Presentation is loading. Please wait.

Presentation is loading. Please wait.

Data Utility : Improvements by Targeting Study Design and Sampling Plans Peg Coleman Senior Microbiologist Syracuse Research Corporation.

Similar presentations


Presentation on theme: "Data Utility : Improvements by Targeting Study Design and Sampling Plans Peg Coleman Senior Microbiologist Syracuse Research Corporation."— Presentation transcript:

1 Data Utility : Improvements by Targeting Study Design and Sampling Plans Peg Coleman Senior Microbiologist Syracuse Research Corporation

2 RAC Charter Through the Risk Assessment Consortium, the agencies will collectively work to enhance communication and coordination between federal agencies and promote the conduct of scientific research that will facilitate risk assessments. Through the Risk Assessment Consortium, the agencies will collectively work to enhance communication and coordination between federal agencies and promote the conduct of scientific research that will facilitate risk assessments. Such research will assist the regulatory agencies in fulfilling their specific food-safety risk management mandates. Such research will assist the regulatory agencies in fulfilling their specific food-safety risk management mandates. Goals of the Risk Assessment Consortium: Goals of the Risk Assessment Consortium: Reduce uncertainties inherent in risk assessment by identifying data gaps and critical research needsReduce uncertainties inherent in risk assessment by identifying data gaps and critical research needs Improve risk assessment researchImprove risk assessment research

3 RAC Data Utility Work Group  Characteristic of a project or assessment with defined scope and explicit applications or decisions  Exercise of multifactorial judgments for inclusion and exclusion of studies  Inherent in process of accounting for data quality, variability and uncertainty

4  Characteristic of a project or assessment Generation of targeted multidisciplinary research to address data gaps and research needs for microbial risk analysis requires current awareness of priority needs of risk assessors within research community Generation of targeted multidisciplinary research to address data gaps and research needs for microbial risk analysis requires current awareness of priority needs of risk assessors within research community Examples of Needs to address bias and interpret confounding effects of extrapolations Examples of Needs to address bias and interpret confounding effects of extrapolations Predictive MicrobiologyPredictive Microbiology Dose ResponseDose Response

5  Characteristic of a project or assessment Examples of Updating Extrapolations Examples of Updating Extrapolations Predictive Microbiology: need to address bias and interpret confounding effectsPredictive Microbiology: need to address bias and interpret confounding effects Stochastic nature of bacterial growth in non-homogeneous compartmentalized shell egg (SERA, FSIS, 2005 revision 1 ) Stochastic nature of bacterial growth in non-homogeneous compartmentalized shell egg (SERA, FSIS, 2005 revision 1 ) Jameson effect in meat, poultry, and fish products (ECRA, FSIS, 2005 revision 2 ) Jameson effect in meat, poultry, and fish products (ECRA, FSIS, 2005 revision 2 ) Effects of initial dose and culture conditions (ECRA, FSIS, 2005 revision 3 ) Effects of initial dose and culture conditions (ECRA, FSIS, 2005 revision 3 ) Marks & Coleman, in press; Coleman Sandburg, Anderson, 2003; Coleman, Tamplin, Phillips, Marmer 2003 1 Marks & Coleman, in press; 2 Coleman Sandburg, Anderson, 2003; 3 Coleman, Tamplin, Phillips, Marmer 2003

6 Experimental Design for Risk Assessment ARS/FSIS/UMES Predictive Microbiology Study  2 x 2 x 3 factorial design  agitation: shaken vs unshaken  initial density: high vs low  incubation temperatures 37, 19, 10°C  Protocol Highlights  Brain Heart Infusion broth  pH 5.5 (typical of ground beef)  duplicate flasks per treatment using staggered inoculum (Oscar, 1998)  triplicate experiments using one strain

7

8

9 Effect of Initial Density 10 C

10  Characteristic of a project or assessment Examples for Future Extrapolations Examples for Future Extrapolations Dose-Response Needs to address bias and interpret confounding effectsDose-Response Needs to address bias and interpret confounding effects Strain variability for campylobacteriosis and salmonellosis (FSIS feasibility study ) Strain variability for campylobacteriosis and salmonellosis (FSIS feasibility study 1,2,3 ) Cellular immunity for campylobacteriosis (RAC Dose- Response Work Group: seminar at Naval Medical Research Center, 2003) Cellular immunity for campylobacteriosis (RAC Dose- Response Work Group: seminar at Naval Medical Research Center, 2003) Mucosal immunity (host variability) and strain variability for cryptosporidiosis (RAC public meeting & ongoing projects at EPA/OW 4 ) Mucosal immunity (host variability) and strain variability for cryptosporidiosis (RAC public meeting & ongoing projects at EPA/OW 4 ) Coleman & Marks, 1998; Coleman & Marks, 2000; Coleman, Marks, Golden, Latimer, 2004; 4 Teunis, Chappell, Okhuysen, 2000a,b; 1 Coleman & Marks, 1998; 2 Coleman & Marks, 2000; 3 Coleman, Marks, Golden, Latimer, 2004; 4 Teunis, Chappell, Okhuysen, 2000a,b; 5 Coleman, Tamplin, Phillips, 2003

11  Some Multifactorial Judgments (Ideal Data Utility) Representativeness Representativeness large dataset from reproducible controlled clinical trial, experimental study, or probabilistic survey representative of populations of interest with study design that accounts for major biological and/or social confounding factors (regional, seasonal, demographic variability, …)large dataset from reproducible controlled clinical trial, experimental study, or probabilistic survey representative of populations of interest with study design that accounts for major biological and/or social confounding factors (regional, seasonal, demographic variability, …) Relevance Relevance Pertinent or predictive of situation of interestPertinent or predictive of situation of interest Robustness Robustness Consistent performance for multiple investigatorsConsistent performance for multiple investigators

12  Some Multifactorial Judgments (Ideal Data Utility) Generalizability or external validity Generalizability or external validity Predictions consistent with other data sources and studiesPredictions consistent with other data sources and studies Soundness of study conclusions or internal validity Soundness of study conclusions or internal validity Acceptable adequacy, completeness, and soundness of conclusionsAcceptable adequacy, completeness, and soundness of conclusions Defensibility Defensibility Use consistent with body of scientific evidence from multiple sourcesUse consistent with body of scientific evidence from multiple sources

13  accounting for data quality, variability, uncertainty Utility involves both getting the right science and getting the science right Utility involves both getting the right science and getting the science right Examples for E. coli O157:H7 Examples for E. coli O157:H7 Jameson Effect for Predictive Microbiology in Food MatricesJameson Effect for Predictive Microbiology in Food Matrices Human and animal clinical studies for dose-responseHuman and animal clinical studies for dose-response

14 Numerical dominance of spoilage flora of ground beef (FSIS baseline survey, n=543, CFU/g in 25-g samples )

15 Ecological Advantage Pseudomonas Rate at 2°C 0.09/hr Rate at 4°C 0.11/hr Rate at 10 °C 0.24/hr E. Coli O157:H7 Below T Minimum 0.028 unshaken,lowN(0) 0.037 unshaken,highN(0)

16 Model Predictions for Baseline Scenario (Coleman, Sandburg, Anderson, 2003) Pseudomonas 95 th percentile 2x10 9 /100-g serving 50 th percentile 6x10 6 /100-g serving 6x10 6 /100-g serving 5 th percentile 6x10 4 /100-g serving 6x10 4 /100-g servingMaximum 4x10 15 /100-g serving 4x10 15 /100-g serving E. Coli O157:H7 000 3x10 3 /100-g serving

17 Questions to be addressed for O157 Dose-Response Modeling Appropriate Surrogates? Appropriate Surrogates? Plausible Model Forms? Plausible Model Forms? Should dose-response models have specific terms for species, strain, host, and matrix effects and interactions?Should dose-response models have specific terms for species, strain, host, and matrix effects and interactions? How do we estimate variability and uncertainty?How do we estimate variability and uncertainty? Low-dose extrapolation? Low-dose extrapolation?

18 Infant Rabbit Dose-Response

19 Infant Rabbit Clinical Study Pai et al., 1986

20 Sampling and Design Issues for Exposure Assessment Heterogeneous distributions in foods Heterogeneous distributions in foods If growth, then clustersIf growth, then clusters Non-monotonic distributions Non-monotonic distributions SE eggsSE eggs Need for not only prevalence (+/-) but levels Need for not only prevalence (+/-) but levels Small, convenience sampling designs unlikely to be sufficiently representative for maximum utility to risk assessors

21 Sampling and Design Issues for Dose-Response Defining curves for dose-response relationships may require higher targets (~10 dose groups), multiple aspects of disease triangle (host, pathogen, matrix) Defining curves for dose-response relationships may require higher targets (~10 dose groups), multiple aspects of disease triangle (host, pathogen, matrix) Include low-dose region Include low-dose region Develop more mechanistic approaches using in vitro, human, and animal data, established dosimetry practices, immunological and physiological data for inter-species extrapolations (RAC Dose Response Work Group)

22 Acknowledgments  Risk Assessment Consortium  USDA Food Safety & Inspection Service, Agricultural Research Service  University of Maryland Eastern Shore

23 Syracuse Research Corporation (SRC) Microbial Risk Assessment Center of Excellence (MRACE) Points of Contact:  Dave L. Colangelo Director, Government Affairs colangelo@syrres.com  Pat McGinnis Associate Director, SRC Environmental Science Center mcginnis@syrres.com  Peg Coleman Senior Microbiologist mcoleman@syrres.com


Download ppt "Data Utility : Improvements by Targeting Study Design and Sampling Plans Peg Coleman Senior Microbiologist Syracuse Research Corporation."

Similar presentations


Ads by Google