A systematic comparison of triple therapies for treatment of Helicobacter pylori infection with proton pump inhibitor/ ranitidine bismuth citrate plus clarithromycin and either amoxicillin or a nitroimidazole
M. J. R. Janssen,
Department of Gastroenterology and Hepatology, University Medical Centre St. Radboud, Nijmegen, the Netherlands
Dr M. J. R. Janssen, Department of Gastroenterology and Hepatology, University Medical Centre St. Radboud, Nijmegen, the Netherlands. E-mail: firstname.lastname@example.org
Triple therapies with proton pump inhibitor/ranitidine bismuth citrate (RBC), clarithromycin (C) and either amoxicillin (A) or a nitroimidazole (I) are widely accepted as treatment for Helicobacter pylori infection. However, it is not clear which of these antibiotic combinations should be preferred.
To evaluate whether there is a difference in efficacy between triple therapies with proton pump inhibitor/RBC, clarithromycin and either amoxicillin or a nitroimidazole.
The literature was examined for randomized trials comparing proton pump inhibitor/RBC-C-A and proton pump inhibitor/RBC-C-I. Studies were grouped according to the type of acid inhibitor used (proton pump inhibitor or RBC) and differences between pooled cure rates were calculated.
Forty-seven studies were identified: seven using RBC, 39 using proton pump inhibitor, one using both. RBC-C-I was somewhat superior to RBC-C-A, although this difference only reached statistical significance in intention-to-treat analysis. Overall, proton pump inhibitor-C-I and proton pump inhibitor-C-A were equally effective, but in nitroimidazole-susceptible strains, proton pump inhibitor-C-I performed better, in nitroimidazole-resistant strains, proton pump inhibitor-C-A performed better. No serious side-effects were reported and pooled drop-out rates were equal.
In general, proton pump inhibitor-C-I and proton pump inhibitor-C-A are equally effective and therefore other factors such as local prevalence of resistant strains, cost of therapy and options for second-line treatment should determine which regimen should be preferred. When using RBC, the RBC-C-I combination is somewhat superior to RBC-C-A.
Helicobacter pylori infection is the main determinant in the pathogenesis of peptic ulcer disease and curing this infection is the treatment goal in patients diagnosed with peptic ulcers.1,2 After curing the infection, ulcer relapse rates are dramatically reduced and ulcer complications are prevented.3–7 The ideal eradication regimen should be simple, safe, cheap and of short duration. It should reach a cure rate of 80% in intention-to-treat and 90% in per protocol analysis.
Triple therapy with a proton pump inhibitor plus clarithromycin (C) and either amoxicillin (A) or a nitroimidazole (I) for at least 1 week is the currently recommended fist-line therapy.8 These regimens have been shown to be sufficiently effective and safe in many populations.9 However, it is not clear which of these combinations of antibiotics should be preferred. Furthermore, triple therapies containing ranitidine bismuth citrate (RBC) have emerged as a valid alternative, which has expanded the number of triple therapies we have to choose from. RBC is a co-precipitate of ranitidine hydrochloride and bismuth citrate. It is more soluble than a simple admixture of these compounds, which enables adequate penetration of the gastric mucous layer.10 A recent review by our group showed that RBC and proton pump inhibitor-based triple regimens were comparably effective when using the clarithromycin–amoxicillin combination, but when using the clarithromycin–nitroimidazole combination RBC was superior to proton pump inhibitor.11
Recent meta-analyses of the entire world literature showed comparable cure rates for triple regimens containing clarithromycin and either amoxicillin or a nitroimidazole.9,12,13 Laheij et al., when analysing data of 53 228 patients, found per protocol cure rates of 86% (proton pump inhibitor-C-A), 87% (proton pump inhibitor-C-I), 88% (RBC-C-A) and 91% (RBC-C-I).12 Pooled results of studies using the aforementioned regimens were used. This might not be the best way to assess differences in efficacy, because many factors are known to influence eradication and the prevalence of these factors differs among populations. The difference between these pooled data does not necessarily reflect the ‘true’ difference because it is also influenced by differences in treatment groups.
This problem does not arise when one only uses results from randomized comparative clinical trials, because such studies compare the efficacy of different regimens in the same population. This means that even if different studies use different patient groups, the pooled treatment groups are similar. Therefore, in this paper we compare clarithromycin/nitroimidazole- and clarithromycin/amoxicillin-containing triple regimens using randomized comparative studies only.
It is well-known that nitroimidazole resistance reduces cure rates of the proton pump inhibitor-C-I regimen.14–16 There is some evidence that using RBC instead of a proton pump inhibitor in combination with clarithromycin and a nitroimidazole improves efficacy in nitroimidazole-resistant strains.17,18 Because of the influence of nitroimidazole resistance, cure rates will be analysed separately in nitroimidazole-susceptible and -resistant sub-groups in this paper. Clarithromycin resistance reduces the efficacy of the proton pump inhibitor-C-I and the proton pump inhibitor-C-A regimen to a similar extent, but data are still scarce.14,19
The aim of this paper is to systematically compare the efficacy of triple regimens with clarithromycin and either amoxicillin or a nitroimidazole by using results from randomized comparative studies only. The results of this literature review should help physicians choose between the different triple therapy regimens presently available for the treatment of Helicobacter pylori infection.
An extensive Medline search was performed to identify reports of clinical studies available to August 2000. A manual review of all abstracts from the following major international meetings was also performed: the Digestive Disease Week of the American Gastroenterological Association (1996–2000), the European Helicobacter pylori Study Group meeting (1995–1999) and the United European Gastroenterology Week (1994–1999). In addition, all papers and reviews were evaluated to identify citations to other studies not yet included in the computerized literature search.
Only randomized trials comparing triple regimens containing proton pump inhibitor/RBC, clarithromycin and either amoxicillin or a nitroimidazole (metronidazole, tinidazole, ornidazole) were included. Extended use of a proton pump inhibitor or H2-receptor antagonist was allowed. Helicobacter pylori eradication had to be confirmed at least 4 weeks after treatment by at least one reliable method (culture, histology or urea breath test). Studies had to have at least an abstract in English, Dutch, German or French. Duplicate reports or studies obviously reporting results from the same population were eliminated and only the most recent abstract or full paper was used. All the studies which we identified are included in this review.
Data were extracted from each study and entered into a computerized database. Pooling was performed by combining the cure rates from intention-to-treat and per protocol analyses for the different treatment modalities. Studies were grouped according to the type of acid inhibitor used: proton pump inhibitor or RBC. Funnel plots were constructed as a crude measure to detect bias in the selection of studies.20 The weighted differences were plotted against study sample size. Differences in the pooled cure rates and drop-out rates between the different treatment modalities and 95% confidence intervals (95% CI) were calculated. The methods used to calculate the pooled differences weighted for effect size were based on the χ2 function for comparative analysis of two rates.21
A total of 70 studies comparing triple therapy regimens with clarithromycin and either amoxicillin or a nitroimidazole could be identified. In case of missing or conflicting data, we tried to contact the authors in order to obtain the necessary data. Fourteen studies were excluded because a prospective, randomized design was lacking;22–35 three because the number of patients in the different treatment groups was not mentioned;36–38 one because it contained conflicting data;39 and five studies because there was no clear statement that randomization was performed.40–44
Only 47 contained valid data. Twenty-five of these were published as a full paper, 22 were published as abstracts only. Some of the abstracts gave only limited information.
Thirty-nine studies compared triple therapy regimens containing a proton pump inhibitor, clarithromycin and either amoxicillin or a nitroimidazole.45–83 Seven studies compared regimens containing ranitidine bismuth citrate, clarithromycin and either amoxicillin or a nitroimidazole.84–90 One study compared all treatments listed above.91
All 47 studies used a randomized comparative design. Eight of these studies used a double-blind design;49,54,56,58,63,77,88,90 nine a single-blind design;45,47,52,60,61,73,76,81,83 nine studies used open-label treatment;46,67,71,78,79,84,87,89,91 and the remaining 21 studies did not mention whether or not a blinding method was used.
The studied populations varied widely: 12 papers studied patients with peptic ulcer;45,49,54–57,63,71,74,76,79,80 six studied patients with non-ulcer dyspepsia;65,81,87–90 and 26 included both ulcer and non-ulcer dyspepsia patients.47,48,50–53,58–61,64,67–70,72,73,75,77,78,82–86,91 In one paper, ulcer and non-ulcer dyspepsia patients were studied separately.46 One study included only patients with insulin-dependant diabetes.65 The remaining study included healthy individuals with Helicobacter infection.62
There were large differences in the number of patients studied in each trial. The median number of patients included in the intention-to-treat analysis was 49 per treatment arm (range: 18–192).
RBC-C-A vs. RBC-C-I
Eight comparative studies could be identified in the literature, comprising 17 treatment arms. The length of therapy was 5 days in one study;84 10 days in another;89 and 7 days in the remaining six studies. All studies used RBC 400 mg b.d. and amoxicillin 1000 mg b.d. Three studies used clarithromycin 500 mg b.d.;84,85,89 one used clarithromycin 250 mg t.d.s.;86 three used clarithromycin 500 mg b.d. in the amoxicillin arm and clarithromycin 250 mg b.d. in the nitroimidazole arm.87,88,90 One used clarithromycin 250 mg b.d. in the nitroimidazole arm and both clarithromycin 250 mg and 500 mg b.d. in the amoxicillin arm.91 All studies used a nitroimidazole dosage of 400 or 500 mg b.d., except for one which used 250 mg b.d.91
Intention-to-treat cure rates were provided by all eight studies, resulting in a pooled cure rate for the amoxicillin arm (RBC-C-A) of 81% (467 out of 574, range: 71–96%). For the nitroimidazole arm (RBC-C-I) the pooled intention-to-treat cure rate was 88% (499 out of 565, range: 78–94%). One study reached a significantly better result for the nitroimidazole containing regimen: Susi found cure rates of 73% (RBC-C-A) and 94% (RBC-C-I), difference 21% (95% CI: 11–31%) in favour of the nitroimidazole-containing regimen.91
The weighted difference between the pooled cure rates for the two treatment arms was statistically significant in favour of RBC-C-I (weighted difference 6% 95% CI: 2–10%) (Figure 1).
Per protocol analysis.
Per protocol cure rates were given for seven out of eight studies. For the amoxicillin arm the pooled per protocol cure rate was 88% (388 out of 441, range 82–100%). For the nitroimidazole arm this was 91% (397 out of 437, range 84–96%). In none of the studies did the difference between the cure rates for the different treatment arms reach statistical significance. The weighted difference between the pooled cure rates (2% in favour of RBC-C-I, 95% CI: – 2–6%) did not reach statistical significance either.
Only one study measured antibiotic susceptibility.85 In this study there was no statistically significant difference in cure rates for nitroimidazole-susceptible and -resistant strains.
Seven studies mentioned the number of drop-outs due to adverse events, but only five formally assessed type and severity of side-effects. It was not possible to analyse these data because all studies measured side-effects differently. None of the studies reported significant differences in the incidence and severity of side-effects between the treatment arms. Side-effects were mild and mostly self-limited, not leading to discontinuation of therapy. Pooled drop-out rates were equal for both regimens.
Proton pump inhibitor-C-A vs. proton pump inhibitor-C-I
Forty comparative studies could be identified in the literature comprising 102 treatment arms. Most studies (34) used a 1-week regimen. Two studies used a 5-day regimen;52,57 two studies used a 10 day regimen;46,70 one study used a 15 day regimen;74 and one study used both seven and 14-day regimens.73
Several proton pump inhibitors were used. Twenty-four studies used omeprazole 20 mg: 21 b.d.; two o.d.;60,64 and one b.d. in the amoxicillin arm and o.d. in the nitroimidazole arm.50 Six studies used lansoprazole 30 mg: in four studies, b.d.;48,52,61,69 in one o.d.;77 and in one b.d. in the amoxicillin arm and o.d. in the nitroimidazole arm.65 Six studies used pantoprazole 40 mg: in five studies b.d.;65,67,71,79,82 in one o.d.51 One study used rabeprazole 20 mg b.d.58 In three studies, more than one comparison between proton pump inhibitor-C-A and proton pump inhibitor-C-I using different proton pump inhibitor regimens was made. Uygun et al. used omeprazole 20 mg b.d. in one comparison and o.d. in the other.46 Paoluzi et al. used omeprazole 20 mg b.d. and lansoprazole 30 mg b.d. in different comparisons.73 Whereas Jonas et al. used both omeprazole 20 mg o.d. and lansoprazole 30 mg o.d.68
Thirteen studies used clarithromycin 500 mg b.d.;45,46,55,47–59,70–73,75,79,82 and one study used it t.d.s.67 Nine studies used clarithromycin 250 mg b.d.;48,51,52,60,61,64,68,69,77 and one used it t.d.s.65 Two studies made several comparisons with clarithromycin 250 mg and 500 mg b.d. regimens.63,91 Eleven studies used clarithromycin 500 mg b.d. in the amoxicillin arm and 250 mg b.d. in the nitroimidazole arm.47,49,50,53,54,56,62,65,76,78,83 Three studies compared several amoxicillin arms using both clarithromycin 500 mg and 250 mg b.d. with one nitroimidazole arm using either clarithromycin 500 mg b.d. or clarithromycin 250 mg b.d.74,80,81
Nearly all studies used amoxicillin 1000 mg b.d. The exceptions were: one study used amoxicillin 750 mg b.d.;75 one used 500 mg b.d.,60 one used 500 mg q.d.s.74 and another used both 500 and 1000 mg b.d.80
Most studies used a nitroimidazole dosage of 400 or 500 mg b.d. Two studies used 250 mg q.d.s.;48,51 two others used 500 mg t.d.s.67,74 and one study used 250 mg b.d.91
Cure rates were given for all studies but one.50 The pooled intention-to-treat cure rate was 79% (2351 out of 2976 range 24–95%) for the amoxicillin arm and 79% (2293 out of 2890, range 42–100%) for the nitroimidazole arm. In five individual studies the difference was statistically significant in favour of the nitroimidazole-containing regimen;52,63,65,68,79 in one study it was in favour of the amoxicillin-containing regimen.62 The weighted difference between the pooled eradication results for the two treatment arms (1% in favour of proton pump inhibitor-C-I 95% CI: – 1–3%) however, was not statistically significant (Figure 2).
Per protocol analysis.
Cure rates were given for all but five studies.65,69,70,75,91 The pooled per protocol cure rate was 83% (2096 out of 2512, range 33–100%) for the amoxicillin arm and 84% (2040 out of 2426, range 52–100%) for the nitroimidazole arm. In six studies, the difference between the cure rates for the different treatment arms reached statistical significance in favour of the nitroimidazole-containing regimen.52,57,63,68,77,79 The weighted difference between the pooled cure rates (1% in favour of proton pump inhibitor-C-I) did not reach statistical significance (95% CI: – 1–3%).
Few studies measured antibiotic resistance patterns. Seven studies described cure rates for the different regimens in nitroimidazole-resistant and -sensitive sub-groups: five studies contained both intention-to-treat and per protocol rates;45,47,51,52,61 one contained intention-to-treat rates only;56 and one per protocol only.58 One study contained these rates (per protocol) but only for the nitroimidazole-containing regimen.60
In the intention-to-treat analysis the pooled cure rate for the amoxicillin arm was 86% (238 out of 277, range: 40–97%) for the nitroimidazole-susceptible sub-group and 88% (138 out of 157, range: 72–100%) for the nitroimidazole-resistant sub-group. For the nitroimidazole arm these figures were 91% (249 out of 274, range 81–95%) and 76% (111 out of 147, range: 43–83%), respectively.
The cure rates of the amoxicillin-containing regimen were comparable in both groups (weighted difference: 1% in favour of the resistant sub-group, 95% CI: – 5–7%). The nitroimidazole-containing regimen however, performed considerably worse in the nitroimidazole-resistant sub-group, compared to the nitroimidazole-susceptible sub-group (weighted difference: 15% 95% CI: 8–22%).
In the nitroimidazole-susceptible group, eradication rates were slightly higher for the nitroimidazole-containing regimen than for the amoxicillin-containing regimen, but this difference did not reach statistical significance (weighted difference: 2% 95% CI: – 3–6%).
In the nitroimidazole-resistant sub-group, the nitroimidazole-containing regimen performed worse than the amoxicillin-containing regimen (weighted difference: 11% 95% CI: 3–20%).
The per protocol analysis (Figure 3) showed a similar pattern to the intention-to-treat analysis. However, in the per protocol analysis the difference between the amoxicillin and nitroimidazole-containing regimens for the nitroimidazole-susceptible sub-group reached statistical significance (weighted difference: 7% in favour of proton pump inhibitor-C-I, 95% CI: 2–13%).
It was not possible to analyse side-effect data because all studies measured side-effects in different ways. Twenty-one studies mentioned the number of drop-outs due to side-effects, 15 formally assessed type and severity of side-effects, but these data could not be analysed statistically because all studies used different methods to record and analyse side-effects. There were no obvious differences in the incidence and severity of side-effects between different treatment arms. Side-effects were mild and self-limited and mostly did not lead to discontinuation of therapy. Pooled drop-out rates were equal for both regimens.
The use of systematic reviews and meta-analyses combining the results of related, randomized clinical trials is becoming increasingly common and has come to play an important role in evidence-based medicine. They provide a framework for research synthesis, increase power and precision, provide an overall estimate and range of effect and identify greater-than-expected variability among study results.92,93
There are several sources of bias in meta-analysis: publication bias (studies showing a significant effect are more likely to be published); location bias (the probability of identifying relevant studies is influenced by their results, for example studies with significant results are more likely to be published in the English language and in journals indexed in one of the major databases); multiple publication bias (different papers representing duplicate representations of the same trial); and poor methodological quality of small studies.94 Although we performed an extensive search of the literature, included studies in other languages than English, checked thoroughly for possible multiple publication and checked studies for methodological weaknesses, some bias cannot be excluded. The construction of a funnel plot is a crude method to detect selection bias.20 In our study, the plot for the proton pump inhibitor group was symmetrical, indicating that the amount of selection bias is limited. In the RBC group, the number and size of the studies were too small to draw any conclusions.
Differences in dose, dosing frequency and therapy duration had to be ignored, because sub-groups would otherwise have got too small for a proper analysis. This could have influenced our results. Clarithromycin 500 mg b.d. gives significantly higher cure rates than clarithromycin 250 mg b.d. for proton pump inhibitor-C-A but not for proton pump inhibitor-C-I.9 Excluding studies using clarithromycin 250 mg b.d. in both arms did not change the results. Proton pump inhibitor dosing o.d. instead of b.d. might give lower eradication results.9,95 Excluding studies using proton pump inhibitor o.d. did not change the results. Furthermore, extending therapy from 1 to 2 weeks increases eradication rates by 5–10%;96 however, because nearly all studies used a 7-day regimen, excluding studies with other therapy durations did not influence the results. The differences between patients included by different studies may have caused variation between the results of different studies but both pooled patient groups had the same composition, because only randomized comparative trials were evaluated.
RBC-C-A vs. RBC-C-I
Our results show that the combination of RBC/clarithromycin with a nitroimidazole gives a higher cure rate than the combination with amoxicillin. In the per protocol analysis however, this difference did not reach statistical significance. This is mainly caused by the fact that the study by Susi, which showed the largest difference between cure rates of the different regimens in intention-to-treat analysis, was not included in per protocol analysis.91 Furthermore, the small number of patients in this analysis make it vulnerable for a type II statistical error. None of the studies reported differences in drop-out rates and type and severity of side-effects. Side-effects were mostly mild and self-limiting. These results are somewhat different from the results of the meta-analysis by Laheij et al., but this can be explained by the fact that we only used randomized comparative studies, as mentioned in the introduction.12
There is evidence that the combination of clarithromycin and a nitroimidazole with RBC performs well, despite the presence of nitroimidazole resistance, whereas the combination with a proton pump inhibitor significantly looses efficacy in case nitroimidazole resistance is encountered.15,17,18,97 Van der Wouden et al. used a 1-week regimen containing RBC, clarithromycin and metronidazole with comparable eradication rates in metronidazole-resistant and -susceptible groups (intention-to-treat: 98%, 59 out of 60, and 95%, 20 out of 21, respectively).18
In a recent review by our group, Van Oijen et al. systematically compared triple regimens with clarithromycin, a nitroimidazole and either a proton pump inhibitor or RBC, and showed that triple therapies with RBC, clarithromycin and a nitroimidazole give better cure rates than triple therapies with a proton pump inhibitor, clarithromycin and a nitroimidazole, although the difference reached statistical significance in intention-to-treat analysis only.11 Unfortunately, few studies analysed in this review measured antibiotic resistance, making statistical analysis of nitroimidazole-resistant/susceptible sub-groups impossible.
Resistance to clarithromycin probably affects both regimens equally, but the prevalence of clarithromycin resistance is still low in most countries and the impact on actual results is therefore limited.98,99
Because of these facts, it would have been better to assess results in this paper for nitroimidazole-resistant/susceptible sub-groups separately, but only one individual study reported cure rates in these sub-groups.85 Sung et al. reported no statistically significant difference in eradication rates for nitroimidazole-susceptible and -resistant bacteria, which is in concordance with the results by Van der Wouden et al.18
Our results support these findings. We show that when using RBC, the nitroimidazole-containing regimen gives better cure rates than the amoxicillin-containing regimen. This picture is much the same as for the proton pump inhibitor triple therapies in the nitroimidazole-susceptible sub-group. Therefore, we believe that the clarithromycin/nitroimidazole combination should only be used with RBC.
Proton pump inhibitor-C-A vs. proton pump inhibitor-C-I
Our results show that the pooled cure rates are equal for triple therapies with a proton pump inhibitor, clarithromycin and either amoxicillin or a nitroimidazole. None of the studies reported differences in drop-out rates and incidence and severity of side-effects. Side-effects were mostly mild and self-limiting. Therefore, in general, we cannot conclude that either regimen is superior to the other. This means that other factors, such as local prevalence of antibiotic-resistant strains, cost of therapy and options for second-line treatment, will determine which regimen should be chosen.
In intention-to-treat analysis, five studies showed a statistically significant difference in favour of the nitroimidazole-containing regimen, and one in favour of the amoxicillin-containing regimen. In per protocol analysis, six studies showed a statistically significant difference in favour of the nitroimidazole-containing regimen. There is, however, a probable explanation for this heterogeneity. These differences could easily have been caused by differences in the prevalence of nitroimidazole resistance in different populations. The prevalence of nitroimidazole resistance varies greatly from country to country. In European countries it is increasing and varies from 7% to 49%.100 In the United States it ranges from 20% to more than 50%, while in Asia and Central Africa it may be as high as 84%.99 Recent reviews show that nitroimidazole resistance decreases the cure rates of triple therapy regimens based on a nitroimidazole by 25–50%.14–16 Therefore, it is interesting to compare triple regimens with a proton pump inhibitor, clarithromycin and either amoxicillin or a nitroimidazole in nitroimidazole-susceptible and -resistant sub-groups.
In seven studies, these subgroups could be analysed. As expected, nitroimidazole resistance reduced the cure rate of the nitroimidazole-containing regimen but did not affect the amoxicillin-containing regimen. Therefore, in the nitroimidazole-resistant group, proton pump inhibitor-C-A gave significantly better results than proton pump inhibitor-C-I. In the nitroimidazole-susceptible group, proton pump inhibitor-C-I gave better cure rates than proton pump inhibitor-C-A, but this latter difference did not reach statistical significance in intention-to-treat analysis.
This means that in areas with a high prevalence of nitroimidazole resistance, the proton pump inhibitor-C-A regimen performs better, while in areas with a low prevalence of nitroimidazole resistance, the proton pump inhibitor-C-I regimen gives higher cure rates.
Data on cure rates in clarithromycin-resistant strains are still scarce, but it probably affects both regimens equally. Because the prevalence of clarithromycin-resistant strains is below 10% in most countries, the impact on actual cure rates is limited.14,45,60,101,102
Therapies based on a nitroimidazole might be cheaper because the clarithromycin dose can probably be reduced from 500 mg to 250 mg b.d. without reducing efficacy.9 This could be an important argument in countries where the cost of therapy plays an important role.
Another argument for making a choice between therapies is the number of options left for second-line therapy. When the first therapy fails, the options for second-line therapy are limited because Helicobacter may have acquired resistance against one or both antibiotics used in the first attempt.103 Amoxicillin resistance is virtually unknown but Helicobacter can be/become resistant against clarithromycin and nitroimidazoles. This means that there is a problem associated with regimens combining both clarithromycin and a nitroimidazole. Although these regimens are very effective, patients who are not cured will develop at least single, and usually double resistance, leaving no empirical second-line therapy.16,103–105 Culture and susceptibility testing are needed in patients who are not cured by this regimen.99 This is especially a problem in areas with high prevalence of nitroimidazole resistance, where more treatment failures will occur by using this regimen.15,16,103 Strains with double resistance cannot easily be eradicated by therapies using either clarithromycin or a nitroimidazole, rendering all standard therapies less suitable. Using a first-line therapy based on clarithromycin and amoxicillin cannot induce double resistance and therefore it will always leave open the (empirical) option of quadruple therapy (proton pump inhibitor–bismuth–tetracycline-I) based on a nitroimidazole as second-line therapy.
Based on the arguments listed above, physicians in different countries will eventually make different choices between these anti-Helicobacter pylori therapies.
In conclusion, our results show that when using a regimen based on RBC, combination with clarithromycin and a nitroimidazole gives higher cure rates than combination with clarithromycin and amoxicillin. When using a proton pump inhibitor-based regimen, no overall difference between cure rates for proton pump inhibitor-C-A and proton pump inhibitor-C-I was found. However, in the nitroimidazole-susceptible sub-group, proton pump inhibitor-C-I gives higher cure rates, and in the nitroimidazole-resistant sub-group, proton pump inhibitor-C-A gives higher cure rates. This means that factors such as local prevalence of resistant strains, cost of therapy and options for second-line treatment will determine which regimen should be chosen. The clarithromycin/nitroimidazole combination is cheaper but carries the risk of inducing double resistance. We recommend that when using clarithromycin and a nitroimidazole, these antibiotics should be combined with RBC instead of a proton pump inhibitor.
Tables summarizing all data used in this paper (including: study, treatment regimen, number of patients treated/cured in intention-to-treat and per protocol analysis) could not be printed because of their size, but interested readers can contact the authors to receive copies of these tables by fax or e-mail.