3.5 Epidemiological information

Epidemiological data can provide valuable insight into the pathogenicity of microorganisms as it applies to the general population. In a sense, outbreaks represent realistic feeding trials with the exposed population often representing a broad segment of society. The doses are essentially real-world levels, and the medium carrying the pathogen represents a range of characteristics (protective, fatty, long residence time, etc.). Ideally, an epidemiological investigation should attempt to collect as much quantitative information as possible in order to lend itself to better characterizing the dose-response relationship for microbial pathogens. In order to refine the dose-response relationship so that it has greater applicability to the general population, various information is required in an epidemiological investigation: the dose, the population exposed, and the number of people exhibiting a response (illness, fever, etc.).

The dose that is suspected to have caused illness in a specific outbreak is often the most difficult measure in an investigation. The lack of dose estimates can be attributed to either the inability to obtain samples of contaminated food or the lack of emphasis being placed on the value of such information. Often, contaminated food is tested and only the presence or absence of the suspected pathogen is reported. This information is often viewed as sufficient to incriminate the food, but it does little to further knowledge of the dose-response relationship.

The attack rate represents the response in a dose-response relationship. In order to estimate the attack rate, an accurate estimate is required of not only the population that was exposed to the contaminated food, but also the number of individuals that became ill. In addition, it is valuable to know the characteristics of the exposed and affected population, in order to account for potential susceptibility issues.

3.5.1 Summary of epidemiological and outbreak information

The following sections present and summarize outbreaks found in the literature that included quantitative information from which the dose and attack rate could be estimated. It is important to note that although these outbreaks include quantitative data, some assumptions had to be made, depending on the nature of the information. In the interest of transparency, the following sections present the information from the original epidemiological reports in as much detail as possible, and, where appropriate, the assumptions that are used are clearly indicated.

In addition, reports that were currently unpublished at the time of drafting this report were received from Japan (Ministry of Health and Welfare, 1999). Although these reports have not been published, and the details of the methods used in the investigations have not been stated (other than through personal communication), they represent a valuable source of information on the real-world dose-response relationship and expand our database of Salmonella pathogenicity considerably. The data in these reports are generated as part of the epidemiological investigations that take place in Japan following an outbreak of foodborne illness. In accordance with a Japanese notification released on March 1997, large-scale cooking facilities that prepare more than 750 meals per day or more than 300 dishes of a single menu at a time are advised to save food for future possible analysis in the event of an outbreak. Thus, 50-g portions of each raw food ingredient and each cooked dish are saved for more than 2 weeks at a temperature below -20ºC. Although this notification is not mandatory, it is also applicable to smaller-scale kitchens with social responsibility, such as those in schools, day care centres and other child-welfare and social-welfare facilities. Some of the local governments in Japan also have local regulations that require food saving, but the duration and the storage temperature requirements vary.

In the evaluation of the outbreak data, whenever sufficient information was available, susceptible and normal populations were separated out of the database to aid in further analysis. Children aged 5 years or younger were considered to be a susceptible population. The criteria or assumptions used to identify potentially susceptible populations are noted in the individual outbreak summaries.

In addition, the uncertainty associated with each of the outbreak parameters are also summarized and defined at the end of each outbreak description. The published reports were used as a basis upon which to derive a reasonable characterization of the uncertainty. However, it should be recognized that since only rarely is sufficient information given upon which to derive a range of uncertainty for the parameters, the uncertainty ranges used are only a crude estimate. In addition, in several reports there is no information whatsoever to use as a basis for uncertainty estimates; in these cases a consistent default assumption was used. To capture the dose uncertainty, 25% over- and under-estimates for the reported concentration and amount consumed were used.

Case Number: 1 - 16

Comments

Composite water samples were collected late in the epidemic (9 days after initial case) and water in the composite samples had been stored for 1 to 4 days at room temperature prior to culturing. Since varying amounts of water, from a few millilitres to as much as 500 ml, were pooled from several sample bottles, it is possible that numbers in some samples were greatly diluted by negative samples. The pooled sample consisted of water from 74 different samples, and only 5 of the 74 samples were actually positive. The concentration of 1000/litre was an order of magnitude estimate following a single isolation made independently from a 1-ml sample (suggesting an order of magnitude of 1000 organisms/litre). The concentration in the water was therefore assumed to range between the two reported concentration estimates (between 50/litre and 500/litre was the range for concentrations), with water consumption of 0.75 litres which results in a dose range of between 37 and 375 cells.

A house-to-house survey was conducted that comprised 8788 people, with 1035 reporting gastroenteric illness. The report also identified attack rates according to age, which was used in the current analysis as an estimate of the potential attack rate for susceptible and normal populations. Children under 5 years (assumed potentially susceptible) were reported to have an 18.9% attack rate, compared with approximately 11% for the rest of the population. The uncertainty in the average attack rate was calculated allowing for 5% under- or over-reporting. Given 1035 people reporting gastroenteric illness, only 983 may have actually been sick, with the other 5% claiming to be sick; alternatively, 1087 people may have actually been sick, with the additional people not reporting sickness. It was assumed that the contamination in the water supply was randomly distributed throughout, such that all 8788 people that reported having drunk water were exposed. It should also be noted that the attack rates listed in this table assume exposure to the pathogen only once during the outbreak.

Outbreak parameter uncertainty

Comments

Samples analysed by CDC, Atlanta, were reported to have an MPN of 1.8 organisms/100 g, while the Food Research Institute in Wisconsin reported an MPN of 0.36 organisms/100 g. According to the restaurant, the serving size was approximately 28 g of cheese per meal. The potentially very low infectious dose for this outbreak was noted by the researchers, and the potential for the occurrence of up to a 99.9% die-off prior to culturing was acknowledged. The concentration in the food at consumption was assumed to range between 108 and 1080 cells per 100 g (99% to 99.9% die-off prior to culture). The dose ingested, based on the nominal amount consumed, was estimated to range between 30 and 300 cells.

The attack rate in this outbreak was reported to range from 28% to 36%. The exposed population was estimated to be 205, consisting of employees in incriminated restaurants, consumers at incriminated restaurants, and employees at restaurants that received contaminated cheese lot shipments and where employee cases existed. The number of positives (57 to 74) was back calculated from the reported attack rate range and the exposed population.

Outbreak parameter uncertainty

Comments

This outbreak involved a susceptible population that consisted of debilitated and aged people, infants and persons with altered gastrointestinal function. Carmine dye capsules are used as a faecal dye marker for such things as the collection of timed stool specimens, gastrointestinal transit time and the demonstration of gastrointestinal fistulas. The number of capsules given to patients ranged from 0.5 to 2; as a result, the dose ingested was assumed to range from 15 000 to 60 000.

There were a total of 21 recognized cases during this outbreak, but 4 were reported to have been infected prior to admission and 5 cases were suspected to have been secondary transmission. Therefore there were 12 confirmed cases directly as a result of carmine dye capsule ingestion. Unfortunately, for attack rate estimation, the total number of exposed individuals was not determined, although the authors of the report note that there were some people who received carmine but were not infected. It was thus inferred that the attack rate was some value less than 100%. As an upper and lower bound, it was assumed that 14 to 20 individuals received dye capsules.

Outbreak parameter uncertainty

Comments

This outbreak occurred in a family consisting of adults and at least two children of grade school age. Smoked ham purchased from a supermarket was taken home and refrigerated for approximately 5 hours. Eight people in the family ate either raw or fried slices of ham, and all 8 experienced acute diarrhoea with gastroenteritis symptoms within 8 to 24 hours. An uneaten portion of ham was obtained and examined in laboratory 2 days after the outbreak occurred. Various bacteria were isolated from the raw ham: total aerobic plate count (268 000 000/g), coliform bacteria (15 000/g), Streptococcus faecalis (31 000 000/g), staphylococci (200 000 000/g) and Salmonella Infantis (23 000/g). Staphylococci were negative for coagulase production and negative for enterotoxin production. Stools from 4 of the 8 persons affected were examined 10 days after the outbreak: mother, father and two grade-school-age sons. S. faecalis was isolated from both parents and one son. S. faecalis var. liquefacies was isolated from another son, and S. Infantis was isolated from both parents but not the sons. The researchers noted that S. Infantis in the ham, stools and the long incubation period implies infection of Salmonella aetiology. However, a mixed infection is a possibility.

The weight of a slice of ham was estimated to range from 50 to 100 g, with 1 to 2 slices consumed. The dose was thus estimated to range from 1 150 000 to 4 600 000 cells of S. Infantis. The exposed and positive populations in this case were quite well established, therefore accounting for uncertainty in these parameters was unnecessary.

Outbreak parameter uncertainty

Comments

This episode involved 14 outbreaks with a total of 3450 people attending various events at which imitation ice cream (chiffonade) was identified as the vehicle of infection. The authors estimated a 52% attack rate based on a survey of persons attending seven of the events. The menus at the various events were relatively extensive, but one of the outbreaks involved a large affair with a limited menu where the authors cite that nearly all those attending had eaten all of the foods offered. Using this outbreak, the attack rate was estimated to be 55% (1400 people attending and 770 people sick).

The chiffonades were stored at -20°C for 1 month before quantitative cultures were done, and the MPN was reported to be 113 or less salmonellae per 75-g serving. The reduction in numbers that could be expected due to freezing was experimentally determined by artificial inoculation of S. Typhimurium into chiffonade and storing of the samples at -20°C. Artificial inoculation experiments indicated that log reductions would have occurred during the storage period, but no more than a 2-log reduction was likely to have occurred during the 1-month storage. As a result, the concentration was estimated to range between 1130 (1 log reduction) and 11 300 (2 log reduction) per serving. In this outbreak, the exposed population was reasonably well established, but the positive population was assumed to have 5% under- and over-reporting (a range of 732 to 809).

Outbreak parameter uncertainty

Comments

The concentration of S. Newport in ground beef was determined from MPN to be between 6 and 23 per 100 g. Accounting for freezing, the authors cite that experimental evidence would indicate a 1- to 2-log reduction due to freezing, which would place the concentration at 60-2300/100 g. However, cooking, even undercooking, is likely to produce a reduction prior to consumption. If the effects of cooking are conservatively assumed to be 1 to 2 log, then the concentration prior to consumption is again estimated to be 6-23/100 g. Assuming consumption of 100 g, the dose that was capable of causing an infection in some people can be estimated to be approximately 6-23 organisms. Unfortunately, this outbreak was geographically widespread and the authors did not report the total number of individuals exposed. The attack rate is therefore undetermined in this outbreak.

Comment

The data in this outbreak is derived from the previous episode described (Case Number 6, reported by Fontaine et al., 1978). However, Fazil (1996) examined the naval outbreak in greater detail through a series of personal communications with the United States Navy, to attempt to determine an attack rate. A total of 21 cases occurred at the naval training centre: 2 were asymptomatic food handlers and 19 were trainees.

The entire complex had a population of 12 483, with the military population listed as 9904 (full time military personnel and trainees). Meals were served at several locations, and included the galley, the staff galley and the exchange cafeteria. The outbreak was reported to have occurred at the "Training Station", which is a separate area within the centre, where training is conducted. There were 7254 recruits who were fed at the galley that serviced the trainees. Therefore, depending on the assumed number of people that ate a contaminated hamburger, an attack rate can be estimated. Assuming 7254 individuals exposed (all present, which is unlikely), the attack rate is estimated to be 0.3% (19/7254). It was assumed that a more likely exposure population was 25% (1813), with an uncertainty range of between 10% (725) and 50% (3627). It was assumed that the positive population was well characterized given the nature of the location. At the naval base the trainees would have had access to convenient medical attention. It should be noted that if there was reporting bias it is more likely to be under-reporting as opposed to over-reporting.

Outbreak parameter uncertainty

Comments

A total of 105 people were reported to have become ill during this episode, which was attributed to ham and pork sandwiches. The sandwiches were suspected to have been contaminated at the restaurant and a refrigerated portion of a pork sandwich from a patient yielded 44×10⁶ S. Newport per gram.

The attack rate that might be inferred from information provided in this report is unknown. The restaurant reported serving approximately 200 people during the period, of whom 105 became ill. However, the information required is an estimate of the number of people that actually ate ham and pork sandwiches and were thus exposed to contaminated food. It can be assumed that not everyone ate the ham and pork sandwiches. If it is assumed that 60% of the people visiting the restaurant ate the contaminated food, then 120 people may have been exposed. At the other extreme, it could also be assumed that only 105 people were actually exposed and 105 became sick, the attack rate then being 100%.

Comments

This outbreak involved a potentially susceptible population and involved chocolate balls. The median age of the cases in this outbreak was 3 years. The attack rate cannot be determined in this case because no information was provided in the report, and the geographically widespread nature of the outbreak makes inferences difficult. The outbreak occurred simultaneously in the United States of America and in Canada. The description of the Canadian portion of the outbreak is described in the next section (Case Number 10).

The New Jersey health department reported a mean concentration of 2.5 salmonellae per gram of chocolate from samples obtained from homes where cases occurred. A bag of the chocolate was reported to be 1 lb or approximately 450 g; therefore the maximum dose causing infection in some people was estimated to be no more than approximately 1000 cells (2.5/g × 450 g). Alternatively, the dose could be as low as 100 cells if only 40 g was consumed (2.5/g × 40 g).

Comments

This outbreak again involved a potentially susceptible population, as 46% of the cases were children aged 1 to 4 years old. There were a total of 95 reported cases. The outbreak was attributed to chocolate balls. Each ball was reported to weigh approximately 10 g, with a bag of chocolate balls containing approximately 50 balls. The contamination of the chocolate balls was estimated to be 2 to 9 salmonellae per chocolate ball. This outbreak was the Canadian part of the outbreak that also occurred simultaneously in the United States of America, and described previously (Case Number 9; Craven et al., 1975).

The dose causing illness in some of the exposed population was estimated by the authors based on the consumption of a bag of chocolate. This estimate, which might be high in view of the assumed consumption of 50 chocolate balls, would place the dose at approximately 100 to 450 cells. Depending on the assumption of the amount of chocolate that was consumed, the dose causing illness could be as low as 2 cells if only 1 ball was consumed at the lowest concentration. However, it is difficult to determine from the information given exactly how much chocolate sick individuals consumed, and what the concentration was in the chocolate that was consumed. The overall attack rate for this outbreak is also difficult to estimate, as for the previous report (Case Number 9; Craven et al., 1975), due to the geographically widespread nature of the outbreak.

Comment

In this outbreak, a total of 56 persons who ate at a Washington D.C. hotel had onset of diarrhoea. The Washington D.C. public health department conducted an investigation into the outbreak and identified hollandaise sauce as the likely vehicle. According to the USDA (USDA-FSIS, 1998), only 39 persons ate the hollandaise sauce, and all 39 became ill, which would imply a 100% attack rate. The attack rate in this case was assumed to be 100%, with a good characterization of the exposed and positive populations.

The actual concentration of S. Enteritidis causing illness in this outbreak was not reported in the publication describing the outbreak (Levy et al., 1996), but the USDA-FSIS (1998) reported the results of some testing. This informal quantitation, which was not performed to extinction, tested a sample of sauce recovered from a patron who had taken it home in a "doggy bag" and refrigerated it for 72 hours. The concentration in this sample was reported to be 10³ per gram. It was assumed that 2 tablespoons (approximately 10 g) were consumed by the patrons of the restaurant, placing the dose at approximately 10⁴ (USDA-FSIS, 1998). To allow for the uncertainty associated with the concentration estimates and the potential underestimate, an additional 1 log was allowed for in the concentration range.

Outbreak parameter uncertainty

Comment

This was an interstate outbreak attributed to ice cream. Hennessy et al. (1996) provide details on the epidemiological characteristics of the outbreak and the concentration of S. Enteritidis found in samples of ice cream, using traditional MPN techniques. The effect of frozen storage was also experimentally investigated. The authors found no evidence of a decrease in numbers during storage at -20°C for 16 weeks, unlike the work of Armstrong et al. (1970), described previously (Case Number 5). A re-analysis of the quantitative MPN results was performed at a later date using alternative statistical tools to better estimate the concentration in the ice cream (Vought and Tatini, 1998). The expected concentration was reported as 0.152 MPN/g at the lower range, and 0.894 MPN/g at the upper range. In addition, a small group of people that were investigated in more detail were reported to have consumed from 65 to 260 g. The uncertainty in the dose was therefore assumed to range from 10 cells to 235 cells.

The outbreak was reported to have affected a large number of people, and from the report by Hennessy et al. (1996) there are details on a smaller cross-section of the group, with whom interviews were conducted. A total of 541 people were interviewed that had purchased the incriminated ice cream, of which 452 were reported to have consumed the product. To allow for some uncertainty in the exposed population, it was assumed that this could be 10% less than the number that reported eating the ice cream. A total of 30 individuals became ill in the population in the cross-section study. The numbers of positives were assumed to have 5% under- and over-reporting.

Outbreak parameter uncertainty

Comments

This outbreak involved a family and one neighbour, and was attributed to home-made ice cream. The ages (years) of the exposed population were: father, 35; mother, 30; sons, 13, 9 and 8; daughters, 6 and 2; and a male neighbour, 22. Ice cream was obtained from the freezer at the farm and found to have 10⁶ salmonellae/ml. One of the sons, aged 13 years, who ate the most ice cream (1000 ml) died from his illness. Various amounts of ice cream, ranging from 100 ml to 1000 ml, were reported to have been consumed by the family members. Since the actual sample of ice cream was obtained from the freezer only a few days after the event, the concentration reported was assumed to reflect that at the time of consumption. The uncertainty in the dose was modelled using the given concentration, and accounting for the different amounts consumed. In the current analysis, the child of 2 years of age was assumed to be potentially more susceptible, while the other individuals were assumed to represent a normal population. In this particular case, there was no uncertainty in the exposed and positive populations.

Outbreak parameter uncertainty

Comments

This outbreak involved more than 1500 people, with cheddar cheese implicated as the vehicle of infection. Cheese samples were obtained from the production plant, as well as from homes of some of the individuals that were ill. The level of contamination in the cheese from the plant was found to be between 0.36 and 9.3 salmonellae per 100 g (D’Aoust, Warburton & Sewell, 1985), while the level of contamination in cheese from individual homes was found to be between 1.5 and 9.1 salmonellae per 100 g (D’Aoust, 1985). The average concentration from cheese plant samples was estimated to be 3.5/100 g while those from homes was estimated to be 4.2/100 g. The authors noted that the number of salmonellae probably did not change substantially during storage, and the levels estimated reflect the levels at the time of consumption. It was estimated that approximately 100 g of cheese was consumed, based on the level of consumption reported for six individuals, ranging from 20 g to 170 g.

The attack rate in this case is again difficult to estimate due to a lack of information on the exposed population and the inability to make reasonable assumptions given the information and the widespread distribution of the outbreak.

Comments

This case involved a susceptible individual (1-year-old child) who developed diarrhoea when treated with pancreatic extract (pancreatin is an extract from the pancreas of mammals, and used to assist in the digestion of food) that was contaminated with S. Schwarzengrund. The pancreatic extract was found to contain 1000 salmonellae per gram, and the child became ill following ingestion of 200 mg. It should be noted that this case involves only one individual and the 100% attack rate quoted for this dose could skew the true attack rate, which could be less for a group of individuals receiving this dose. For example, it could be possible that this one individual might be the only one that got sick if 20 similarly susceptible individuals were given the same dose. In that hypothetical situation, the attack rate would be estimated to be only 5%.

Comments

This case involved a single susceptible individual (9-month-old child with cystic fibrosis), who was fed pancreatin contaminated with S. Schwarzengrund. The pancreatin was found to be contaminated at a level of 8 salmonellae per gram. The child was given approximately 700 mg with each 6-hourly feed for the first 10 days, increasing to approximately 1.4 g in the 36 hours before the onset of symptoms. The authors note that the child had therefore ingested less than 22 organisms per day initially and less than 44 organisms per day in the last 36 hours. If the dose is not cumulative over 24 hours, then the infective dose would be approximately 44 organisms (24 hours, fed every 6 hours, which translates to 4 feedings; each feeding is 1.4 g, which translates to 5.6 g. 5.6 g × 8/g = approximately 44 cells). The points raised about one individual exposed and the attack rate estimates in the previous case (Number 15; George, 1976) also apply in this case.