This is the first community-based study in Vietnam to explore the prevalence of antibiotic resistance and their determinants in commensal E. coli. It increases the limited knowledge of the prevalence of resistance to different antibiotics that are found in a rural population in a lower middle income country. A study of 113 children from rural areas around Hanoi conducted between 1996-1999 found similar resistance prevalences for co-trimoxazole, ampicillin and ciprofloxacin . Comparisons between these studies are limited by differences in methodologies (we used CLSI interpretative breakpoints, whereas the authors in the previous study followed standards set by the Swedish Reference Group for antibiotics), and that the majority of children in this previous study had diarrhoea, whereas most children in our study did not (95% vs. 1% presence of diarrhoea).
A more recently published study in urban Hanoi examined resistance exclusively in diarrhoeagenic serotypes of E. coli in a mixture of children with diarrhoea in hospital and healthy children from daycare and healthcare centres, with no differences found in resistance prevalences between the two groups . Compared with our study, that study reported higher prevalences of resistance to co-trimoxazole (89% vs. 68%), ampicillin (86% vs. 65%), chloramphenicol (77% vs. 40%), and ciprofloxacin (4% vs. < 1%), but a lower prevalence of resistance to nalidixic acid (19% vs. 27%). These differences may reflect higher levels of access and use of antibiotics in the urban setting. The prevalences identified in children in FilaBavi are considerably higher for most antibiotics than the few reported rural community studies in other areas of Southeast Asia, including Thailand and Indonesia [10, 11]. Reports from community studies in rural areas in India and Peru have also presented lower prevalences of resistance in commensal E. coli than identified here [15, 16].
Compared with children aged 24-60 months, isolates taken from children aged 6-23 months had higher prevalences of resistance to co-trimoxazole, ampicillin and chloramphenicol. This is in keeping with a few previous reports which have suggested that age can act as a risk factor for carriage of resistant bacteria, independent of increased antibiotic use of younger children [13, 33]. The higher prevalence of ampicillin resistance in the highland area may be partly explained by a higher proportion of children taking beta lactam antibiotics in this area than in the other areas. Living in the lowland areas was independently associated with isolates having a lower prevalence of tetracycline resistance, a drug which should not be administered to children. This may reflect previously lower use of tetracyclines within the lowland area, or differences in exposure of E. coli to tetracylines from other sources, such as agriculture.
The majority of antibiotic use in our study was for the treatment of acute respiratory tract illnesses . Analysis of specific antibiotic use showed that the antibiotics most commonly taken by the children, beta-lactams and sulphonamides, were strong risk factors for carriage of ampicillin and co-trimoxazole resistant isolates, respectively. This highlights the importance of individual antibiotic use in selecting for carriage of resistant E. coli. Some studies have failed to find associations between antibiotic use and resistance in commensals in community settings, which may reflect important differences in methodologies employed, such as not analysing by use of specific antibiotic classes [16, 34, 35]. Our data show a trend towards higher prevalence of co-trimoxazole and ampicillin resistance in isolates from children who took sulphonamide and beta-lactam antibiotics, respectively, in the seven days prior to sampling, compared with children who only took these antibiotics between eight and twenty-one days prior to sampling. Both of these subgroups remained significantly higher than children unexposed to these antibiotics during the study period.
CLSI criteria are used to identify bacteria that demonstrate resistance to antibiotics in a clinical setting, whereas ECOFF values are designed to identify bacteria that show reduced susceptibility to an antibiotic, as compared with the wildtype population of bacteria. Analysis of decreased susceptibility according to EUCAST ECOFF values produced comparable results for most antibiotics to those reached through the use of the CLSI clinical breakpoints. This reflects the similarity in the CLSI clinical breakpoint values and the EUCAST ECOFF values for E. coli. However, a considerable difference was observed for ciprofloxacin. Whereas 0.2% and 2.7% isolates were resistant or intermediately susceptible, respectively, using CLSI clinical breakpoints, 74% of isolates were found to have decreased susceptibility compared with the wildtype population, according to EUCAST ECOFF value. As few studies have thus far presented comparisons between the prevalence of decreased susceptibility using ECOFF values and the prevalence of clinical resistance using CLSI breakpoints, interpretation of this result must be made cautiously. It is possible that this represents an emerging decrease in susceptibility of E. coli in the study area to fluoroquinolones. A study by de Jong et al. also found significant differences between clinical resistance and decreased susceptibility to ciprofloxacin in commensal E. coli isolates from healthy food-producing animals in Europe, however the magnitude of the difference was less than that found in our study . The EUCAST ECOFF value for ciprofloxacin in E. coli is based on 17877 observations , from which a similar definition of the wildtype population was reached using a normalized resistance interpretation method .
We identified a prevalence of EHEC carriage of 5%, which meant it was the dominant diarrhoeagenic E. coli subtype in our study. Two studies have found similar prevalences of EHEC carriage in asymptomatic individuals, one in workers from meat processing plants in Switzerland , and another in families living on dairy farms in Canada . This high level of carriage might come from the large number of farm animals present in Bavi, or from other environmental sources, which we have not attempted to investigate in the present study. It may be that this high prevalence is due to a clonal spread, which could be explored in further studies. EHEC carriage was not associated with presence of any gastrointestinal symptoms, and its clinical significance remains unclear.
Our community study of resistance in commensal bacteria has several strengths, including its size and high response rate. The daily reporting of antibiotic use, validated by weekly follow up interviews, have enabled analysis of individual drug use to be correlated with antibiotic resistance, and suggested a trend towards higher resistance when antibiotics were consumed close to the date of sampling. There are also a number of limitations to our study. Importantly, only one E. coli isolate from each child was tested for antibiotic susceptibility, whilst it is appreciated that multiple E. coli strains usually co-exist in an individual's gastrointestinal tract [17, 41]. The true prevalence of carriage of resistant E. coli is consequently likely to be higher than the figures presented here.
A further limitation is that the analysis on antibiotic use counted all days on which antibiotics were taken as equivalent, assuming that full doses had been taken. It has been noted elsewhere that sub-therapeutic dosing may be common in low- and middle-income countries . Our study did not assess the contribution of certain potential exposures to antibiotics and antibiotic resistant bacteria remarked on elsewhere: antibiotic use by household members, day care attendance, exposure to contaminated water sources, contact with animal microflora. We have not investigated the prevalence of different mechanisms underlying the resistant phenotypes presented here.