A quantitative risk assessment on Cryptosporidium in food and water

Raw water contamination

Some data can be found in literature. However, differences occur in detection method, amount of water analysed and source of water analysed. Based on Dutch data collected from water coming from a river to the `Biesbosch' storage reservoirs, the contamination level of water with oocysts is assessed. From 39 samples of 200 litre, 35 were positive with numbers ranging from 1-76 per 200 litre (mostly 10-15 per 200 litre). This data is simplified to 0.005 (best-case), 0.1 (mean) and 1 (worst-case) per litre to be used for the Dutch situation. In other studies averages of 0.1 - 1 oocysts per litre surface water are reported for the USA, UK and Germany (RIWA 2001). Incidentally high numbers in water are reported (around 100 oocysts per litre). However, this also includes water which is suspected to be contaminated with animal faeces and irrigation water of poor quality (Thurston-Enriquez, 2002). An initial contamination of 0.1 and 1 oocyst respectively per litre water are used for single point relatively worst-case estimates.

LeChevallier reported recoveries of Cryptosporidium oocysts detected in river water (spiked samples). From these results a recovery of 10-30% could be observed (LeChevallier et al., 1991; Teunis et al., 1999). This is also taken into account in the risk assessment model.

Cryptosporidium is not able to multiply outside a host. Although, the protozoan is able to survive for a long period in the environment, the viability and infectivity may decrease in time. Research has been described in which the morphological characteristics are investigated and taken as a measure for viability. It was concluded that 30-50% of detected oocysts were regarded as `viable'. In the model a minimum of 10%, a mode of 30% and a maximum of 100% viability is used. The decrease in viability might also be important further in the water chain. However, no data are available. For now, this factor is only taken into account at this risk factor.

Reduction in viable numbers during storage prior to purification

From research of water from the `Biesbosch' storage reservoirs it was concluded that the numbers of viable oocysts decreased during storage. Since the mean storage time is 5 months this is a factor of importance. After storage the numbers were low (mostly 0-3 oocysts in 1000 litre). Therefore, a reduction of 1-2 log can be expected. For the worst-case scenario no reduction is assumed.

Reduction during physical purification (coagulation, flotation, filtration)

In the Netherlands most water is treated physically: coagulation, flotation, filtration. This will result in a removal of Cryptosporidium oocysts. Because the number of oocysts are expected to be very low, it is not effective to analyse the treated water. However, in literature a relation between the removal of spores of sulphite reducing Clostridium (SSRC) and the removal of Cryptosporidium oocysts is described. For SSRC in most cases a reduction of more then 2-3 log was observed. However, in some cases the removal was only 1 log. These data are used for Cryptosporidium oocysts: minimum 1 log, mode 3 log and maximum 4 log inactivation.

In Ireland, the larger purification plants use coagulation, rapid gravity filtration or slow sand filtration. Smaller plants sometimes do not use a physical purification step.

It is known that purification systems can fail temporary. This may have been the most important risk factor during several outbreaks, a.o. Scotland 2002 outbreak. The impact of this situation is assessed for the worst-case scenarios.

Reduction during chemical purification

In the Netherlands no chlorine is added to drinking water. In Ireland all purification plants use chlorination. However, this will not reduce Cryptosporidium. In some Dutch plants the water is treated with ozone or UV-light. In literature a reduction of 0-2 log is described (mode 1 log).

Amount of consumption

Only water consumed without boiling is taken into account, since boiling (even a treatment with the intensity equal to pasteurisation for 15 seconds at 72°C) will eliminate Cryptosporidium oocysts (Harp et al., 1996).

A rough estimation of the daily amount of consumed tap water is made. It was assumed that the consumption is between 100 and 500 ml a day per person (mode 150 ml), based on research by Teunis et al. (1997). Gale (2001) reports a consumption amount of 47 - 322 ml a day per person in the UK (accounts for 95% of the consumers). For the worst-case scenarios a daily consumption of 250 ml is assumed. The currently used 2 litre daily in microbiological risk assessment for drinking water in the USA is an unrealistic over-estimate (Gale, 2001).

Risk characterisation for Cryptosporidium in water

In Table 1 and 2 the results of the quantitative risk assessment are presented.

¹95% resp. 99.5% of the outcomes have a value below this value, should be interpreted as worst-case

The worst-case scenario with a failed purification or chlorination only results in a daily risk of ~ 1:200-2000. In some outbreaks there have been hundreds of infected people, so this confirms that this worst-case scenario can incidentally occur.

The Dutch Ministry of Housing, Spatial planning and the Environment has made a policy position stating that a risk of infection of a maximum of 1 per 10,000 persons a year due to consumption of tap water is acceptable. The same level of risk was described by the US Environmental Protection Agency as a negligible risk for infection through drinking water (RIWA, 1991).

Looking at the different scenario's the 95% confidence point of the risk estimate will meet this risk level. The probability of reaching this risk level will be 3 % based on this rough model (calculation from the probability model). Also, when the reduction due to storage of water and purification is around 4-5 log units, a high contamination level is expected to result in an acceptable risk.

In the UK, in the Water supply Regulations 1999, a criterion of less then 1 per 10 litre drinking water is stated (Campden, 2000). In practice 500 litre sample is filtered with a criterion of a maximum of 50 oocysts. Predictions of the risk of illness using this contamination level, would result in a yearly risk of a few per 10,000 to a few per million depending on the purification (3 to 5 log). However, as it has been noted by the UK Group of Experts, there is no increase in cases of Cryptosporidiosis since the implementation of the regulation and monitoring.

The risk of Cryptosporidium by drinking tap water seems to be low for average conditions. The most important risk factors that can be controlled are `storage of water before purification' and `physical water treatment'. It can also be expected that the viability and infectivity will be reduced during storage and treatments. However, no data is available about those risk factors in the water chain. Off course the initial contamination is the most relevant factor, since when no Cryptosporidium oocysts are present, there is no risk. This factor is however hard to control. Potential points of control could be to minimise the draining of highly contaminated water near water collection sources and a stricter treatment of water known to be highly contaminated. In the current analysis method no account is made for viability. This means that non-viable oocysts are also counted. In the light of risk assessment it is recommended to assess the viability as well as the number of oocysts during testing of water.

Exposure assessment of Cryptosporidium in raw fruits and vegetables

The food chain is divided into five parts: cultivation, harvesting, transport and storage, industrial processing and food preparation by the consumer. The risk factors in every part are discussed below.

Risk characterisation for Cryptosporidium in raw fruits and vegetables

Very little data was found in the literature regarding the risk factors, Therefore, only a rough exposure assessment was possible.

If sludge contains 10³ oocysts per kg and 10 g is adhering to 1 kg crop, the contamination will be 10 oocysts per kg crop.

If manure or slurry contains 10⁴ oocysts per kg and 10 g is adhering to 1 kg crop, the contamination will be 100 oocysts per kg crop.

If sewage water or other dirty water contains 10 oocysts per litre and 10 ml is adhering to 1 kg crop, the contamination will be 0.1 oocysts per kg crop.

The cumulative contamination of fruits and vegetables is dependent on the relative frequency of contamination of the water, slurry, sludge, manure, etc.

For a better underpinned risk assessment it would be recommended to obtain more data about the specific contamination sources. In addition, information about the transfer of these sources to the fruits and vegetables should be obtained using surrogate microorganisms. At a later stage, monitoring data of the fruits and vegetables itself can be used for verification and/or optimisation of the risk assessment.

It seems that faecal contamination of fruits and vegetables is the most important risk factor, since faecal material can be highly contaminated with Cryptosporidium oocysts. It is unknown how much faecal material will adhere to the product. Because of the potentially high contamination level of manure with Cryptosporidium oocysts, it is recommended that the treatment of manure with respect to the inactivation of oocysts be validated.

The structure of the fruits and vegetables is also expected to be an important risk factor. It is unknown how much of the contamination of the raw crops will be removed by normal preparation and washing of the fruits and vegetables. It is expected that the probability of contamination during industrial processing and at the stage of the consumer is negligible compared to the possibility of contamination during primary production.

The washing process might result in an inactivation due to removal of debris and dirt. Storage of fruits and vegetables might also result in a reduction in viability of Cryptosporidium oocysts. It is recommended that more quantitative information be gathered on the effect of storage conditions (time, temperature, food matrix) in relation to survival of Cryptosporidium oocysts.

Exposure assessment of Cryptosporidium in meat products

Firstly, the risk determining factors in the meat chain are identified. The risk factors for the slaughterhouse result in the occurrence of Cryptosporidium parvum in raw meat (trimmings). Together with the risk factors in the production process this will result in the occurrence in meat products bought by the consumer. The risk factors at the consumer level finally determine the intake (exposure) of Cryptosporidium parvum through the consumption of meat products.

Risk characterisation for Cryptosporidium in meat products

From some surveillance data, it can be concluded that the prevalence of Cryptosporidium oocysts in young livestock is higher than in older livestock. The prevalence of Cryptosporidium is likely to be higher in calves. McEvoy et al (2003) (paper 3 this proccedings) tested faces of cattle post slaughter and Cryptosporidium spp. were isolated from 20 /288 (6.9%) faecal samples at a level of 50 - 37,500 g^-1. No Cryptosporidium was found on carcass meat (n=288)

Calf meat is in general always cooked before consumption, which will inactivate Cryptosporidium. This also accounts for poultry and game meat. However, if the heating process is done by the consumer, there is still a probability of cross-contamination from the raw meat to the kitchen (chopping board, dresser, etc.). The risk of cooked meat products is considered to be negligible. Some pig and beef products are consumed raw. These products are often fermented, salted, dried or smoked. The effect of these treatments on the viability of Cryptosporidium parvum oocysts is not known, but is considered to be important in order to classify the risk of these products in relation to Cryptosporidium parvum.

General conclusions

A lot of assumptions have been made in the risk assessment for water. Although from an epidemiological perspective, water is highlighted as the most important route of exposure to Cryptosporidium, there is relatively little quantitative data availble in the literature. Therefore, the results of the risk assessment should not be interpreted quantitatively but qualitatively. Calculating different scenario's are used to assess the impact of certain risk factors. In general, the most important advantage of risk assessment is to identify and prioritise measures for improvement. In most countries there is no routine testing for Cryptosporidium and if there is, no account is made for viability of the observed oocysts. On the other hand it may be adequate to determine the performance of different purification systems. A reduction of 3-4 log may be sufficient for an acceptable risk.

From the limited data on Cryptosporidium in fruits and vegetables it can be concluded that Cryptosporidium oocysts are present in a small percentage of products and at low numbers per gram product. In general, the same risk factors and therefore potentially the same control measures apply for Cryptosporidium as for Salmonella and E. coli O157:H7. A reduction in viability during storage may be worth studying in more depth.

When products are randomly sampled the probability of getting a positive sample is expected to be low. This should be realised in the context of using the data for risk assessment. Supposing 10% of the crops had a contamination level of 100 oocysts per kg. When a 5 gram sample is investigated the probability of finding a positive sample is ~4% (calculated with a Poisson distribution). When analysing a sample of 20 gram this probability is ~9%. This example shows that a relatively high contamination level is hard to observe when performing surveillance research. This is also expected to be the case for sampling of (beef) carcasses.

All data gathered from relatively worst-case points in the food chain might give more positive results. These points are important to consider or to evaluate (new) control measures. However, the data will overestimate the risk when these data will be used for risk assessments. This should be taken into account, especially when this data is communicated to risk managers.