The Relationship among Previous Antimicrobial Use, Antimicrobial Resistance, and Treatment Outcomes for Helicobacter pylori Infections FREE

–The Editors

Helicobacter pylori is a common pathogen of the gastric mucosa, infecting up to 40% of persons in developed countries and up to 90% of individuals living in developing nations (1–2). Infection with H. pylori has been shown to be a major cause of gastric and duodenal ulcers and is also associated with chronic gastritis, mucosa-associated lymphoid tissue lymphoma, and adenocarcinoma of the stomach (3–6). Eradication of H. pylori has been reported in up to 95% of patients treated with a combination of antimicrobial agents (7–9). In the United States and elsewhere, antimicrobial resistance to metronidazole and clarithromycin is increasing; resistance to amoxicillin and tetracycline remains uncommon (10–12). Compared with persons with susceptible isolates, persons infected with clarithromycin- or metronidazole-resistant H. pylori have lower cure rates when treated with regimens containing these antimicrobial agents (12–15). In studies that have addressed risk factors associated with resistant H. pylori infection, none to our knowledge have evaluated prediagnosis antimicrobial use as a risk factor for resistance or treatment failure (16–17).

Alaska Natives—persons of Eskimo, Indian, and Aleut descent—have high rates of H. pylori infection, with an overall seroprevalence of 75% for specific antibodies (18). We are conducting a study in urban Alaska Native, rural Alaska Native, and urban non-Native adults to determine and compare reinfection rates after successful eradication of H. pylori. The study in urban Alaska Native adults is completed, and in this group, we sought to determine whether past antimicrobial use was associated with antimicrobial resistance among H. pylori isolates obtained through diagnostic endoscopy. We then determined whether H. pylori antimicrobial resistance affected the outcome of H. pylori treatment.

From September 1998 through June 2002, the Arctic Investigations Program of the U.S. Centers for Disease Control and Prevention and the Alaska Native Medical Center (ANMC) conducted a study to determine reinfection rates after successful eradication of H. pylori infection among Alaska Natives living in Anchorage. The ANMC is a 150-bed referral hospital that provides outpatient primary care services for Alaska Natives living in the Anchorage area. The institutional review boards of the Centers for Disease Control and Prevention, the Alaska Area Tribal Health Consortium, and the Indian Health Service approved the study, as did the Alaska Native Health Board. Written informed consent was obtained from all participants.

From September 1998 through June 2002, we attempted to recruit all Alaska Natives 18 years of age or older from the Anchorage area who had no history of an immunodeficiency condition or were not taking immunosuppressive medications (such as corticosteroids or cancer chemotherapeutic agents) and were scheduled for esophagogastroduodenoscopy. Alaska Natives who use ANMC for surgical procedures are very likely to receive most of their care from this facility, since no copayment is assessed to persons eligible for care at ANMC. A study nurse recruited 293 persons during the study period, 149 of whom had a positive culture for H. pylori. Since 27 September 1990, patient care information from all outpatient health care visits and outpatient pharmacy encounters at ANMC has been entered into a computerized records system. For this analysis, we recorded all antimicrobial prescriptions for 10 years before diagnosis of H. pylori infection by consulting outpatient records of study participants for whom at least 8 years of pharmacy records were available. We also recorded antimicrobial use during all hospitalizations and emergency department visits at ANMC for each participant in the same 10-year period. An antimicrobial “course” was defined as a prescription for an antimicrobial drug regardless of dose, duration, and frequency.

All participants had up to 3 gastric biopsy specimens taken for culture, antimicrobial susceptibility testing, and histologic examination for H. pylori. Gastric biopsy specimens stored in cysteine freeze medium at −80 °C were ground in a sterile tissue grinder with heat-inactivated fetal bovine serum and inoculated to 3 types of solid media: blood agar (tryptic soy agar with 5% sheep blood); chocolate agar; and brucella agar containing 10% horse blood, trimethoprim, vancomycin, and polymyxin B. All cultures were incubated at 37 °C under microaerophilic conditions and high humidity (12% CO₂, 98% humidity) for up to 10 days. Positive cultures were usually identified after 3 to 5 days of incubation. Isolates were identified as H. pylori on the basis of positive catalase, oxidase, and urease reactions; typical uniform, small, translucent colonies; curved gram-negative bacilli on Gram-stained smears; susceptibility to cephalothin (30 µg); and resistance to nalidixic acid (30 µg).

Minimum inhibitory concentrations (MICs) for clarithromycin, amoxicillin, metronidazole, and tetracycline were determined by using agar dilution. Helicobacter pylori isolates were defined as susceptible if the MIC was within the following limits: less than or equal to 0.25 µg/mL for clarithromycin, less than or equal to 0.25 µg/mL for amoxicillin, less than 8 µg/mL for metronidazole, and less than 2 µg/mL for tetracycline. Helicobacter pylori isolates with MICs above these limits were classified as resistant (19). In participants who had many cultures of their H. pylori isolates, the highest MIC determined from all of the cultures was used for analysis. Culture results were not available to providers before initiation of treatment.

Each participant's attending physician decided whether to initiate treatment and selected the treatment regimen. Patients who were treated received a 2-week course of a combination of 2 or 3 antibiotics plus lansoprazole. A study nurse called each patient approximately every 3 days to document adherence. Successful eradication of H. pylori was defined as negative results on a urea breath test (BreathTek UBT, Meretek Diagnostics, Inc., Nashville, Tennessee) 8 weeks after initiation of treatment.

Statistical analysis was performed by using StatXact 5 (Cytel Software Corp., Cambridge, Massachusetts) and SAS software (SAS Institute, Inc., Cary, North Carolina). Confidence intervals for binomial proportions were computed by using the Casella procedure (20). Bivariate associations were examined by using the chi-square test or the Fisher exact test for dichotomous variables. The Wilcoxon rank-sum test was used for comparisons of continuous variables. Logistic regression was used to test univariable dose-response relationships and multivariable associations with antimicrobial resistance. Variables were considered for the multivariable models if their univariable P value was less than 0.25, with the exception of sex, which was included in all models. The numbers of courses of macrolides, clarithromycin, erythromycin, azithromycin, and metronidazole were entered into the multivariable models as indicator variables (≥ 1 course or 0 courses), while the number of all other courses of antibiotics was entered as an interval variable. Variables were considered confounders and remained in the model if their exclusion changed the value of the coefficients of interest by more than 15%. All P values are two-sided, and confidence intervals are exact when appropriate.

One hundred forty-nine persons with culture-confirmed H. pylori infection were enrolled in the study. Of these, 125 who had documented health encounters for at least 8 years before enrollment (84%) were included for analysis. The median age of participants was 46.5 years (range, 22.2 to 88.7 years). Eighty-two (66%) were women, and all were Alaska Natives. Pharmacy records were available for a median of 8.6 years (range, 8.0 to 9.6 years). A median of 11 (range, 0 to 68) antimicrobial courses was prescribed during the 8 to 10 years before enrollment (mean, 1.52 courses per year). For β-lactam antimicrobial agents and macrolide antimicrobial agents, the median number of courses prescribed was 5 (range, 0 to 31) and 1 (range, 0 to 11), respectively. The median number of courses for all other classes of antimicrobial agents prescribed (including metronidazole) was 3 (range, 0 to 30). Thirteen patients (10%) had received previous treatment for H. pylori infection in the 10 years before enrollment. Clinical symptoms included heartburn (75%), nausea (72%), vomiting (40%), and hematemesis (11%). Endoscopic findings included hemorrhagic or superficial ulcerations of the gastric mucosa in 56 patients (45%), duodenal ulcers in 4 (3%), and gastric ulcers in 8 (6%).

Two specimens of the gastric mucosa were obtained for culture from 105 participants (84%) at the time of diagnosis. Seventeen patients (14%) had a single specimen submitted for culture, and 3 (2%) had 3 or more specimens. Among the 125 participants, 83 (66%) were found to have H. pylori isolates resistant to metronidazole, 37 (30%) were found to have H. pylori isolates resistant to clarithromycin, 7 (6%) were found to have H. pylori isolates resistant to amoxicillin, and 2 (2%) were found to have H. pylori isolates resistant to tetracycline. Twenty-nine patients (23%) had H. pylori isolates resistant to both clarithromycin and metronidazole, including 3 isolates that were also resistant to amoxicillin.

Persons with clarithromycin-resistant H. pylori isolates and those with clarithromycin-susceptible isolates were similar in age, sex, and number of years of available medical records (Table 1). Of the 37 persons with H. pylori isolates resistant to clarithromycin, 34 (92%) had been prescribed a macrolide antimicrobial agent in the 10 years before diagnosis, compared with 50 of 88 persons (57%) with H. pylori isolates susceptible to clarithromycin (P < 0.001) (Table 1). The percentage of infections with clarithromycin-resistant strains increased with more previous courses of macrolide treatment (Table 2). A similar dose-response relationship between clarithromycin resistance and number of antimicrobial courses was observed for clarithromycin, erythromycin, and azithromycin (data not shown). Macrolide use in the relatively distant past was also associated with clarithromycin resistance. After we excluded persons with any macrolide use in the 5 years before diagnosis, 6 of 9 patients (67%) with a macrolide-resistant infection had been prescribed a macrolide during the 6 to 8 years before diagnosis, compared with 7 of 45 patients (16%) with susceptible strains (relative risk, 4.3 [95% CI, 1.6 to 10.4]).

Persons with metronidazole-resistant H. pylori infection and those with metronidazole-susceptible isolates were similar in age and number of years of available medical records, but the former were more likely to be women (76% vs. 45%; P < 0.001) (Table 3). However, after we controlled for metronidazole use, sex was no longer associated with metronidazole resistance (P = 0.21). In addition, 60% (50 of 83) of those with resistant isolates had been prescribed metronidazole compared with only 10% (4 of 42) of those with susceptible isolates (P < 0.001) (Table 3). Metronidazole use in the distant past was also associated with resistance. After we excluded persons with any metronidazole use in the 5 years before diagnosis, 9 of 42 persons with resistant isolates (21%) had been prescribed metronidazole in the 6 to 8 years before diagnosis compared with 1 of 39 persons with susceptible isolates (3%) (relative risk, 8.4 [CI, 1.5 to 108.3]).

Ninety-six patients were tested for cure of H. pylori infection, and 68 (71%) were cured, as indicated by a negative result on a urea breath test 8 weeks after antimicrobial treatment was started. Among the initial 125 participants, 4 were lost to follow-up, 1 died, 9 did not adhere to treatment or declined all of their medication and left the study before the test of cure, and 15 were not treated by their physicians. Among the 87 persons (91%) who were tested at 8 weeks and reported that they had taken the full course of treatment, 64 (74%) were cured.

Treatment consisted of a 14-day course of a clarithromycin-based regimen (n = 53), a metronidazole-based regimen (n = 26), or regimens with both clarithromycin and metronidazole (n = 7); 1 person took tetracycline only. The second antimicrobial agent was amoxicillin in 94% of patients with clarithromycin-based therapy (50 of 53) and tetracycline in 96% of patients with metronidazole-based therapy (25 of 26). Eighty-five of these 87 persons (98%) were also prescribed lansoprazole. Overall, 5 of 26 persons who received metronidazole-based treatment and 15 of 53 persons who received clarithromycin-based treatment experienced treatment failure (19% vs. 28%; P > 0.25).

Treatment failure differed according to pretreatment antimicrobial resistance. Fifty-three persons were treated with clarithromycin-based regimens, and treatment failed in 77% (10 of 13) of persons with clarithromycin-resistant H. pylori compared with 13% (5 of 40) of those with clarithromycin-susceptible isolates (relative risk, 6.2 [CI, 1.9 to 37.1]; P < 0.001). Therefore, 84% of treatment failures among patients who received clarithromycin were associated with clarithromycin resistance. In contrast, among the 26 persons treated with metronidazole-based therapy, treatment failed in 11% (2 of 18) of those with metronidazole-resistant isolates and 38% (3 of 8) of those with metronidazole-susceptible isolates (P > 0.25). For the subset of persons whose pretreatment H. pylori culture showed an MIC greater than or equal to 128 µg/mL for metronidazole, only 2 of 11 (18% [CI, 3% to 50%]) had treatment failure with metronidazole-based therapy.

In our study, we found that a high proportion of Alaska Native adults were infected with H. pylori resistant to clarithromycin (30%) and metronidazole (66%). We demonstrated a statistically significant relationship between a patient's previous use of metronidazole or a macrolide antimicrobial agent and subsequent isolation of H. pylori resistant to these medications. Infection with clarithromycin-resistant H. pylori was associated with a 6-fold increased risk for treatment failure in persons receiving a clarithromycin-based regimen. To our knowledge, our study is the only study that links previous antibiotic use to H. pylori resistance and subsequent treatment outcomes.

While the association between previous antibiotic use and antibiotic resistance has been observed for other bacterial pathogens (21–24), only one previous report linked previous metronidazole use to metronidazole resistance among H. pylori(25). High rates of metronidazole resistance have been reported elsewhere for women (16–17). We demonstrated that in women, metronidazole resistance was associated with previous metronidazole use.

Clarithromycin and azithromycin are widely used alone or as empirical treatments (26–28). Use of clarithromycin as a single agent has been shown to eradicate H. pylori in only 15% of patients (29). Thus, incidental macrolide treatment could exert selective pressure on an existing H. pylori infection, selecting for strains that are macrolide resistant and influencing H. pylori susceptibility for years after treatment (30–31). Our data indicate that previous macrolide use, even 6 or more years before diagnosis of H. pylori infection, was associated with clarithromycin-resistant H. pylori. We estimate that among patients in our study who had clarithromycin-resistant infections and received clarithromycin-based treatment, 84% of treatment failures were attributable to clarithromycin resistance. Antimicrobial resistance among nontargeted pathogens is an unintended negative effect of antimicrobial use and underscores the importance of using antimicrobial agents judiciously to preserve antimicrobial susceptibility (32–33).

High rates of treatment failure have been reported among persons with clarithromycin-resistant H. pylori infections (34). In contrast, in our study, the outcome for persons who had metronidazole-resistant H. pylori and were given metronidazole-based treatment was not significantly different from the outcome for those with metronidazole-susceptible isolates, although our sample of metronidazole-susceptible infections was small. Regarding the treatment effectiveness of metronidazole, conflicting results have been found for resistant H. pylori infections (13, 35). A recent randomized study showed that when a proton-pump inhibitor was added to metronidazole, tetracycline, and bismuth, H. pylori was eradicated in 87% of those whose organisms were resistant to metronidazole compared with only 55% of those who received the same regimen without the proton-pump inhibitor (36). All but 1 of our patients treated with metronidazole-based regimens received a proton-pump inhibitor. This may explain why metronidazole resistance did not affect treatment outcome in our patients.

Antimicrobial resistance to H. pylori isolates ranges widely by geographic location. In studies from the United States and Europe published during the time patients were recruited for our study, the prevalence of clarithromycin-resistant H. pylori isolates ranged from 0.8% to 16.9% and the prevalence of metronidazole-resistant strains ranged from 11.9% to 90% (10–15, 37–43). The mean antibiotic prescription rate in our patients (1.52 courses per person per year) was higher than the U.S. rate reported in a large 1992 national survey that included both managed care and private physician practices (0.44 course per person per year, all ages) (44). This could explain the higher prevalence of resistance among H. pylori isolates in our study sample.

Several issues should be considered when interpreting or generalizing our findings. First, this was not a population-based study. Enrollment occurred sequentially among persons about to undergo endoscopy because of gastrointestinal symptoms. Second, we probably underestimated actual antibiotic use since we did not collect data from medical providers other than those at the ANMC. Also, because we recorded available antibiotic prescription data since 1990, we could not evaluate the relationship between very distant use and resistance. However, since Alaska Natives can obtain health care and prescriptions free of charge from ANMC, our record review probably captured most antibiotic use by participants during the study period. It is unlikely that differential capture of data among participants with differing antimicrobial susceptibilities occurred.

Our study has several important clinical implications. First, with a limited number of antimicrobial agents showing activity against H. pylori, increasing antimicrobial resistance will further limit the choices of effective treatments (9). Some current practices could worsen this situation, such as managing nonulcer dyspepsia by empirically treating H. pylori infection despite no evidence of substantial benefit (45–46) and treating H. pylori-infected dyspeptic patients on the basis of noninvasive serologic testing (the “test and treat” strategy) (47–49). Liberal use of antimicrobial agents for H. pylori infection could also increase antimicrobial resistance among other colonizing organisms, such as Streptococcus pneumoniae. A National Institutes of Health Consensus Statement recommends that treatment of H. pylori should be reserved for persons with peptic ulcer disease or mucosa-associated lymphoid tissue lymphoma (50).

Data from our study suggest 3 ways to improve success of treatment for H. pylori infections. First, patient-specific data on antimicrobial susceptibility can be very useful in guiding the choice of antibiotic agents. However, H. pylori culture and susceptibility testing are not widely available. The use of published H. pylori susceptibility data may mislead clinicians seeking guidance for an individual patient, since antimicrobial resistance among H. pylori in the United States appears to vary considerably by region (16, 32, 38). A community-based approach to surveillance for H. pylori antimicrobial resistance may be useful. Second, if a detailed history reveals past macrolide or metronidazole use, the possibility of H. pylori resistance should be considered when prescribing therapy. Third, in areas where drug-resistant H. pylori infections are common, providers should consider conducting a post-treatment test for cure, such as the urea breath test, to confirm successful eradication. Finally, this study again reemphasizes the importance of the appropriate use of antibiotics to decrease antimicrobial resistance.