Evaluation of clinical and socio-demographic risk factors for antibacterial resistance of Helicobacter pylori in Bulgaria

Primary Helicobacter pylori resistance to antibacterial agents is an important factor for failure of eradication regimens; however, resistance rates may vary between patient groups according to the age, sex, disease, demographic and other factors (Megraud, 2004; Osato et al., 2001; Koivisto et al., 2004). It is important that the previous use of antibiotics can increase the primary H. pylori resistance to the agents (McMahon et al., 2003). Relatively few authors have evaluated birthplace and comorbidity, e.g. diabetes, as possible risk factors for primary H. pylori resistance (Katelaris et al., 1998; Elviss et al., 2005; Ojetti et al., 2005). To the best of our knowledge, other patient characteristics, e.g. the duration of the symptoms or the patient's education level and type of profession, were not discussed in studies on H. pylori resistance status. The aim of the present study was to assess the influence of the hospital centre and patient characteristics such as age, sex, disease, place of residence, place of birth, presence of diabetes, non-steroidal anti-inflammatory drug (NSAID) use, and type of profession and education level, as possible risk factors for primary H. pylori resistance among untreated Bulgarian patients. Patients and specimens. In total, 266 consecutive H. pylori strains, isolated from untreated patients with upper gastrointestinal complaints who agreed to fill in a questionnaire (Table 1), were evaluated in 2004–2008. Participation was voluntary, and informed consent was obtained in each case. The patients were aged 18–40 years (n=73), 46–65 years (n=138) or 66–80 years (n=55). The classification of patients' diseases was based on endoscopic findings and clinical signs. The patients had duodenal ulcers (n=80), gastric ulcers (n=20), ulcer scars (n=17), chronic gastritis (n=101), gastro-oesophageal reflux disease (n=47) and hiatus hernia (n=1). In the questionnaire, the duration (in years) of the symptoms, use of aspirin or other NSAIDs, presence of diabetes, places of residence and birthplace, type of present or past profession, and education level were also recorded. Data about previous use of antibiotics by the patients were not available. The patients' specimens were taken in four hospital centres named as centre A (64 patients), centre B (102), centre C (46) and centre D (54). Two gastric biopsy specimens per patient were taken. The specimens were transported in Stuart transport medium (Becton Dickinson). The present study was approved by the Ethical Committee of the Medical University of Sofia.

Table 1. Resistance of H. pylori to antibacterial agents according to the patient group

Microbiology. The specimens were used for direct Gram staining, in-house rapid urease test and culture as described previously (Boyanova, 2007). Briefly, culture was performed on Columbia agar (Becton Dickinson) with 10 % sheep blood (non-selective agar), and the same medium with 1 % IsoVitaleX (Becton Dickinson) and the following agents: 10 µg vancomycin ml^–1, 5 µg trimethoprim ml^–1, 5 µg cefsulodin ml^–1 and 5 µg amphotericin B ml^–1 (selective medium). Plates were incubated at 35 °C for up to 11 days in a microaerophilic atmosphere (Campy Pak; Becton Dickinson). Identification of H. pylori was made by Gram staining of the colonies, lack of aerobic growth on blood agar plates, and testing for the presence of urease, oxidase and catalase.

Susceptibility testing. Susceptibility of the strains was evaluated by a breakpoint susceptibility test as previously described (Boyanova et al., 2008). Briefly, about 30–60 µl H. pylori suspension, with a density corresponding to a McFarland turbidity standard of 2–3, were inoculated on Mueller–Hinton blood agar plates (National Center of Infectious and Parasitic Diseases, Bulgaria), containing one of the following drugs and concentrations: 8, 16 or 32 µg metronidazole ml^–1; 0.25, 0.5, 1 or 2 µg clarithromycin ml^–1; 0.5, 1 or 2 µg amoxicillin ml^–1; 4 µg tetracycline ml^–1. The plates were incubated microaerophilically at 35 °C for 2 to 3 days. Non-selective Mueller–Hinton blood agar plates were used for a control of strain viability. The breakpoints for resistance were: >8 µg metronidazole ml^–1, ≥1 µg clarithromycin ml^–1, >0.5 µg amoxicillin ml^–1 and >4 µg tetracycline ml^–1 (Megraud et al., 1999; CLSI, 2000; Megraud & Lehours, 2007).

Statistical analysis. Statistical analysis was performed using SPSS 15.0 statistical software. The chi-square test and Fisher's exact test of independence were used to compare the following variables of interest: sex, disease, age, place of residence, birthplace, type of present or past profession, educational level, duration of symptoms, use of aspirin or other NSAIDs, presence of diabetes and hospital centre. Logistic regression was used for selection of significant predictor variables and to estimate odds ratios (ORs) of the same significant variables, as well as, if possible, to predict outcome. Receiver operating characteristic (ROC) curves were performed to test the ability to classify the cases as resistant or sensitive.

Overall H. pylori primary resistance rates were: 1.1 % for amoxicillin, 24 % for metronidazole, 15.4 % for clarithromycin, 4.9 % for tetracycline, and 4.9 % for both metronidazole and clarithromycin (Table 1). No significant differences in the resistance rates were found between men and women (P≥0.251), and between patients aged >65 years and younger ones (P≥0.207). The results are similar to those reported by Elviss et al. (2005) and Kato et al. (2000), although other authors have found the female sex and age to be risk factors for metronidazole or clarithromycin resistance (Chisholm et al., 2007; Osato et al., 2001). Clarithromycin resistance has been associated with one or several of the following factors: age, female sex, geographical region, race, the patient's disease status and previous use of macrolides (Broutet et al., 2003; Megraud, 2004; Toracchio & Marzio, 2003; Seow & Chew, 2004; McMahon et al., 2003; Chisholm et al., 2007; Duck et al., 2004) (Table 2).

Table 2. Some reported clinical and socio-demographic risk factors for antibacterial resistance in H. pylori in adult dyspeptic patients

H. pylori virulence factors have been associated with both antibacterial resistance rates and success of eradication. Cure rates have been higher for patients with cagA⁺/vacA s1 strains compared to those with less virulent strains (Van Doorn et al., 2000). Clarithromycin resistance has been more common in non-ulcer patients than in peptic-ulcer patients (Megraud, 2004). Possible explanations could be the higher prevalence of cagA-positive strains in the ulcer patients, the shorter generation time of the cagA-positive strains and their closer contact with gastric cells, which can render these strains more accessible to antibacterials, or a higher antibiotic consumption in non-ulcer compared to ulcer patients (Megraud, 2004). In addition to antibacterial resistance, other factors can also affect the eradication success of H. pylori infection. Following treatment by lansoprazole, amoxicillin and clarithromycin, the eradication of clarithromycin-susceptible strains has been influenced by differences in S-mephenytoin 4'-hydroxylase (CYP2C19) genotype (Kawabata et al., 2003).

In the present study, the clarithromycin resistance was associated with hospital centre. The resistance rate was higher (23.4 %, 15 of 64 patients) in the hospital centre A than in the other centres (12.9 %, 26 of 202, P=0.041). Centre A is a large multi-profile hospital for active treatment and emergency medicine. For this reason, in centre A, there were more ulcer patients evaluated (70.3 %, 45 of 64) than in the other centres (27.2 %, 55 of 202, P=0.0001). However, unlike the patients from other hospital centres, those from centre A involved many (29.7 %, 19 of 64) cases with complicated ulcer disease or a stomach operated on for peptic ulcer, including patients with bleeding ulcers (7), perforated ulcers (2), an operated stomach (9), and both perforated ulcer and operated stomach (1). Probably, the complex characteristics of the patients from the different hospital centres and/or the previous antibacterial use could be associated with the antibacterial resistance rates in H. pylori. According to several authors, the regional and inter-hospital differences in the rates of H. pylori resistance to antibacterials are not completely explained, but could be due to variations in either prescribing practice or previous use of antimicrobial agents for other infections (Elviss et al., 2004; Kato et al., 2000).

For clarithromycin resistance, the omnibus test of model coefficients was of borderline significance (chi-square=3.86, P=0.05). The Nagelkerke R-square value was 0.025. Logistic regression analysis showed that the patients from hospital centre A were two times more likely to have clarithromycin resistance than those from other hospital centres. Hospital centre may be a risk factor [OR 2.07, 95 % confidence interval (95 % CI) 1.02–4.21] for clarithromycin resistance, but because of the borderline significance, further studies are required to verify this suggestion (Table 3).

Table 3. Risk factors for H. pylori resistance to metronidazole and clarithromycin among Bulgarian symptomatic patients (n=266) confirmed by logistic regression analysis

Metronidazole resistance has been associated by several authors with the previous use of nitroimidazoles and female sex (in women younger than 40–45 years), probably owing to the use of nitroimidazoles to treat gynaecological infections, as well as with ethnicity and place of residence (Megraud & Lehours, 2007; McMahon et al., 2003; Meyer et al., 2002; Chisholm et al., 2007). The previous use of nitroimidazoles for dental infections and, especially in the tropical areas, for parasitic diseases can also increase the primary H. pylori resistance to the agent (Megraud, 2004).

In the present study, metronidazole resistance was found in less (17.0 %, 17 of 100 patients) peptic-ulcer patients than in the rest of the patients (28.3 %, 47 of 166, P=0.037). Birthplace was also associated with H. pylori resistance to metronidazole. Metronidazole resistance was more common in patients born in towns (27.6 %, 56 of 203) than in those born in villages (12.7 %, 8 of 63, P=0.016). Another study has also detected non-ulcer patients to be associated with a higher metronidazole resistance rate (Lui et al., 2003). An inverse association has been reported in Spain (Ferrero et al., 2000).

For metronidazole resistance, the omnibus test of model coefficients indicated significance for the overall model (chi-square=11.0, P=0.004). The Nagelkerke R-square value was 0.061. Logistic regression revealed that the ulcer patients had a two times lower risk for metronidazole resistance (OR 0.51, 95 % CI 0.27–0.96) than the non-ulcer patients. The patients born in towns had OR value of 2.64 (95 % CI 1.18–5.93) of metronidazole resistance compared to those born in villages. Both the non-ulcer disease and birthplace in towns were important independent predictors of metronidazole resistance. Likewise, birthplace has been associated with metronidazole resistance, but not with clarithromycin resistance, in a study from the UK (Elviss et al., 2005).

The influence of birthplace on antibacterial resistance rates of H. pylori could be associated with the local usage of antibacterial agents during childhood. It is well known that H. pylori infection is mainly acquired in childhood (Halitim et al., 2006). In a study from India, metronidazole resistance has been also associated with geographical differences (Thyagarajan et al., 2003). Another study from the USA, however, has not reported regional differences as a risk factor for H. pylori resistance to antibacterials (Osato et al., 2001).

The higher metronidazole resistance rate in the non-ulcer patients compared to those with ulcers may be due to the lower incidence of virulent (cagA⁺/vacA s1) strains, and the resulting slower strain proliferation and lower penetration of antibiotics in the less inflamed gastric mucosa (Van Doorn et al., 2000; Megraud, 2004; Broutet et al., 2003). Gisbert et al. (2001) have suggested that H. pylori eradication therapy could be more effective in peptic ulcer than in non-ulcer patients because of different infecting H. pylori strains, different severities of inflammation or patient compliance, age and gender, although another study did not confirm this difference (Huang et al., 2005). Other factors that may be associated with the higher metronidazole resistance rate in the non-ulcer patients compared to the ulcer patients are poorer patient compliance of the non-ulcer patients, or higher antibiotic consumption in this patient group or both.

Combined resistance to clarithromycin and metronidazole has been associated with one or several of the following factors: female sex, older age and ethnicity (Seow & Chew, 2004; Meyer et al., 2002). In the present study, the difference in rates of double resistance to clarithromycin and metronidazole between the patients born in towns (5.9 %, 12 of 202 patients) and those born in villages (1.6 %, 1 of 64, P=0.312) did not reach a significant level. Logistic regression analysis did not identify any independent predictor of combined resistance to clarithromycin and metronidazole.

From the ROC curves, the area under the curve (AUC) for clarithromycin was 0.57, SE=0.5, P=1.13 (95 % CI 0.47–0.67); the AUC for metronidazole was 0.62, SE =0.04, P=0.003 (95 % CI 0.55–0.70), and the AUC for both metronidazole and clarithromycin was 0.50, SE =0.04, P=1.00 (95 % CI 0.42–0.58). The results indicate that the ability to predict which patient will harbour a resistant H. pylori strain failed for clarithromycin, and for both metronidazole and clarithromycin, and was poor for metronidazole (Fig. 1).

(12K): Fig. 1. ROC curve for metronidazole resistance in H. pylori strains from 266 symptomatic patients. AUC=0.62 (95 % CI 0.55–0.70). Solid line, ROC curve; broken line, random distribution.

Aspirin has been shown to enhance the in vitro susceptibility of H. pylori to amoxicillin, clarithromycin and metronidazole (Wang et al., 2003). In the present study, the use of aspirin or other NSAIDs, a duration of symptoms of >10 years and the level of education were not identified as risk factors for primary H. pylori resistance.

According to Pilotto et al. (2000), concomitant diseases and a histological activity of chronic gastritis have not been related to antibiotic resistance. However, diabetes has been associated with H. pylori infection in several studies (Ojetti et al., 2005; Sargn et al., 2003). The lower eradication success in diabetic patients compared to controls has been explained by changed gastric microvasculature and a possible reduction of antibiotic absorption, possible gastroparesis and the common use of antibiotics for recurrent bacterial infections (Ojetti et al., 2005). In the present study, diabetic patients had only a slightly higher (37.5 %, 9 of 24 patients) metronidazole resistance rate than the rest of the patients (22.7 %, 55 of 242, P=0.106); however, the difference was not significant possibly because of the small sample size of the diabetic patients.

In conclusion, the risk factors for metronidazole resistance identified by logistic regression were non-ulcer disease and a birthplace in a town. The hospital centre may be a risk factor for clarithromycin resistance but additional studies are necessary to confirm this association. Knowledge of the risk factors for H. pylori resistance to antibacterials could facilitate the treatment choice for H. pylori eradication and the recommendation of culture and susceptibility testing of strains for patients of the at-risk groups.

This study was supported by a grant from the Medical University of Sofia (Council of Medical Science, project no. 20/2007, grant no. 22/2007).