections 3 μm thick were then cut and stained with hematoxylin and eosin (H&E).

Specimens of submucosal invasive GC were used as controls, as definite submucosal invasion is the most acceptable definition of cancer for all pathologists. Submucosal invasive GC is diagnosed as “cancer” by both the Western and Japanese diagnostic criteria [15]. Exclusion criteria were as follows: (1) no pretreatment biopsy specimen obtained at SCCH; (2) previous chemotherapy and/or radiation therapy for GC.

All patients provided written informed consent for gastrectomy or ESD, and this study was approved by the institutional review board at our hospital (institutional code number: 24-J81-24-1-3).

Pathological review

Before pathological review, the defined histological findings of invasion (described below) were reconfirmed by four clinicians (M.Y., T.Sh, K.K., T.Su) through the use of training slide sets showing typical invasive lesions. All the pretreatment biopsy specimens were independently reviewed by two experts in gastrointestinal pathology (K.K., T.Su) who were blinded to the clinical information. These pathologists reviewed only H&E-stained slides and no immunohistochemically stained materials. In each case, biopsy specimens were diagnosed according to the Western and the Japanese diagnostic criteria. Biopsy specimens were also reviewed for the presence of the muscularis mucosa. After individual review, the two pathologists arrived at a final diagnosis through discussion.

Western diagnostic criteria

Histological evidence of invasion is required for diagnosis of GC [1, 4]. There are no universally accepted criteria for invasive growth into the lamina propria [16]. Therefore, histological findings of invasion are defined according to the World Health Organization (WHO) classification and European management guidelines as follows: invasion into the muscularis mucosa or submucosa; lymphovascular invasion; infiltration of the stroma by single cells; small clusters of cells (small nest); small gland formation, and marked glandular crowding; excessive branching, budding, and a trabecular pattern; fused glands, cribriform glands; stromal response (desmoplasia) [6, 8]. For objective evaluation of invasion in this study, the Vienna Classification was used (Table 1) [9]. Category 5.1 or 5.2 was diagnosed as cancer because of evident invasion. Category 4.3 was also diagnosed as cancer because the lesion was suspected to be invasive. Categories 1, 2, 3, and 4.1 were diagnosed as noncancerous lesions because of lack of any evidence of invasion. Category 4.2 was not used in this study because it was created to understand the Japanese diagnosis of noninvasive carcinoma (carcinoma in situ) from a Western point of view.

Table 1 Vienna classification of gastrointestinal epithelial neoplasia
Full size table

Japanese diagnostic criteria

Structural and cytological abnormalities are necessary for diagnosis of GC regardless of the presence of invasion [5, 6]. Features of cytological abnormality included variation in nuclear size and shape; presence of hyperchromatic, large, spherical, vesicular nuclei; irregularly clumped chromatin; increased frequency of mitotic figures; pseudostratified nuclei; poor cellular differentiation; increased nuclear-to-cytoplasm (N/C) ratio; and loss of nuclear polarity. Features of structural abnormality included increased crypt complexity with crowding, branching glandular epithelium, fused glands, budding, a cribriform pattern, and variability of crypt size and shape [17, 18]. As an objective method of defining different levels of abnormality, the Group Classification of the Japanese Gastric Cancer Association was used (Table 2) [19]. Group 5 was diagnosed as cancer. Group 4 was defined as neoplasia or suspected cancer. Groups 1, 2, and 3 were diagnosed as noncancerous lesions.

Table 2 Japanese group classification
Full size table

Statistical analysis

The final pathological diagnosis based on resected specimens, which was submucosal invasive GC in all cases, was defined as the standard. Diagnostic agreement between pretreatment biopsy specimens and the corresponding resected specimens, and the interobserver agreement for each of the criteria, were evaluated.

Counts and percentages were calculated for all categorical variables. Interobserver agreement was calculated as the number of concordant lesions/total number of lesions. Accuracy was calculated as the number of cases diagnosed as carcinoma/total number of lesions. Statistical significance was defined as a P < 0.05 with two-tailed tests. Comparison of accuracy and interobserver agreement between the Western and Japanese criteria was performed using χ 2 test. The data were analyzed using the SPSS statistical analysis software package (IBM SPSS Statistics, version 21).

Results

Interobserver agreement

The diagnoses submitted by the two pathologists are detailed in Tables 3 and 4. Use of the Vienna Classification yielded 172 concordant cases and the Group Classification 225 concordant cases, the interobserver agreement being 73.8 % and 96.5 %, respectively (Table 5). There was a statistically significant difference in interobserver agreement between the Vienna Classification and the Group Classification (P < 0.001). There were 29 cases of disagreement between Categories 4.1 and 5.1 in the Vienna Classification, compared with 6 such cases between Groups 4 and 5 in the Group Classification.

Table 3 Interobserver agreement based on the Western criteria
Full size table
Table 4 Interobserver agreement based on the Japanese criteria
Full size table
Table 5 Comparison of interobserver agreement
Full size table

Accuracy of diagnosis of gastric cancer

The submitted final diagnoses are detailed in Tables 6 and 7. The pretreatment biopsy diagnosis was noncancerous in 44 lesions based on the Western criteria, and in only 1 lesion based on the Japanese criteria, and the diagnostic accuracy based on biopsy was 81.1 % and 99.5 %, respectively, the difference being significant (P < 0.001). In disagreement cases between diagnosis based on the Western criteria and Japanese criteria, 43 lesions were diagnosed as Category 4.1 in the Vienna Classification and Group 5 in the Group Classification (Fig. 1). Among the biopsy specimens, 187 (80.2 %) contained the muscularis mucosa, and invasion into the submucosa was detected in 25 (10.7 %).

Table 6 Final diagnosis based on the Western criteria and Japanese criteria
Full size table
Table 7 Diagnostic accuracy of the Western and Japanese criteria
Full size table
Fig. 1
figure 1

Disagreement case between the Western and Japanese diagnoses. a, b Pretreatment biopsy specimen. This lesion was diagnosed as high-grade dysplasia (Category 4.1) based on the Western criteria, and definite cancer (Group 5) based on the Japanese criteria. c, d The corresponding resected specimen. The lesion was submucosal invasive gastric cancer

Full size image

Discussion

A tumor is considered a “cancer” only if its cells have acquired the ability to invade into surrounding tissue [20]. Therefore, pathologists always attempt to identify morphological features that suggest such invasiveness. The aim of the present study was to clarify those diagnostic criteria that would be more accurate indicators of invasive ability.

The Japanese diagnostic criteria had a significantly higher accuracy rate. Overall, histological discrepancy between diagnosis based on biopsy and that based on the corresponding resected specimens was seen in 44 cases (18.8 %) when the Western criteria were used, and in only 1 case (0.4 %) when the Japanese criteria were employed. Other recent studies have reported that the discrepancy rate based on the Western criteria is 2.7–44.5 % [21].

One hundred eighty-seven cases (80.2 %) contained the muscularis mucosa in preoperative biopsy specimens, and submucosal invasion was detected in only 25 cases (10.7 %). Most of the biopsy specimens were thus mucosal. Based on the Japanese criteria, pathologists were able to detect GC irrespective of biopsy status. Forty-three cases (18.4 %) were diagnosed as HGD. Thus, in the West, many patients with submucosal GC might be diagnosed as HGD on the basis of biopsy.

Use of the Japanese criteria also produced better interobserver agreement. It might not be fair to compare the Vienna Classification and the Group Classification because the Vienna Classification had more categories than the Group Classification. However, the disagreement cases of Japanese criteria were seen in only eight cases. Six of these disagreement cases were Group 4 (suspected carcinoma) versus Group 5 (definite carcinoma). Diagnosis of Group 4 was the result of the small and fragmented state of the biopsy specimens in all cases.

The prevalence of GC has been historically high in Japan. In the 1960s, the age-adjusted mortality rate from GC (annual number of newly recorded deaths adjusted by the standard world population, per 100,000 people) was approximately 65 for males and 35 for females, and GC was the most frequent cause of cancer death [22]. On the basis of the differences seen between biopsy and resection specimens, Japanese gastrointestinal pathologists established the Japanese diagnostic criteria [23, 24]. Japanese pathologists consider that structural and cytological abnormalities are indicative of invasive ability. With regard to this point, there is an essential difference between the West and Japan.

Advances in medical equipment and implementation of the Japanese pathological diagnostic system have led to early detection of GC and relatively good therapeutic results [15, 25–27]. In Japan, the number of new GC patients in 2006 was estimated to be 116,911 [28]; more than half of the newly diagnosed lesions were early GC, i.e., carcinoma limited to the mucosa or submucosa [29, 30], for which endoscopic curative treatment is indicated [31–33]. Early detection and accurate diagnosis have a strong impact on cancer care. In fact, the mortality rate from GC in Japan has been decreasing dramatically [10, 11, 28]; in 2006, the age-adjusted mortality rate was 21.7 for males and 8.4 for females [22].

Our results suggest that use of the Western criteria made it difficult to detect evidence of invasion in biopsy specimens, and in fact diagnostic discrepancy was evident between biopsy and resected specimens in many cases. A number of previous studies have suggested that low-grade dysplasia (LGD) progresses to HGD [34–37], and that HGD becomes invasive GC within several months [38–43]. In 20–40 % of cases, diagnostic discrepancy between dysplasia evident at biopsy and GC observed in the resection specimen has been reported [34, 37, 44]. In most cases, the lesions were thought to be cancer at the time of initial biopsy [1, 44–47].

From a Western point of view, the final diagnosis rests on examination of the surgical or ESD specimen, because there is some possibility of cancer even if the diagnosis is LGD or HGD [21, 23, 36, 46–48]. Some studies have indicated that LGD and HGD should be treated, because genomic instability may already be present [49–52]. However, the European guidelines recommend reassessment with biopsy sampling and surveillance at 6-month to 1-year intervals for patients with HGD in the absence of endoscopically defined lesions [8]. According to whether there is evidence of invasion, there is a potential risk of underdiagnosing a lesion as LGD or HGD, even if it is submucosal or advanced GC. Moreover, there is a higher possibility of LGD if only the WHO classification is used for diagnosis because the features of invasion are more strictly limited in this classification. Pretreatment biopsy drives therapeutic decision making, and if diagnoses are inconsequent, clinicians run the risk of misleading patients with regard to prognosis and treatment.

The limitation was that findings of invasion were limited. However, as there is no complete consensus about these features among pathologists worldwide, we did not expect complete consensus because interobserver agreement was insufficient [53]. Basically, it is crucial to bear in mind that some cases are impossible to diagnose as GC if evidence of invasion is the only criterion.

In conclusion, the Japanese diagnostic criteria are considered to be significantly more accurate for detection of GC than the Western criteria. Early detection and accurate diagnosis offer patients a much higher chance of cure. These diagnostic criteria should contribute to reducing mortality caused by GC in the era of minimally invasive treatment.