According to the preferred reporting items for Systemic Reviews and Meta-Analyses (PRISMA) protocols, we searched PubMed, EMBASE, KoreaMed, and Cochrane Library for studies comparing CT with ST for H. pylori infection.6 Combinations of the following key words were used in the search: “Helicobacter pylori”, “concomitant therapy”, “sequential therapy”, “disease eradication”, “therapeutics”, and “drug therapy”. The results of the search were reviewed by two authors (HJ Bae and JS Kim).

Studies were eligible if they included both CT and ST as the first-line therapy, and met the following criteria: (1) diagnosis of H. pylori infection by at least one of the following methods: rapid urease test, urea breath test, histology, or culture; (2) confirmation of eradication of infection by an appropriate follow-up test; (3) CT consisting of a PPI, amoxicillin, clarithromycin, and metronidazole given concomitantly for 7-14 days; and (4) ST consisting of PPI and amoxicillin for the first 5-7 days, followed by triple therapy, including PPI, clarithromycin, and metronidazole for the remaining 5-7 days. Case reports, letters, editorials, commentaries, reviews, and abstracts were excluded.

A manual for data extraction was made ahead of time, and two authors (HJ Bae and JS Kim) collected the data independently. The following items were extracted from the selected articles: study design; number of patients enrolled each study in total and in respective treatment groups; diagnostic methods of testing H. pylori infection before enrollment and after the end of treatment; drug regimens; number of patients who were successfully eradicated for corresponding regimens; and the incidence of adverse events. The extracted data were sorted out independently by the two authors, and disagreements were resolved by consensus. We tried to contact the original authors if we need additional information.

We used two different risk assessment tools according to the study designs. The risk of bias for randomized controlled trials (RCTs) was assessed using the tool developed by Cochrane collaboration.7 The criteria were randomization (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and other biases. Each part was recorded as yes (Y), no (N), or unclear (U), and any disagreements were resolved by consensus. The risk of bias for non-randomized studies was assessed by Newcastle-Ottawa scale.8 It was categorized to three parts: selection (composed of four individual questions), comparability of study groups (of two questions), and ascertainment of exposure or outcome (of three questions). Each question was scored for one point, and the studies receiving more than six points were considered as high quality.

The pooled odds ratio (OR) with 95% confidence intervals (CIs) were calculated. The eradication rates were analyzed on both intention-to-treat (ITT) population and per-protocol (PP) population. Full application of ITT analysis can only be performed for RCTs; thus, we excluded retrospective studies for ITT analysis. We evaluated the heterogeneity between the pooled estimates using Cochran's Q test and I² statistics. When there was high heterogeneity (p<0.1 or I²>50), the random-effect model should be used to combine the effect sizes of the included studies. However, a nonsignificant test only represents a scarcity of evidence for heterogeneity, as there may have been insufficient power to detect any heterogeneity. Therefore, we decided to take the pooled OR on a conservative approach and used a random-effects model. Subgroup analysis and meta-regression were also performed when needed. To evaluate the discrepancies of small scale studies and publication bias, we also conducted Egger's test and formed a funnel plot when necessary. Statistical analyses were executed by the aid of Comprehensive meta-analysis software version 2 (Biostat Inc., Englewood, NJ, USA).

Literature review was initially conducted on 218 studies. After excluding 193 studies after reviewing the summaries, 25 studies were close examined. Thereafter, 6 studies, including 4 RCTs 9101112 and 2 observational studies,1314 met our inclusion criteria for meta-analysis (Fig. 1).

The main characteristics of the studies are presented in Tables 1 and 2. Four studies were single-centered, whereas the remaining two studies were multi-centered design. The enrollment period ranged between 2010 and 2015. Two studies used pantoprazole, three studies used rabeprazole, and the remaining one study used lansoprazole for CT and ST regimens. Two studies used metronidazole 500 mg three times a day, while the rest used metronidazole two times a day. Treatment durations were 7, 10, and 14 days for CT, and 10 and 14 days for ST.

Tables 3 and 4 show summaries of the quality assessment of the included studies . Among the four prospective studies, allocation concealment was adequate in only one study. The two authors both assessed the literature determined that the two observational studies were relatively well-designed.

A total of 1897 Korean patients were included; of whom 917 were treated with CT and 980 were treated with ST. The pooled ITT eradication rates were 82.53% (95% CI: 71.0-94.1) for CT and 77.54% (95% CI: 64.8-90.2) for ST. The pooled PP eradication rates were 85.50% (95% CI: 78.6-92.4) for ST and 91.50% (95% CI: 83.8-99.2) for CT. The pooled OR of ITT eradication rates was 1.382 (95% CI: 1.031-1.853, p=0.031) (Fig. 2). The pooled OR of PP eradication rates was 2.114 (95% CI: 1.502-2.974, p<0.001) (Fig. 3). The pooled OR of PP eradication rates for RCTs was 2.102 (95% CI: 1.346-3.282, p=0.001).

There was some funnel plot asymmetry (Fig. 4), which was compatible with publication bias; however, statistical analysis using Egger's test showed no significance (p=0.426). These results should be interpreted with caution because the asymmetry test is probably underpowered as there were only six studies included in this meta-analysis.

No significant heterogeneity was observed among the studies for adverse events (p=0.1401, I²=42.21%). The pooled OR for adverse events was 1.171 (95% CI: 0.883-1.554, p=0.274) and the corresponding rates were 39.50% (95% CI: 30.3-48.7) for CT and 37.79% (95% CI: 31.7-43.9) for ST (Fig. 5). Low heterogeneity was observed among the studies for compliance to the medications (p=0.7056, I²=0.0%). The pooled OR for compliance was 0.734 (95% CI: 0.407-1.322, p=0.303) and the corresponding rates were 96.78% (95% CI: 95.1-98.4) for CT and 96.95% (95% CI: 94.0-99.9) for ST.

Except for one study, all other studies included a sequential therapy of 10 days. However, the duration of CT varied, from 7, to 14 days. Hence, we performed a subgroup analysis based on the duration of CT. Three studies compared 10 days of concomitant therapy with 10 days of sequential therapy. The pooled OR of eradication rates were 2.617 (95% CI: 1.484-4.615, p=0.001) favoring CT. Two studies compared 7 days of concomitant therapy with 10 days of ST. The pooled OR of eradication rates were 2.079 (95% CI: 1.073-4.029, p=0.030) favoring CT. We did not perform a subgroup analysis if the duration of CT was longer than the duration of ST. Moreover, because all but one study tested a sequential therapy lasting 10 days, a subgroup analysis according to the duration of ST was not performed.