Our present meta-analysis shows that the presence of iceA1
was significantly associated with PUD. Although several studies failed to show a positive association between the iceA
status and clinical outcomes, this meta-analysis supported the original report from 1998 
. However, this association was not very strong.
The mechanism of the development of PUD induced by iceA
remains unclear. iceA1
demonstrated sequence homology with a gene from Neisseria lactamica
IIIR, which encodes a CTAG-specific restriction endonuclease 
. However, studies on the genetic organization of the iceA
locus indicated that a full-length nla
IIIR-like open reading frame has only been observed in 10 (20.4%) of 49 iceA1 H. pylori
strains and only full-length iceA1
was a functional endonuclease gene 
. These data indicate that mutations in iceA1
are common, resulting in protein products with poor or no endonuclease activity. It remains to be determined whether iceA1
plays a role other than encoding an nla
III-like endonuclease or not.
Most isolates with an iceA2
allele could be divided into 2 types according to the presence of repeated sequences of 105 nucleotides and the size of the PCR products (229 bp for iceA2-
1 or 334 bp for iceA2
. Ashour et al.
reported that no association was observed between the size of the iceA2
amplicon and diseases 
. On the other hand, Kidd et al.
reported that the 334-bp iceA2
amplicon was more prevalent in strains from patients with PUD 
. These changes may translate into differential binding or function of the protein. The function of iceA2
The prevalence of mixed infection by iceA1
was also different in each study. The strains of mixed infections were excluded in several studies 
. Figueiredo et al.
reported that 36.7% of strains were positive for both iceA1
, and 53.8% of these strains also contained multiple vacA
. In our previous study, the rate of both iceA1
positivity was significantly lower in the United States than in Korea, Japan, and Columbia (4.3% vs. 20.0, 17.0, and 22.4%, respectively) 
. Multiple genotypes indicate the presence of multiple strains because there is no full-sequenced strain containing both iceA1
genes in Genbank (data not shown). It may be speculated that more than 1 strain may be acquired in childhood, especially in countries with a high prevalence of H. pylori
. Mixed infection by more than 1 strain in the same individual may reflect the capacity of H. pylori
to evolve genetic variations during long-term colonization from childhood 
. The high prevalence of mixed iceA
-type strains may obscure any potential relationship between the allele and clinical outcomes.
Fourteen studies combined DU with GU as PUD. However, DU and GU are linked to entirely different patterns of gastric inflammation, such that it would seem they should be examined separately 
. Tham et al.
indicated that in patients with H. pylori
infection, those with DU have a higher degree of acute and chronic inflammation in the gastric antrum and higher H. pylori
density than those with GU 
, which illustrates the different pathogenic processes of DU and GU. Therefore, a study to examine the roles of virulence factors needs to be conducted according to DU and GU, respectively. Furthermore, we should pay attention to the fact that patients with only gastritis at the time of endoscopy may develop ulcer diseases later in life and, therefore, may have been misclassified in the present study 
The association between the iceA
status remains unclear. Several virulence factors of H. pylori
correlated with the presence of cagA 
. In this study, we found that 15 studies examined the association between the iceA
status. As a result, most of them showed no association, indicating that iceA1
might be a risk factor for PUD, independent of cagA
. In addition, our recent meta-analysis showed that dupA
, which induces DU and has a suppressive action on GC, was significantly associated with DU 
. The association between dupA
has not been clarified yet; however, 1 study showed that there was no association 
. To confirm the significance of iceA
, it is better to perform a multivariate analysis adjusted for the cagA
status and other risk factors for peptic ulcer. However, unfortunately, we could not obtain the raw data from each study. It is difficult to perform a multivariate analysis adjusted for these factors without the raw data. In addition, most studies did not consider other risk factors in their papers. Further study is necessary to examine which factors are true virulence factors and which are just confounding factors. However, it might be better to hypothesize that these factors interact synergistically with each other and induce serious diseases than to discuss which of these factors is the most virulent. A recent study showed that groupings by multi-locus sequence typing (MLST) using 7 housekeeping genes were associated with the prevalence of GC 
, although we reported a problem of this interpretation 
. It may be better to classify H. pylori
according to the structure of the bacteria instead of each virulence factor.
iceA1 was weakly, but significantly associated with PUD especially DU, whereas iceA2 was inversely associated with PUD. A relationship between iceA and GC and GU was not found in this meta-analysis. It is possible that iceA is a discriminating factor for PUD, independent of cagA.