Meta-analysis: the effect of antibiotic resistance status on the efficacy of triple and quadruple first-line therapies for Helicobacter pylori

Authors


Dr L. Fischbach, University of North Texas Health Science Center, School of Public Health, 3500 Camp Bowie Blvd, CBH – 355 Fort Worth, TX 76107, USA.
E-mail: lfischba@hsc.unt.edu

Summary

Background  Information regarding the effects of drug resistance on therapies for Helicobacter pylori is limited.

Aims  To determine the effect of drug resistance on the efficacy of first-line treatment regimens for H. pylori and identify the most efficacious treatments in the presence of drug resistance.

Methods  We searched for studies using the keywords: ‘Helicobacter pylori’,‘resistance’ and ‘treatment’ or ‘therapy’. Multilevel meta-regression models were used to determine the effect of drug resistance on treatment efficacy.

Results  We analysed data from 93 studies with 10 178 participants. For triple therapies, clarithromycin resistance had a greater effect on treatment efficacy than nitroimidazole resistance. Metronidazole resistance reduced efficacy by 26% in triple therapies containing a nitroimidazole, tetracycline and bismuth, while efficacy was reduced by only 14% when a gastric acid inhibitor was added to the regimen. Quadruple therapies containing both clarithromycin and metronidazole were the most efficacious; >80% of H. pylori infections were consistently eradicated with these regimens.

Conclusions  Drug resistance was a strong predictor of efficacy across triple therapies for the eradication of H. pylori in adults. Resistance to either clarithromycin or metronidazole, but not both simultaneously, may be overcome by using quadruple therapies, especially those containing both clarithromycin and metronidazole.

Introduction

Over 50% of the global population is infected with Helicobacter pylori,1, 2 and in many adult populations in developing countries, the prevalence can be over 90%.3, 4 To combat this global disease, the latest Maastricht Consensus has recommended the first-line use of clarithromycin with metronidazole or amoxicillin, combined with a proton pump inhibitor (PPI), to eradicate H. pylori.5 However, the efficacy of clarithromycin-based therapies to treat H. pylori has decreased over time,6 while the prevalence of clarithromycin resistance has increased over time in some populations.7–11 Of added concern is the prevalence of metronidazole-resistant strains, which is estimated to be 30–40% in the US and Europe12–14 and up to 80% or 100% in the developing world.15–19 Although relatively few studies have been performed in developing countries where the burden of infection is greatest, treatment regimens are less efficacious in developing nations than in industrialized nations.6 This lower efficacy may result from the higher prevalence of metronidazole-resistant strains in developing countries,6, 15–19 since the efficacy of nitroimidazole regimens decreases as the prevalence of nitroimidazole resistance increases.6 To make matters worse, antibiotic treatment failure may be the primary risk factor for the occurrence of resistant strains of H. pylori.20–22 The purpose of this meta-analysis is to determine the effect of pre-treatment nitroimidazole resistance and/ or clarithromycin resistance on the efficacy of first-line treatment regimens to eradicate H. pylori in adults, and to determine the most efficacious treatments in the presence of nitroimidazole and/or clarithromycin resistance. We also identified other sources of variation in treatment efficacy within resistant strains for each treatment regimen.

Materials and methods

Data sources and study selection

Inclusion criteria were established a priori to minimize selection bias. We restricted our search to original studies in English or Spanish on the efficacy of triple or quadruple therapies aimed at eradicating H. pylori in adult humans. Mono and dual therapies were excluded as they have been previously shown to be ineffective in most adult populations.6 Study eligibility was also limited to those that reported triple or quadruple treatment efficacy within strata of subjects who were either all sensitive or all resistant to clarithromycin and/or a nitroimidazole. Additionally, treatment arms were only selected if they had at least four participants with H. pylori at baseline and followed a consistent, well-defined protocol regarding drug type, dosage, frequency and duration of treatment. The minimum number of four participants was chosen as simulations suggest that this is the smallest number of observations which can produce a distribution for the estimated measure of effect which approximates normality.23

We began searching for eligible papers by performing a Medline and a PubMed search with the following keywords: ‘Helicobacter pylori’,‘resistance’ and ‘treatment’ or ‘therapy’ for all years up to February 2007. We also searched for eligible studies using reference lists from previous meta-analyses and reviews on H. pylori and antibiotic sensitivity.

Potential sources of variation

Data identified as potentially affecting H. pylori eradication were extracted from the study report. This data consisted of date of published paper; type of sensitivity test; treatment regimen; drug names and dosing regimens; duration of treatment; geographic location; diagnosis; mean age; gender and type of analysis (per-protocol or intent-to-treat). Treatment efficacy was recorded separately by drug resistance status, and was defined as the proportion of H. pylori infections eradicated in participants tested at least 1 month after the end of the study.

Statistical methods

All analyses were performed with the treatment arm as the unit of analysis. We began by assessing the heterogeneity of treatment efficacy across treatment arms using graphs of the proportion eradicated after treatment and Pearson’s χ2 for independence to test for homogeneity, both within each regimen and stratified on drug resistance, as described previously elsewhere.6 When sample sizes were small, we used exact tests for homogeneity.

A treatment regimen was defined as a group of arms that used a similar combination of medications. For example, metronidazole and tinidiazole are similar drugs which can be classified as nitromidazoles. Likewise, omeprazole, lansoprazole, ranitidine and other PPIs and H2-receptor antagonists can be classified as gastric acid inhibitors. Therefore, the combination of metronidazole, clarithromycin and omeprazole, and the combination of tinidazole, clarithromycin and ranitidine both would be included in the regimen containing a nitroimidazole, amoxicillin and a gastric acid inhibitor.

Overall, we performed weighted multilevel meta-regression similar to our previous meta-analyses to determine how drug resistance affected treatment efficacy; sample size was used as the weight.6 We also used weighted multilevel meta-regression models to identify factors which influenced treatment efficacy within subject groups resistant to metronidazole or clarithromycin; level 1 corresponded to the treatment arm and level 2 corresponded to the study.24, 25 We calculated summary estimates of efficacy with weighted means within homogenous groups, or reported the entire range of efficacy when residual heterogeneity existed. Funnel plots and unweighted meta-regression models with sample size as a predictor were run to identify publication bias.26

Results

Overall results

We screened 1369 reports, of which 172 reported H. pylori treatment efficacy stratified on drug resistance status. Of these 172 reports, 16 dealt with second-line therapies, 54 had insufficient information, two were reviews, two treated children and five contained monotherapies or dual therapies, leaving a total of 93 original studies eligible for this analysis. None of the 172 reports were excluded due to the language the report was written in. These 93 studies contained 10 178 participants, eight treatment regimens, and 313 treatment arms that were homogeneous with regard to drug-resistant status (there were 155 drug-resistant and 158 drug-sensitive arms; Table 1).

Table 1.   Treatment regimens – number of studies – number of subjects
RegimenNumber of studiesNumber of treatment armsNumber of subjects P-value for the test of homogeneity overallP-value for the test of homogeneity for metronidazole-resistant armsP-value for the test of homogeneity for clarithromycin-resistant arms
  1. A, amoxicillin; B, bismuth; G, gastric acid inhibitor; C, clarithromycin; M, metronidazole; N, nitroimidazole; T, tetracycline; R, ranitidine.

  2. * Fewer than three arms were included; † Some studies included more than one regimen.

ANB27–32614319<0.00010.0002 
NTB27, 33–451431711<0.00010.0104 
ANG33, 40, 41, 46–6624571945<0.0001<0.0001 
CNG22, 31, 53, 54, 57, 60–64, 67–9034993128<0.0001<0.00010.1440
ACG38, 59, 62, 63, 67, 70, 78, 84, 87, 91–10524642556<0.0001 0.0512
MTBG34, 38, 39, 58, 99, 106–1161630889<0.0001<0.0001 
ACMG82, 117263650.5267**
CMBR31, 67, 111, 113, 118512265<0.00010.7995*
Total93†31310 178   

The studies were conducted in 36 different nations across five continents. Most of the treatment arms came from studies conducted in Europe (171 arms, 55%). There were 83 treatment arms from developed countries in Asia (27%), 27 from North America (9%), 16 (5%) from Australia or New Zealand and 16 (5%) from developing countries.

Most treatment arms (68%) came from randomized clinical trials and unblinded studies (77%). The average proportion of subjects with peptic ulcer disease in the study population was 67%, while an average of 30% of patients had non-ulcer dyspepsia. Antibiotic resistance status was determined primarily by an E-test (59%) and by agar dilution (19%), with the remainder by disc diffusion, antibiogram, dry plate method, oxoid disc, Stokes method or not reported. Out of all eight treatment regimens, six contained a gastric acid inhibitor such as a PPI or H2-receptor antagonist. The test for homogeneity revealed heterogeneity of overall efficacy in all regimens except for the one containing amoxicillin, clarithromycin, metronidazole and a gastric acid inhibitor (Table 1). Funnel plots and unweighted meta-regression models did not indicate publication bias by sample size for any of the regimens within drug-resistant strata.

The effect of nitroimidazole resistance on triple therapies containing nitroimidazoles

Amoxicillin, nitroimidazole and bismuth A total of 319 subjects were treated with amoxicillin, nitroimidazole and bismuth (ANB) in studies conducted in Europe (five studies, 12 arms) and Australia (one study, two arms).27–32 Treatment duration ranged from 3 to 16 days. Nitroimidazole was given as metronidazole in all but two of the arms, with tinidazole used in the remaining two. Nitroimidazole was given in dosages of 400–500 mg, two to five times per day.

Amoxicillin, nitroimidazole and bismuth was the least efficacious treatment in the presence of metronidazole resistance. Efficacy was heterogeneous across metronidazole-resistant arms (= 0.0002), where it ranged from 0% to 71% (Table 2). Metronidazole resistance lowered efficacy by 38% [95% confidence interval (CI): 15.1–61.7%; Table 2]. The weighted mean efficacy was 44% (95% CI: 7.8–80.4%) in the presence of metronidazole resistance, and 83% (95% CI: 67.7–97.3%) when only metronidazole-sensitive strains were present.

Table 2.   Percentage reduction in treatment efficacy due to resistance
Treatment regimenPercentage reduction in treatment efficacy due to resistance (95% CI)
Model without dual resistanceModel with dual resistance
Nitroimidazole overallClarithromycin overallNitroimidazole onlyClarithromycin onlyDual resistance
  1. A, amoxicillin; B, bismuth; C, clarithromycin; G, gastric acid inhibitor; C, clarithromycin; M, metronidazole; N, nitroimidazole; T, tetracycline; R, ranitidine.

  2. ANB,27–320, NTB,27, 33–45 ANG,33, 4038, 41, 46–66 CNG,22, 31, 53, 54, 57, 60–64, 67–90 ACG,38, 59, 62, 63, 67, 70, 78, 84, 87, 91–105 MTBG,34, 38, 39, 58, 99, 106–116 ACMG82, 117 and CMBR.31, 67, 111, 113, 118

Triple therapies
 ANB38.4 (15.1–61.7)    
 NTB26.0 (14.2–37.8)    
 ANG 30.0 (21.8–38.2)    
 CNG 18.2 (13.1–23.3) 35.4 (25.4–45.4)18.7 (13.4–24.0)33.9 (29.4–48.6)13.1 (−22.4 to 48.8)
 ACG  66.2 (58.2–74.2)   
Quadruple therapies
 MTBG 14.0 (5.4–22.6)    
 CMBR 1.5 (−1.7 to 4.7)12.7 (−27.0 to 1.6)0.9 (−2.0 to 3.8)3.0 (−2.0 to 8.0)44.5 (37.5–51.5)
 ACMGNegligibleNegligibleNegligibleNegligibleNegligible

The strongest predictor of variation for arms resistant to metronidazole was the daily nitroimidazole dose (Table 3). Within metronidazole-resistant strains, the weighted mean efficacy was 64% (56.2–71.8%) when at least 1500 mg of a nitroimidazole was given daily, while efficacy was only 17% (0.0–59.5%) when the daily dose was below 1500 mg (Table 4). Efficacy in the presence of nitroimidazole resistance was homogeneous across ANB treatment regimens containing at least 1500 mg of nitroimidazole daily (= 0.9248).

Table 3.   The effect of other sources of variation within nitroimidazole-resistant strains*
RegimenPredictors of efficacy within metronidazole resistant armsPercentage change in efficacy (95% CI)
  1. A, amoxicillin; B, bismuth; G, gastric acid inhibitor; C, clarithromycin; M, metronidazole; N, nitroimidazole; T, tetracycline; R, ranitidine.

  2. * The efficacy of CNG was homogeneous within clarithromycin-resistant strata, and there were no significant predictors of variation within clarithromycin-resistant strata.

  3. ANB,27–32, NTB,27, 33–45 ANG,33, 40, 41, 46–66 CNG,22, 31, 53, 54, 57, 60–64, 67–90 ACG,38, 59, 62, 63, 67, 70, 78, 84, 87, 91–105 MTBG,34, 38, 39, 58, 99, 106–116 ACMG82, 117 and CMBR.31, 67, 111, 113, 118

ANBDaily dose of nitroimidazole >1500 mg47.3 (29.2–65.4)
NTBNorway or the Netherlands39.8 (23.1–56.5)
Use of metronidazole instead of tinidazole48.1 (15.5–8.1)
ANGUse of lansoprazole−40.2 (−57.8 to −17.2)
Use of omeprazole−14.8 (−28.9 to −0.7)
CNGArms were also clarithromycin resistant−23.8 (0.015 to −0.49)
Clarithromycin and nitroimidazole 2 times/day17.3 (5.1–29.5)
ACGNone
MTBGNortheast Asia27.3 (9.4–45.2)
Lansoprazole or ranitidine−30.1 (−45.1 to −15.1)
<7 days of GAI−31.7 (−72.2 to 8.8)
ACMGNone
CMBRMacrolide resistance−47.8 (−63.7 to −31.9)
Table 4.   Mean percentage eradicated
RegimenImportant predictor groups within metronidazole-resistant strains Mean percentage eradicated (95% CI*)Strata within clarithromycin-resistant strainsMean percentage eradicated (95% CI*)
  1. A, amoxicillin; B, bismuth; G, gastric acid inhibitor; C, clarithromycin; M, metronidazole; N, nitroimidazole; T, tetracycline; R, ranitidine.

  2. * Ranges are given where residual heterogeneity existed.

  3. ANB,27–32 NTB,27, 33–45 ANG,33, 40, 41, 46–66 CNG,22, 31, 53, 54, 57, 60–64, 67–90 ACG,38, 59, 62, 63, 67, 70, 78, 84, 87, 91–105 MTBG,34, 38, 39, 58, 99, 106–116 ACMG82, 117 and CMBR.31, 67, 111, 113, 118

ANBNitroimidazole >1500 mg/day64.0 (56.2–71.8)  
Nitroimidazole ≤1500 mg/day16.7 (0.0–59.5)  
NTBWith metronidazole in Norway or the Netherlands87.9 (67.1–100.0)  
With metronidazole elsewhere50.4 (39.3–61.4)  
With tinidazole elsewhere0.0  
ANGWith famotidine84.6 (62.6–100)  
With omeprazole or lansoprazole27.0, 88.2  
CNGClarithromycin resistant and clarithromycin 3 times/day0.0Overall51.0 (41.5–60.4)
Clarithromycin resistant and clarithromycin 2 times/day16.7, 77.8  
Clarithromycin sensitive and clarithromycin 3 times/day25.0, 77.0  
Clarithromycin sensitive and clarithromycin 2 times/day46.9, 100.0  
ACG  Overall0.0, 50.0
MTBGTreatment with lansoprazole or ranitidine and given for >4 days in Northeast Asia76.0, 100.0  
Treatment with lansoprazole or ranitidine and given for >4 days elsewhere58.7 (44.8–72.5)  
Treatment with other proton pump inhibitors and given for >4 days elsewhere88.8 (80.7–96.9)  
Treatment with other proton pump inhibitors and given 4 days elsewhere 57.1  
ACMGOverall85.0, 92.6Overall95.6
CMBRClarithromycin sensitive97.0 (93.9–100.0)Metronidazole sensitive94.0
Clarithromycin resistant50.0Metronidazole resistant50.0

Nitroimidazole, tetracycline and a bismuth compound Treatment arms containing a nitroimidazole, tetracycline and bismuth (NTB) were given to 711 subjects from Europe (10 studies, 19 arms), Asia (one study, two arms), Australia/New Zealand (two studies, four arms) and the Middle East (one study, six arms).27, 33–45 The treatment duration for all arms ranged from 4 to 15 days. Nitroimidazole again was given as metronidazole in all but two of the treatment arms; tinidazole was given in the remaining two arms.

Nitroimidazole, tetracycline and bismuth performed slightly better than ANB overall, as well as across metronidazole-resistant treatment arms. NTB efficacy was heterogeneous across metronidazole-resistant arms (= 0.0104). The estimated weighted mean efficacy when no nitroimidazole resistance was present was 84% (95% CI: 75.0–92.7%). NTB efficacy was reduced by an estimated 26% (95% CI: 14.2–37.8%) in the presence of nitroimidazole resistance (Table 2).

Within metronidazole-resistant arms, efficacy was higher in Norway and the Netherlands and improved by using metronidazole instead of tinidazole (Table 3). Studies conducted in Norway or the Netherlands reported an efficacy 39.8% higher (95% CI: 23.1–56.5%) than other locations; NTB efficacy in Norway and the Netherlands ranged from 46% to 97% in the presence of nitroimidazole resistance. Nitroimidazole-resistant treatment arms that used metronidazole were 48% (95% CI: 15.5–80.1%) more efficacious than those which used tinidazole (Table 3); NTB efficacy ranged from 32% to 75% when metronidazole was used outside of Norway or the Netherlands, and was 0% when tinidazole was used (Table 4).

Amoxicillin, nitroimidazole and a gastric acid inhibitor Amoxicillin, a nitroimidazole and a gastric acid inhibitor (ANG) were used to treat 1945 subjects in North America (one study, two arms), Asia (seven studies, 17 arms), Australia (one study, two arms) and Europe (15 studies, 36 arms).33, 40, 41, 46–66 Most treatment arms used metronidazole (88%) as the nitroimidazole, and omeprazole (67%) as the gastric acid inhibitor; 18% used lansoprazole and 11% used famotidine. The duration of treatment with the antibiotics ranged from 5 to 15 days. Efficacy was heterogeneous across nitroimidazole-resistant ANG arms (< 0.0001; Table 1). Nitroimidazole resistance reduced efficacy by 30% (95% CI: 21.8–38.2%; Table 2). The estimated weighted mean efficacy for ANG was only 63% (95% CI: 55.6–70.2%) in the presence of nitroimidazole resistance.

The strongest predictor of efficacy variation for nitroimidazole-resistant arms was the type of gastric acid inhibitor used (Table 3). Although over 1/3 of all ANG treatments against nitroimidazole-resistant strains failed, treatment with famotidine, an H2-receptor antagonist, was somewhat more efficacious. Omeprazole lowered efficacy by an estimated 15% (95% CI: 0.7–28.9%), and lansoprazole lowered it by 40% (95% CI: 17.2–57.8%; Table 3). The estimated weighted mean efficacy for ANG arms containing famotidine was 85% (95% CI: 62.6–100%) in the presence of nitroimidazole resistance, while the resistant arms containing omeprazole or lansoprazole eradicated 27–88% of H. pylori infections (Table 4).

Clarithromycin, nitroimidazole and a gastric acid inhibitor The most commonly tested regimen in our analysis treated 3128 subjects with clarithromycin, a nitroimidazole and a gastric acid inhibitor (CNG; Table 1). CNG treatment arms came from studies conducted in North America (two studies, 11 arms), South America (one study, two arms), Europe (21 studies, 58 arms), Asia (nine studies, 24 arms) and Australia (one study, four arms).22, 31, 53, 54, 57, 60–64, 67–90 The duration of CNG treatment ranged from 5 to 14 days. Clarithromycin and a nitroimidazole were simultaneously given twice daily in 91% of the treatment arms.

Efficacy was heterogeneous across nitroimidazole-resistant CNG arms (< 0.0001; Table 1). Within CNG regimens, nitroimidazole resistance lowered efficacy an estimated 18% (95% CI: 13.1–23.3%; Table 2). The estimated weighted mean efficacy for CNG was 72% (95% CI: 66.0–76.2%) in the presence of nitroimidazole resistance compared to 91% (95% CI: 88.1–93.5%) in the absence of nitroimidazole resistance.

Within nitroimidazole-resistant arms, concomitant resistance to clarithromycin and the frequency of clarithromycin and nitroimidazole administration were the strongest predictors of variation in the efficacy of CNG (Table 3). Between 0% and 78% of infections were eradicated when the strain was simultaneously resistant to metronidazole and clarithromycin (Figure 1, Table 4). Efficacy increased by 17% (95% CI: 0.5–29.5%) when clarithromycin and metronidazole were both given twice daily (Table 3). In the presence of metronidazole resistance, but not clarithromycin resistance, the estimated weighted mean efficacy ranged from 47% to 100% for studies using both clarithromycin and metronidazole twice daily (Table 4).

Figure 1.

 Comparison between outcomes of antibiotic-sensitive strains, nitroimidazole-only-resistant strains, clarithromycin-only-resistant strains and dual-resistant strains, all treated with clarithromycin, a nitroimidazole and a gastric acid inhibitor (CNG). References are reported in the order they are presented in the figure. CNG-sensitive;22, 31, 35, 62, 64, 67, 72, 84, 89 CNG-macrolide resistant;31, 62, 67, 80, 80, 84, 86, 89 CNG-nitroimidazole resistant22, 31, 62, 64, 67, 72, 80, 80, 84, 86, 89 and CNG-dual resistant.64, 80, 80, 89

The effect of clarithromycin resistance on triple therapies containing clarithromycin

Clarithromycin, nitroimidazole and a gastric acid inhibitor The efficacy of CNG was reduced more by clarithromycin resistance than by metronidazole resistance (Table 2, Figure 1). Metronidazole resistance reduced CNG efficacy by 18% while clarithromycin resistance reduced it by an estimated 35% (95% CI: 26.4–46.4%; Table 2). The estimated weighted mean efficacy for CNG in the presence of clarithromycin resistance was only 51% (95% CI: 41.5–60.4%, Table 4).

Amoxicillin, clarithromycin and a gastric acid inhibitor Treatment with amoxicillin, clarithromycin and a gastric acid inhibitor (ACG) was given to 2556 subjects (Table 1). ACG treatment arms came from North America (two studies, six arms), South America (one study, two arms), Europe (10 studies, 21 arms), Australia/New Zealand (one study, two arms) and Asia (10 studies, 33 arms).38, 59, 62, 63, 67, 70, 78, 84, 87, 91–105

Efficacy was heterogeneous across clarithromycin-resistant arms (P = 0.0512). Clarithromycin resistance reduced the efficacy of ACG by an estimated 66% (95% CI: 58.2–74.2%, Table 2). We were unable to identify sources of variation within clarithromycin-resistant strains treated by ACG; efficacy ranged from 0% to 50% (Table 4).

The effect of metronidazole resistance on quadruple therapies containing metronidazole

Metronidazole, tetracycline, bismuth and a gastric acid inhibitor Quadruple therapies containing metronidazole, tetracycline, bismuth and a gastric acid inhibitor (MTBG) were tested on 889 subjects in North America (four studies, eight arms), Europe (eight studies, 14 arms), Asia (two studies, four arms), Australia/New Zealand (one study, two arms) and multiple locations (one study, two arms) (Table 1).34, 38, 39, 58, 99, 106–116 Treatment duration ranged from 1 to 14 days. Omeprazole, lansoprazole and pantoprazole were each used in 20% of the treatment arms, ranitidine was used in 33%, and rabeprazole was used in 7%.

Efficacy was heterogeneous across metronidazole-resistant arms (< 0.0001, Table 1). Metronidazole resistance reduced the efficacy of MTBG regimens by 14% (95% CI: 5.4–22.6%), with an estimated weighted mean efficacy of 79% (95% CI: 68.3–88.7%) against metronidazole-resistant strains (Table 2).

The strongest predictors of variation in MTBG efficacy for arms resistant to metronidazole were geographic location, duration of treatment and type of gastric acid inhibitor assigned (Table 3). The estimated efficacy across resistant strains was 27% (95% CI: 9.4–45.2%) higher for studies conducted in Mainland China or Hong Kong, and was 30% (95% CI: 45.1–15.1%) lower when lansoprazole or ranitidine was used (Table 3). Efficacy was only 57% when MTBG was given for only 4 days in the presence of metronidazole resistance (Table 4). Efficacy did not appear heterogeneous across 7- to 14-day metronidazole-resistant MTBG arms outside of Northeast Asia when omeprazole, rabeprazole or pantoprazole was used (= 0.3675), or with lansoprazole or ranitidine (= 0.5142). The estimated weighted mean efficacy in the presence of metronidazole resistance was 59% (95% CI: 44.8–72.5) when either lansoprazole or ranitidine was used (Table 4). On the other hand, the use of omeprazole, pantoprazole or rabeprazole increased efficacy to almost 90% (Table 4).

Amoxicillin, clarithromycin, metronidazole and a gastric acid inhibitor The two studies containing amoxicillin, clarithromycin, metronidazole and a gastric acid inhibitor (ACMG) were conducted in 1999, with 365 subjects from Japan (four arms) and the UK (two arms). ACMG was given for 5 and 7 days. We did not detect heterogeneity in efficacy measures (= 0.5267; Table 1). Overall, the estimated weighted mean efficacy was 94% (95% CI: 88.2–99.9%; Table 4).82, 117 There were not a sufficient number of drug-resistant treatment arms to examine funnel plots.

Clarithromycin, metronidazole, bismuth and ranitidine A total of 265 subjects were treated with clarithromycin, metronidazole, bismuth and ranitidine (CMBR) in Europe (three studies, seven arms) and Northeast Asia (two studies, five arms; Table 1). For all treatment arms, subjects were assigned to 400–500 mg of metronidazole and 250–500 mg of clarithromycin, with ranitidine and a bismuth compound twice daily for 7 days.

Heterogeneity was not detected across arms resistant to metronidazole (P = 0.7995), and metronidazole resistance lowered the estimated efficacy by only 2% (−1.7% to 4.7%; Tables 1 and 2). When strains were sensitive to clarithromycin and resistant to metronidazole, the estimated weighted mean efficacy was 97% (95% CI: 93.9, 100.0%; Table 4).31, 67, 111, 113, 118 There was not a sufficient number of drug-resistant treatment arms to examine funnel plots.

The effect of clarithromycin resistance on quadruple therapies containing clarithromycin

Amoxicillin, clarithromycin, metronidazole and a gastric acid inhibitor The effect of clarithromycin resistance on the efficacy of ACMG regimens was negligible, with 95% efficacy in the one clarithromycin-sensitive arm, and 96% in the one clarithromycin-resistant arm.

Clarithromycin, metronidazole, bismuth and ranitidine When strains were sensitive to metronidazole, clarithromycin resistance lowered the efficacy of CMBR by 13% (95% CI: −1.6% to 27%; Table 2). For metronidazole-sensitive strains, efficacy appeared homogeneous (= 0.7239) regardless of clarithromycin resistance. The mean weighted efficacy was estimated to be 98% (95% CI: 95.6–99.6%) for metronidazole-sensitive arms.

The effect of dual clarithromycin-nitroimidazole resistance on therapies containing clarithromycin and a nitroimidazole

Clarithromycin, a nitroimidazole and a gastric acid inhibitor Simultaneous resistance to metronidazole and clarithromycin lowered the estimated efficacy of CNG by 13% (95% CI: −22.4% to 48.8%; Table 2). In one treatment arm, Realdi observed that none of the subjects with dual resistance to metronidazole and clarithromycin had their H. pylori infection eradicated after treatment with CNG, while Wong observed eradication in only one of six subjects (17%).64, 89 The estimated weighted mean efficacy of CNG was 63% (95% CI: 52.9–73.6%) when subjects had strains resistant to only metronidazole or clarithromycin, but not to both.

Clarithromycin, metronidazole, bismuth and ranitidine The efficacy of CMBR was 50% when subjects had strains simultaneously resistant to metronidazole and clarithromycin (Table 4).89 When subjects were only resistant to only metronidazole or clarithromycin, but not both, the weighted mean efficacy was estimated to be 97% (93.9–99.2%).31, 63, 112, 114

Discussion

This meta-analysis shows that pre-treatment metronidazole resistance, and to a greater extent clarithromycin resistance, affect triple therapy efficacy. Metronidazole resistance lowered the efficacy of metronidazole-based triple therapies by 18–38%. Clarithromycin resistance had an even stronger effect, lowering efficacy by 35% and 66% in regimens containing clarithromycin, a gastric acid inhibitor, and either a nitroimidazole or amoxicillin, respectively. We observed a greater reduction in efficacy when treatment with amoxicillin and a gastric acid inhibitor was combined with clarithromycin in the presence of clarithromycin resistance than when amoxicillin and a gastric acid inhibitor was combined with a nitroimidazole in the presence of nitromidazole resistance (Figure 2). No triple therapy consistently eradicated H. pylori in at least 80% of metronidazole-resistant subjects or in at least 50% of clarithromycin-resistant subjects.

Figure 2.

 Comparison between outcomes of nitroimidazole-resistant strains treated with amoxicillin, a nitroimidazole and a gastric acid inhibitor (ANG) and clarithromycin-resistant strains treated with amoxicillin, clarithromycin and a gastric acid inhibitor (ACG). References are reported in the order they are presented in the figure. ANG33, 40, 41, 47, 47–51, 51–55, 55, 56, 56–61, 61–66and ACG.38, 62, 67, 70, 70, 91, 91–96, 96, 97, 97, 99, 100, 100–102, 102, 102–104, 104, 104, 105

We also observed that resistance to metronidazole or clarithromycin could be overcome to a great extent with quadruple therapies given for 5 or more days, especially those containing metronidazole and clarithromycin concomitantly. The efficacy of the metronidazole, tetracycline and bismuth regimen was improved by adding a gastric acid inhibitor. Triple therapies typically eradicated H. pylori in fewer than 80% of subjects with a resistant strain, while quadruple therapies containing both drugs eradicated the infection in >90% of subjects on average in the presence of single-drug resistance. The CMBR regimen showed a mean efficacy of over 95% when used against metronidazole-resistant strains, which was more than 20% higher than clarithromycin, metronidazole and a gastric acid inhibitor alone. Megraud suggested that CMBR's efficacy may be due to a synergistic effect between bismuth, ranitidine and the antibiotics.14 Dual resistance to clarithromycin and metronidazole was rare for quadruple regimens, and resulted in eradication in 50% or fewer subjects.

Prior to this meta-analysis, two other meta-analyses by van der Wouden and colleagues and Dore and colleagues, examined the effect of drug resistance on the efficacy of H. pylori treatments.118, 119 Both studies extracted data from the literature prior to 1998. The van der Wouden study looked solely at the effect of nitroimidazole resistance on the efficacy of two groups of nitroimidazole-based triple therapies, and nitroimidazole-based quadruple therapies overall.118 Consistent with our findings, they reported quadruple therapies to be the most efficacious treatment option in the presence of metronidazole resistance. The Dore study examined the effect of pre-treatment clarithromycin and metronidazole resistance on the efficacy of double and triple therapies overall, and not by regimen type.119 Dore and colleagues found metronidazole resistance and clarithromycin resistance to lower efficacy by 38% and 55%, respectively. Besides drug resistance, Dore et al. was not able to identify sources of variation in efficacy. Neither study reported efficacy estimates for each regimen.

Although current consensus recommendations include the use of triple therapies containing clarithromycin, amoxicillin or a nitroimidazole, and a PPI as first-line therapy,5 clarithromycin resistance resulted in an extreme reduction in efficacy, with reduced efficacy ranging from 34% to 66% (Table 2). Megraud observed a similar reduction in efficacy,120 as did McLoughlin.121 However, data from both our study and the Dore study suggest that use of clarithromycin-based triple therapies as a first-line treatment is inadvisable outside of populations where the current prevalence of clarithromycin resistance is very low, the person is at low risk of harbouring a clarithromycin-resistant strain, or susceptibility testing has confirmed clarithromycin sensitivity.119 Routine resistance testing before first-line treatment is still considered controversial, with cost still a primary consideration, along with the difficulties involved in ensuring continued H. pylori sample viability during transport and testing.120, 121

Likewise, nitroimidazole-containing triple therapies are best used when nitroimidazole prevalence is low or when nitroimidazole sensitivity has been confirmed. However, the prevalence of nitroimidazole-resistant strains of H. pylori ranges from 20% to 40% in developed countries, and is much higher in developing nations, ranging from 50% to 80%,15–19 and testing for nitroimidazole resistance may be unavailable, not feasible or yield inaccurate results, making the decision to use a nitroimidazole-containing triple therapy problematic.13, 14

Although the efficacy of nitroimidazole-based triple therapies against nitroimidazole-resistant strains is affected by the type of gastric acid inhibitor (ANG), complexity of therapy (CNG), strength of nitroimidazole (ANB), type of nitroimidazole (NTB) and concomitant resistance to clarithromycin (CNG) there appears that little can be done to improve the efficacy of clarithromycin-based therapies in the presence of clarithromycin resistance. If the prevalence of nitroimidazole resistance and clarithromycin resistance were equal, metronidazole-based therapies would undoubtedly be the better option. In most populations, however, clarithromycin resistance usually occurs less frequently than nitroimidazole resistance, but both can vary greatly between geographic locations.13–18

Given that quadruple regimens perform well even in the presence of drug resistance, quadruple therapies should be considered as a first-line therapy when previous studies have shown the prevalence of resistance to metronidazole and/or clarithromycin to be high in the local population. The topic of first-line quadruple therapies has been extensively discussed elsewhere in a meta-analysis by one of the authors (LF).122 Briefly, the results from that analysis revealed that quadruple therapies were typically similar in cost, adverse events and compliance relative to the recommended triple therapies.122

As with any meta-analysis, there are limitations that may affect validity and limit the generalizability of the findings. The primary limitation of this and other meta-analyses of H. pylori treatments is the scarcity of studies conducted in the geographical regions that are most affected by this infection, thus greatly limiting the generalizability of the findings. Although the vast majority of people living in the developing world are infected with H. pylori123–125 and metronidazole resistance tends to be higher in these populations,15–19, 126 only 5% of studies examining efficacy in the presence of drug resistance were conducted in a developing country. Future studies and meta-analyses should include subjects and data from these neglected populations.

The current meta-analysis is also limited by the small number of available trials. For example, only 23 of the 93 (25%) studies and only 48 (15%) of the treatment arms included a quadruple treatment. Furthermore, the ACMG regimen was only examined in two studies. Nevertheless, no previous meta-analysis has estimated the effect of clarithromycin resistance on the efficacy of quadruple therapies. Future studies may contradict the findings reported in this analysis, and therefore global recommendations may be premature.

Another limitation of all meta-analyses is publication bias, whereby larger studies with positive effects are more likely to be published than smaller studies or negative findings. We examined the symmetry of funnel plots and unweighted regression models for the effect of sample size on efficacy, and did not find an obvious publication bias, presumably because the estimated efficacy in the presence drug resistance, whether high or low, is an important finding and not likely to influence publication. Publication bias may still exist; however, if null or small effects of drug resistance on efficacy are not included in publications due to the perception that these findings would not be of interest to editors and readers. If this bias did occur it could explain, at least in part, the large effect of clarithromycin resistance on triple therapy efficacy observed in this analysis.

Conclusion

As the prevalence of nitroimidazole and clarithromycin resistance rises, standard triple therapies may no longer be adequate to eradicate H. pylori infection. Quadruple regimens, ordinarily used as a second line of defence, need to be considered as viable first-line options, particularly in areas of high drug resistance. Quadruple therapies that administer metronidazole and clarithromycin concomitantly are especially promising treatments in the presence of either clarithromycin or metronidazole resistance; however, further studies, especially in developing countries, are needed to confirm this finding.

Acknowledgements

Declaration of personal and funding interests: None.

Ancillary