Developments in microbiological risk assessment for drinking water

Summary, 191 2. Introduction, 192 2.1 Overview of the risk assessment approach, 192 3. Pathogen exposures through drinking water, 192 3.1 Pathogen exposures under non-outbreak conditions, 193 3.1.1 Variation in micro-organism counts within large volume samples, 193 3.1.2 Effect of treatment on the spatial distribution of micro-organisms, 193 3.1.3 Implication for Cryptosporidium exposures to drinking water consumers under non-outbreak conditions, 195 3.2 A model for Cryptosporidium oocyst concentrations in drinking water during an outbreak, 196 4. Dose–response curves. Estimating the risk to humans from ingesting low pathogen doses, 197 4.1 Cryptosporidium parvum, 197 4.1.1 Experiments with salmonellas in mice suggest that pathogens act independently and do not co-operate during infection, 198 4.1.2 Acquired protective immunity for Cryptosporidium parvum, 199 4.1.3 Virulence of different strains of Cryptosporidium parvum, 199 4.2 Human rotavirus, 199 4.3 Bovine spongiform encephalopathy, 200 4.3.1 The human oral ID50, 200 4.3.2 Evidence that bovine spongiform encephalopathy prions may act co-operatively, 200 4.3.3 Estimating the risks from ingestion of minute subfractions of an ID50 through drinking water, 200 4.4 The need for more dose–response data for waterborne pathogens, 201 5. Integrating pathogen exposures and dose–response curves, 201 5.1 Risk prediction for an outbreak of cryptosporidiosis, 201 5.2 Risk prediction for rotavirus, 202 5.3 Risk prediction for highly infectious pathogen (ID50=1 micro-organism), 202 6. Risk prediction for bovine spongiform encephalopathy, 202 7. Discussion, 202 8. References, 204 This paper considers the development of microbiological risk assessment models for pathogenic agents in drinking water with particular reference to Cryptosporidium parvum, rotavirus and bovine spongiform encephalopathy (BSE). The available evidence suggests that there is potential for considerable variation in exposures to C. parvum oocysts through drinking water, during both outbreak and non-outbreak conditions. This spatial/temporal heterogeneity arises both from variation in oocyst densities in the raw water and fluctuations in the removal efficiencies of drinking water treatment. In terms of risk prediction, modelling the variation in doses ingested by individual drinking water consumers is not important if the dose–response curve is linear and the oocysts act independently during infection. Indeed, the total pathogen loading on the population as represented by the arithmetic mean exposure is sufficient for risk prediction for C. parvum, BSE and other agents of low infectivity, providing the infecting particles (i.e. oocysts or BSE prions) are known to act independently. However, for more highly infectious agents, such as rotavirus, ignoring the variation and just using the arithmetic mean exposure may over-estimate the risk by a factor of about threefold. If it were to be shown that pathogens co-operate with each other during initiation of infection, such that the dose–response relationship is non-linear, then modelling the variation in doses ingested by individual consumers would be very important. Possible mechanisms for co-operation of pathogens during infection are considered. Simulations show that acquired protective immunity for C. parvum reduces the risk of infection during outbreak conditions by over 10-fold. Variation in virulence between strains of C. parvum is a further source of uncertainty. Pathogens can and do gain entry into drinking water supplies even in well-developed countries. Break-through during treatment and ingress through cracked pipes are well-documented causes of waterborne outbreaks of cryptosporidiosis (6) and Escherichia coli O157 (42), respectively. Microbiological risk assessment (MRA) is the emerging method to predict the risks to public health from those pathogens. Models have been developed for a range of waterborne pathogens in drinking water, including C. parvum (45), Giardia (37) and enteric viruses (21). Most of the models represent non-outbreak conditions and model the endemic levels of infection through drinking water. Recently, a model for C. parvum has been developed for conditions representative of a waterborne outbreak of cryptosporidiosis (13). The risk assessment approach has been used as a guide for microbial standards (39). It be used to such as more be if of the drinking water treatment and by public health be if is of is to a for and emerging agents in the of a risk assessment for the of BSE to humans through drinking water developed This is different to those for well-documented waterborne pathogens in the that there is evidence to the of of BSE from humans to of bovine to in human in is the The between BSE in and the of in humans has been data are available for the of bovine in The of the for BSE were to the BSE in the a risk to those drinking water consumers by water to and to drinking water supplies were The not to and risks of infection. risk assessment models are on the is the of the protective are of the the are to and the infectious the is to the of drinking water treatment and of the in BSE acquired protective immunity in C. parvum infection and in infection be in the of the data available for the and In there be In the variation in pathogen exposures through drinking water is and then of and variation in dose–response are considered. are then used to different dose–response with pathogen exposure models the of micro-organisms, including and of and pathogens, show considerable variation in drinking water supplies This variation can be not between different water within different of the ingress of water through a cracked in pathogen such an for the waterborne outbreak of coli O157 in in of and The variation in densities can be in pathogen in the in pathogen in the raw and fluctuations in the of removal by drinking water treatment the spatial/temporal variation in pathogen densities within the The models for drinking water are by the variation in pathogen exposures the of the individual drinking water This variation is with reference to exposure to Cryptosporidium oocysts through drinking water supplies under both non-outbreak and outbreak conditions. Pathogen counts are in large volume from drinking water for the of Cryptosporidium in water are not do not an of has been of the or of the pathogens under non-outbreak conditions Indeed, of from drinking water supplies to oocysts of However, even during non-outbreak conditions, of Cryptosporidium oocysts do through drinking water treatment and into the This is drinking water treatment is even Indeed, drinking water pathogen a of oocysts in a volume of water. In the oocyst densities of to have been in drinking water or outbreak in the The oocysts in those may have oocysts and not infectious to It is that oocyst densities in drinking water supplies during non-outbreak conditions show large a counts This both spatial/temporal within different of the and between different supplies to different raw water that of in the to in the between and water that of the water in the is not and that of between and water doses individual drinking water consumers in is on pathogens are within those the of ingested by individual a volume of drinking water with Cryptosporidium models for drinking water that each (21). be with each drinking of that oocysts are to the distribution and each just oocysts are and between and oocysts and oocysts are to the distribution in the oocysts and the other are not of pathogens to the there is evidence for to The available evidence suggests that the may be between and from the volume of water and a considerable range of over The counts not the distribution and with more about The data the of and within the volume of water. evidence that in the of of by the of drinking water treatment. The treatment by an drinking water treatment The that drinking water treatment not of the the spatial of the the counts in within large of raw water. In the counts of in water variation in This is shown in for counts within a volume of water from a the mean counts of or counts from the water be by the Indeed, the counts in as a It is that the distribution not the of in as a of volume of water counts and and for counts in from water data from the drinking water treatment by a more counts were not in of the water In the of of into the raw water to the counts were within water However, spatial heterogeneity of the in the water in other This is shown for a of of the in 4. of the or counts and the arithmetic mean and This heterogeneity not be by the Indeed, of from a distribution with a mean of would between and are for the spatial heterogeneity in counts within water counts and distribution for counts in from water from a Variation in counts of in from within water The with counts are to or However, it be that were more in the water and in the raw Indeed, counts in the raw in are by and during drinking water treatment in a from particles during treatment of the with to microbiological of different of or within the In there is the of the is (i.e. show spatial do In to the of spatial heterogeneity of total and by drinking water treatment in Indeed, counts were to be within of water those in It be that water supplies do not and a distribution of Cryptosporidium oocysts in an drinking water data are available on the of spatial heterogeneity of C. parvum oocysts within large of water. If the are representative of C. parvum oocyst within large volume drinking water under conditions, oocysts be under other conditions, spatial heterogeneity The total of in the in the of Cryptosporidium oocysts within an volume be in such that would or individual would oocysts and It is even under non-outbreak conditions, the that consumers doses of just or oocysts in a It be that oocysts are during non-outbreak conditions for to However, oocyst counts of and in have been other with oocyst counts have been and would have been by more It be that the of the population oocysts each during non-outbreak conditions, the of oocysts in water supplies used a approach to exposures to Cryptosporidium oocysts to individual consumers drinking water for conditions representative of a waterborne The approach to a distribution to the doses of oocysts in the volume ingested by consumers drinking from the The of individual consumers ingesting doses of oocysts are in 1. to the the of consumers do not oocysts each even during an This is with the that of not oocysts during the outbreak of cryptosporidiosis and that oocysts are not in outbreaks The that about of those consumers were ingested just a oocyst each However, a of consumers were to between and oocysts in consumers to very doses the ID50 of about oocysts for C. parvum in human with acquired protective immunity ID50 is the of pathogens to each and of a of that dose–response curve by to Cryptosporidium parvum human data for on the of evidence of infection with C. parvum The curve is shown for of were with a of oocysts curve through The arithmetic mean for the oocyst exposures oocysts using and a distribution are in 1. The models with the distribution and the large variation in the exposures would predict that a of the population is to oocysts with that those are more about oocysts each ignoring the heterogeneity in would predict that a would be to a the ID50 for C. parvum even during an It is important to that the total loading of oocysts on the population is the oocysts for both exposure models in 1. The models in those oocysts are the Dose–response data are for Dose–response data from human have been for waterborne pathogens including rotavirus and C. parvum of models may be to the data to a dose–response curve for in are the and curves. The in in pathogens act independently or during initiation of infection. The risks by the of curve the low pathogen exposures through drinking water and the of model the The of a of C. parvum by in human A total of were each a of between and The each is in 5. in suggest even under outbreak conditions, drinking water consumers oocysts and of those consumers to oocysts just a the doses by are for the of modelling risks through water. the risk from ingestion of just a C. parvum a of low of dose–response is a dose–response curve to the C. parvum human data to model the of by ingesting a of pathogens is is a for the pathogen and the population and the of the ingested pathogens that to This dose–response relationship that the oocysts act independently during infection and do not act co-operatively, for in the protective Indeed, low doses, a linear relationship between the ingested and of infection. The risks of infection by for low oocyst doses are in 2. The risk of infection with low the risk from oocysts is that from ingestion of just a If pathogens during infection, by in or of the such that oocysts a of infection, low doses, the of infection would be to the model a dose–response relationship is the curve This model is by and and is shown in 5. It is that the and do not over the range of doses in the However, are on to the low doses of oocysts to drinking water consumers are to be The risk of infection by the model from ingestion of a oocyst is that by the model to the ingestion of oocysts a very risk to humans and oocysts a considerable risk of infection ingested in doses (i.e. This the that if oocysts then low doses would to the the risk of infection from a of oocyst is from ingestion of doses of about the risks by the and dose–response are The for a dose–response relationship is the or The of a or is in it suggests risk of infection very low pathogen This not to be the for C. has that models doses of or oocysts do not the data in and to dose–response curves. is evidence for a and a oocyst or other waterborne pathogen not be to infection, with very low in Indeed, that of the from the can a infection and that the of such an the of an a that the for is just oocyst of C. to C. parvum is very infectious to and the ID50 is The ID50 for C. parvum in human with exposure is about oocysts and oocysts to risks to human to The for may in the of the protective In have are known to infection by enteric pathogens such as and It has been for that coli the of in the relationship between and between and for to mice by that the mean with doses the ID50 to for doses the are with the independently such doses the ID50, a the of This of the of a linear dose–response curve in However, the pathogens were by and not by oral in those This is an important in models for drinking water consumers are through oral Indeed, the oral more to infection by and the of a low potential and by the that the of a of to of a of mice is or by a or more the are by that from to mice by from a very of or just a that act independently not However, it be if infection through the oral were a then are to the and that the more with a of infection by is the of the on the ingested Indeed, the were by the oral for a of from to If infecting salmonellas were to co-operate with each other by the or by the then a linear dose–response relationship may not be a of mechanisms for including to the of agents, and In or a of in in the are by of to the of other It be that in an oral of of those salmonellas mechanisms to the the of a infection. would a for microbial infection. with salmonellas in mice that infection through an a in a of the is by a The is it be that act during that to and to the in a and causes If then exposure to low pathogen doses would risks those by the or dose–response and a model would be more are data on C. parvum oocysts act independently or during infection. It is that oocysts act independently and that the dose–response curve is a model for infection on and The used in the of were on the of evidence of infection with C. that endemic Cryptosporidium infection, for through water from through acquired protective on the of in drinking water and the protective is for development and of models for C. parvum in drinking water. In a infection of humans with C. parvum is protective The were of the used in the of 8. with a of oocysts in infection of of the a dose–response curve through curve in suggests that exposure the ID50 from to oocysts and that the risk of infection from a oocyst in is from to with parvum with doses of The ID50 It be that the protective of acquired immunity on infection important Indeed, that infection and were with the the of C. parvum The ID50 for infection oocysts and of infection from ingestion of oocysts in to the dose–response relationship are and waterborne pathogens such as salmonellas and the dose–response curve is more the curve (39). The dose–response curve is on a is for a The dose–response curve that is not a is by a to the the of infection, from ingesting a is and the dose–response the model to the model the data for human rotavirus in ingested doses from to of The of each are in 6. The dose–response curve by is in with the curve with an ID50 of The curve is the linear model (37) and risks for the an exposure of is into a risk of infection of by the model by the model in It is from that are very infectious to humans and the ID50 is just However, it be that the of the rotavirus in the may have been over that in of the of that may have a would on pathogen are an important factor risks of rotavirus infection through water and dose–response curve by to rotavirus data of model with ID50 is shown for Risk assessment models for BSE in drinking water have been developed by the human oral ID50 on the of data from mice in is the of the to the the of it that the oral ID50 for humans is just bovine suggest that is of BSE be to a large in the BSE were to into a drinking water drinking water consumers would be to doses minute subfractions of an ID50, even over the of a human are not have been for a or a by The is on there a for a of prions such that individual infection. The that of into the through a between individual within the of the on the mechanisms of there is evidence for a from data in from show that the to of with the to of the for a of bovine for and for This the as to the is for doses, If the of with then for very doses the may the of a bovine even a such that doses risk if ingested However, it be that is not of a that the mean for mice by with with doses the ID50 to for doses the Indeed, that independently such doses the ID50, a the of that BSE infection is through infectious Evidence for a is to by data show evidence of of BSE in would infection through and dose–response to data for bovine in mice The doses between and for the C. parvum dose–response data the were over the range of doses for data were However, the in for exposures of of an This is the for the of infection is The curve a linear between risk and as if the prions were to act independently In the risk by the curve the doses would be with the prions in a Bovine spongiform encephalopathy dose–response from and dose–response to BSE data in mice BSE data from Dose–response data available for waterborne pathogens in from with or data are available for the more in the population such as the and the is on acquired protective immunity the dose–response curve not just for C. parvum for other waterborne pathogens such as coli O157 and It is for that waterborne pathogens such as and viruses act independently during infection and do not co-operate or However, evidence on there is a for BSE prions is of for the risks of BSE to humans and through of Cryptosporidium oocyst exposures to the and It is important to that the arithmetic mean exposure oocysts is the for both the total oocyst loading on the population as a each is the for both The models in those oocyst doses are to individual consumers within the The of in variation in pathogen exposures on risk prediction is using the dose–response for C. parvum and rotavirus the risks of infection by the C. parvum dose–response for the oocyst exposures in 1. The risk using the dose–response curve is by the variation in exposures is or just the arithmetic mean exposure is used in the both the and of a risk of about is Indeed, the risk is by using the arithmetic mean of oocysts in This is the dose–response relationship is linear low doses It has been shown if oocysts act independently during infection, then the arithmetic mean exposure is sufficient for prediction of the risk of infection In the risk of infection is to the total of oocysts in the drinking water and is not by spatial in the with oocysts in oocysts in the other consumers or consumers each oocyst in the risk by is for the and for the The variation in exposures important for the dose–response Indeed, ignoring the variation in oocyst exposures and just using the distribution of exposures with the dose–response curve the risk of infection by a factor of This is the distribution of pathogen exposures not predict exposures to pathogen doses the ID50 It exposures to low doses, have a risk to the curve the arithmetic mean exposure of oocysts in the dose–response curve a risk with using the exposure model with the dose–response curve The of acquired protective immunity in the Cryptosporidium risk assessment model is important. with acquired protective immunity the model for both the and exposures This is a infection that for with evidence of infection with C. the risks from enteric viruses in drinking water and that the risk of from a exposure to waterborne may be as as in is The risk using a of exposure on the arithmetic mean exposure from the data of The on risk prediction of ignoring the variation in densities in drinking water and just using the arithmetic mean exposure is by the Cryptosporidium oocyst exposures in represent the arithmetic mean exposure of in the dose–response curve in In the variation in exposures in ignoring the variation in exposures through drinking water in the of the risk the risk by a factor of This is to the that ignoring the variation that more consumers are to each in a dose–response curve for the ID50 is just This would represent a highly infectious the variation in exposures in with by using the arithmetic mean exposure of oocysts in 1. by ignoring the variation in the risk assessment would the risk by threefold. The arithmetic mean exposure through the drinking of of water from an by is ID50 the arithmetic mean ingested by each is ID50 over a of This into a risk of using the If the BSE prions were to act co-operatively, to the dose–response then and of the BSE prions in drinking water would drinking water as a of to the the risk of from ingesting an of ID50 in a exposure is This is a factor of that by the linear dose–response The available evidence suggests that there is the potential for considerable variation in pathogen exposures through drinking water. arises from variation in oocyst in the raw and fluctuations in the removal efficiencies of oocysts by drinking water treatment A for oocyst exposures during a waterborne outbreak of cryptosporidiosis that consumers on the are not to oocysts each a doses approach the ID50 for C. parvum in a during non-outbreak conditions, mechanisms may by a of consumers be to doses in a the variation in pathogen densities in drinking water is important for for heterogeneity in pathogen densities on to the pathogen loading in the drinking water and the risk to public health (13). the in variation of pathogen densities in water to the raw water to to the arithmetic mean pathogen to a in the water in the raw This in an of the pathogen removal by treatment and an of the risk (13). the of variation the of the pathogen doses ingested by individual consumers This paper the of modelling the variation in pathogen exposures through drinking water in the of risk prediction for C. parvum and in individual exposure are by the doses of BSE ingested through of and through drinking water from an by The dose–response relationship that the infecting agents or BSE prions) act independently during initiation of infection and that there is In the of the infectious agents such as C. parvum BSE modelling the variation in exposures to individual consumers is not if dose–response relationship is This is from the in show that the risks using the dose–response are the variation in exposures is using the distribution or the arithmetic mean exposure is used Indeed, that the arithmetic mean oocyst exposure is the of exposure for C. However, in the of more highly infectious agents, such as rotavirus or an with ignoring the variation in pathogen exposures and just using the arithmetic mean the risk by a factor of about threefold. The variation in pathogen densities in drinking water is an important for the of to the for an outbreak of cryptosporidiosis of volume the arithmetic mean oocyst and the risk of infection. in more of the volume the arithmetic mean oocyst in the drinking water by a factor of over 10-fold. in the dose–response relationship in microbiological risk assessment are the pathogens act independently during infection and the of acquired protective immunity in the drinking water It is from that acquired protective immunity has a on the risk of Cryptosporidium infection in an outbreak the of the available dose–response data for acquired protective immunity the risk prediction by about on C. parvum oocysts act independently during infection, or co-operate in to be important for the cryptosporidiosis outbreak model Indeed, using the distribution of exposures in the prediction of with the dose–response model is a factor of the prediction of with the dose–response curve However, the more the doses, or oocysts It is low doses that consumers would be to even during outbreak conditions if the oocysts were the model for oocyst exposures in with the dose–response curve the risk of infection by a factor of in the outbreak model The risk from ingestion of doses of oocysts is by the dose–response model by the model models predict exposures to very low oocyst doses that of would be to over-estimate the risk by more if there were a between oocysts in the during initiation of infection. mechanisms can be for However, it is that oocysts act independently during infection. The dose–response model is used in the risk assessment models for C. parvum It is that modelling the variation in pathogen exposures would be of considerable for risk assessment if pathogens were to act during infection. if there were and as by then microbiological risk assessment models need to the variation in exposures to the of consumers ingesting doses those on the risks from exposure to low doses of BSE prions is for the risks of BSE to both and humans through the In the of evidence for a between risk assessment models a linear dose–response The risk for drinking water consumers from an with prediction on the that the is a factor of (i.e. humans are more to BSE between individual humans in of the human to be a factor in the of the the risk assessment developed the data available were for humans with the of the were data available for humans with the of the data that humans of are more to BSE humans with the It that the risk assessment that the and that there protective for This would the risk of BSE to humans through drinking water by a factor of to would be for a on a for BSE prions would have to about the of the providing it have been shown that BSE prions are in drinking water. The of in the risk of BSE through drinking water to the or dose–response curve is used is Indeed, the curve may not be in terms of it is to for the a of ID50 co-operate during initiation of infection. The may through The to be in is is the the the human In the of such risk assessment models for BSE in drinking water on the have the