10 10 10 mean dose

Figure 4.2. Fits of different dose-response models to nontyphoid Salmonella.

This problem (of differences between low-dose extrapolated risks) has arisen in the context of chemical risk assessment (for example, see Brown and Koziol, 1983). The use of a biologically plausible dose-response model may add reassurance that the extrapolation is reasonable. In the case of microbial risk assessment, low-dose extrapolation can be supported by validation against attack rates noted during outbreaks (this, in fact, is an avenue that is not realistically available in the case of chemical risk assessment). Hence, the validation of the estimated dose-response relationship forms an important step in confirming the adequacy of the chosen model.


Validating Models

The task of validating a dose-response model involves obtaining information on actual human exposure during an outbreak (e.g., average number of organisms ingested) and information on the attack rate. The exposure information is then used to compute an expected attack rate based on the dose-response curve (computed from feeding studies), and the coherence with the measured attack rate is examined.

For example, the best-fit beta Poisson dose-response parameters for non-typhoid Salmonella are a = 0.3126, N50 = 2.36-104 (Fazil, 1996). In 1975, there was an interstate outbreak of human salmonellosis that was attributed to the ingestion of raw or undercooked hamburger. The outbreak occurred in Colorado, Maryland, and Florida (Fontaine et al., 1978). The outbreak in Florida occurred at the U.S. Naval Air Training Station in Orlando; as a result, it is of particular interest because of the presence of a "captive" audience.

For this portion of the outbreak, there were 21 reported cases due to S. newport between September 24, 1975 and October 11, 1975. Two of the cases were asymptomatic food handlers. Of the remaining 19 cases, 13 occurred over a four-day span from September 24 through September 28 (Fontaine et al., 1978). By personal communication with base personnel, it was ascertained that the potential exposed population consisted of 7,254 recruits who were fed at the galley.

On the basis of the attack information, the total attack rate was 0.00289 (= 21/7254). Assuming that the exposure occurred over a four-day period, the daily risk is computed (from Eq. 4.2) as 7.2 • 10~4.

The analysis of the contaminated hamburger detected an MPN of 6-23 organisms per 100 g (Fontaine et al., 1978). The probable inoculum size according to Fontaine, et al. (1978), taking into consideration a 1- to 2-log reduction after freezing, would still place the infecting concentration between 60 and 2,300 organisms per 100 g. Cooking, even undercooking, would further reduce the number of organisms. Salmonella newport has a decimal reduction time at 140 F of approximately 1.5 min (Mitscherlich and Marth, 1984). If we assume the meat was undercooked as was described by Fontaine et al. (1978), this would still result in a 1- to 2-log reduction in the number of organisms. The probable inoculum size after cooking would thus be approximately 6-23 organisms per 100 g.

To complete the comparison it is necessary to determine the concentration of the organisms consumed in the hamburger. In 1975, the average hamburger consumption was 30.5 lb per year (American Meat Institute, 1994). The daily consumption can thus be estimated as 37.85 g. Therefore, the daily estimated ingestion of Salmonella during the outbreak is estimated as 2.3-8.7 organisms.

Note that this is almost four orders of magnitude below the lowest administered dose in the human feeding trials.

Using this estimate for dose, in conjunction with the best-fit parameters for the dose-response relationship, the expected daily risk is computed to be 2.5 ■ 10~4. This is about 1/3 of the observed attack rate. Given the uncertainties in the epidemiological measurement (underreporting of cases, duration of exposure) and exposure assessment (measurement of Salmonella and estimation of consumption and losses from freezing and cooking), the expected attack rate and the observed attack rate are in concordance.

Available Dose-Response Parameters

To date, numerous dose-response parameters have been estimated for bacteria, viruses, and protozoa transmitted by the fecal-oral route. A number of these have been summarized by Haas, Rose et al. (1999).


The field of microbial dose-response modeling remains an active and fertile one for future work. There are a number of areas in which progress is being and will be made.

It is likely that pathogens will emerge for which human dose-response information is unavailable and may not become available. In these cases there may be a necessity to rely upon animal models. In the case of E. c.oli 0157:H7 (Haas et al., 2000) and Listeria monocytogenes (Haas, 1999), the use of animal data to make inferences with respect to human potency appears realistic. Further studies are needed with other organisms to gain experience with trans-species extrapolation for microbial risk assessment.

As noted above, the assumptions made for treating multiple exposures assume independent behavior. This must be critically examined, probably by using animal models for particular agents. Animal model studies for the assessment of changes in infectivity with host status (immune competency, nutrition, and age) are also needed.

0 0

Post a comment