Risk assessment and the future of animal antibiotics

Tony Cox, Cox Associates, Denver, CO. tony@cox-associates.com

The future of animal antibiotic use (AAU) in the United States depends partly on facts about the human health risks and benefits from AAU and partly on what risk assessment questions policy-makers ask. If the question asked is only whether banning AAUs will decrease resistance in foodborne bacteria, then the US will very likely follow Europe in withdrawing approval for many routine uses of animal antibiotics. The animal and human health consequences of such a change are uncertain. They may include reduced prevalence of resistant bacteria in animals and in healthy people; transient increases and higher new equilibrium levels of animal illnesses (e.g., necrotic enteritis and airsacculitis in live chickens) that can be partly offset by increased therapeutic drug use in animals; increased fecal contamination and pathogenic microbial loads in processed chickens; and higher rates of foodborne illnesses per year and per capita-year in the human population, resulting in increased antibiotic use and accelerated spread of antibiotic resistance in humans. If the question asked is whether quantitative human health risks from AAU are acceptably small or whether quantitative human health benefits from discontinuing AAUs are large enough to be worthwhile, then withdrawal of AAUs is less likely. If the question asked is how the human health benefits from continued prudent use of animal antibiotics compare to the potential human health risks, then in many cases, prudent AAU will probably continue. The most informative question for policy-makers is how best to coordinate allocation of limited antibiotic resources between animal uses that reduce human risks vs. treatment of human illnesses once they occur, taking into account that both prevention and treatment applications can increase resistance in bacteria. It is likely that AAU will remain important in any allocation that optimizes human health benefits.

Human health impacts of different AAU patterns can be quantified in terms of population risks, defined as expected numbers of various adverse human health outcomes (e.g., number of deaths, illness-days by severity category, QALYs lost, etc.) per year; and individual risks, expressed as expected adverse human health outcomes per capita-year. These numbers can be estimated from available data using a Rapid Risk Rating Technique (RRRT) developed in an Animal Health Institute (AHI) project in 2003 to build on and add details to the FDA CVM’s Draft Guidance #152 framework. The RRRT achieves fast, easy, accurate, quantitative risk and uncertainty estimates using existing data organized within the following simple causal risk assessment framework:

Decision/act ® DExposures ® D Illnesses ® D Health Consequences ® D QALYs (optional)

(D AAU)

behaviors susceptibility treatment

In this framework, estimated human health impacts per year (or per capita-year), expressed as change in fatalities, illness-days by severity class, QALYs, etc. for different AAU decisions, can be quantified via the formula:

D health consequences = (D use)[S_paths(Dconsequences/D illness)(D illnesses/Dexposure) (D exposure/ D use)] = (D use)[S _paths(consequence factor)(exposure-response factor)(exposure factor)]

These three multiplicative factors (for consequences, exposure-response, and exposure) are all estimated from available data on consumption, illness, and resistance rates using upper bounds to manage uncertainties. They are estimated and multiplied for each causal path, i.e., each combination of bacterial strain (including susceptible or resistant, if it affects health effects), antibiotics (including drugs affected by co-selection and cross-resistance, if any), food commodity (e.g., chicken, hamburger, pork), and human subpopulation (e.g., immunocompetent vs. immunocompromised, male vs, female, community vs. ICU). Summing over all paths and time periods of interest gives the total human health risks and benefits for proposed changes in animal antibiotic use.

Applying this framework to antibiotics commonly used in both human and animal medicine, including fluoroquinolones, macrolides, and streptogramins, suggests that human health risks from continued animal antibiotic use may be orders of magnitude smaller than human health benefits, especially for drugs that reduce pathogen loads in processed food. For example, withdrawing macrolides gives estimated human health benefits of at most 0.25 QALYs/year in the US (mainly from reduced illness-days of severe macrolide-treated campylobacteriosis) but increases estimated human health risks by more than 268 additional QALYs lost/year (due to increased human campylobacteriosis cases per year from airsacculitis-positive flocks). If such estimates are even approximately correct, then prudent use of animal antibiotics should continue to play an important role in promoting human health.