Serological assessment of gastric mucosal atrophy in gastric cancer

Patients' data was verified using the electronic patient documentation system of the Department of Gastroenterology, Hepatology and Infectious Diseases at the University Hospital of Magdeburg, Germany. Patients were included in case of complete records and availability of adequately stored serum samples for serological analysis (-80°C). In total, 118 patients met these criteria.

Only patients with either intestinal (n = 59) or diffuse type tumors (n = 59) according to the Laurén classification were included for further analysis, mixed type tumors were excluded (n = 12). Exclusion of mixed type tumors was done to analyse only clearly distinct tumor entities. Since "Laurén type" was considered as a potential confounding factor for the serological tests, we aimed for a strict comparison between intestinal and diffuse type gastric cancer.

Tumors were classified according to the location of the main tumor mass into carcinomas of the antrum, the corpus and the cardia. Cardia tumors were further subclassified according to the AEG-classification as proposed by Siewert and Stein in 1998 (AEG = adenocarcinoma of the esophagogastric junction) [19]. Supracardial carcinomas (AEG1) with a high likelihood of the distal esophagus being the site of origin have been excluded, as well as AEG2 with adjacent Barrett's mucosa. Tumors have been stratified into "proximal" and "distal" carcinomas of the stomach according to the location of the main tumor mass as described previously [7]. Tumors in the region of the esophagogastric junction were clearly classified as "proximal" and antrum-tumors as "distal". The group of corpus-carcinomas was subdivided. Carcinomas with the main tumor mass in the fundus and in the proximal third of the corpus were classified as "proximal", all more distally located corpus carcinomas (lower two thirds) as "distal".

Tumor allocation was assessed either by the primary endoscopy report including endoscopic ultrasound or by the pathological report of the resected specimen in case of gastrectomy. This classification of tumor allocation was chosen by convention of the authors. Tumor allocation was stratified prospectively to any statistical analysis.

Histopathological alterations of the gastric mucosa (IM and glandular atrophy) were assessed according to the updated Sydney system [20].

Biopsies had been taken in duplicate from the pre-pyloric antrum and the corpus (lesser and greater curvature). Cardia-derived samples were obtained directly below the Z-Line at the proximal end of the gastric folds. Additional samples were taken from the tumor itself and in some cases from the surrounding mucosa.

In patients that were referred to our hospital for gastrectomy with no further endoscopic examination being performed, histology was assessed on the surgical specimen.

Serum was prepared from 5-7 ml venous blood by centrifugation at 7.000x g at 4°C for 15 min, aliquoted in three individual cryotubes (each 1-1.5 ml) within 3 hours after taking blood. Samples were stored at -80°C until analysis. The concentration of anti-H. pylori IgG antibodies were analyzed using the Pyloriset EIA-G III (Orion Diagnostica, Finland) according to the manufacturer's instructions. According to the validation of the kit, a positive result was defined as ≥ 30 EIU, a negative result as < 30 EIU. The prevalence of anti-CagA antibodies was investigated using undiluted sera and the Helicobacter pylori Vira blot test kit (Viramed Biotech AG, Lich, Germany) according to the instructions by the manufacturer. Patients with positive CagA-status were classified as "H. pylori positive" even when anti-H. pylori IgG was below the cut-off level also including patients with "a serological scar". It has to be mentioned that by this definition not only patients with actual infection were classified as H. pylori-positive, but also patients with prior eradication therapy or those in which bacteria have disappeared during the progression of histological alterations. Since the so-called "point of no return" from which on eradication cannot prevent further progression of premalignant conditions is not defined yet, we believe that inclusion of all patients with a past history of H. pylori infection is justified.

The general characteristics of the study population are described in table 1.

Patients with intestinal type tumors were significantly younger than those with diffuse type GC (p = 0.029), and showed a higher proportion of female patients (p = 0.037). There was no significant difference in the distribution of proximal and distal GC between the two Laurén types. However, there were more corpus carcinomas in the group with diffuse type GC (p = 0.013). Both H. pylori and CagA status were similar between the two Laurén types. There was a higher prevalence for IM and glandular atrophy in case of intestinal type GC compared to diffuse type tumors (p = 0.002, p < 0.001, resp.). Table 2 gives an overview about for the distribution of mucosal changes in both Laurén types.

Patients with diffuse type GC presented with 1.56-fold higher PG1 levels compared to those with intestinal type tumors (127.9 ng/ml vs. 81.9 ng/ml). This difference was not statistically significant (p = 0.062; additional file 1). Values of the PG1/2-ratio were significantly higher in case of diffuse type carcinomas (10.4 vs. 6.8; p = 0.003; table 1; additional file 2). There was no difference concerning the serum levels of G17 and PG2 between GC of the intestinal and diffuse type.

Generally, serum levels of PG1, PG2 and G17 were similar between H. pylori-positive and -negative subjects. A trend towards lower PG1 serum levels (94.6 ng/ml vs. 141.6 ng/ml; p = 0.058) was observed for patients with positive CagA status compared to negative ones (table 1; additional file 1).

There was no difference in serum parameters between carcinomas with proximal and distal location.

Patients with presence of atrophy showed significantly lower levels of PG1 (p < 0.001) as well as a lower PG1/2-ratio (p > 0.001) than those without atrophy (Table 3; additional file 2). This was confirmed for patients with intestinal type tumors for both PG1 (p = 0.003) and the PG1/2-ratio (p = 0.028).

Patients with IM in the stomach demonstrated both lower PG1 values (p = 0.02) and a lower PG1/2-ratio (p = 0.006) than patients without metaplasia (additional file 2). Laurén specific comparison could confirm this result for PG1 levels in case of diffuse type carcinomas (p = 0.029).

There was no significant influence of the mucosal changes on the absolute levels of PG2 and G17 in the serum.

Serum levels of PG1 correlated inversely with the degree of gastric glandular atrophy (p < 0.001; r = -0.385). This relation was also demonstrated for the PG1/2-ratio (p = 0.001; r = -0.344) (additional file 3). According to the updated Sydney-classification, mucosal changes have been graded independently for antrum and corpus. Location-specific correlation-analyses revealed similar results for patients with atrophic foci in the antrum and the corpus (data not shown).

PG1 (p < 0.001, r = -0.351) and the PG1/2-ratio (p = 0.002; r = -0.285) correlated inversely with the degree of IM (additional file 3). Subgroup analysis did not show any difference between IM in antrum and corpus (data not shown).

There was no correlation of absolute serum values for PG2 with the degree of any mucosal alteration. G17 showed an inverse correlation with the degree of IM (not with atrophy) in the antrum (p = 0.044; r = -0.191; data not shown).

Both the degree of IM and the degree of atrophy correlated with the patients' age (IM: p < 0.001, r = 0.341; atrophy: p = 0.003, r = 0.272) (additional file 4).

All parameters mentioned above (Laurén type, localisation, IM, atrophy, H. pylori status, CagA status) have been evaluated by univariate ANOVA-analysis. Only the presence of atrophy (p = 0.022) and a positive CagA status (p = 0.013) had an influence on the serum levels of PG1. For none of the parameters analysed a significant influence on the PG1/2-ratio could be identified. Age was respected in this analysis but did not show an influence.

To analyse the subgroup characteristics of patients with pathological pepsinogen test, a cut-off below 70 μg/ml for PG1 in the serum, as well as a cut-off below 3.0 for the PG1/2-ratio was defined as pathological. Patients with positive pepsinogen serum test showed a higher prevalence of glandular atrophy (p = 0.003 and 0.022, respectively). Differences in further parameters did not show statistical significance. Please see table 4 for details.

The low incidence of glandular atrophy in our population is mainly due to the strict diagnostic criteria as mentioned in the methods section and is in line with data in the literature [21]. It has to be mentioned that we preferred applying the updated Sydney classification over more recently suggested staging systems for gastric mucosal alterations like OLGA or OLGIM [22, 23]. The latter systems are still under evaluation in clinical practice and the Sydney-scaled assessment is widely accepted.

Our data confirm the general association between gastric atrophy and decreased PG1 levels in the serum as well as a decreased PG1/2-ratio [13]. The same is seen in case of IM in the stomach which can partly be explained by the strong association between the presence of IM and atrophy. The consistence of our data is confirmed by the inverse correlation between the degree of both atrophy and IM and the serum levels of PG1 and the PG1/2-ratio which has been previously reported [24].

Variable cut-off-values for PG1 and the PG-ratio were applied in former studies [25], but levels below 70 μg/ml and 3.0, respectively, found most acceptance resulting in a sensitivity of 66.7-84.6% and a specificity of 73.5-87.1% for the detection of atrophic gastritis [12, 26‐28]. Although some authors presented similar values for the detection of GC [27‐29], especially the sensitivity for GC screening is significantly lower (36.8%-62.3%) compared to the assessment of gastric atrophy [15, 30, 31]. In our study, patients with a pathological pepsinogen-test revealed a prevalence of mucosal atrophy that was twice as high as in patients with normal serum pepsinogens.

It has been reported recently, that the results of the PG-test method are independent from the Laurén type [32]. In the present study, only for the PG1/2-ratio a significant difference between intestinal and diffuse type carcinomas was documented which was reported by other authors [7, 33]. However, in patients with diffuse type GC pepsinogen values in the serum gave no adequate information about the presence and degree of mucosal atrophy. There was no significant difference neither for PG1 nor for the PG-ratio between patients with and without atrophy. This could be an effect of the low prevalence of IM and atrophy in patients with diffuse type GC. Another explanation might be the diffuse infiltration of teh gastric mucosa by malignant tissue formations.

Serum analysis of PG1, PG2 and G17 can yield information about the distribution of IM and atrophy in the stomach [11, 29]. It is still under debate if the location of a gastric adenocarcinoma has a significant influence on the results of these serological tests. Pepsinogen serum values are not different between patients with proximal and distal GC [32, 34‐38]. But previous studies showing no association to cardia cancer lack of a precise distinction between adenocarcinoma of the distal esophagus and proximal GC with only the latter being associated to chronic atrophic changes [31, 39].

According to the updated Sydney-system the degree of glandular atrophy and IM was scaled separately for each antrum and corpus. Each localisation was evaluated separately in addition to the overall evaluation. The results of these analyses were similar so that location dependent bias is excluded in our study.

The influence of H. pylori infection on serum pepsinogens was mostly evaluated in patients without gastric malignancy. The altered pepsinogen levels were interpreted as reaction to the inflammation process [11, 18, 40, 41]. In the present study, only carriers of the more aggressive CagA-positive strains of H. pylori were affected in their PG1-levels. This influence is in line with studies that investigated risk-factor patterns that are associated with more progressive gastric disease [42]. It has been shown, that a positive CagA status is an independent risk factor that increases the H. pylori related odds ratio for the development of chronic atrophic gastritis [43, 44]. The CagA antigen as major trigger of the inflammatory response might have a further influence on pepsinogen levels at different stages of mucosal transition, even without overt histopathological changes. It has been demonstrated that the CagA status is an individual indicator for atrophic changes in the gastric mucosa independent from pepsinogen levels in the serum [21, 30, 45, 46]. Concerning the effect of H. pylori itself on the pepsinogen tests, a statistical bias cannot be excluded since 83.9% of the study population were H. pylori positive compared to only 16.1% H. pylori negatives.

The influence of CagA was confirmed by the univariate ANOVA-test analysing the influence of the pathobiological characteristics on the pepsinogen values. Besides the presence of atrophic changes only a positive CagA status revealed an independent influence on PG1 levels in the serum. Laurén type and location of the tumor have no influence on the serum-pepsinogen tests as reported before [47].

It is notable that in the present study H. pylori status was assessed only by serum-based methods. Therefore, not only patients with actual H. pylori infection were classified as H. pylori-positive but also patients with prior infection. It has been shown in studies on patients with atrophic gastritis that bacteria are cleared during the development of atrophy [48]. Thus, the infection cannot be detected anymore although the related histomorphological changes were initially caused by H. pylori-driven inflammation. Inclusion of patients with a "serological scar" of prior infection prevents underestimation of H. pylori-associated processes.

G17 and absolute PG2 serum values were not influenced by GC-related factors. It has been reported, that PG2 is upregulated inflammation-dependent in case of H. pylori induced gastritis, whereas the data about more severe stages of gastric disease are scarce and only reflected by changes in the PG1/2-ratio [11, 31, 45]. A recent study on a population in Shanghai reported a three-fold increased risk for GC in case of PG2 levels above 6.6 ng/l in the serum [34].

At the stage of gastritis with or without focal atrophic changes, G17-levels in the serum correlate not only with the degree of the mucosal damage but also with their location in the stomach. However, it is questionable if these associations can be transferred to the stage of invasive GC. Hansen et al. reported that there was no difference in G17-levels in patients with proximal and distal GC [34].