In this study, by evaluating a large series of patient, we demonstrated that those infected by CagA-positive H. pylori strains possessing more than one EPIYA C motif are at thrice-fold increased risk for developing gastric cancer. The strengths of our study include adequate sample size, identification of strains by means of PCR with confirmation by sequencing and adjustment for confounding factors by logistic regression analysis.
CagA is considered to be an important bacterial virulence factor associated with both gastric adenocarcinoma and duodenal ulcer disease [2, 5, 11, 12, 26]. The number and pattern of phosphorylation motifs seem to further stratify the risk associated with individual strains [18, 27].
It has been demonstrated that H. pylori CagA EPIYA patterns have a significant geographic variability and closely follow patterns of historical human migrations. EPIYA D is a characteristic Asian EPIYA pattern that virtually does not occur in the Western H. pylori strains . The Brazilians form an unique Western population because, despite the multiple origins and the consequent wide diversity of phenotypic appearance, there has been a substantial degree of inter-ethnic breeding and thus most individuals cannot be ascribed to any of the founding groups on the basis of genetic background, rather they carry about 33% of genes from each of the major races that historically composed the country (Caucasians, Africans and Amerindians) . With this background, it would be expected to find some CagA EPIYA D in our H. pylori strains, as it has been detected among Amerindians (in keeping with the theory that initially people from Asia populated the Americas migrating from the East Asia), but we did not detect any EPIYA D in our population.
Unfortunately, there are few studies in respect to the association between EPIYA C number and H. pylori associated diseases in Western populations with discordant results among them. Basso et al.  showed that higher number of EPIYA C segments was associated with gastric carcinoma in a Caucasian population from Italy, similarly to the results of Yamaoka et al.  evaluating American patients from Texas. Otherwise, no association was observed when Colombian patients were evaluated in the Yamaoka's study  in accordance with the results obtained by Acosta et al. , whereas Sicinschi et al.  observed associations between increased EPIYA C segments and precancerous lesions. Also, non-conclusive results published by Argent et al.  evaluating 44 strains from African patients the authors showed tendency of association between CagA with two or more EPIYA C segments and gastric cancer.
These differences may be explained by different study designs, sample size, populations and geographical diversity of H. pylori markers of pathogenicity, in respect to the CagA EPIYA pattern, highlighting the need of studying different geographical regions.
The results of this study showed that higher number of EPIYA C segments is associated with gastric cancer and with pre-malignant lesions, atrophy and intestinal metaplasia of the corpus mucosa in the group of patients with gastritis. In agreement with these findings, we also demonstrated that serum concentration of PGI was twice decreased in the patients infected by cagA-positive strains with two or three EPIYA C motifs. Because PGI is secreted by chief cells and mucous neck cells in the corpus and fundic glands, serum PGI levels reflect the functional and morphologic status of the oxyntic gastric mucosa; thus, gastric atrophic/metaplastic changes lead to decreased PGI serum levels [30, 31]. Although the corpus mucosa of patients with H. pylori associated duodenal ulcer is either mildly or not inflamed, the PGI serum levels were also decreased in duodenal ulcer patients infected by strains containing higher number of EPIYA C segments.
The results of the present study strengthen the potential role of CagA polymorphism in the development of gastric cancer in agreement with the results of the previous studies [18, 19]. However, we can not exclude the possibility that the genetic constitution of the host, more than the bacterium strain, might predispose to atrophic gastritis and the H. pylori strains carrying increasing numbers of EPIYA C repeats would have an advantage over other strains in colonizing the new gastric environment or alternatively a more complex interplay of both mechanisms.
In respect to duodenal ulcer, also the results of the studies are discordant [19, 25]. Our results are in agreement with those reported by Basso et al.  who also did not find association between the disease and the number of EPIYA C segments in an Italian population. Notably, none patient with duodenal ulcer of our cohort was colonized by CagA possessing three EPIYA C segments. As suggested by Yamaoka et al. , it is possible that strains with higher number of EPIYA C segments may be less resistant to the acid .
We also evaluated whether colonization by different strains (mixed infection) could be associated with disease outcomes. We found that gastric cancer patients were significantly more often colonized by mixed strains, whereas patients with duodenal ulcer had a trend toward less mixed strain colonization. One possibility is that patients with gastric cancer would have areas of gastric mucosa showing cancer transformation, alternating with areas of atrophy, intestinal metaplasia, dysplasia, and normal mucosa, each of them representing microenvironments that would be selectively advantageous to mixed infections [32, 33].