Quantitative Microbial Risk Assessment

By Charles N. Haas, Joan B. Rose and Charles P. Gerba

Provides the latest QMRA methodologies to determine infection risk cause by either accidental microbial infections or deliberate infections caused by terrorism
• Reviews the latest methodologies to quantify at every step of the microbial exposure pathways, from the first release of a pathogen to the actual human infection
• Provides techniques on how to  gather information, on how each microorganism moves through the environment, how to determine their survival rates on various media, and how people are exposed to the microorganism
• Explains how QMRA can be used as a tool to measure the impact of interventions and identify the best policies and practices to protect public health and safety
• Includes new information on genetic methods
• Techniques use to develop risk models for drinking water, groundwater, recreational water, food and pathogens in the indoor environment

• Language: English
• Publisher: Wiley
• Release date: Jun 9, 2014
• ISBN: 9781118910023

Book preview

CONTENTS

Cover

Title page

Copyright page

Preface

Chapter 1: Motivation

Prevalence of Infectious Disease

Prior Approaches

Scope of Coverage

Potential Objectives of a QMRA

Secondary Transmission

Outbreaks versus Endemic Cases

References

Chapter 2: Microbial Agents and Transmission

Microbial Taxonomy

Clinical Characterization

Microorganisms of Interest

Transmission Routes

References

Chapter 3: Risk Assessment Paradigms

Chemical Risk Assessment: National Academy of Sciences Paradigm

Ecological Risk Assessment

Approaches for Assessing Microbial Risks

Development of the QMRA Framework and Processes

References

Chapter 4: Conducting the Hazard Identification (HAZ ID)

Identifying and Diagnosing Infectious Disease

Health Outcomes Associated with Microbial Infections

Sensitive Populations

Women during Pregnancy, Neonates, and Young Babies

Diabetes

The Elderly

The Immunocompromised

Databases for Statistical Assessment of Disease

ICD Codes

Waterborne and Foodborne Outbreaks

Epidemiological Methods for Undertaking HAZ ID

Controlled Epidemiological Investigations

HAZ ID Data Used in the Risk Assessment Process

Recommendations for Updating Quantitative Data for HAZ ID Information

References

Chapter 5: Analytical Methods and the Qmra Framework

Introduction

Approaches for Developing Occurrence and Exposure Databases

Overview of Methodological Issues

Specific Techniques for Bacteria, Protozoa, and Viruses

Molecular Techniques

References

Chapter 6: Exposure Assessment

Conducting the Exposure Assessment

Characterizing Concentration/Duration Distributions

Consumption Distributions

Afterword

Appendix

References

Chapter 7: Predictive Microbiology

Objective

Basic First-Order Processes and Deviations

Physical Removal

Types of Decay Processes

Types of Growth Processes

Data Sources

References

Chapter 8: Conducting the Dose–Response Assessment

Plausible Dose–Response Models

Framework for Mechanistic Dose–Response Relationships

Empirical Models

Fitting Available Data

Potential Impacts of Immune Status

Relationship between Dose and Severity (Morbidity and Mortality)

Reality Checking: Validation

Use of Indicators and Other Proxy Measures in Dose–Response

Advanced Topics in Dose–Response Modeling

Appendix

References

Chapter 9: Uncertainty

Point Estimates of Risk

Terminology: Types of Uncertainty

Sources of Uncertainty

Sources of Variability

Variability That Is Uncertain

Approaches to Quantify Parametric Uncertainty

Applications

Combining Parametric Uncertainty from Multiple Sources

Overall Risk Characterization Example

Second-Order Methods

Model Uncertainty and Averaging

References

Chapter 10: Population Disease Transmission

Introduction: Models for Population and Community Illnesses

REFERENCES

Chapter 11: Risk Characterization and Decision Making

Introduction

Valuing Residual Outcomes

Decision Making

Other Aspects Entering into a Decision

References

Index

End User License Agreement

List of Tables

Chapter 01

Table 1.1 Comparison of Five-Year Averages for Common Foodborne Reported Outbreaks

Table 1.2 Comparison of Laboratory Isolations and Outbreak Cases in England and Wales, 1992–1994

Table 1.3 Effect of Different Hypothetical Policy Options on Distribution of Risk Among Communities (for a Fixed Total Risk)

Table 1.4 Summary of Secondary Case Data in Outbreak Situations

Chapter 02

Table 2.1 Some Characteristics of Environmentally Transmitted Parasites

Table 2.2 Selected Bacteria of Medical Importance Transmitted through the Environment

Table 2.3 Classification of Animal Viruses

Table 2.4 Some Human Enteric Viruses

Table 2.5 Some Respiratory Viruses

Table 2.6 Host Factors That Can Influence Susceptibility and Severity of Disease

Table 2.7 Incubation Time for Common Enteric Pathogens

Table 2.8 Concentration of Enteric Pathogens in Feces

Table 2.9 Some Human Enteroviruses and Parechoviruses and Clinical Illness

Table 2.10 Diseases/Outcomes Caused by Coxsackievirus

Table 2.11 Drinking Water Outbreaks Associated with Enteroviruses

Table 2.12 Classification of Human Adenoviruses

Table 2.13 Summary of Pathogenic E. coli Incidence and Epidemiology

Table 2.14 Clinical and Epidemiological Characteristics of Legionnaire’s Disease and Pontiac Fever

Table 2.15 Routes of Transmission

Table 2.16 Sources of Microbial Aerosols

Chapter 03

Table 3.1 Definitions Used in Risk Analysis

Table 3.2 Steps in Risk Assessment to Address Human Health Effects Associated with Chemicals

Table 3.3 Some Agencies Involved in Risk Assessment

Table 3.4 Statutory Mandates on Risk

Table 3.5 What Is Acceptable Risk?

Table 3.6 Examples of Hazards and CCPs for Specific Foods

Chapter 04

Table 4.1 Methods for Diagnosing Infections and Disease by the Medical Community

Table 4.2 Morbidity Ratio for Salmonella (Nontyphi)

Table 4.3 Acute and Chronic Outcomes Associated with Microbial Infections

Table 4.4 Hospitalizations Associated with Microorganisms Responsible for Waterborne Outbreaks Reported in the United States, 1971–1992

Table 4.5 Enteric Bacterial Illnesses and Associated Hospitalizations and Mortality in Children in the United States

Table 4.6 Sensitive Populations in the United States

Table 4.7 Case Fatality Ratios for Enteric Pathogens in Nursing Homes versus General Population

Table 4.8 Mortality Ratios among Specific Immunocompromised Patient Groups with Adenovirus Infection

Table 4.9 Summary of Notifiable Diseases in the United States for Selected Years

Table 4.10 Summary of Patients Discharged from Hospitals by Category of Disease in the United States, 1990

Table 4.11 Historical Reported (US) Cases by Organism

Table 4.12 Various Estimates of Cases, Hospitalizations, and Deaths Associated with Foodborne Diseases

Table 4.13 Prediction of Waterborne Cryptosporidiosis in New York City in AIDS Patients Compared to the General Population

Chapter 05

Table 5.1 Method Attributes, Usefulness, and Limitations

Table 5.2 Microorganisms, Methods, and Units for Measurement

Table 5.3 Types of Data Needed for Exposure Assessment in Drinking Water

Table 5.4 Examples of qPCR Assays for the Detection of a Variety of Microbial Hazards

Chapter 06

Table 6.1 Count Data for Example 6.1

Table 6.2 Count Data for Example 6.2

Table 6.3 Schematic Layout of Dilution Assay

Table 6.4 Structure of Count Data Experiments from a Goodness-of-Fit Point of View

Table 6.5 Sample Volumes (L) in Which Various Numbers of Oocysts Were Found

Table 6.6 Frequency Distribution of Coliform Counts in 1 ml Samples in Lake Erie [28]

Table 6.7 Computations for Example 6.7

Table 6.8 Significance Level for the KS Test

Table 6.9 Count Data for Example 6.8

Table 6.10 Computations for Example 6.8

Table 6.11 Set Densities and Corresponding Likelihoods for Example 6.8

Table 6.12 Assay Data for Example 6.9

Table 6.13 Results for Example 6.9

Table 6.14 Computations for Example 6.10

Table 6.15 Computed Likelihoods for Example 6.10

Table 6.16 95% Confidence Limits to Poisson Counts for Small Total Counts (Exact Values)

Table 6.17 Results for Example 6.12

Table 6.18 Parameter Combinations of the GIG Distribution Yielding a Mean of 2 and a Variance of 10

Table 6.19 Results for Example 6.13

Table 6.20 Results for Example 6.14

Table 6.21 Comparison of Distributions for Example 6.14

Table 6.22 Computation of AIC and BIC for Fits in Table 6.20 (note nobs = 55)

Table 6.23 Results for Example 6.15

Table 6.24 Presumptive Giardia Counts in a Raw Water Supply

Table 6.25 Results for Example 6.16

Table 6.26 Data for Example 6.17

Table 6.27 Mean and Standard Deviation of Density from Various Mixing Distributions

Table 6.28 Point Estimates for Consumption/Contact Factors

Table 6.29 Consumption Frequency (U.S. Average) for Raw Foods (1989–1990) [78].

Table 6.30 Reported Distributions for Exposures

Table 6.31 Transfer of Bacteria between Skin and Hands

Table 6.32 Effect of Level of S. enteritidis in Fresh Eggs on the Probability of Acquiring Salmonella on Fingertips upon Breaking the Egg

Table 6.33 Variability of Food Consumption Patterns by Region and by Ethnic Group [78]

Table 6.34 Age-Specific Probability Distribution Parameters (LN) for Water Consumption (ml/day) in U.S. Populations

Chapter 07

Table 7.1 Data on Microorganism Concentration versus Distance

Table 7.2 Data on Growth of L. monocytogenes

Table 7.3 Summary of Fits of Alternative Logistic Models to Listeria Data

Table 7.4 Summary of Fits of Gompertz Model to Listeria Data

Table 7.5 Types of Interactions between Two Populations

Table 7.6 Die-off of V. splendidus in Water

Table 7.7 Data for Growth of Listeria in Milk

Table 7.8 Growth of P. aeruginosa in Water in the Presence of Background Microorganisms

Chapter 08

Table 8.1 Taxonomy of Potential Mechanistic Dose–Response Models

Table 8.2 Empirical Dose–Response Functions

Table 8.3 Schematic Layout of Dose–Response Assay

Table 8.4 Human Dose–Response Study of Ward et al.

Table 8.5 Best-Fit Parameters for Example 8.1

Table 8.6 Results for Example 7.2. Salmonella

Table 8.7 C. parvum (Iowa Strain) Infectivity in Human Volunteers by Oral Dosing

Table 8.8 Experimental Data Sets for E. coli

Table 8.9 Results for Example 8.4

Table 8.10 Hypothetical Data Sets Indicating Lack of Fit

Table 8.11 Optimal Parameter Estimates for Hypothetical Data Sets

Table 8.12 Experimental Data on Human Response to V. cholerae Inaba 569B Ingested with Bicarbonate

Table 8.13 Dose–Response for Infection of Human Volunteers by S. anatum Strain I

Table 8.14 Results for Example 8.7

Table 8.15 Summary of Mortality Ratios for Various Pathogens [55]

Table 8.16 Summary of Incubation Time Distribution Information

Table 8.17 Families of Time Dependency Models

Chapter 09

Table 9.1 Example Bootstrap Replicates for Cryptosporidium Data in Table 8.7

Table 9.2 Trials for Example 9.3

Table 9.3 Results for Example 9.3

Table 9.4 Modifications of the Chicken Problem to Include Variability

Table 9.5 Hypothetical Data Set (x1 and x2) with Ranked Values

Table 9.6 Rearranged Hypothetical Data

Table 9.7 Areas of Regions in Figure 9.19 in Terms of Copulas

Table 9.8 Results for Example 9.7

Table 9.9 Good Risk Assessment Principles

Table 9.10 Summary Statistics for Salmonella Risk Characterization

Table 9.11 Computations for Model Averaging Solution of Example 9.8

Chapter 10

Table 10.1 Comparison of Fits of Alternative Distributions to the Duration of Illness

Table 10.2 Microbial Daily Intakes Corresponding to Reported U.S. Waterborne Risk

Table 10.3 Confidence Limits to Number of Observed Illnesses as a Function of Community Size (constant per capita risk of 0.000294/person)

Chapter 11

Table 11.1 Costs (1981 dollars) of Eagle-Vail Waterborne Gastroenteritis Outbreak

Table 11.2 Economic Costs of Luzerne County, PA, Giardiasis Outbreak (1984 dollars)

Table 11.3 Cost of Illness per Case for Various Infectious Agents (in 2009 dollars)

Table 11.4 Willingness to Pay for Different Severities

Table 11.5 Original Classification of Disability Weights ([13]).

Table 11.6 Derivation of Average Cost per Case for Example 11.1

Table 11.7 Computation of Solution to Example 11.1 (years 6–26 omitted)

List of Illustrations

Chapter 01

Figure 1.1 Percentages of outbreaks associated with public water systems (n = 680) by time period 1971–2006 that had unknown etiologies based on data from Ref. [6].

Figure 1.2 Relationship between exposure, level of technological protection, and microbial risk. The middle curve indicates the best estimate. The other two curves indicate the upper and lower confidence regions.

Figure 1.3 Schematic of disease occurrence in a hypothetical community (Modified from Ref. [47]).

Figure 1.4 Weekly count of reported organism isolations in England and Wales: (a) rotavirus, (b) Clostridium difficile, (c) Salmonella derby, (d) Shigella sonnei, (e) influenza B, and (f) Salmonella typhimurium DT 104 (From Ref. [48]).

Chapter 02

Figure 2.1 Outcomes of exposure to a microbial infection.

Figure 2.2 Detection or notification of Cryptosporidium cases represents only a small proportion of all cases.

Figure 2.3 Infections of enteric viruses.

Figure 2.4 Routes of acquisition of rhinovirus infections modified from Ref. [39].

Figure 2.5 Routes of enteric microorganism transmission through the environment.

Figure 2.6 Environmental routes of pathogen transmission in the household.

Figure 2.7 Location of rhinovirus acquisition modified from Ref. [126].

Figure 2.8 Transport and fate of enteric viruses in the marine environment.

Chapter 03

Figure 3.1 The risk assessment and management elements as described in the Red Book 1983

Figure 3.2 Interactive risk assessment and management framework

Figure 3.3 (a and b) The ILSI framework for QMRA and analysis phase

Figure 3.4 Use of risk assessment for setting criteria in HACCP

Chapter 04

Figure 4.1 Outcomes of the infection process: possible groups for quantification in populations.

Figure 4.2 Quantifiable health outcomes associated with E. coli O157:H7 exposures and infections during a foodborne outbreak. *HUS is hemolytic uremic syndrome, affecting the kidneys, which leads to death or possible kidney failure and there is a need for a transplant or long-term dialysis.

Figure 4.3 Quantifiable health outcomes associated with assessment of Campylobacter-associated Guillain-Barré syndrome (GBS) a neurological disorder causing paralysis.

Figure 4.4 Hospitalizations in the elderly before and during a waterborne outbreak of cryptosporidiosis

Figure 4.5 Recreational waterborne outbreaks (n = 188) in natural waters, 1985–2008 in the United States by HAZ ID

Figure 4.6 Recreational waterborne cases (n = 7717) of disease from natural waters, 1985–2008 in the United States by HAZ ID

Figure 4.7 Attack rates for drinking water and recreational waters by HAZ ID (etiological agent)

Chapter 05

Figure 5.1 Steps involved in the collection, detection, and quantification of pathogens.

Figure 5.2 Procedure for concentration and detection of enteric viruses from water.

Figure 5.3 Detection of enteroviruses by PCR.

Chapter 06

Figure 6.1 Likelihood function versus mean density for Example 6.2.

Figure 6.2 First-order bias for various dilution experiments. Volumes are 10, 1, 0.1, and (for the four-dilution experiment) 0.01 ml, 5 or 50 tubes per dilution. Bias is estimated by the method of Salama.

Figure 6.3 Confidence region for Example 6.11.

Figure 6.4 pdf for Example 6.12.

Figure 6.5 Distribution functions for Poisson and NB distributions at fixed μV = 2.

Figure 6.6 Distribution functions for Poisson and PLN distribution at fixed μV = 2.

Figure 6.7 Distribution function for Poisson and PIG distributions at fixed μV = 2.

Figure 6.8 PGIG distribution with constant mean (μV = 2) and variance.

Figure 6.9 Confidence regions for the NB parameters of the fit to data in Example 6.14. Numbers on contours indicate the probability that true parameters are included in the region.

Figure 6.10 Cumulative distribution of mean densities estimated from data of Example 6.16 with 95% confidence limits based on uncertainties in distribution.

Figure 6.11 Density function for a triangular distribution.

Figure 6.12 Screen snapshot showing numerical solution of Example 6.9.

Figure 6.13 Screen snapshot for formulae to solve Example 6.9.

Figure 6.14 Setup of Solver for solution of Example 6.9.

Figure 6.15 MATLAB functions for solution of Example 6.9.

Figure 6.16 Output from execution of MATLAB solution for Example 6.9.

Figure 6.17 R functions for solution of Example 6.9.

Figure 6.18 Output from execution of R solution for Example 6.9.

Chapter 07

Figure 7.1 Graph of microbial concentrations from Example 7.1.

Figure 7.2 Chick’s Law and its deviations. Rate of inactivation (top) and survival curve (bottom).

Figure 7.3 Locus of points satisfying the conditions of Example 7.2.

Figure 7.4 Spectrum of electromagnetic radiation.

Figure 7.5 Conceptual batch microbial growth curve.

Figure 7.6 Logistic growth curves for different values of N0. Parameters are k = 2 day−1 and K = 10⁶.

Figure 7.7 Gompertz growth curve. Parameters: a = 18.816, b = 3 days, and c = 2 day−1.

Figure 7.8 Plot of theta-logistic model (Eq. 7.21). In all cases, k = 2 and K = 10⁶.

Figure 7.9 Plot of the Ross–Savageau generalization of the logistic (Eq. 7.23). Parameters: k = 0.3/day, K = 10⁶, and θ1 = 0.7, θ2 = 1.3.

Figure 7.10 R code to compute generalized logistic growth equation.

Figure 7.11 R listing for fitting Listeria data.

Figure 7.12 Fitting of Gompertz and three-theta generalized logistic to L. monocytogenes example.

Figure 7.13 Time course of microorganism and substrate concentration for simple Monod kinetics with exponential growth rate (N0 = 10, S0 = 500, μm =2, Ks = 30, Y = 0.7).

Figure 7.14 Time course of microorganism and substrate concentration for Monod kinetics with lag (N0 = 10, S0 = 500, μm = 2, Ks = 30, Y = 0.7, α = 0.2).

Figure 7.15 Monod model with lag and inhibitor-produced decay (N0 = 10, S0 = 500, μm = 2, Ks = 30, Y = 0.7, α = 0.05, kI = 0.001, kd = 0.1).

Chapter 08

Figure 8.1 Development of consequences from E. coli O157:H7.

Figure 8.2 Comparison of exponential and beta-Poisson models on log–log (top) and arithmetic (bottom) scales.

Figure 8.3 Behavior of threshold models as a function of kmin. All models computed to have a median infectious dose of 10 organisms.

Figure 8.4 Dose–response model for constant host–microorganism interaction ratio (r) and truncated Poisson distribution for kmin. All curves fixed at a median infectious dose of 200.

Figure 8.5 Log-probit, log-logistic (logit), and Weibull models. All curves have median infectious dose of 200 and 10% infectious dose of 100 organisms in average dose.

Figure 8.6 Comparison of model fits to rotavirus data. Inset graph shows log–log plot of best-fit models.

Figure 8.7 Comparison of fit of rotavirus dose–response to exact and approximate beta-Poisson equations. Inset on log–log scale.

Figure 8.8 Comparison of best-fit (solid line) exponential dose–response relationship to data on oral infectivity of Cryptosporidium to humans (Table 8.6). Dashed lines give 95% confidence limits.

Figure 8.9 Deviance residuals for hypothetical data sets.

Figure 8.10 Comparison of beta-binomial and binomial distributions (n = 25, π = 0.4).

Figure 8.11 Sample trajectories for microbial dynamics in vivo (six trajectories shown).

Figure 8.12 Results of 500 simulations for conditions in Figure 8.11.

Figure 8.13 Exponential dose–response model with inverse power TPI dependency (curves) compared to observed mortalities against TPI (points) from the study of Rogers et al. [90]. (○) 2 CFU, (▵) 8 CFU, (+) 26 CFU, (×) 74 CFU, (◊) 257 CFU.

Figure 8.14 Program listing in R to compute best-fit parameters.

Figure 8.15 Output of the optimization for fitting the data from Table 8.4 to the beta-Poisson model.

Chapter 09

Figure 9.1 Point versus interval estimate of risk.

Figure 9.2 Taxonomy of uncertainty. After Reference [4].

Figure 9.3 Schematic of the logic of the bootstrap process. Modified from Reference [12].

Figure 9.4 Example informative and uninformative priors (parameter range is restricted to the interval <0,1>).

Figure 9.5 R code for bootstrap analysis of mean density for data in Table 6.5.

Figure 9.6 Bootstrap cdf for Cryptosporidium density estimate (Table 6.5).

Figure 9.7 R code for determining the posterior distribution for the negative binomial distribution parameters fit to data in Example 6.5.

Figure 9.8 Contours (numbers indicate values of the density) of the posterior distribution of the negative binomial parameters.

Figure 9.9 Code snippet for generating samples from posterior for negative binomial example.

Figure 9.10 Random samples from the posterior distribution.

Figure 9.11 Determination of likelihood ratio confidence interval for Cryptosporidium data.

Figure 9.12 Likelihood-based confidence limits for the rotavirus beta-Poisson dose–response models. Labels indicate confidence levels.

Figure 9.13 Best-fit and 95% likelihood-based confidence limits for rotavirus dose–response using the beta-Poisson model.

Figure 9.14 Bootstrap distribution of rotavirus parameter estimates (beta-Poisson model). Of the 1000 trials, 12 (not plotted) had α > 10. The central marker indicates the maximum likelihood estimate from the underlying data.

Figure 9.15 Distribution of average microbial exposure from Monte Carlo trials: CCD (top) and full distribution (bottom).

Figure 9.16 Rank correlation coefficient of inputs to the estimation of average microbial dose for the Salmonella-on-chicken problem.

Figure 9.17 Distribution of derived function from correlated and relatively uncorrelated data.

Figure 9.18 Relationship between Spearman correlation and the parameter in Frank’s copula (for positive values). Points indicate exact computations of numerical integral. Curve shows the approximate numerical equation.

Figure 9.19 Definition sketch for evaluation of bivariate probability in a rectangular region. The probability in region IV is sought.

Figure 9.20 Scatterplot of 3000 random variates drawn from four different copulas, each with a bivariate rank (Spearman) correlation of 0.86.

Figure 9.21 Effect of number of trials per Monte Carlo run on mean output and standard deviation of output (eight runs per sample size) for Example 9.3.

Figure 9.22 Bootstrap sampling distribution of Salmonella dose–response curves (1000 bootstrap iterations). Central crosshairs indicate maximum likelihood values.

Figure 9.23 Cumulative (b) and complementary cumulative (a) distributions for the estimated risk in the Salmonella case study.

Figure 9.24 Sensitivity chart for Salmonella case study.

Figure 9.25 Two-dimensional Monte Carlo presentation of Salmonella case study. The x-axis is the variability dimension. The lines provide ranges based on uncertainty (the upper and lower 2.5 percentiles).

Figure 9.26 Schematic two-dimensional Monte Carlo plots indicating results with low (left) and high (right) importance of uncertainty.

Chapter 10

Figure 10.1 Basic SIR model.

Figure 10.2 Time course of a simple outbreak (single 1-h exposure at t = 0–1 h).

Figure 10.3 Time course of outbreak (two 1-h exposures at t = 0–1 and t = 24–25 h).

Figure 10.4 Excretion of rotavirus by children subject to onset of diarrhea. The straight line is the best fit exponential. Error bars are standard deviations from binomial sampling.

Figure 10.5 Dynamics of outbreak consideration durations in diseased and asymptomatic states.

Figure 10.6 Effect of dose on incubation curve from exponential dose response with inverse exponential time dependency.

Figure 10.7 Effect of theta on incubation curve from exponential dose–response with inverse exponential time dependency.

Figure 10.8 Comparison of cumulative distribution functions for the Nipah virus duration to observed data.

Figure 10.9 Excretion of rotavirus in children prior to symptomatic illness. Error bars are one standard deviation of the observed proportion.

Figure 10.10 Outbreak dynamics with secondary infections.

Figure 10.11 Probability of detection of clusters given baseline values from Table 10.3.

Chapter 11

Figure 11.1 Concept of minimizing social costs.

Figure 11.2 Patterns of disability weights for diseases with different types of progression.

Figure 11.3 Example disability weights. .

Figure 11.4 Impact of including ozone on hypothetical water risks in an exemplar Dutch scenario.

Figure 11.5 Hypothetical cumulative distribution function of estimated net benefits.

Figure 11.6 Hypothetical set of alternatives evaluated by risk reduction (benefit) and technological implementation costs. *, Potential alternatives.

Second Edition

Quantitative Microbial Risk Assessment

Charles N. Haas

Joan B. Rose

Charles P. Gerba

Wiley Logo

Copyright © 2014 by John Wiley & Sons, Inc. All rights reserved

Published by John Wiley & Sons, Inc., Hoboken, New Jersey

Published simultaneously in Canada

No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 7508400, fax [phone], or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, [phone], fax [phone], or online at http://www.wiley.com/go/permission.

Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 7622974, outside the United States at (317) 5723993 or fax [phone].

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com.

Library of Congress Cataloging-in-Publication Data:

Haas, Charles N.Quantitative microbial risk assessment / Charles N. Haas, Joan B. Rose, Charles P. Gerba. – Second edition.

p. cm

Includes bibliographical references and index.

ISBN 978-1-118-14529-6 (cloth : alk. paper)

1. Communicable diseases–Epidemiology–Methodology. 2. Health risk assessment. 3. Infection--Mathematical models. 4. Environmental health–Mathematical models. I. Rose, Joan B. II. Gerba, Charles P., 1945- III. Title.

RA643.H22 2014

615.9'02–dc23

2014002690

Preface

In the 14 years since we prepared the first edition, there has been an explosion in knowledge of and need for quantitative microbial risk assessment (QMRA). While our motivation for the first edition stemmed from concerns (principally in water) about enteric bacteria, viruses, and protozoa, the motivation has now exploded to new domains and agents. SARS, influenza, biothreat agents, and zoonotic pathogens have all become of greater concern.

The 2001 anthrax letters have highlighted the need for risk assessment of inhaled agents. Both biothreat agents and emergence of new strains of virulent contagious organisms have raised concern for modeling pathogen dynamics in populations.

In this edition, we have retained the fundamental approach of the risk assessment methodology as a central paradigm. We have added new material on modern pathogen analytical methods, predictive microbiology (of pathogen growth and decay), dynamic risk models (explicitly considering incubation time), and disease propagation models in populations. Of necessity we have removed some material—it is no longer possible to present comprehensive tables of microbial dose–response parameters.

In the years since the first edition, the authors have gained experience in teaching this material to generations of students—in the form of formal classes, tutorials, independent studies, and short courses. We know this book can be valuable in instructing advanced students in environmental sciences, environmental engineering, public health, and microbiology. It is also a useful reference for practitioners and regulatory personnel. Some prior statistical background would be useful in approaching the material, but not necessary; the key requirement for any risk assessor is the absence of fear from mathematical constructs and concepts.

The three of us have been on a QMRA journey for almost 30 years. We have learned that doing high-quality risk assessments is of necessity a team sport, requiring individuals with different skills and interests. We have learned a tremendous amount from each other, from our students, from our collaborators, and from the problems that we have sought to approach. Practitioners of the art of quantitative microbial risk assessment should be advised to cast a wide net with respect to colleagues and collaborators to perfect their craft.

We encourage comments and feedback from users of this work, and look forward to observing and participating in developments in coming years, and ultimately to handing the baton off to our students and their students.

sign1

Charles N. Haas

sign2

Joan B. Rose

sign3

Charles P. Gerba

November 2013

Chapter 1

Motivation

The prevention of infectious disease transmission from human exposure to contaminated food, water, soil, and air remains a major task of environmental and public health professionals. There are numerous microbial hazards, including exposure via food, water, air, and malicious release of pathogens that may arise. Indeed, some have argued that the property of virulence of human pathogens is one which is favored by evolutionary interactions between pathogens and host populations and therefore will always be of important concern [1]. To make rational decisions in preparing, responding, and recovering from exposures to such hazards, a quantitative framework is of high benefit.

The objective of this book is to comprehensively set forth the methods for assessment of risk from infectious agents transmitted via these routes in a framework that is compatible with the framework for other risk assessments (e.g., for chemical agents) as set forth in standard protocols [2, 3].

In this chapter, information on the occurrence of infectious disease in broad categories will be presented, along with a historical background on prior methods for assessment of microbial safety of food, water, and air. This will be followed by an overview of key issues covered in this book.

Prevalence of Infectious Disease

Outbreaks of infectious waterborne illness continue to occur, although it remains impossible to identify the infectious agent in all cases. For example, in 1991, a waterborne outbreak in Ireland resulting from sewage contamination of water supplies infected about 5000 persons. However, the infectious agent responsible for this outbreak could not be determined [4]. In the United States, it has been estimated that 38 million cases of foodborne infectious disease occur annually with unidentified agents [5].

In the United States, there have typically been three to five reported outbreaks per year in community drinking water systems involving infectious microorganisms, with perhaps up to 10,000 annual cases [6]. The 1994 Milwaukee Cryptosporidium outbreak with over 400,000 cases [7, 8] was a highly unusual event among these statistics. As shown in Figure 1.1, there has been an increasing ability to identify microorganisms responsible for waterborne diseases, and it is expected that with advances in molecular biology, this will increase.

c1-fig-0001

Figure 1.1 Percentages of outbreaks associated with public water systems (n = 680) by time period 1971–2006 that had unknown etiologies based on data from Ref. [6].

There are substantially more outbreaks and cases of foodborne infectious diseases than are reported. Table 1.1 summarizes reports of U.S. cases of principal microbial infectious foodborne illnesses for two 5-year periods (1988–1992 and 2002–2006). There is a mix of causal agents, including bacteria, virus, and protozoa. It is noteworthy that (as in the case of waterborne outbreaks) the frequency of outbreaks of unknown etiology has dramatically decreased but the frequency of outbreaks associated with norovirus has dramatically increased. These changes are due in part to the ability to better identify causal agents (e.g., via molecular methods).

Table 1.1 Comparison of Five-Year Averages for Common Foodborne Reported Outbreaks

Source: From Refs. [9, 10].

a Include both Shiga toxigenic and enterotoxigenic.

It is generally recognized that reported outbreaks, either of water- or foodborne infectious disease, represent only a small fraction of the total population disease burden. However, particularly in the United States, voluntary reporting systems and the occurrence of mild cases (for which no medical attention is sought but nevertheless are frank cases of disease) have made it difficult to estimate the total caseload.

In the United Kingdom, comparisons between the number of confirmed cases in infectious disease outbreaks and total confirmed laboratory illnesses (occurring in England and Wales) have been made (Table 1.2). This suggests that the ratio of reported outbreak cases to total cases that may seek medical attention may be from 10 to 500:1, with some dependency on the particular agent.

Table 1.2 Comparison of Laboratory Isolations and Outbreak Cases in England and Wales, 1992–1994

Source: Modified from Ref. [11].

Colford et al. [12] developed estimates for the total disease burden associated with acute gastroenteritis from drinking water. This relies on combining the reported outbreak data with interventional epidemiologic studies. Based on their analysis, the total U.S. disease burden is estimated to be 4.26–11.69 million cases per year in the United States, which is substantially in excess of the reported outbreaks. In the case of foodborne illness, there are an estimated 14 million cases per year [13].

Drinking water and food are by no means the only potential routes of exposure to infectious agents in the environment. Recreation in water (either natural or artificial pools) containing pathogens can produce illness [14].

Indoor air transmission can be a vehicle of infection. Legionella transmitted through indoor environments has been a concern since the 1970s [15]. The multinational epidemic of severe acute respiratory syndrome (SARS), caused by a coronavirus, was abetted at least in one location in Hong Kong by indoor aerosol transmission between apartments of infected individuals and susceptible individuals [16]. A broad spectrum of other respiratory pathogens including influenza, rhinoviruses, and mycobacteria can be transmitted by this route [17].

The deliberate release of Bacillus anthracis spores in 2001 (the Amerithrax incidents) brought widespread awareness to the potential for indoor releases (as well as releases in other venues) of bioterrorist agents to cause risk [18]. Therefore, of necessity, microbial risk assessors may need to consider the impact of malicious activity in certain applications.

Prior Approaches

Concerns for microbial quality of food, water, and other environmental media have long existed. In the early twentieth century, the use of indicator microorganisms was developed for the control and assessment of the hygienic quality of such media and the adequacy of disinfection and sterilization processes. The coliform group of organisms was perhaps first employed for this purpose [19–21]. Indicator techniques have also found utility in the food industry, such as the total count for milk and other more recent proposals [22]. Other indicator groups for food, water, or environmental media have been examined, such as enterococci [23–25], acid-fast bacteria [26], bacteriophage [27–29], and Clostridia spores [29–31].

The use of indicator organisms was historically justified in because of difficulty in enumerating pathogens. However, with the increasing availability of modern microbial methods, for example, PCR, immunoassay, etc., for direct pathogen assessment, this justification has become less persuasive. In addition, in order to develop health-based standards from indicators, extensive epidemiologic surveillance is often necessary. The use of epidemiology has limitations with respect to detection limits (for an adverse effect) and is also quite expensive to conduct. Indicator methods are also limited in that many pathogens are more resistant to die off in receiving environments or source waters than indicators or have greater resistance to removal by treatment processes than indicators [26, 28, 29, 32]. Thus, the absence of indicators may not suffice to ensure the absence of pathogens. Even after a century of use, the indicator concept remains imperfect [33].

The use of quantitative microbial risk assessment (QMRA) will enable direct measurements of pathogens to be used to develop acceptance/rejection guidelines for food, water, and other vehicles that may be the source of microbial exposure to human populations. The objective of this book is to present these methods in a systematic and unified manner.

Scope of Coverage

QMRA is the application of principles of risk assessment to the estimate of consequences from a planned or actual exposure to infectious microorganisms. In performing a QMRA, the risk assessor aims to bring the best available information to bear in understanding the nature of the potential effects from a microbial exposure. Since the information (such as dose–response relationships, exposure magnitudes) is almost invariably incomplete, it is also necessary to ascertain the potential error involved in the risk assessment. With such information, necessary steps to mitigate, control, or defend against such exposures may be developed.

At the outset of performing a risk assessment, a scoping task should be undertaken. This task should set forth the objectives of the analysis and the principal issues to be addressed. Items such as consideration of secondary cases, individual versus population risk, agent or agents to be examined, exposure routes, and/or accident scenarios must be stipulated. However, this scoping may be changed during the course of a QMRA, to reflect the input derived from the risk manager(s) and other stakeholders.

Potential Objectives of a QMRA

There may be diverse objectives for a QMRA. These objectives relate to the rationale for the performance of the assessment, as well as the methods to be employed. Broadly, the different objectives reflect different scales at which a risk assessment may be performed. The step of problem formulation is critical to any risk estimate [34]. It is necessary that the problem be formulated to meet the needs of the risk managers and stakeholders; indeed, it is now recognized that the successful practice of risk analysis requires frequent interchange with manager and stakeholders [3]. In general, the problems posed are of several types.

Site-Specific Assessment

The simplest type of QMRA that may be performed involves one site or exposure scenario. The following are typical of the questions that might be asked:

If a water treatment plant is designed in a certain way (with given removals of pathogens), then what is the risk that would be placed upon the population served?

A swimming outbreak (in a recreational lake) has just occurred. I believe that it resulted from a short-duration contamination event. What pathogen levels would be consistent with the observed attack rate?

Microbial sampling of a finished food product has found certain pathogens. What level of risk does this pose to consumers of the product?

A certain amount of infectious agent has been released into a room. What is the immediate danger to occupants, and how stringent should cleanup levels be?

Note that there are certain other contrasts in the objectives of the risk assessments to be posed. In (1) and (3), a before-the-fact computation is desired, while in (2) and (4), an after-the-fact computation is described. Also in (1), (3), and (4), pathogen levels are available (or somehow are estimated), while in (2), an inverse computation is needed given an observed attack rate.

In performing this risk assessment, the relationship between an exposure or technological metric and a risk measurement must be ascertained and then the particular point of correspondence determined (Fig. 1.2). In cases (1), (3), and (4), for a known (or assumed) exposure (on the x-axis), the corresponding range of risks on the y-axis is sought. In cases (2), for known or assumed risks (on the y-axis), the corresponding range of exposures (or level of technological protection) is to be determined (on the x-axis).

c1-fig-0002

Figure 1.2 Relationship between exposure, level of technological protection, and microbial risk. The middle curve indicates the best estimate. The other two curves indicate the upper and lower confidence regions.

Ensemble of Sites

A somewhat more complex situation occurs if the risk for a set of events or sites must be estimated. Basically, this now includes the necessity to incorporate site-to-site factors into the assessment. Some examples of this are as follows:

If I desire keeping the risk to a population served by multiple water treatment plants at a given level (or better), then what criteria should I use (microbial levels)?

For a food product subject to contamination by pathogens, what would be an acceptable treatment specification (e.g., heating time, holding period) to ensure microbial acceptability?

I am designing a water quality standard for recreational bathing waters. If a uniform (e.g., national) standard is to be developed, what standard would ensure that average risk was acceptable with keeping the risk of a large cluster of illnesses low?

In addition to incorporating a measure of ensemble average risk, in general, it is also desired to ensure that no single member of the ensemble be unacceptably extreme. For example, consider the evaluation of three options of disease control among three communities, as indicated in Table 1.3.

Table 1.3 Effect of Different Hypothetical Policy Options on Distribution of Risk Among Communities (for a Fixed Total Risk)

This table indicates the number of cases, and the rate, among the three communities. The three policy options yield the same number of expected cases. However, there are differences in the allocation of risk among the communities of different sizes. In option A, all communities have an identical level of estimated risk. In option B, the risk increases as community size decreases, while in option C, the risk increases as community size increases. This distribution of risk among affected subsets of the ensemble being considered adds an additional dimension for consideration by a risk manager—which may be termed risk equity.

Secondary Transmission

Infectious microbial diseases are different in terms of risk to a population than are chemical agents in that an individual who may become infected (with or without illness) can then proceed to infect additional individuals. These secondary (tertiary, quaternary, etc.) cases may be persons who had no direct contact with the initial vehicle of exposure, but nevertheless in fairly accounting for the public health impact, they should be considered.

Secondary cases may arise by a variety of mechanisms. Particularly among close family members, household secondary cases can arise by direct or indirect (e.g., surface contamination) contact; this is particularly so when the primary case or one household secondary case is a child [35–37]. Table 1.4 summarizes secondary case statistics obtained from a variety of outbreaks. As will be discussed in Chapter 10, the secondary case rate is a complex factor involving (among other things) the nature of the venue and contact patterns when infected and susceptible individuals intermingle.

Table 1.4 Summary of Secondary Case Data in Outbreak Situations

N/A, information not available.

a Ratio of secondary cases to primary cases.

b Proportion of households with one or more primary cases who have one or more secondary cases.

c Proportion of persons in contact with one or more primary cases who have a secondary case.

Presumably, secondary cases may also arise from close contact with an asymptomatic individual (in the carrier state). This is well known for highly acute and (now) uncommon illnesses (such as typhoid). Excretion of Norwalk virus following recovery (and resulting in additional cases) has been documented to occur for as long as 48 h post recovery [44].

Outbreaks versus Endemic Cases

As noted previously, there may be a substantial difference between reported outbreak cases and total disease burden in a community. In order for a disease case to receive recognition by the public health authorities, the following specific and sequential steps must occur [47]:

An ill person must seek medical care.

Appropriate clinical tests (e.g., blood, stool) must be ordered by the attending physician.

The patient must comply with obtaining the sample.

The laboratory must be capable of detecting the relevant pathogens.

The clinical test must be positive.

The test result must be reported to the health agency in a timely manner.

If any of the links in this sequential chain are broken, then a disease case will not enter the records maintained by health authorities. For example, with increasing controls on medical care, stool samples may not be obtained from mild cases of illness. Some organisms may only be present sporadically, or may be difficult to test, in stool or blood sample. Patients may not seek medical attention for mild cases of illness. Furthermore, in the United States in particular, the surveillance of environmentally induced disease is done on a passive basis, and hence, the number of actual illness clusters that are actually compiled into recorded statistics is only a small fraction of such clusters of illness that occur [47].

From a more fundamental point of view, an outbreak of illness is generally defined as occurrence of the illness at a level greater than normal or anticipated. This definition recognizes that there is a level of illness (endemic) that may exist under usual circumstances. The detection of such outbreaks poses a particular challenge. The problem is illustrated conceptually in Figure 1.3.

c1-fig-0003

Figure 1.3 Schematic of disease occurrence in a hypothetical community (Modified from Ref. [47]).

Additional complications arise from the different patterns of illness in a community, including definite periodicities, as well as temporal trends, and from the presence of reporting lags associated with laboratory analysis and time for patients to seek medical attention. Figure 1.4 illustrates the different patterns of illness in the case of six pathogens for England and Wales [48].

c1-fig-0004

Figure 1.4 Weekly count of reported organism isolations in England and Wales: (a) rotavirus, (b) Clostridium difficile, (c) Salmonella derby, (d) Shigella sonnei, (e) influenza B, and (f) Salmonella typhimurium DT 104 (From Ref. [48]).

In the case of waterborne and foodborne illnesses, it is highly likely that the level of such endemic illnesses is substantially greater than those occurring during outbreaks (even accounting for unrecognized outbreaks).

As a result, there are often many cases of environmentally caused (water, air, food) infectious disease that are unrecognized. One example of this is Campylobacter. There has been an average of about 200 cases per year of water- and foodborne illness in outbreaks of this organism, and yet estimates of the disease burden suggest about 2,100,000 cases per year, that is, approximately 10,000 cases per case of detectable outbreak illness. Therefore, it will be important to assess the factors that may influence outbreak detection. These issues will be discussed in subsequent chapters.

References

1. Levin, B. R. 1996. The Evolution and Maintenance of Virulence in Microparasites. Emerging Infectious Disease 2:93–102.

2. National Academy of Sciences. 1983. Risk Assessment in the Federal Government: Managing the Process. National Academy Press, Washington, DC.

3. National Research Council. 2009. Science and Decisions: Advancing Risk Assessment. National Academies Press, Washington, DC.

4. Fogarty, J., L. Thornton, and R. Corcoran. 1995. Illness in a Community Associated with an Episode of Water Contamination with Sewage. Epidemiology and Infection 114:289–295.

5. Scallan, E. 2011. Foodborne Illness Acquired in the United States—Unspecified Agents. Emerging Infectious Diseases 17, 16–22.

6. Craun, G. F., J. M. Brunkard, J. S. Yoder, V. A. Roberts, J. Carpenter, T. Wade, R. L. Calderon, J. M. Roberts, M. J. Beach, and S. L. Roy. 2010. Causes of Outbreaks Associated with Drinking Water in the United States from 1971 to 2006. Clinical Microbiology Reviews 23:507–528.

7. Edwards, D. D. 1993. Troubled Waters in Milwaukee. ASM News 59:342–345.

8. MacKenzie, W. R., N. J. Hoxie, M. E. Proctor, M. S. Gradus, K. A. Blair, D. E. Peterson, J. J. Kazmierczak, K. R. Fox, D. G. Addias, J. B. Rose, and J. P. Davis. 1994. Massive Waterborne Outbreak of Cryptosporidium Infection Associated with a Filtered Public Water Supply, Milwaukee, Wisconsin, March and April 1993. New England Journal of Medicine 331:161–167.

9. Anonymous. 2010. Surveillance for Foodborne Disease Outbreaks—United States, 2007. Morbidity and Mortality Weekly Reports 59:973–979.

10. Bean, N. H., J. S. Goulding, C. Lau, and F. J. Angulo. 1996. Surveillance for Foodborne-Disease Outbreaks—United States, 1988–1992. Morbidity and Mortality Weekly Reports 45:1–66.

11. Wall, P. G., J. de Louvois, R. J. Gilbert, and B. Rowe. 1996. Food Poisoning: Notifications, Laboratory Reports and Outbreaks—Where do the Statistics Come From and What Do They Mean? Communicable Disease Report Review 6, R93–R100.

12. Colford, J. M., S. Roy, M. J. Beach, A. Hightower, S. E. Shaw, and T. J. Wade. 2006. A Review of Household Drinking Water Intervention Trials and an Approach to the Estimation of Endemic Waterborne Gastroenteritis in the United States. Journal of Water and Health 4:71.

13. Mead, P. S., L. Slutsker, V. Dietz, L. F. McCaig, J. S. Bresee, C. Shapiro, P. M. Griffin, and R. V. Tauxe. 1999. Food Related Illness and Death in the United States. Emerging Infectious Disease 5:607–625.

14. Dziuban, E. J., J. L.