Re-examining the causes of campylobacteriosis

The development of fluoroquinolone (FQ)-resistant strains of Campylobacter (CP) in chickens is favored by enrofloxacin, which is used to combat respiratory disease in chicken broilers. Thus, it seems plausible that eating chicken treated with FQs could increase the risk of FQ-resistant CP illness. We call this hypothesis 'Causal Model 1'. It is the dominant paradigm driving current regulatory efforts in the USA to ban FQ use in chickens. Although plausible, Causal Model 1 does not explain why several recent data sets indicate that eating chicken (and even touching raw chicken) at home can reduce the risk of CP illness. An alternative, Causal Model 2, instead attributes the risk of sporadic domestic CP cases primarily to commercial cooking of hamburgers, chicken, and other meats. This paper re-examines case-control data collected by the Centers for Disease Control (CDC), and re-analyzes previously published case-control data to assess the evidence for Causal Models 1 and 2. We apply causal graph, classification tree, conditional independence and Bayesian Model Averaging (BMA) methods of data analysis to identify potential causal relationships of interest in the case-control data. Available case-control data are generally consistent with Causal Model 2, but not Causal Model 1, because they indicate that chicken (and other meats) are protective against campylobacteriosis risk when prepared at home, and are only risky when consumed in commercial establishments such as restaurants. This suggests that poor hygiene in some restaurants may be a predominant cause of campylobacteriosis in humans, while chicken-borne CP plays at most a minor (statistically undetectable) role in human health risk. Both domestic and international data support this model. If true, Causal Model 2 implies that current regulatory efforts in the USA to protect human health by focusing on chicken-borne CP may be ineffective, conferring no detectable human health benefit. Instead, improving restaurant hygiene for a small minority of high-risk restaurants (as well, perhaps, as reducing over-use of FQs in human medicine) are the interventions most likely to create significant human health benefits, based on currently available data.