WHO Global Salm-Surv has conducted annual regional training courses for national reference laboratories since 1999 and has managed an external laboratory quality assurance program since 2000 to facilitate a standard approach to isolating and serotyping salmonellae (10). National reference laboratories can become WHO Global Salm-Surv members and share Salmonella serotype data with other members through the country databank. The country databank is a Web-enabled Oracle database that is password protected for data entry and accessible for public viewing at http://www.who.int/salmsurv. Each year, a designated national reference laboratory representative enters into the country databank the number of Salmonella isolates serotyped from human, animal, food, environmental, and feed sources and the 15 most frequently identified serotypes.

Descriptive analysis was conducted by using Microsoft Excel (Microsoft Corp., Redmond, WA, USA) on data from all countries that submitted data for 2000, 2001, or 2002 as of June 2004. Analyses for trends over time were conducted on data from 2000 to 2002. More detailed analyses, including ranking of serotypes, comparison of human to nonhuman isolates, and regional comparisons are presented for 2002 data only, the year in which the most countries participated.

Before 2001, nonhuman isolates were grouped together. Since 2001, countries have been able to submit food, animal, environmental, and feed data separately. For comparison purposes, all nonhuman data were combined in this analysis.

Data were grouped into regions approximately corresponding to 6 geopolitical continents: Africa, Asia, Latin America and the Caribbean, Europe, North America, and Oceania. To accommodate local epidemiologic characteristics, New Caledonia was incorporated into Asia, and Israel was incorporated into Europe. For years in which a single country contributed data for a region, regional data are not presented, but the data are included in the overall results. A region-specific serotype was defined as a serotype that, for each of the years of the study period, was among the 15 most commonly reported serotypes and for which >90% of the isolates were from that region.

S. Enteritidis is the most common Salmonella serotype in humans globally but especially in Europe, where it accounts for 85% of Salmonella cases, Asia (38%), and Latin America and the Caribbean (31%). The S. Enteritidis pandemic was first noted in the late 1980s and has been attributed to contaminated eggs (11). The proportion of Salmonella infections associated with this serotype seems to have increased over time. In 1995, 36% of salmonellae worldwide were S. Enteritidis, compared to 65% in 2002 (8).

S. Typhimurium has been 1 of the 2 most frequent serotypes in humans since 1990 (8). Since S. Enteritidis and S. Typhimurium are so common, additional subtyping methods, including phage typing, antimicrobial susceptibility testing, and pulsed-field gel electrophoresis (PFGE), are needed to identify clusters of infection from the same source. WHO Global Salm-Surv includes antimicrobial susceptibility testing training in all regional courses and has introduced phage typing in the Eastern European region course. The country databank could include data from such subtyping efforts. PFGE subtyping data are exchanged in North American between PulseNet USA and PulseNet Canada (12). PulseNet International is an affiliate member of WHO Global Salm-Surv, and the networks are coordinating their efforts to ensure synergy.

S. Typhi is a pathogen of concern in the developing world, especially Asia (13). However, in our analyses, S. Typhi was the ninth most frequent serotype in Asia in 2002. The Asian countries that contributed to the country databank did not include many of the developing countries in south-central and Southeast Asia, where S. Typhi is still highly prevalent. S. Typhi was the sixth most frequent serotype globally in 1995 and was decreasing in relative importance (8). That trend seems to have continued; S. Typhi ranked 14th globally in 2002. S. Typhi has no animal reservoir, which makes it susceptible to improvements in hygiene and sanitation seen in many regions of the world, such as Latin America and the Caribbean.

The distribution of nonhuman serotypes is more heterogeneous than that of human serotypes. The same serotypes appear among the top 5 in human and nonhuman sources, although in a different order. S. Enteritidis is only the third most common serotype among nonhuman sources. In 2002, it was not reported at all among the 10 most common nonhuman serotypes from North America. This finding partly reflects the capacity of S. Enteritidis to contaminate eggs in low numbers and the difficulty of isolating it from food or the environment. Moreover, in North America, few samples from eggs are submitted for routine testing. For example, in the United States, routine testing of eggs is not required, whereas routine testing for salmonellae is required of meat and poultry plants. As eggs are frequently used in foods that do not undergo heat treatment (e.g., pastries, homemade ice cream, and mayonnaise) and are widely distributed, this food contamination has a substantial effect on public health.

The country databank contains far fewer nonhuman than human serotypes, possibly because more participating laboratories are human national reference laboratories, fewer countries have formal nonhuman surveillance, and some countries may be less likely to share nonhuman data because of trade concerns. In 2001 and 2002, 15 of 22 countries reported nonhuman isolates by source. Food serotypes were reported from most countries (11 in 2001 and 12 in 2002), followed by animal serotypes (7 countries in 2001 and 10 in 2002). Most isolates serotyped were from animals (66%), followed by food (29%), feed (3%), and the environment (2%). The reporting of S. Weltevreden from the environment, feed, animals, food, and humans in Southeast Asia is an example of how the country databank can be used to track Salmonella serotypes along the food chain.

Many serotypes are restricted to a single region of the world. This finding may reflect an ecologic niche or a local food source that is not exported. A number of such examples have been reported in the past, such as S. Marina associated with marine iguanas from South America found in the United States and S. Tilene in imported African pygmy hedgehogs in the United States and Canada (14–16). The country databank is uniquely placed to allow countries to observe this phenomenon. Investigators have reported infections of S. Javiana associated with exposure to wild amphibians in a confined area in the southeastern United States (17). According to the country databank, S. Javiana is only reported among the 15 most common serotypes in the United States. In 2001, the WHO Global Salm-Surv country databank helped confirm that S. Weltevreden was largely restricted to Southeast Asia. A survey of Southeast Asian laboratories showed that items most frequently associated with this serotype include seafood, water, and Asian vegetables (18). In the same region, S. Rissen has increased in both human and nonhuman sources (19). The country databank allows countries to become aware that a common serotype in their country may be rare elsewhere in the world, leading to hypothesis generation in outbreaks and studies to understand the sources of disease. Countries that report a large number of isolates to the country databank, such as North American and European countries, typically do not report rare serotypes because these would not rank in their top 15, thus limiting the ability to track rare serotypes in these countries.

Countries with fewer resources may lack complete antisera kits necessary to identify certain serotypes, which would lead to underreporting. For example, although we assume that S. Enteritidis human infections occur globally, a number of countries in Asia, Africa, and Latin America and the Caribbean did not report this serotype in their top 10 (Table 2). Lack of resources can also cause misclassification of serotypes. For example, S. Paratyphi B was reported to be among the most common serotypes in Latin America and the Caribbean. However, some countries in the region lack the capacity to differentiate between S. Paratyphi B and S. Paratyphi B tartrate var. Java.

In general, industrialized countries are more likely to regularly contribute to the country databank and to report more isolates. The results are therefore biased towards the industrialized world. However, the country databank lacks data from many Western European countries. Twenty-four European countries report human Salmonella serotype results annually to Enter-net, a European-based surveillance network for gastrointestinal infections, as compared to only 14 reporting to WHO Global Salm-Surv from 2000 to 2002. A review of recent Enter-net data confirms that S. Enteritidis is by far the most commonly isolated serotype in Europe but at a lower proportion than that reported to WHO Global Salm-Surv; 68%–71% of Enter-net Salmonella isolates were S. Enteritidis from 2000 to 2002 (Ian Fisher, pers. comm.) (20). A recent agreement between WHO Global Salm-Surv and Enter-net will lead to routine electronic sharing of data between the 2 systems to improve efficiency and representativeness (Henrik Wegener, pers. comm.).

Serotypes reported by a region are not necessarily circulating locally and may have been imported through travel or traded foods. Intraregional comparisons are limited by the fact that case definitions and surveillance systems vary between countries. The country databank does not collect the source of isolation. Some countries may report salmonellae isolated from both blood and stool and others from stool only. The low number of isolates and countries reporting nonhuman data and the pooling of food, animal, environmental, and feed sources hamper further analysis of nonhuman data. Some regional results may not be representative, since some regions have few countries reporting data to the country databank (e.g., Africa).

Some countries may not submit data to the country databank because of concern regarding international trading of food. Others do not have the supplies or training necessary to conduct serotyping. WHO Global Salm-Surv training courses were launched in Southeast Asia in 1999 and expanded to South America and the Middle East in 2000; China, Central America, and the Caribbean in 2001; and West Africa and Eastern Europe in 2002 (http://www.who.int/salmsurv). All participating countries were initially provided with antisera to conduct serotyping. WHO Global Salm-Surv established an external quality assurance system (EQAS) in 2000 to assess the accuracy of serotyping and antimicrobial susceptibility testing among member national reference laboratories. From 2000 to 2002, the number of laboratories participating in EQAS increased from 44 to 117, and the capacity to correctly serotype 8 Salmonella isolates improved from 76% to 90% of participating laboratories (21). The increased reporting of certain serotypes during the study period may be due to capacity improvement and increased participation in the country databank as well as real changes in the epidemiologic features of salmonellae. For example, S. Montevideo was first reported in Africa in 2001. In 2002, Tunisia participated in the country databank for the first time and reported S. Montevideo in its top 15 (increased participation). Oceania started reporting S. Montevideo in its top 15 in 2002 for the first time since participating in the country databank (change in epidemiologic features).

Data are not entered into the country databank in a timely enough way to detect international outbreaks. However, regularly monitoring the data can allow emerging trends in regional and international Salmonella epidemiology and region-specific serotypes to be detected. This information in turn leads to hypothesis generation, studies, and international collaboration to improve control of salmonellae in the long term. Examples of such work include the surveillance of Salmonella serotypes and antimicrobial drug resistance in South America, China, and the Democratic Republic of Congo; the assessment of risk factors for and drug resistance of S. Weltevreden in Southeast Asia; and the molecular characterization of S. Corvallis isolates from Bulgaria, Thailand, and Denmark (18,22–26, F. Aarestrup, pers. comm.).

The WHO Global Salm-Surv Country Databank is a valuable resource for international Salmonella surveillance. Past attempts to characterize Salmonella serotype distribution globally have either not been widely accessible or relied on irregular surveys of laboratories (8,11). Trends in global Salmonella epidemiology can now be updated and followed across regions and over time. In an era when most national institutions have access to the Internet, using a Web-based data collection tool is both feasible and practical. The data are immediately and publicly accessible for viewing and analysis (http://www.who.int/salmsurv). The results have several limitations in terms of representativeness and comparability but can be used to follow trends, generate hypotheses, and assess the effect of major interventions. This surveillance is a step toward improving the understanding and control of salmonellae worldwide.