The Changing Epidemiology of Salmonella: Trends in Serotypes Isolated from Humans in the United States, 1987–1997
Sonja J. Olsen1,2,
Richard Bishop3,
Frances W. Brenner1,
Thierry H. Roels1,
Nancy Bean3,
Robert V. Tauxe1 and
Laurence Slutsker1
+ Author Affiliations
1Foodborne and Diarrheal Diseases Branch, Division of Bacterial and Mycotic Diseases, National Center for Infectious Diseases,
2Epidemic Intelligence Service, Division of Applied Public Health Training, Epidemiology Program Office, and
3Biostatistics and Information Management Branch, Division of Bacterial and Mycotic Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia
Reprints or correspondence: Dr. Sonja J. Olsen, Centers for Disease Control and Prevention, Foodborne and Diarrheal Diseases Branch, 1600 Clifton Rd., MS A-38, Atlanta, GA 30333 (
[email protected])
Next Section
Abstract
Salmonellosis is a major cause of illness in the United States. To highlight recent trends, data for 1987–1997 from the National Salmonella Surveillance System were analyzed. A total of 441,863 Salmonella isolates were reported, with the highest age-specific rate among infants (159/100,000 infants at 2 months). Annual isolation rates decreased from 19 to 13/100,000 persons; however, trends varied by serotype. The isolation rate of Salmonella serotype Enteritidis increased until 1996, whereas declines were noted in Salmonella serotypes Hadar and Heidelberg.
Overall, serotypes that increased in frequency were significantly more likely than those that decreased to be associated with reptiles (P=.008). Salmonella infections continue to be an important cause of illness, especially among infants. Recent declines in food-associated serotypes may reflect changes in the meat, poultry, and egg industries that preceded or anticipated the 1996 implementation of pathogen-reduction programs. Additional educational efforts are needed to control the emergence of reptile-associated salmonellosis
Salmonellosis is an important cause of diarrheal illness in humans, causing ∼1.4 million illnesses and 600 deaths annually in the United States [1]. Much of what is known about the epidemiology of salmonellosis comes from outbreak investigations. These investigations have determined that most human infections result from the ingestion of foods of animal origin that are contaminated with Salmonella species [2–5]. Other vehicles, including nonanimal foods, such as fresh fruits and vegetables [6], water [7], reptiles [8–13], and direct person-to-person transmission [14–16], also have been implicated. However, most Salmonella infections do not occur in recognized outbreaks, but rather as sporadic infections [17]
There are 2449 known serotypes of Salmonella [18]. Serotyping is a useful classification scheme that allows for trends in Salmonella surveillance data to be followed over time. For example, from the late 1800s to the mid-1900s, Salmonella serotype Typhi was the leading cause of Salmonella infection in the United States [19]. Improvements in sanitation nearly eliminated Salmonella Typhi infections, and in the 1950s, nontyphoidal Salmonella infections began to increase in importance. More recently, specific serotypes have been linked with certain foods or exposures. For example, outbreaks of Salmonella serotype Enteritidis have been repeatedly associated with raw or undercooked eggs [2, 3, 20–22], and Salmonella serotype Marina infection has been associated with exposure to reptiles [12, 23]
Surveillance for Salmonella species in the United States began in 1962 and is conducted jointly by the Council of State and Territorial Epidemiologists, the Association of Public Health Laboratories, and the Centers for Disease Control and Prevention (CDC). The objectives of the surveillance system are fourfold: to define endemic patterns of salmonellosis, to identify trends in disease transmission, to detect outbreaks, and to moni-tor control efforts. These surveillance data are periodically summarized and published; the last report reviewed data from 1984 through 1986 [24]. This report reviews the trends in Salmonella infections in the United States from 1987 through 1997
Previous SectionNext Section
Materials and Methods
Salmonella surveillance system.Salmonella surveillance is a passive, laboratory-based surveillance system conducted in all 50 states and the District of Columbia. Clinical laboratories are requested (and, in some states, required) to forward clinical isolates of Salmonella species to their state public health department laboratory for serotyping. The state health department completes a report that contains information on patient sex, age, race, and county and state of residence, as well as the specimen source, serotype, and date of collection. Because additional clinical or epidemiologic information is not included in the report, we do not know the proportion of isolates that are cultured from persons without gastrointestinal symptoms; however, we suspect that it is low, because only ∼10% of specimens are from sources other than stool samples. Reports are then sent to the CDC. Since 1994, reports from all states have been sent electronically through the Public Health Information System. Summaries of the reported isolates are tabulated annually (available at
http://www.cdc.gov/ncidod/dbmd/phlisdata/salmonella.htm)
Exact duplicate entries are discarded on a weekly basis. When the data are closed out for a given calendar year, all entries for a given name within a state are compared (except for Oklahoma, which provides initials only). Duplicate entries for a name that also match on age, state, and specimen date within the same or 2 consecutive months are discarded. Entries that match on these criteria but differ in the specimen source are not discarded; however, because most specimens were obtained from stool samples, this number should be small. For Salmonella Typhi, only the first occurrence in a year for a person is kept
A reptile-associated serotype was defined as a Salmonella serotype in the non-human Salmonella database (National Veterinary Services Laboratory, Ames, IA) that was isolated from reptiles in ⩾50% of the isolates tested. The nonhuman Salmonella database was begun in 1981 and has >20,000 isolates. In addition, all Salmonella subspecies II, III, IV, V, or VI were considered to be reptile associated, because reptiles have historically been the source in >50% of these isolates reported to the CDC [25]
SerotypingIsolates are serotyped according to the Kauffman and White scheme, using somatic (O) and flagellar (H) antigens [18]. Most clinical laboratories use only the O antigen to serogroup salmonellae. Complete serotyping is available through state public health laboratories. The National Salmonella Reference Laboratory is located at CDC
Statistical analysisWe calculated annual isolation rates, using US Census Bureau data and intercensal population estimates. To calculate age-specific isolation rates by month of age for infants <1 year old, we used a denominator of one-twelfth the total population of infants. To calculate isolation rates by geographic region, we grouped states into the 9 divisions established by the US Census Bureau. To examine annual trends among serotypes with ⩾100 reports per year, we performed least squares regression and calculated a 95% confidence interval around the slope, using SAS software (version 6.12; SAS Institute). To compare the proportion of reptile-associated serotypes among those serotypes that were significantly increasing or decreasing in frequency over time, we calculated 2-tailed Fisher’s exact P values, using Epi Info (version 6; CDC)
Previous SectionNext Section
Results
From 1987 to 1997, 441,863 Salmonella isolates were reported to CDC; 92% had a known serotype (table 1). The number of isolates with an unknown serotype increased from 3.3% in 1987 to 7.3% in 1991 and then decreased to a low of 1.1% in 1997. The annual isolation rate was highest in 1987 (19.1/100,000 persons) and decreased to a low of 12.9/100,000 persons in 1997; this decrease occurred in all age groups. During the entire 11-year surveillance period, the top 5 reported Salmonella serotypes (Salmonella Typhimurium, Salmonella Enteritidis, Salmonella Heidelberg, Salmonella Newport, and Salmonella Hadar) accounted for 66% of all isolates with a known serotype
During the study period, Salmonella Typhimurium was the most frequently reported Salmonella serotype, accounting for 24% (100,713) of all isolates with a known serotype. The annual incidence of Salmonella Typhimurium remained fairly stable over the study period, as did the proportion of isolates that were Salmonella Typhimurium
The next most frequently reported serotype was Salmonella Enteritidis, accounting for 22% (91,280) of all isolates with a known serotype. From 1994 to 1996, Salmonella Enteritidis was the most frequently reported serotype in the United States. The annual incidence rate rose from 2.9/100,000 persons in 1987 to 3.9/100,000 persons in 1995. In addition, from 1987 to 1997, the proportion of all Salmonella isolates with a known serotype that were Salmonella Enteritidis increased from 16% in 1987 to 27% in 1994 and remained >20% through 1997. Data from 1996 and 1997 suggest that Salmonella Enteritidis infections are decreasing. Between 1996 and 1997, the annual isolation rates decreased from 3.6 to 3.0/100,000 persons
Salmonella Heidelberg was the third most common serotype during the 11-year surveillance period; this serotype&ranked third each year, except in 1995, when it&ranked fourth. The total number of reported Salmonella Heidelberg isolates, 35,840, accounted for 9% of all isolates with a known serotype. The proportion of isolates with a known serotype that were Salmonella Heidelberg decreased steadily over the study period. In 1987, this proportion was 13%, and by 1997 it had decreased to 6%
Age and sexInformation on patient age was available for 83% of reported isolates. The highest incidence rate was for infants, at 121.7/100,000 persons per year; this rate peaked at 159.5/100,000 persons in the second month of life (figure 1). Rates decreased abruptly after infancy, remained relatively constant through the adult years, and increased slightly among persons ⩾70 years old. Boys <15 years old had a slightly higher age-specific isolation rate than did girls (data not shown). In contrast, the isolation rate for women (10.2/100,000 persons per year) exceeded that for men (8.8/100,000 persons per year) between the ages of 20 and 74 years. The age- and sex-specific isolation rates for Salmonella Typhimurium and Salmonella Heidelberg were similar to that for all serotypes combined, whereas the rate of Salmonella Enteritidis infection was higher among adults than were those of other serotypes
Figure 1
View larger version:
In this page
In a new window
Download as PowerPoint Slide
Figure 1
Isolation rates by age and sex of patient and year for all Salmonella and Salmonella serogroups Typhimurium, Enteritidis, and Heidelberg in the United States, 1987–1997. Inset is isolation rate for all Salmonella species among infants, by month of age and sex
For isolates from patients with known age, 15% were from infants, 33% were from children 1–19 years old, 34% were from persons 20–49 years old, 15% were from persons 50–79 years old, and 3% were from persons ⩾80 years old (table 2). Persons with Salmonella Typhimurium infection were most likely to be 1–19 years old. In contrast, persons with Salmonella Enteritidis or Salmonella Typhi infection were more likely to be 20–49 years old
Figure 2
View larger version:
In this page
In a new window
Download as PowerPoint Slide
Figure 2
Percentage of reported Salmonella isolates by month of specimen collection and selected serotype in the United States, 1987–1997penta center-figure-legend>
Figure 3
View larger version:
In this page
In a new window
Download as PowerPoint Slide
Figure 3
Isolation rate of all Salmonella species (A) and of Salmonella serotypes Typhimurium (B) and Enteritidis (C) by geographic region in the United States, 1987–1997
Figure 4
View larger version:
In this page
In a new window
Download as PowerPoint Slide
Figure 4
Number of human isolates by year among selected animal-associated serotypes of Salmonella in the United States, 1987–1997. A, Salmonella Marina (reptile associated). B, Salmonella Kentucky and Salmonella Heidelberg (broiler chicken associated). C, Salmonella Hadar and Salmonella Heidelberg (turkey associated). D, Salmonella Derby and Salmonella Typhimurium (swine associated)
Table 1
View larger version:
In this page
In a new window
Download as PowerPoint Slide
Table 1
Annual number of reported Salmonella isolates from humans and their&ranking: 15 most frequent serotypes, United States, 1987–1997
Table 2
View larger version:
In this page
In a new window
Download as PowerPoint Slide
Table 2
Distribution, by age group and serotype, of the 15 most frequently reported serotypes of Salmonella isolates from humans, United States, 1987–1997
SeasonalityOverall, the largest percentages of isolates were reported in August (12%) and September (12%), and the smallest percentage was reported in February (5.5%; figure 2). Although this general pattern was consistent for most serotypes of Salmonella some serotypes demonstrated slight variations (e.g., Salmonella Hadar peaked earlier). Salmonella Newport exhibited the greatest seasonal variation. The trend toward more isolations in late summer did not vary by age group (data not shown)
Site of isolationA known clinical source was reported for 384,266 isolates (87%; table 3). Of these, 89% were isolated from stool samples. Among the top 15 serotypes, except for Salmonella Typhi, the most frequent site of isolation was stool samples, ranging from a high of 94% for serotypes Salmonella Newport, Salmonella Braenderup, and Salmonella Muenchen to a low of 83% for Salmonella Oranienburg. Of all isolates, 22,979 (6.0%) were from blood, 10,750 (2.8%) were from urine, and 258 (<1%) were from cerebrospinal fluid. Of those serotypes with >100 isolates reported from 1987 to 1997, those most commonly isolated from blood were Salmonella Paratyphi A (442 [64%] of 694), Salmonella Choleraesuis (467 [63%] of 736), Salmonella Typhi (2565 [62%] of 4147), and Salmonella Dublin (350 [52%] of 671). The isolation rate of Salmonella species from sterile sites (i.e., blood or cerebrospinal fluid) was 6.5/100,000 infants, 0.5/100,000 persons 1–19 years old and 20–49 years old, 0.6/100,000 persons 50–79 years old, and 1.5/100,000 persons ⩾80 years old. Serotypes most commonly isolated from urine were Salmonella Cubana (63 [18%] of 355), Salmonella Tennessee (185 [12%] of 1509), Salmonella Meleagridis (42 [11%] of 393), and Salmonella Senftenberg (141 [9%] of 1598)
Table 3
View larger version:
In this page
In a new window
Download as PowerPoint Slide
Table 3
Reported number of Salmonella isolates from humans, by source of isolation and serotype: 15 most frequent serotypes, United States, 1987–1997
Regional distributionFrom 1987 to 1997, the rates of isolation of Salmonella species were highest in the New England states (figure 3A). Rates of Salmonella Typhimurium were also highest in New England, except from 1991 to 1993, when the rates in the Pacific region surpassed those in New England (figure 3B). In the early 1990s, rates of Salmonella Enteritidis (figure 3C) nearly doubled in a single year in both New England (5.7/100,000 persons in 1992 to 10.2 in 1993) and the Pacific region (3.2/100,000 persons in 1993 to 6.7 in 1994). In the Mountain region, rates of Salmonella Enteritidis increased more gradually, from 1.2/100,000 persons in 1992 to 4.5 in 1996. In contrast, in the Mid-Atlantic region, rates of Salmonella Enteritidis decreased 49% from a high in 1989 of 10.5/100,000 persons to a low of 5.4/100,000 in 1997. These decreases were also seen in the Pacific (34% decline from 1994–1997) and New England (57% from 1995–1997) regions. In all other regions, the rate of Salmonella Enteritidis remained relatively stable over time
Animal-associated serotypesWe examined trends in serotypes known to be associated with animals, including the reptile-associated serotype Salmonella Marina [12], as well as the top 2 serotypes identified by a 1998–1999 US Department of Agriculture (USDA) survey of broiler plants (Salmonella Kentucky and Salmonella Heidelberg), turkey plants (Salmonella Hadar and Salmonella Heidelberg), and swine plants (Salmonella Derby and Salmonella Typhimurium) [26]. Fewer than 10 human isolates of Salmonella Marina were reported until 1991, at which time there was a dramatic increase, which peaked at 81 isolates in 1996 and then decreased to 36 in 1997 (figure 4A). The number of reported Salmonella Kentucky isolates from humans decreased from 66 in 1987 to 31 in 1992, generally increased to reach 80 in 1995, and then decreased to 60 by 1997 (figure 4B). Salmonella Heidelberg steadily decreased from 6107 isolates in 1987 to 2104 in 1997 (figure 4B). The number of human isolates of Salmonella Hadar peaked at 2442 in 1988 and then steadily decreased to 543 in 1997 (figure 4C). Overall, the number of Salmonella Derby isolates recovered from humans declined over time from 412 in 1987 to 152 in 1997, and the number of Salmonella Typhimurium isolates decreased from 10,719 in 1987 to a low of 7950 in 1992 and then increased to 9116 in 1997 (figure 4D)
Emerging serotypesThe serotype with the greatest average annual increase in the number of isolates from 1987 to 1997 was Salmonella Stanley, followed by Salmonella Paratyphi B, and S. Marina (table 4). Of the top 20 increasing serotypes, 7 (35%) were common reptile-associated Salmonella serotypes (Salmonella Marina, Salmonella Flint, Salmonella Kintambo, Salmonella Wassenaar, Salmonella Ealing, Salmonella Carrau, and Salmonella Abaetetuba), compared with none of the 20 decreasing serotypes (P=.008, 2-tailed Fisher’s exact test). During the same period, the serotypes with the greatest average annual decreases in the number of isolates were Salmonella Heidelberg, Salmonella Hadar, and Salmonella Infantis (table 4)
Table 4
View larger version:
In this page
In a new window
Download as PowerPoint Slide
Table 4
Top 20 increasing and decreasing Salmonella serotypes, United States, 1987–1997
Previous SectionNext Section
Discussion
For the 11-year period, 1987–1997, 441,863 culture-confirmed Salmonella infections were reported to CDC, a mean of 40,169 each year. During this period, the isolation rate decreased from 19.1 to 12.9/100,000 persons, a decrease of ∼30%. This is in marked contrast to the earlier trend seen from 1976 to 1986, when the isolation rate increased from 10.7 to 18.1/100,000 persons, an increase of ∼70% [24]
Although the recent overall downward trend in rates of Salmonella infections is encouraging, it should be noted that this trend varies by serotype. Our findings document a general continued increase in Salmonella Enteritidis infections in the United States during the first 10 years of the reporting period that is consistent with previous studies both in the United States [24, 27] and Europe [27–30]. Furthermore, from 1988 to 1996, the percentage of Salmonella outbreaks that were caused by Salmonella Enteritidis increased from 47% to 55%, and Salmonella Enteritidis was the most common cause of all foodborne disease outbreaks in the United States during this same period [5, 17]. Many of these outbreaks have been epidemiologically linked to the consumption of undercooked eggs [2–5]. Case-control studies of sporadic Salmonella Enteritidis infections have also implicated raw or undercooked eggs [31–34] and, to a lesser extent, chicken [34, 35] as vehicles for transmission. In the early 1990s, efforts were made to control the spread of Salmonella Enteritidis through eggs by traceback of eggs implicated in outbreaks and diversion of infected eggs to pasteurization [36]. In addition, the egg industry implemented quality-assurance programs in several states, in an effort to reduce Salmonella Enteritidis in eggs and flocks [37]. The recent decline in the number of human Salmonella Enteritidis isolates in 1996 and 1997 may reflect these improvements; data from 1998 suggest that this decline has continued [38]
Salmonella isolation rates were highest for infants <1 year old. Although there may be some detection bias because of a greater likelihood that samples from an ill infant would be cultured, this most likely reflects a real increased rate among infants. The reasons for this are largely unknown but may include host susceptibility and exposure differences. The finding that infants have a 4–13-fold higher rate of invasive disease than do other age groups is intriguing and highlights the need for prevention
Salmonellosis demonstrated a marked pattern of seasonality, increasing in the warm summer months. This pattern is similar to other foodborne pathogens, such as Escherichia coli O157:H7 [39]. The reasons for this pattern are not entirely known; however, it may be related to infection trends in animal hosts or to holding food at an insufficiently cool temperature and food mishandling during the warmer months
Several O serogroups, A, C1, and D1, more commonly cause invasive infection and are frequently isolated from blood. Although Salmonella virulence factors are not well understood, differences in the O-side chain of the lipopolysaccharide appear to be important [40, 41]. Similar to findings from a recent study in California [42], the serotypes most frequently isolated from urine included O serogroups G, C1, E1, and E4
Overall, the rate of Salmonella species isolation was greatest in the New England states. This high rate is largely driven by the high rate of Salmonella Enteritidis, which persisted in this region until 1995. High rates of Salmonella Enteritidis in New England, the Mid-Atlantic region, and later in the Pacific region, probably reflect high infection rates among chicken flocks in these areas [43]
Although some serotypes, such as Typhimurium, are common in many different animal species, a number of Salmonella serotypes have specific animal reservoirs that are thought to contribute to disease in humans. For example, the animal reservoir for Salmonella Marina is a marine iguana that has become increasingly popular as a pet [12]. Salmonella Marina and other reptile-associated serotypes appear to be increasing over time. The recent decline of Salmonella Marina infections beginning in 1997 may reflect increased edu-cation for pediatricians, veterinarians, and pet store owners about the risk of reptile-associated salmonellosis [11, 44]
In 1996, the USDA implemented a rule, the Pathogen Reduction and Hazard Analysis and Critical Control Point (HACCP) Systems, aimed at reducing pathogens in our food. As part of this system, in 1998, USDA began routine testing for Salmonella species in large, federally inspected raw meat and poultry plants [26]. Of interest, 2 of the serotypes most commonly isolated from poultry demonstrated significant decreases in the number of human isolations from 1987 to 1997. During this same period, per capita consumption of chicken and turkey increased (chicken, from 39.1 to 50.9 lbs; turkey, from 11.6 to 13.9 lbs), and consumption of swine decreased from 47.7 to 45.6 lbs [45]. Although the decreases seen in human illness preceded the formal implementation of the HACCP system, industry-wide changes had been taking place for a number of years. For example, in the late 1980s, turkey manufacturers began providing chlorinated drinking water to turkeys to improve overall health (Peter Poss, poultry consultant, Wilmar, MN; personal communication). These changes may have contributed to the precipitous decline in the number of human isolates of 2 poultry-associated serotypes, Salmonella Hadar and Salmonella Heidelberg
From 1987 to 1997, there was a steady increase in the number of Salmonella Stanley isolations, and a peak occurred in 1995 during a large outbreak associated with alfalfa spouts [6]. Another serotype that increased during this period was Salmonella Paratyphi B, which is difficult to distinguish from Salmonella Java; the primary biochemical difference between Salmonella Paratyphi B and Salmonella Java is tartrate (positive in Salmonella Java), and a test for this is not routinely done in most state public health laboratories. Although the reasons for these increases are not known, both Salmonella Stanley and Salmonella Java have been found in reptiles [46]
The completeness of data reported through this system appears to be increasing. The percentage of isolates with an unknown serotype decreased from 6.2% in 1992 to 1.1% in 1997, which suggests that state public health laboratories may have improved serotyping capability. Furthermore, compared with previous years, patient age was more frequently reported on the surveillance forms (79% in 1984–1986 [24] vs. 85% in 1987–1997). Despite improvements, these data have several limitations. This surveillance system is a passive system that relies on clinical laboratories to send Salmonella isolates to the state laboratory. Only persons who were ill, sought care, had a stool culture, and had the Salmonella isolate forwarded to the state could have been captured in this system. For these reasons, it is recognized that the burden of illness caused by salmonellosis is greatly underreported in the United States [1]
Future surveillance will be enhanced by 2 new tools, the Salmonella outbreak detection algorithm [47] and PulseNet [48]. As described in detail elsewhere, these 2 tools will greatly aid epidemiologists in the detection of outbreaks. Serotyping remains an important laboratory tool that helps public health researchers better understand and define the epidemiology of salmonellosis in the United States. Measuring trends in serotypes over time can provide information about emerging serotypes and about the efficacy of prevention and control measures