Helicobacter pylori Eradication for Prevention of Metachronous Recurrence after Endoscopic Resection of Early Gastric Cancer

Literature search

MEDLINE (through PubMed), EMBASE, and the Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library were searched using common keywords related to H. pylori eradication after endoscopic resection of EGC in the development of metachronous gastric cancer (from inception to July 2014). Medical Subject Headings (MeSH) were used because all 3 databases permit searches using MeSH terminology. The keywords included 'Helicobacter' and 'metachronous neoplasms' using Boolean operators. Only publications on human subjects were sought, and the bibliographies of relevant articles were also reviewed to identify additional studies. The language of the publications was not restricted.

Selection criteria

We included randomized or non-randomized studies that met the following criteria: 1) the study was designed to evaluate the occurrence of metachronous gastric cancer after endoscopic resection of EGC in the target or control group, or 2) the study included a group that was given medicine for the eradication of H. pylori and a comparison group that was given placebo or no antibiotics. The exclusion criteria were as follows: 1) incomplete data, 2) review articles, or 3) abstract-only studies.

Selection of relevant studies

Two of the authors (C.S.B. and G.H.B.) independently evaluated the eligibility of all studies retrieved from the databases based on the predetermined selection criteria. The abstracts of all identified studies were reviewed to exclude irrelevant articles. Full-text reviews were performed to determine whether the inclusion criteria were satisfied by the remaining studies. Disagreements between the two evaluators were resolved by discussion or by consultation with a third author (D.J.K.).

Assessment of methodological quality

The methodological quality of the enrolled studies was assessed using the Risk of Bias table (RoB) for randomized studies and the Newcastle-Ottawa Scale (NOS) for non-randomized studies. The RoB was assessed as described in the Cochrane handbook by recording the method used to generate the randomization sequence, allocation concealment, the determination of whether blinding was implemented for participants or staff, and whether there was evidence of selective reporting of the outcomes (10). Review Manager version 5.3.3 (Revman for Windows 7, the Nordic Cochrane Centre, Copenhagen, Denmark) was used to generate the RoB table. The NOS is categorized into three parameters: the selection of the study population, the comparability of the groups, and the ascertainment of the exposure or outcome. Each parameter consists of subcategorized questions: selection (n=4), comparability (n=1), and exposure or outcome (n=3) (11, 12). Stars awarded for each item serve as a quick visual assessment for the methodological quality of the studies. A study can be awarded a maximum of nine stars, which indicates the highest quality. Two of the authors (C.S.B. and G. H.B.) independently evaluated the methodological quality of all studies, and any disagreements between the two evaluators were resolved by discussion or by consultation with a third author (D.J.K.). Subgroup analyses were performed according to the results of the methodological quality assessment.

Primary and modifier-based analyses

We investigated the efficacy of H. pylori eradication on the development of metachronous gastric cancer after endoscopic resection of EGC. Efficacy was defined as non-occurrence of metachronous gastric cancer (non-development of new gastric cancer at a previously uninvolved site in the stomach) in the follow-up endoscopic examinations. We also performed subgroup analyses based on the study design (randomized/non-randomized/non-randomized, prospective/non-randomized, retrospective studies), study setting (multicenter vs. single center studies), location of the studies (Korea vs. Japan), the definition of metachronous recurrence, and the methodological quality of the enrolled studies. In randomized studies (9, 13), which enabled both intention-to-treat (ITT) and per-protocol (PP) analysis, a meta-analysis based on the ITT analysis was performed, whenever possible. However, a separate meta-analysis was also performed, and both a cumulative analysis and a one-study-removed analysis were performed.

Statistics

Comprehensive Meta-Analysis (CMA) software (version 2.2.064, Biostat; Borenstein M, Hedges L, Higgins J and Rothstein H. Englewood, NJ, USA) was used for this meta-analysis. We calculated the RRs with 95% confidence intervals (CIs) using 2×2 tables based on ITT analysis, and whenever possible, from the original articles in order to compare the efficacy of H. pylori eradication in the prevention of metachronous recurrence after endoscopic resection of EGC. Heterogeneity was determined using the I² test developed by Higgins, which measures the percentage of total variation across studies (14). I² was calculated as follows: I² (%)=100×(Q-df)/Q, where Q is Cochrane's heterogeneity statistic and df signifies the degree of freedom. Negative values for I² were set to zero, and an I² value over 50% was considered to be of substantial heterogeneity (range: 0%-100%) (15). Pooled-effect sizes with 95% confidence intervals (CIs) were calculated using a random effects model and the method of DerSimonian and Laird because of methodological heterogeneity (16). These results were confirmed by the I² test. A fixed effects model using the inverse variance-weighted (Woolf's) method was used in the subgroup analyses, including cumulative and one-study-removed analyses, based on the assumption of a common effect size shared by the studies within each subgroup (17, 18). Significance was set at P=0.05 in both models. Publication bias was evaluated using Begg's funnel plot, Egger's test of the intercept, Duval and Tweedie's trim and fill, and Begg and Mazumdar's rank correlation test (19, 20, 21, 22, 23).

Ethics statement

This study was approved by the institutional review board of Chuncheon Sacred Heart hospital (2015-14). Informed consent was waived by the board.

Identification of relevant studies

Fig. 1 illustrates a flow diagram of how relevant studies were identified. A total of 225 articles were identified by a search of 3 core databases and a manual search of relevant bibliographies. In all, 49 duplicate studies and an additional 128 studies were excluded during the initial screening through a review of the titles and abstracts. The full texts of the remaining 48 studies were then thoroughly reviewed. Among these studies, 38 were excluded from the final analysis. The reasons for study exclusion during the final review were as follows: review article (n=10), incomplete data (n=12), abstract-only article (n=14), letter (n=1), or case study (n=1). The remaining 10 studies (2 randomized and 8 non-randomized studies) were included in the final analysis.

Fig. 1
Flow diagram for identification of relevant studies.

• Click for larger image
• Download as PowerPoint slide

Characteristics of studies

Within the 10 studies (9, 13, 24, 25, 26, 27, 28, 29, 30, 31), we identified a total of 5,914 patients with EGC or dysplasia. The clinical characteristics of the patients in the included studies are shown in Tables 1 and 2. The included studies were published between 1997 and 2014. Two randomized studies (9, 13) and 8 non-randomized studies, 2 prospective (24, 26) and 6 retrospective studies (25, 27, 28, 29, 30, 31) were included. All the studies were conducted in Asia, 5 studies (13, 28, 29, 30, 31) in Korea and 5 studies (9, 24, 25, 26, 27) in Japan. Four studies were conducted in a multicenter setting (9, 25, 26, 27), whereas the other studies were conducted in a single center setting (13, 24, 28, 29, 30, 31). All the included studies were written in English.

Table 1
Clinical data of included studies

• Click for larger image
• Click for full table
• Download as Excel file

Table 2
Clinical data of included studies (continued)

• Click for larger image
• Click for full table
• Download as Excel file

The age of enrolled patients ranged from 59 to 72 yr (median). Nearly all of the enrolled patients were diagnosed with EGC. However, 2 studies (13, 29) featured patients with EGC or dysplasia, and thus a separate analysis of only patients with EGC was impossible. To minimize the heterogeneity and to confirm the robustness of the main analysis, a subgroup analysis that excluded studies that included patients with dysplasia was performed. The endoscopic treatment method for EGC varied from strip biopsy and ablation to EMR or ESD. The duration of the follow-up ranged from 2 to 5 yr. Five studies (13, 27, 29, 30, 31) reported the definition of metachronous recurrence, whereas the remaining studies did not. Among the studies that reported the definition of metachronous recurrence, 3 studies (27, 30, 31) commonly defined the standard occurrence time as at least 1 yr after endoscopic resection of EGC, 1 study (29) defined the time course as at least 6 months after endoscopic resection of EGC, while the remaining 1 study (13) did not define the time standards. With respect to the H. pylori eradication regimen, the combination, dose, and duration varied. However, a proton-pump inhibitor with clarithromycin and amoxicillin for 1 or 2 weeks was the most frequently prescribed regimen (Table 2).

Efficacy of H. pylori eradication on the development of metachronous gastric cancer after endoscopic resection of EGC

The overall efficacy of H. pylori eradication on the development of metachronous gastric cancer after endoscopic resection of EGC showed a RR of 0.467 (95% CI: 0.362-0.602, P<0.001) in a random effect model-based meta-analysis of 10 studies (Fig. 2). According to the methodological quality assessment of the randomized studies (9, 13), the risk of random sequence generation and allocation concealment was low in both of the studies. However, blinding was not accomplished (open-label studies) in either of the studies (Fig. 3). In the non-randomized studies, the mean value of the awarded star was 7.375, 6 (2 studies), 7 (2 studies), 8 (3 studies), and 9 (1 study) (Table 3).

Fig. 2
Total efficacy of H. pylori eradication for the prevention of metachronous recurrence after endoscopic resection of EGC. H. pylori, Helicobacter pylori; EGC, early gastric cancer; ITT, intention-to-treat. The size of each square is proportional to the study's weight. Diamond is the summary estimate from the pooled studies (random effect model, including studies with ITT analysis).

• Click for larger image
• Download as PowerPoint slide

Fig. 3
RoB table for the assessment of methodological quality for randomized studies. RoB, risk of bias. (+) denotes low risk of bias, (-) denotes high risk of bias, (?) denotes unclear risk of bias.

• Click for larger image
• Download as PowerPoint slide

Table 3
Methodological quality of included studies measured by Newcastle-Ottawa scale

• Click for larger image
• Click for full table
• Download as Excel file

Subgroup meta-analysis

The cumulative meta-analysis of the enrolled studies in the order of the year published showed a consistent result (Appendix 1). A study by Naomi Uemura et al. (24) demonstrated a relatively low RR compared to other studies. However, the methodological quality of this study was the lowest among the included studies (Table 3) and this was the only study which did not prescribe a triple eradication regimen (omeprazole 20 mg/day with clarithromycin 400 mg/day for 2 weeks). Moreover, the definite follow-up duration was unclear (more than 3 yr and not exceeding 4 yr) (Table 2). The one-study-removed meta-analysis of the included studies in the order of the year published showed a consistent result and showed no specific outlier (Appendix 2). Among the included studies, 2 studies were randomized trials (9, 13), which enabled both an ITT and a PP analysis. The primary analysis was based on the result of the ITT analysis. Thus, a separate meta-analysis based on the size effects from the PP analysis was performed. The overall efficacy of H. pylori eradication on the development of metachronous gastric cancer after endoscopic resection of EGC showed a RR of 0.464 (95% CI: 0.360-0.599, P<0.001) in a random effect model-based meta-analysis of 10 studies (Appendix 3), which was not different from the result based on the ITT analysis (Fig. 2).

The subgroup analysis based on the study design showed consistent results (Appendix 5, 7, and 8), except for the analysis of the non-randomized prospective studies (Appendix 6). The subgroup analysis of the randomized studies showed a RR of 0.473 (95% CI: 0.277-0.809, P=0.006) (Appendix 5), and the subgroup analysis of the non-randomized retrospective studies showed a RR of 0.463 (95% CI: 0.346-0.622, P<0.001) (Appendix 7). The subgroup analysis of all non-randomized studies (prospective and retrospective studies) showed a RR of 0.465 (95% CI: 0.348-0.621, P<0.001) (Appendix 8). However, the analysis of non-randomized prospective studies showed a RR of 0.511 (95% CI: 0.101-2.583, P=0.417), which is not statistically significant (Appendix 6). However, non-randomized prospective studies, including the studies by Naomi Uemura et al. (24) and Akiko Shiotani et al. (26) showed the lowest methodological quality among the included studies (Table 3). In addition, a relatively small number of patients was evaluated in these studies (Table 1).

Among all of the included studies, 2 (13, 29) did not feature patients with only EGC but instead had a combined population with EGC or dysplasia (low grade or high grade). However, a subgroup analysis that excluded these studies that featured patients with dysplasia demonstrated a consistent result (RR: 0.422, 95% CI: 0.306-0.582, P<0.001) (Appendix 9).

Among the included studies, 3 studies (13, 29, 30) reported metachronous recurrence as not only metachronous cancer but also as metachronous dysplasia. However, a subgroup analysis that excluded these studies gave a consistent result (RR: 0.415, 95% CI: 0.287-0.599, P<0.001) (Appendix 10).

A subgroup analysis based on the study setting (multicenter vs. single center studies) also demonstrated consistent results (multicenter study: RR: 0.481, 95% CI: 0.321-0.721, P<0.001; single center study: RR: 0.457, 95% CI: 0.330-0.634, P<0.001) (Appendix 11 and 12).

Furthermore, a subgroup analysis based on where the studies were performed (Korea vs. Japan) showed consistent results (Japanese studies: RR: 0.465, 95% CI: 0.311-0.693, P<0.001; Korean studies: RR: 0.468, 95% CI: 0.337-0.651, P<0.001) (Appendix 13 and 14).

A subgroup analysis based on the definition of metachronous recurrence (studies that gave a definition/studies that did not give a definition/studies that did not define the time standard of at least 1 yr after endoscopic resection of EGC) showed consistent results (RR: 0.503, 95% CI: 0.360-0.702, P<0.001/RR: 0.422, 95% CI: 0.285-0.624, P<0.001/RR: 0.479, 95% CI: 0.304-0.755, P=0.002) (Appendix 15, 16, and 17).

A subgroup analysis based on the methodological quality (among the non-randomized studies, defined as high quality by NOS >7 vs. low quality for NOS ≤7) showed consistent results (RR: 0.503, 95% CI: 0.360-0.702, P<0.001 / RR: 0.370, 95% CI: 0.209-0.656, P=0.001) (Appendix 18 and 19).

Publication bias

A funnel plot for the included studies is illustrated in Fig. 4 (including studies that comprised the ITT analysis) and in Appendix 4 (including studies that comprised the PP analysis). These plots show a symmetrical shape. In the publication bias analysis that consisted of studies included in the ITT analysis, Egger's regression test revealed that the intercept was -0.91 (95% CI: -2.37-0.55, t-value: 1.43, df: 8, P=0.09 (1-tailed) and P=0.19 [2-tailed]). A trim and fill analysis showed that no study was missed or trimmed. The rank correlation test indicated a Kendall's tau of -0.27 with a continuity correction (P=0.14 [1-tailed] and P=0.28 [2-tailed]).

Fig. 4
Funnel plot for publication bias (including studies with ITT analysis). ITT, Intention-to-treat. Funnel plot of studies. The line in center is the natural logarithm of pooled RR, and 2 oblique lines are pseudo 95% confidence limits. RR: risk ratio.

• Click for larger image
• Download as PowerPoint slide

In the analysis that consisted of the studies included in the PP analysis, Egger's regression test revealed that the intercept was -0.91 (95% CI: -2.37-0.55, t-value: 1.44, df: 8, P=0.09 [1-tailed] and P=0.19 [2-tailed]). A trim and fill analysis showed that no study was missed or trimmed. The rank correlation test indicated a Kendall's tau of -0.27 with a continuity correction (P= 0.14 [1-tailed] and P=0.28 [2-tailed]).

Overall, there was no evidence of publication bias in this analysis.