AMR risk assessment project Project team: Colorado State University - Keith Belk & Paul Morley Kansas State University - Mike Apley & Katie Hope University of Nebraska-Lincoln - Bing Wang & Sapna Chitlapilly USDA-ARS-MARC - Terry Arthur, John Schmidt, & Tommy Wheeler EpiX Analytics - Solenne Costard, Huybert Groenendaal, Francisco Zagmutt Presenter: Francisco J. Zagmutt www.epixanalytics.com
Objectives and scope 1. Pilot QRA to assess: Does Tylosin and chlorotetracycline (CTC) use in beef cattle affect antimicrobial resistance in humans? If so, how much, and how does it compare with other sources? 2. Identify key data gaps in the knowledge of the ecology of antimicrobial resistance and beef production 3. Create framework potentially useful to evaluate other food safety antimicrobial resistance risks Outcome: # foodborne AMR resistant cases (attributable to beef) Antibiotic resistance: macrolides and tetracyclines Pathogens: STEC, Salmonella spp., Campylobacter spp. EpiX Analytics LLC
Also useful to quantify benefits The risk analysis process Hazard identification Risk assessment Risk management Or to assess the tradeoffs between risks and benefits Risk communication EpiX Analytics LLC Modified from: Copyright 2013 by Sidney Harris.
http://www.weakstream.us/wordpress/wp- Typical mechanics microbial FS model Proportion of contaminated meals Bacterial load/meal (load + mixing) Total illnesses = Binomial(Total meals, Pill) 1.2 1 P(Ill Contam) - > DR model P(Inf Dose) 0.8 0.6 0.4 0.2 Data Logistic Beta-Poisson Beta-Binomial Probit 0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 Log dose Probability of illness/meal (Pill)
QRA approaches: bottom-up, farm-to-fork (FTF) Explicitly describe and quantify every key step from production to consumption, and then from consumption to illness and AMR Advantage: granular prediction of food risk changes with mitigations at individual steps in the FTF chain e.g. If we reduce 1 log of Salmonella sp. contamination at evisceration, and increase cooking temperature by 10⁰F, we will observe x% reduction of salmonellosis in humans Disadvantage: Very data- and assumption-intensive. P(Inf Dose) 1.2 1 0.8 0.6 Data Logistic 0.4 Beta-Poisson Beta-Binomial 0.2 Probit 0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 EpiX Analytics LLC Rosenquist et al (2003) Quantitative risk assessment of human campylobacteriosis associated with thermophilic Campylobacter species in chickens. International Journal of Food Microbiology. 83, 87 103 Log dose
Results QRA Campy. chickens -> humans Rosenquist et al (2003). Detailed FTF model for Campy in Denmark The simulations showed a linear relationship between the flock prevalence and the fraction of positive chickens leaving the slaughterhouse (Fig. 6A), and between the flock prevalence and the incidence of campylobacteriosis (Fig. 6C). Rosenquist et al (2003) Quantitative risk assessment of human campylobacteriosis associated with thermophilic Campylobacter species in chickens. International Journal of Food Microbiology. 83, 87 103 EpiX Analytics LLC
Alternative QRA approach: top-down models (taken for this project phase) Still model logic of FTF continuum, but simplify/eliminate some steps between production and consumption (e.g. by using linearity discussed earlier) Advantage: require less data, allow model calibration using national (foodborne) disease surveillance data Disadvantage: potentially less granular prediction of intervention strategies e.g. If we reduce prevalence of Salmonella sp. By 10%, we will observe an x% reduction of foodborne salmonellosis in humans EpiX Analytics LLC Otto et al. CID 2014 Estimating the Number of Human Cases of Ceftiofur-Resistant Salmonella enterica Serovar Heidelberg in Québec and Ontario, Canada http://cid.oxfordjournals.org/content/early/2014/07/31/cid.ciu496.full
A little analogy: modeling a cow Source: http://hilobrow.com/wp-content/uploads/2011/04/cow.jpg Source: http://img13.deviantart.net/0812/i/2012/309/0/9/skeleton_cricket_by_skeletoncricket-d5k4pcd.jpg EpiX Analytics LLC Source: http://physiolgenomics.physiology.org/content/physiolgenomics/23/2/217/f6.large.jpg Unless I plan to predict what happens if I change each node above, perhaps the cow can be approximated with
AMR due to beef consumption AMR in beef environment AMR in natural environment Tylosin-CTC use Animals ARBs Inter- microorganism transmission Pathogen* resistance in animal Other pathogen resistance in animal ARBs ABX Feedlot surface (feces and soil blended) Rain runoff Carcass contamination with gut pathogens Airborne or direct contact of animal and meat ARB reselection Manure/ slurry Beef related Illness Human consumption No Foodborne illness Rain runoff underground water (irrigation, recreational, consumption) AMR Transfer AMR to other pathogens Non-beef consumption Crop/vegetable/fruit Illness AMR Outcome: # treatment failures due to AMR Text Text boxes with dashed lines indicate elements of the conceptual model not included in the initial analysis * Pathogen = STEC, Salmonella and Enterococci ARB = Antibiotic resistance bacteria; ARG = Antibiotic resistance genes; ABX = Antibiotics
Relationship between enteric illnesses with AMR and ABX in cattle: Conceptual model and data sources # DOMESTIC (US) ENTERIC ILLNESSES STEC, Salmonella, Campy (adjusted for under-reporting and under-diagnosis, and catchment area) FoodNet data 2014: 1996-2014 yearly incidence rates per pathogen Scallan et al 2011 (Table 3): 2005-2008 data Ebel et al 2012 (Table 2) # DOMESTIC (US) ENTERIC ILLNESSES WITH AMR PATHOGENS NARMS data 1997-2013: % AMR per pathogen and drug Scallan et al 2011 (Table 2): % foodborne illnesses FDA data 2009-2011; usage in humans by drug class Van Boeckel et al 2014 Beef Food Human treatment Environment Other Painter et al 2013: single estimates per pathogen for 1998-2008 Relationship? Abx used to produce amount of beef meat consumed Abx use per cattle Beef Dairy Meat consumption Ground Whole muscle USDA Beef consumption 1960-2014, national level U.S. meat case retail scanner data 2008-2014: +/- 70% of retail sales by cut and region Foodservice Volumetric data 2003-2014, ground vs whole muscle, national level Note: 55% beef via foodservice, 45% via retail NAHMS / Dairy: (1996), 2002 & 2007 estimates of % cattle using abx (drug class) Recommended drug use (e.g. drugs.com) NAHMS / Feedlots: 1999 & 2011 estimates of % cattle using abx (drug class), # treatments / cattle and average duration in days Recommended drug doses (e.g. drugs.com)
Hypothetical example of environmental pathway output Exposure of antibiotic resistant enteric bacteria contributed by the food and environmental pathways 23% Food pathway 30% 28% 32% Water Air Farm/production handling Environmental Pathway 0 15 30 45 60 Prepared by Drs. Bing Wang & Sapna Chitlapilly, UNL
Current work Conceptual models for food and environmental pathways developed Identification and processing of key data currently available, including Group level (e.g. macrolides) abx use in humans Group level (e.g. macrolides) abx use in animals AMR in samples from human foodborne illnesses Per-capita beef consumption Some information/data still in process/missing: Representative product-level ABX use in humans, and also in animals Individual foodborne illness-level data (food source, AMR) State-level stratification (increase our statistical power) Report of pilot project due this spring: will highlight current analysis limitations, data gaps, and future data collection efforts EpiX Analytics LLC
Foodnet reported outbreak and sporadic cases (lab confirmed) Note: not accounting for under-reporting, under-diagnosed, catchment area (10 us sites), foreign travel
Salmo and STEC - Very limited amount of data for Macrolides - 3 years of data, almost no cases Campy Azithromycin since 1998, Erythromycin 1997, Telithromycin 2005 Results from subset of Foodnet and other samples tested. Generated from data provided by Julian Grass, CDC
US: 3 rd largest consumer of abx (22 units per person in 2010) Note: standard unit = pill/capsule/ampoule per person Between 2000 and 2010: slight decrease (annual growth rate: -1.25 to -2.50 )
In US, human use of ABX stable (Tetracyclines) or slightly reduced (Macrolides) between 2000 & 2010 What about # illnesses with AMR to these abx (NARMS)? Pathogen / Abx 2000: % samples with AMR* 2010: % samples with AMR* * Mean and 90%CrI ** Data on % of samples with AMR to Azithromycin not available for Salmo. and STEC in 2000 and 2010 Confidence of change from 2000 to 2010? Campy. / Azithromycin** 9.1(6.6-11.8)% 3.1(2.4-4.0)% 99% Campy. / Tetracycline 40.9(36.4-45.5)% 44.8(42.5-47.1)% 89% Salmo. / Tetracycline 17.6(16.0-19.2)% 9.5(8.6-10.4)% 99% STEC / Tetracycline 7.3(5.3-9.6)% 5.2(2.8-8.3)% 84% Per animal production type?
Some data challenges Difficulty to retrieve and combine historical data on abx use in humans and animals: Few studies / sources over few years, with little overlap Not same units of measurements: prescriptions, standard units, kg or lbs. E.g. 2010 humans, FDA: 3,278,904kg, Van Boeckel et al. (IMS): 6 8 10⁹ standard units Different sources for estimation => comparable? E.g. environmental advocacy vs coalition of drug makers IMS very expensive thus not many estimates exist Study from Van Boeckel et al 2014 provides only 2 point estimates (2000 & 2010), and exact estimates for US to be requested from author
Data gaps # DOMESTIC (US) ENTERIC ILLNESSES STEC, Salmonella, Campy (adjusted for under-reporting and under-diagnosis, and catchment area) Non-enteric illnesses and AMR # DOMESTIC (US) ENTERIC ILLNESSES WITH AMR PATHOGENS Historical AMR data for Macrolides (only 2011-2013 available) Dataset with link between case of illness, food source and AMR status Beef Food Human treatment Environment Other Usage data over time (FDA 2009-2011 only) Usage by specific drug (rather than class) or relative proportions of drugs within class over time Relationship? Abx used to produce amount of beef meat consumed Abx use per cattle Meat consumption Feedlots & Dairy: usage over time (only few point estimates currently available), and by specific drug rather than class Changes in recommended drug doses over time Beef Dairy Ground Whole muscle Historical beef consumption per state / region
Next steps / Improvements # DOMESTIC (US) ENTERIC ILLNESSES STEC, Salmonella, Campy (adjusted for under-reporting and under-diagnosis, and catchment area) # DOMESTIC (US) ENTERIC ILLNESSES WITH AMR PATHOGENS Feasibility of requesting state-level illness data? Request access to FoodNet data & obtain and/or calculate yearly adjustment estimates for: proportion of enteric illnesses that are foodborne catchment area, under-reporting, under-diagnosis, and domestic vs travel Contact Van Boeckel et al 2014 re estimates for years between 2000 & 2010 (study based on IMS data) Beef Food Human treatment Environment Other Account for historical abx use estimates for non-cattle animals (livestock, poultry and pets) Request NORS data Calculate yearly estimates for food source allocation, and use improved methodology (vs Painter) to account for outbreak vs sporadic cases Relationship? Abx used to produce amount of beef meat consumed Abx use per cattle Meat consumption Beef Dairy Ground Whole muscle Use expert opinion to inform trends in abx usage (specific drugs within class) over time (rather than linear trend)
Next steps This phase will help us establishing baseline risk and how reduction/increase in abx use in beef production can change it. Need to put numbers in context. E.g. compare with other human health risks, other sources of AMR Also, consider data needed for a more holistic approach (e.g. weighing the tradeoffs between reduction in AMR resistance in humans vs potential increase in non-amr food illnesses if ABX not used + economic and ecological impact of alternative mitigation strategies Fine-tune mitigation strategies: do we need to aim towards a more bottom-up FTF model? More tight integration food and environmental pathways, with feedback dynamics Leveraging whole genome sequencing data Explicitly model resistance and health effects in non-food pathogens?
Acknowledgments The project team Dr. Mandy Carr (NCBA) Dr. Betsy Booren (NAMIF) Julian Grass (CDC) Elaine Scallan (UC Denver) Dana Cole (USDA-APHIS-CEAH) Project funded in part by the Beef Checkoff through NCBA, and NAMIF EpiX Analytics LLC
Thanks! If you have any ideas, data to share, criticisms (or better pictures of crickets), please contact me or team members EpiX Analytics LLC Dr. Francisco J Zagmutt Managing partner EpiX Analytics LLC Fzagmutt@EpiXAnalytics.com