Does the use of antibiotics in food animals pose a risk to human health? An unbiased review?

Sir, The review by Phillips et al. is an insight into the reasoning of the scientific advisors of the animal drug industry. We hope that the industry discounts this bad advice. It would be alarming if preventive action to control resistant human pathogens in the food supply should be taken only when sufficient numbers of humans had been harmed, to enable undisputable epidemiological proof of a ‘devastating clinical effect’. Phillips et al. state that there might be disadvantages to human health and to animal health in the discontinuation of the use of antimicrobial growth promoters (AGP), but do not make a convincing plea. Their argument that in Denmark the ban of AGP may have caused an increase in the incidence of salmonellosis and campylobacteriosis is not correct. The pathogen load has decreased (Salmonella) or remained at a constant level (Campylobacter). The AGP ban in Denmark in 1998 in no way affected the yearly increase in the incidence of human campylobacteriosis, of 20% in 1993–1999 (except for 1997), 5% in 1999–2002 and a decrease of 17% in 2003. The authors repeatedly refer to a small increase in human salmonellosis in 2001, but fail to mention that the incidence has decreased markedly since 1997. This contradictory argument compromises the credibility of the paper. Based on a Danish finding of a higher prevalence of resistance in Campylobacter jejuni isolates in human infections than in Danish chicken isolates, the authors falsely conclude ‘chicken is not the major source of Campylobacter in humans’. In Denmark, some 27% of the poultry products are imported. The high level of resistance in the imported products explains the higher levels of resistance in Campylobacter isolated from humans, as compared with domestic products. Yet another example of misquoting and misinterpreting scientific results is in Table 2; only one of the references has contributed and it does not support the heading—that there is international evidence of chicken consumption being a protective factor for C. jejuni illness. Selectively citing the finding of ‘eating chicken legs’ as a ‘protective’ factor—when all other chicken cuts came out as risk factors in the same study—is hardly scientific! Neimann et al. do not support the statement ‘Case-control studies that fail to consider alternative hypotheses frequently find chicken consumption to be a major risk factor’. Chicken meat remains the most consistently identified risk factor for campylobacteriosis throughout the scientific literature. The authors state, ‘The realization that VRE infections are largely confined to clinical units in which glycopeptides are heavily used. . . suggested that such usage in humans might be the driving factor’. Agreed, but VRE do not emerge spontaneously in hospitals; they are introduced from external sources. Numerous studies show that VRE is common in the general population, in food, animals and in the environment because of the use of avoparcin for growth promotion. Enterococci are highly heterogeneous bacteria, and the probability of isolating identical DNA subtypes of VRE from animals and humans is consequently low in population-based surveys. The authors interpretation of this is that transmission between animals and humans is uncommon. This is false. While the isolation of identical types provides strong evidence of transmission, the rate of transmission cannot be assessed. Dendrogramatic analyses of amplified fragment length polymorphism (AFLP) patterns show that human strains and strains causing infections in hospitals belong to the same clusters as strains in animals, food and healthy humans in the community. It has been shown experimentally that animal enterococci can transiently colonize the gut, and studies suggest that gene transfer can take place in the intestine following transient colonization. Nevertheless, Phillips et al. conclude ‘The truth about gene transfer from animal isolates . . . remains beyond our grasp’, misinterpreting the study by Sørensen et al. and ignoring Berchieri. Phillips et al. point out that no transconjugants were detected in the ingestion experiment of Enterococcus faecium and suggest that it is ‘not a common event in vivo’. This study was not designed to select for transconjugants. With a transfer rate of 10 3 transconjugants/donor, 1609 isolates should have been tested to obtain an 80% power to find at least one transconjugant. The authors repeated call for good old ‘shoe leather epidemiology’ is amusing considering the difficulty the industry advisors have in accepting the current strong epidemiological evidence, which rests on the four solid pillars of microbial epidemiology: bacterial isolation, bacterial typing, descriptive epidemiology and analytical epidemiology.