Salmonella enterica contamination of foodstuffs is a considerable human health problem with more than 170 000 human cases reported in the EU in 2005[1]. It is well known that Salmonella contamination of animal feed can be disseminated to food producing animals and further down the food chain, causing big economic losses and most importantly be a threat to animal and public health [2, 3]. The presence of Salmonella in animal feed and feed ingredients is not unusual [1] and cases of human illnesses have been reported where the source of infection were found to be contaminated animal feed [4]. Investigations by the US Food and Drug Administration (FDA) showed that protein-based animal feed are frequently contaminated with Salmonella enterica [2]. A major outbreak of human salmonellosis in the United States, United Kingdom, Israel and the Netherlands back in 1970, caused by the, at the time, uncommon serotype S. Agona, was traced back to Peruvian fish meal used as an ingredient in the feed manufacturing [2]. Since then S. Agona has caused considerable human illness every year in United States [5]. In a study by Davies et al. [6], a strong link between Salmonella contamination in feed mills and infections in chickens that received feed from the contaminated feed mills could be established. Sauli et al. [7] pointed out that contaminated animal feed is a significant source of infections in pigs. In 2003, an outbreak of S. Cubana in Sweden occurred in a large number of pig farms where the source of infection was traced back to a feed mill that produced pelleted pig feed, indicating the potential effects of feed contamination for further dissemination of Salmonella in the food chain [8]. It was found that the source of infection was a contaminated cooler for the pelleted feed where multiplication of Salmonella occurred in the humid and warm coatings inside the cooler.

Salmonella control of food producing animals has a long history in Sweden and started already in the late 1950s [9]. The present Swedish control program for feed, based on hazard analysis of critical control point (HACCP) principals in the feed mills, was initiated in 1991 by the feed industry [10, 11]. Scrapings and dust samples from critical control points (CCPs) in the processing line are analysed for Salmonella on a weekly basis giving a rapid alert if Salmonella is detected. Imported feed raw materials are sampled for salmonella according to a sampling programme and must remain in quarantine until the analytical results are completed. Salmonella positive ingredients are treated with organic acids and re-tested for Salmonella before they can be used in the feed manufacturing.

The reported prevalence of Salmonella in different feed materials varies considerably between different countries which is most likely due to differences in the sampling and isolation methods used [1]. As pointed out by Gardner [12], lack of comparability may induce bias in the risk estimates if the results from different studies are treated as if the tests and sampling schemes were the same. Animal feed and feed raw materials are usually dry products with a low water activity and the Salmonella cells present are strongly dehydrated. For that reason isolation methods for Salmonella in feed must be able to regenerate the multiplication of dehydrated and stressed bacterial cells. Geue and Schluter [13] showed that five-fold fractional enrichment of feed and faeces samples resulted in an increased number of Salmonella isolations. Validation studies of isolation methods are mainly focused on food and only very few feed materials are usually included in the studies. In a study by Salomonsson et al. [14] the levels of competing microflora in different feed materials including scrapings were found in the range of 10²–10⁷/g.

Animal feed are produced in large quantities, usually as bulk materials in a batch wise production in feed mills. The numbers of Salmonella cells in feed are usually low and their distribution may not be even. According to Maciorowski et al. [3] it is crucial to give injured and stressed Salmonella cells a chance to recover and multiply in the enrichment in order to successfully isolate Salmonella from animal feed.

The international standard method for detection of Salmonella, EN ISO 6579:2002, consists of non-selective pre-enrichment in Buffered Peptone Water (BPW), selective enrichment in Rappaport-Vassiliadis with soy broth (RVS) and Müller-Kauffmann tetrathionate-novobiocin broth (MKTTn), plating on the selective solid medium Xylose Lysine Deoxycholat agar (XLD) and a second selective solid medium such as Brilliant Green agar (BGA) and a final serological and biochemical confirmation [15].

The same procedure, without the MKTTn step, is used in the NMKL71 method, which is the standard method for Salmonella detection in the Nordic countries for food but also for feed [16].

The Modified Semisolid Rappaport Vassiliadis (MSRV) method, Draft Annex D of EN ISO 6579:2002 [17] is based on migration of motile Salmonella through the selective medium [18]. It has been shown in several studies of naturally infected or artificially contaminated food or faecal samples that MSRV is, in most cases, more sensitive than the standard methods [18–21]. However, there are also studies showing that MSRV is less sensitive compared to other enrichment media [19, 22]. In the recent EFSA report about microbial risks in feedingstuffs it is recommended that MSRV or alternative methods should be validated also for use in feed [23].

The standard cultural methods are primarily developed for food materials and only very few feed materials have been included in the validations studies. For a bibliography of isolation methods for foodborne Salmonella see Schonenbrucher et al. [24]. Despite this, methods developed for food are the standard methods used for most feed materials. The levels of Salmonella in feed are usually low and an estimate of accuracy based on high spiking levels is thus not relevant [25] and it is important to validate the methods at realistic levels before they are adopted as official standard methods for feed.

The aim of the present study was to evaluate the NMKL71, MSRV and the EN ISO 6579:2002 methods in terms of their relative ability to detect low levels of Salmonella in artificially contaminated feed and feed ingredients and also to investigate the accuracy, sensitivity and specificity of respective method. To our knowledge, no such comparisons have so far been performed with feed materials. It was also of interest to find out if there were any differences in detection levels of some serotypes often found in feed materials.

In Figure 1 a flow-chart diagram of the different steps of respective method are presented. For all methods 3–4 days are required for isolation of Salmonella. The a_w of the feed materials used in the study were < 0.66 before spiking and after spiking < 0.68. All non-spiked feed materials were negative for Salmonella with the NMKL71 method, which was also later confirmed by the EN ISO 6579:2002 and MSRV methods.

The detection level was slightly different in each feed type and the lowest spiking level of 0–1 cfu/25 g did not generate any Salmonella positive samples with any of the serotypes (Tables 2, 3, 4). One XLD plate from MKTTn was positive at 10-10² cfu/25 g for wheat spiked with S. Yoruba that was not detected by the other selective media. MKTTn failed to detect Salmonella in two out of five replicates at 10-10² cfu/25 g for rape seed meal spiked with S. Yoruba.

The RVS in comparison with the MKTTn medium produced fewer false negative isolates on BGA plates (Tables 2, 3, 4). The XLD and BGA plates were more contaminated with competing microflora from the enrichment in MKTTn broth than from the RVS broth. The intrinsic flora was more abundant on XLD and BGA plates from scrapings compared to other feed materials tested and selective enrichment of scrapings in MKTTn, plated on BGA plates, did not generate any typical Salmonella colonies in any experiment.

On the MSRV plates no additional positive results, after incubation for another 24 h, was observed. The XLD and BGA plates plated from MSRV contained considerably less intrinsic flora compared to plates from MKTTn or RVS broth. Serological agglutination of isolated colonies confirmed the presence of added serotypes.

The objective of the present study was to investigate the relative performance of common cultural methods for Salmonella in animal feed and feed ingredients using artificially contaminated samples. By stressing the Salmonella bacteria before isolation attempts were made to simulate the more natural conditions prevailing in feed materials. Because of the low levels of Salmonella potentially present in animal feed, low levels of Salmonella were added to the feed samples and for that reason the calculated accuracy and sensitivity data are lower than reported in other studies [20].

The volume of the bacterial cultures added to the respective feed material did not generate any significant change of the a_w, clearly indicating that available moisture did not allow any bacterial growth before the BPW was added. Salmonella was not detected in any of the non-spiked samples by any method and serotyping of isolated strains verified that no other than the added Salmonella serotypes were present in the feed materials.

The detection levels observed with MSRV, MKTTn or RVS were similar with the feed materials tested. However, the detection levels varied for the different feed and feed ingredients investigated and for palm kernel meal it was higher for all serotypes compared to the other feed materials tested. At the lowest spiking level, stochastic events will result in an uneven distribution of cells between individual samples, and it is possible that some samples contained no or only few cells. This fact may have contributed to the low sensitivity observed at these spiking levels. The reasons why low levels of Salmonella are not detected in palm kernel meal is presently unknown. However, the most likely explanation might be poor growth of Salmonella during the enrichment in the presence of palm kernel meal. The higher detection levels in palm kernel meal might leave Salmonella positive consignments undetected when screening of feed ingredients are carried out.

According to several studies, a low pH value during the pre-enrichment might affect the viability of Salmonella bacteria [19, 26] and according to the EN ISO 6579:2002 method the pH value should stay above 4.5. In the present study the lowest pH-value was detected for rape seed meal (4.9). For palm kernel meal, with the highest detection level of Salmonella, the pH-value was 6.0, indicating no correlation between a low pH value and a low isolation rate. Other factors that might affect the re-isolation of Salmonella could be the presence of potential growth inhibitors in the feed materials or competing bacterial flora.

Positive BGA plates were more difficult to interpret when inoculated from MKTTn, most likely because of high levels of competing microflora and few characteristic Salmonella colonies. D'Aoust [27] showed that members of Citrobacter, Enterobacter, Escherichia and Proteus could be resistant to tetrathionate, resulting in reduced selectivity for the MKTTn broth.

Salomonsson et al. [14] showed that wheat and scrapings contained high numbers of aerobic microorganisms and also Escherichia coli. According to Van Schothorst and Renaud [28] a high level of intrinsic flora in the sample, after enrichment, may reduce isolation of Salmonella bacteria on BGA. A comparison between XLD plates and BGA-plates showed that XLD is superior to BGA, due to higher selectivity [22, 29], which is in agreement with the results in the present study. Re-incubation of MSRV plates for an additional 24 h did not increase the number of positive samples and an uneven halo was typically observed, not representing growth of Salmonella.

The cultural methods compared in this study are primarily developed to detect Salmonella in different food samples and since the level of intrinsic flora is usually high in animal feed [14] the isolation of Salmonella from feed might be affected. Another drawback is that the cultural methods are time consuming which might delay possible control measures. MSRV is unable to detect non-motile Salmonella bacteria, that represent < 1.0% of the isolates from animal feeds [30]. However, this study shows that selective enrichment on MSRV results in lower level of competing flora on the XLD and BGA plates then does enrichment in other media.

In conclusion, the cultural methods in the present study were shown to be surprisingly equivalent in terms of accuracy, sensitivity and specificity for different feed materials and serotypes, respectively. An interesting observation was the differences in detection levels between different feed and feed ingredients.

NMKL71:5, the international standard method EN ISO 6579:2002 and the Draft Annex D of EN ISO 6579:2002, using the MSRV as a selective enrichment medium, were compared, in relation to their accuracy, specificity and sensitivity for different feed ingredients and feed materials.