During the course of the study, HARP consisted of 11 hospital study sites across the United States. The first five patients who sought treatment each month for esophago-gastro duodenal endoscopy (EGD) as part of treatment for H. pylori infection, confirmed by gastric biopsy, were enrolled at each site. When fewer than five eligible study participants were seen at a study center, those eligible were enrolled. Written, signed consent was obtained from all participants or their guardians. Culture of biopsied material for H. pylori was performed at the study site or at the Atlanta Veterans Affairs Hospital. If primary isolation was performed at the HARP study site, isolates were placed in trypticase soy broth (TSB) supplemented with glycerol and stored at –70°C. Isolates were sent to the Centers for Disease Control and Prevention for antimicrobial resistance testing. Samples forwarded to the Atlanta Veterans Affairs Hospital for isolation were packed in 6 pounds of dry ice and maintained at –70°C to –80°C until cultured (up to 1 month). Each specimen was accompanied by a standard case report form containing 145 variables of demographic, clinical, and epidemiologic information.

Of 317 enrolled HARP study participants for whom complete demographic information was available, 205 (65%) were male, 116 (37%) were white, 172 (54%) were black, 14 (4%) were Asian, 3 (1%) were Native American, and 12 (4%) were of other ethnic backgrounds. The median age was 57 years (range 3–94 years). Among males, 97 (31%) were white, 90 (28%) were black, and 9 (3%) were Asian, 1 (0.3%) was Native American, and 8 (3%) were of other ethnic backgrounds. Among females, 19 (6%) were white, 82 (26%) were black, 5 (2%) were Asian, 2 (0.6%) were Native American, and 4 (1%) were of other ethnic backgrounds. The most common endoscopic diagnoses for conditions of HARP participants were gastric erosions, gastritis, duodenal and gastric ulcers, and esophagitis (Table 1). H. pylori resistance was not statistically associated with endoscopic findings.

Among 347 H. pylori isolates submitted to HARP from 1998 to 2002, 118 (34%) were resistant to >1 antimicrobial agent, 101 (29%) H. pylori isolates were resistant to one agent only, and 17 (5%) H. pylori isolates were resistant to more than one antimicrobial agent. Three isolates were resistant to amoxicillin (1%), 45 were resistant to clarithromycin (13%), 87 were resistant to metronidazole (25%), and no isolate was found to be resistant to tetracycline. Multiple-agent resistance was observed for clarithromycin and amoxicillin (1 isolate, 0.3%) and clarithromycin and metronidazole (16 isolates, 5%). A test of trend showed no significant trend for resistance to metronidazole, but a significant trend (R² = 0.76) was noted for a decline in resistance to clarithromycin during the study period (Figure).

Most isolates were submitted by centers in the Northeast region, followed by the South, Midwest, and the West. Submitted H. pylori isolates from the Northeast region had the highest frequency of single- and dual-agent resistance (Table 2). The Midwest region submitted H. pylori isolates with the second-highest single agent resistance rate, while the southern region had the second-highest dual agent resistance rate. The highest proportion of H. pylori isolates resistant to clarithromycin, metronidazole, and amoxicillin was in the Northeast (Table 2).

The four sites reporting the highest proportion of isolates resistant to one agent were Indianapolis, Indiana; Galveston, Texas; Philadelphia, Pennsylvania; and Washington, DC (Table 3). Dual agent resistance was most prevalent in Detroit, Michigan. Indianapolis had the highest overall resistance: 50% of all isolates submitted were resistant to at least one agent. The Atlanta site reported the highest proportion of clarithromycin-resistant isolates (Table 3). The highest proportion of metronidazole resistance was reported from the Indianapolis site. Isolates resistant to amoxicillin were found in Nashville, Tennessee, Philadelphia, and Washington, DC.

Demographic, clinical, and epidemiologic data were analyzed for association with infection by a resistant H. pylori strain. Results of univariate analysis of risk factors for antimicrobial resistance are summarized in Table 4. On univariate analysis, significant risk factors for H. pylori resistance included past treatment of H. pylori infection with clarithromycin, treatment of past H. pylori infection for a second time in the past 5 years, treatment of past H. pylori infection with clarithromycin more than once, and treatment of past H. pylori infection with a proton pump inhibitor more than once. No protective factors for resistant H. pylori resistance approached significance.

On multivariate analysis, the most significant risk factor for H. pylori resistance was previous treatment of H. pylori infection within the past 5 years (Table 5). Protective factors for H. pylori infection were treatment with an antacid (Mylanta) and submission of specimens from New York, Palo Alto, California, Atlanta, or Nashville. However, a high degree of colinearity was found between geographic location and race in this model. For example, 84% of Washington, DC, participants were black, compared to 20% of participants from Nashville. Accordingly, a second multivariate model was constructed in which geographic location was controlled by conditioning. In this model, black race was the only risk factor significantly associated with infection with a resistant H. pylori strain (Table 6).

The emergence of antimicrobial resistance in H. pylori represents a serious public health challenge because of the prevalence of infection and incidence of severe sequelae. A 12-fold increase in resistance to antimicrobial agents such as clarithromycin has been documented in cases in which antimicrobial therapy regimens do not eliminate H. pylori (9). Once resistance to clarithromycin appears, cross-resistance to all other macrolides also occurs, which eliminates the potential for their use in salvage therapies. In the past 5 years, amoxicillin and tetracycline resistance has also been described (6,10).

H. pylori resistance is amply documented in a variety of study populations. However, most published studies have been based on cross-sectional designs, not prospective population-based surveillance of resistance. A meta-analysis of 20 nationwide clinical trials estimated H. pylori resistance to clarithromycin, metronidazole, and amoxicillin to be 10.1%, 36.9%, and 1.4%, respectively (6). These findings were also reflected in smaller single-site studies (11). Resistance rates may be higher in pediatric populations; for example, a study in Michigan demonstrated that clarithromycin resistance in pediatric H. pylori strains was 2.5 times higher than that reported in adults in the same region and elsewhere in the United States (12).

Single-site and multicenter surveillance network studies in Britain, Croatia, the Czech Republic, Turkey, Greece, Poland, Russia, Slovenia, Turkey, Estonia, Italy, Bulgaria, Germany, and Belgium have observed primary metronidazole and clarithromycin resistance with rates as high as 40.3% and 10.6%, respectively (13–17). A large prospective surveillance project in northeastern Germany showed primary metronidazole and clarithromycin resistance rates of 26.2% and 2.2%, respectively, in 1,644 clinical H. pylori isolates collected from 1995 to 2000 (17). High resistance rates to metronidazole and clarithromycin have been reported in Portugal, Brazil, Hong Kong, Saudia Arabia, and Lebanon (18–22). Overall, H. pylori resistance to metronidazole among single and multicenter surveillance networks is greater than all other resistant antimicrobial agents combined.

HARP is the only multicenter network providing ongoing prospective antimicrobial resistance and associated risk factor data for H. pylori in North America. The first 4 years of data show that resistance to antimicrobial drugs commonly used to treat H. pylori infections is widespread, though rates varied from year to year; this finding was true for both clarithromycin and metronidazole. We speculate that the precipitous decrease in metronidazole resistance in 2002 could be attributed to the reduction in isolates received by HARP, which limited the power and representiveness of our sample.

We performed a two-phased multivariate analysis on risk factors significantly associated with resistance on univariate analysis. In the first-phase, multivariate model, antimicrobial resistance was significantly associated with previous antimicrobial treatment for H. pylori, and antacid use was protective. However, in this model, geographic location was highly collinear with race, which necessitated the second-phase multivariate model, in which the contribution of race was assessed by conditioning on geographic location, previous antimicrobial treatment for H. pylori infection, and antacid use. This model clearly demonstrated an association between black race and infection with an antimicrobial-resistant H. pylori. This association may indicate persistent transmission of resistant strains or in vivo induction of resistance. The association between H. pylori infection and blacks in the United States has been amply documented (2,3). However, the association between incidence of H. pylori resistance and black race is new. The biologically plausible association between previous antimicrobial use and resistance approached significance in the second-phase model; however, many study participants could not recall whether they were previously treated. This model does not rule out an independent association between geographic location and resistance. Enrollment of additional participants in HARP may sufficiently enhance the statistical power of the study to show such an association, while enrollment in sites serving a more racially diverse population may help clarify the contributions of geographic location as a risk factor for resistant infection.

This study has several limitations. While drawn from various regions in the continental United States, the 11 HARP sites represent a convenience sample of academic medical centers that is not truly representative of the U.S. population. Some participating sites enrolled more patients per month than others, so that cumulative data are biased towards individual study sites. Most of the participating hospital sites serve adult populations, and thus the results largely reflect resistance incidence towards adult patients. Patients undergoing endoscopy for H. pylori-related illness constitute a minority of infected persons, whose clinical condition and H. pylori-resistance status may not be representative of the total population with underlying H. pylori infection. Although this trend has not been documented, many public health analysts suspect that the proportion of H. pylori infections diagnosed and treated in academic medical centers is decreasing compared with those in community practices. We suspect that this trend may have accounted for the reduced number of isolates submitted to HARP over time. Although these limitations primarily affected our sample size and how the sample represented the population, they did not invalidate the epidemiologic significance of the results generated from this study.

Given the limited population-based surveillance data on H. pylori resistance, HARP offers the best available data on H. pylori resistance in the United States. The accuracy and usefulness of this prospective, multicenter network can be enhanced in a number of ways. Increasing the number of HARP sites will improve geographic and demographic representation. Given the fact that only 12 study participants <12 years of age were included, enrollment of pediatric-care sites will improve the age representation of the HARP cohort. This feature is particularly important considering the increasing number of reports that describe higher resistance rates in pediatric strains compared to those isolated from adults, particularly resistance to macrolides (12,23). Enrolling study participants in community practices will provide data on resistance rates in H. pylori isolates from infected persons not treated at academic medical centers, a population in which resistance rates in H. pylori are unknown and may differ significantly from the current HARP cohort. Future HARP findings may serve as the basis for specific therapeutic recommendations and pretreatment antimicrobial susceptibility testing in high-risk populations.

In summary, we have shown that antimicrobial resistance in clinical H. pylori isolates is extensive, that it varies from year to year, and that resistant isolates are more common among blacks. Ongoing, prospective surveillance of H. pylori resistance is essential to assure [or ensure?] that appropriate data are available to guide the choice of therapy, particularly in high-risk populations.