Introduction
There has been an increased consumer demand for drinking raw milk [1, 2], and dairy products made from unpasteurised milk (including raw and thermised milk) such as cheese, butter, cream, ice-cream and kefir. There has also been an increase in the popularity of consuming raw milk and associated products produced from non-bovine species such as goat, sheep, horse, donkey and camel [Reference Faye and Konuspayeva3, Reference Hazeleger and Beumer4].
The consumption of raw milk and unpasteurised dairy products can present health risks from contamination by a variety of pathogenic micro-organisms, with more risks being reported in association with cows' milk as compared to other species such as goats and sheep [5, Reference Willis6]. However, there is comparatively little current data available for other unpasteurised dairy products consumed in the UK [Reference Lewis7–Reference Little and De Louvois10].
Food on sale in England is regulated as part of the Official Controls [11] and, because raw cow's milk has been categorised as a ‘risky food’, there are restrictions for the sale of cow's milk for drinking [1, 12]. Raw cow's milk for drinking can only be sold at farms and farm shops at the point of production, which is also extended to local deliveries and farmer's markets. Sale is prohibited at town markets, village fetes, school fetes, pop concerts, horse events, car boot sales, agricultural shows or laybys on the side of a road [12]. These restrictions do not apply to milk from other species or other dairy products made from unpasteurised milk.
The availability of published microbiological criteria for the interpretation of results in these products is variable. European Regulation EC 2073/2005 (as amended [13]) sets limits for Listeria monocytogenes in all ready-to-eat foods, as well as Escherichia coli levels in cream and butter made from unpasteurised milk, and coagulase positive staphylococci (CPS) in cheese made from unpasteurised milk. Raw drinking milk is not covered in this regulation but criteria for hygiene indicator tests (coliforms and aerobic colony count (ACC)) are specified in English legislation (Food Hygiene [England] Regulations, 2013). The Health Protection Agency Guidelines for ready-to-eat foods [14] indicate the need for pathogen and indicator organism testing and provide interpretations for a range of bacterial species. The Specialist Cheesemakers have also provided interpretations for E. coli, Enterobacteriaceae and Shiga toxin producing E. coli (STEC) in raw cheeses for manufacturers (http://www.specialistcheesemakers.co.uk/). Guidance and microbiological criteria are also available for local authority for raw cheese enforcement from the Scottish Food Enforcement Group, particularly for the control of STEC (https://www.foodstandards.gov.scot/downloads/Guidance_for_Local_Authorities_–_Cheese_made_from_Unpasteurised_Milk_-_May_2019.pdf).
Public Health England (PHE) manages a network of Food Water and Environmental Microbiology (FW&E) Laboratories which routinely examines dairy products for the presence of a range of bacterial pathogens and indicator organisms. Food samples are collected by Environmental Health Practitioners as part of their responsibility to enforce food safety legislation, monitoring food business operators or investigating cases of potential foodborne illness. Samples are also submitted to the PHE FW&E laboratories by producers who wish to perform their own verification checks on their products. In 2013, the FW&E laboratory network invested in and implemented a single laboratory information management system (LIMS). By mid-2019, at the time of writing (August 2019) this LIMS had microbiological results and associated data on more than 240 000 food samples. This database represents a resource for outbreak investigation, as well as for hazard analysis, risk assessment, risk management and risk communication [Reference Willis6, Reference McLauchlin15]. The aim of this study was to review results held within this LIMS which were obtained from testing both raw drinking milk and dairy products made with raw or thermised milk examined between 2013 and 2019.
Materials and methods
Sample collection
Data on the testing of all samples of raw drinking milk and other unpasteurised milk products tested during 2013–2019 were extracted from the FW&E LIMS. This dataset included samples collected from the point of sale as well as the point of manufacture and were sampled for routine monitoring, in response to hygiene concerns as well as part of outbreak investigations. Results of testing of finished product either at the point of production, on retail sale or in catering environments were included: product collected during maturation or food ingredients (e.g. raw milk used for cheese making or cheese curds) were not included in this analysis. In some instances, routine monitoring was carried out at production sites associated with incidents, but these were only considered as ‘routine’ following at least two sets of satisfactory clearance samples and at least 1 month after the completion of the incident investigation. Results on testing raw bovine milk for drinking that had been examined between 2014 and 2016 were excluded since these have been published previously [Reference Willis6].
Samples were collected in England and transported in accordance with the Food Standards Agency Food Law Practice Guidance [16] and examined in one of the network of PHE Official Food Control Laboratories located in Birmingham, London, Preston, Porton and York.
Data were collected on each individual sample which, together with the original data collected at the time of sampling, was annotated by internet searches. Cheeses made from raw or thermised milk were classified by type (hard, semi hard, semi soft, soft, blue and fresh) according to the Codex General Standards for cheese [17]. Additional data on cheese type, ingredient milk species and the range of products sold by individual dairies were added following manual searches of manufacturer's web sites as well as data from the Alphabetical List of Cheese (https://cheese.com/alphabetical/), the Specialist Cheese Makers Association (http://www.specialistcheesemakers.co.uk/) and the FSA's register of Raw Drinking Milk Premises in England, Wales and Northern Ireland (https://data.gov.uk/dataset/f6706084-9c82-4a50-a781-41e0e6229948/raw-drinking-milk-premises-in-england-wales-and-northern-ireland, 1st August 2019 update).
Results from testing a total of 2529 samples of raw milk and products made from unpasteurised milk and collected in England between February 2013 and April 2019 were identified in the PHE LIMS database. This dataset included: 719 raw cow's milks (2017–2019), 584 raw milks from non-bovine animals (2013–2019); 100 cream, two ice-cream, 37 butter, 24 kefir and 1063 cheeses (all from 2013–2019). The mean annual total was 360 samples examined and varied between 823 tested in 2017 and 101 in 2019 (January to April only).
The characteristics of the 2529 samples and the sampling settings for each of the various dairy products is shown in Table 1. Amongst all the samples, 56 bovine milks and 79 cow or goats milk cheeses were associated with six incidents or outbreaks of infection which are summarised in Table 2.
Table 1. Characteristics of the 2529 samples and sampling settings for raw milk and dairy products made from unpasteurised milk
Table 2. Summary of microbiological results of testing samples of raw milk and unpasteurised milk cheeses associated with incidents and outbreaks of infection
ACCs, aerobic colony counts; CPS, coagulase positive staphylococci; MLST, multilocus sequence type; SNP, single-nucleotide polymorphism type; ST, sequence type; STEC, Shiga toxin-producing E. coli.
Microbiological examination
A 10−1 homogenate of each sample was prepared in either maximum recovery diluent, dipotassium hydrogen phosphate buffer or buffered peptone water according to ISO 6887-1:1999 (International Organisation for Standardisation, 1999) and tested using standardised methods (Table 3). All presence/absence tests were performed on single 25 g/ml samples. Overall, of the 2529 samples tested, 7% were tested in Birmingham FW&E laboratory, 27% in London, 44% in Porton, 3% in Preston and 19% in the York laboratory.
Table 3. Test methods used for the various microbiological parameters
ACC, aerobic colony counts; CPS, coagulase positive staphylococci; STEC, Shiga toxin-producing E. coli.
Microbiological results were interpreted using: Commission Regulation (EC) No. 2073/2005 [13], the Food Hygiene (England) Regulations 2013 [23] and the HPA guidelines for assessing the microbiological safety of ready-to-eat foods placed on the market [14] (Table 4). Data on the detection of STEC stx genes (in the absence of the isolation of STEC organisms) were interpreted as satisfactory but is also included in this analysis.
Table 4. Criteria for the interpretation of microbiology results
Characterisation of isolates was performed in GBRU using a variety of methods [Reference Dallman25–Reference Nastasijevic28]. Results were compared with isolates from clinical cases as part of national surveillance. Descriptions of incidents were derived from local investigations.
Descriptive and statistical analysis of the data was undertaken using Excel 2010 (Microsoft, Redmond, Washington). Relative proportions were compared using the Fisher's exact test (GraphPad Software, San Diego, California). A probability value of less than 5% was defined as significant.
Results and discussion
Amongst all 2529 samples tested, 69% were classified as of satisfactory microbiological quality, 10% were borderline, 16% were unsatisfactory and 5% were unsatisfactory and potentially injurious to health due to the presence of pathogens (Table 5). As previously reported [Reference Willis6], results of statutory hygiene indicator tests for raw drinking milk do not correlate well with the presence of pathogens (see text later). Results of microbiological testing of samples collected during the investigation of incidents and outbreaks of foodborne illness (cow's drinking milk and cheese only) in Table 2 showed a higher overall proportion interpreted as unsatisfactory and unsatisfactory/potentially injurious to health: 44% as compared to 20% for those taken for routine monitoring (Table 5).
Table 5. Microbiological quality of samples of raw milk, and dairy products (cream, ice-cream, butter, kefir and cheese) made from unpasteurised milk which were collected in England during 2013–2019
Results are presented from routine monitoring which was either performed for the purpose of evaluating the hygiene of foods to support their routine food inspection process (in close collaboration with regulatory authorities) or directly for food manufacturers to support the validation of their food hygiene management systems. For all types of products collected for routine monitoring, similar results to those reported here (Tables 5–8) were found in previous studies in England of raw milk and unpasteurised dairy products [Reference Willis6–Reference Little and De Louvois10]. Although comparisons between studies should be interpreted with some caution in that sampling may be carried out for different purposes and not strictly co-ordinated within a rigid study design (e.g. based on market share). However, these results show many similarities and identify the same trends as those generated using more formal study designs [Reference Jørgensen24], and we are increasingly utilising this type of data and recognising its risk-based value to provide useful microbiological information from routinely collected food data [Reference Willis6, Reference McLauchlin15].
Table 6. Microbiological results from routine monitoring of raw drinking milk
NA, not applicable; STEC, Shiga toxin-producing E. coli; CPS, coagulase positive staphylococci.
a Includes unsatisfactory: potentially injurious to health.
b STEC detected but not isolated from a further two samples.
Table 7. Results from routine monitoring of cream, ice-cream, butter, kefir and yoghurt prepared from unpasteurised milk
NA, not applicable; CPS, coagulase positive staphylococci.
a Includes unsatisfactory: potentially injurious to health.
Table 8. Microbiological results from routine monitoring of cheese prepared from unpasteurised milk
NA, not applicable; STEC, Shiga toxin-producing E. coli; CPS, coagulase positive staphylococci.
a STEC detected but not isolate from a further 10 samples.
b Includes unsatisfactory: potentially injurious to health.
The application of microbiological criteria for the interpretation of results can be problematic with this group of products, with no single guidance document or statutory instrument. In this report, we interpreted the microbiological quality of ready-to-eat foods using legislative criteria [13, 23] and the HPA Guidelines [14]. The HPA guidelines were designed to be applied to ready-to-eat foods placed on the market and should therefore be used with some caution for products collected during production, as is the case here. However, we consider that the use of the HPA Guidelines is appropriate here since only end products were included, and not those during manufacture or food ingredients. Furthermore, the legislative requirements for raw cow's drinking milk require that the point of sale is the same as the point of production. Other dairy products (particularly butter and cream but sometimes cheese) are produced by the same businesses as selling raw milk for drinking (Table 1). Finally, for all types of unpasteurised dairy products, the point of sale can be directly from the manufacturer (including through postal or internet sales), as well as through farm shops which can be co-located with the point of production.
Raw milk for drinking
The FSA advice recommends business owners to test raw bovine milk for the following: indicator bacteria (E. coli, Listeria spp., ACCs, coliforms) and pathogenic bacteria (Salmonella, STEC, Campylobacter, CPS and L. monocytogenes [Reference Willis6]). For raw cow's drinking milk, results reported here for samples collected between 2017 and 2019 were similar to those previously reported for 2014–2016 and these two studies provide further baseline data and interpretation for subsequent monitoring of raw cow's drinking milk. There is no evidence to support an improvement in microbiological quality of raw cow's milk for drinking despite the efforts by the Food Standards Agency [1].
For raw milks for drinking which were collected for routine monitoring (Tables 5 and 6), cow's milk were generally of poorer microbiological quality than goat's milk or sheep's milk, both for the presence of indicators as well as for pathogens. Amongst the raw cow's drinking milk tested for routine monitoring purposes, results from 24 samples (4%) were interpreted as unsatisfactory: potentially injurious because of the presence and levels of pathogens (Table 5), no illness were detected as linked to consumption of these products: Campylobacter spp. were isolated from 18 of the cow's milk samples (Table 6), 13 of which came from only three producers: results for other parameters were all satisfactory from 13 of the samples, and in remaining five, Salmonella enterica serovar Mbandaka was isolated from one, unsatisfactory levels of coliforms were detected in three, and unsatisfactory ACCs were detected in the final sample. In one cow's milk sample there was an unsatisfactory level of L. monocytogenes detected (6.8 × 102 cfu/ml): levels of coliforms and ACC were also unsatisfactory for this sample. Salmonella was detected in three samples: S. Mbandaka was recovered in the presence of Campylobacter (see above) and in the remaining two, S. enterica serovar Dublin was isolated. Both the samples where S. Dublin was detected were collected on different occasions from the same dairy. The levels of coliforms detected were unsatisfactory for both samples while all other microbiological parameters were satisfactory. In the remaining three cow's milk samples categorised as unsatisfactory/potentially injurious to health, STEC was isolated. Two of the isolates both came from different samples collected from the same farm and were both identified as STEC O113:H4, ST10 (stx2d; eae-negative), the final isolate was identified as STEC O15:H16; ST: 325 (stx2g; eae-negative); all other microbiological parameters were satisfactory for all three samples.
Two of the goat's milks and none of the milks from sheep or other animals were categorised as unsatisfactory and potentially injurious to health. The unsatisfactory goat's milk samples had high levels of CPS, together with unsatisfactory levels of ACCs and coliforms: both samples came from the same farm and were collected in the same year.
Cream, ice-cream, butter and kefir
Cream, butter and kefir made from unpasteurised milk are niche products and unlike cheese, are manufactured at the same locations as raw milks (Table 1). Overall, results of microbiological testing classified the majority of the samples of cream, ice-cream, butter and kefir as satisfactory (Table 5). There were no results interpreted as unsatisfactory/potentially injurious to health except for two samples of kefir (Table 7) due to the presence of CPS (104 cfu/g): one was prepared from cow's milk and the other from goat's milk.
There are some difficulties in interpreting the results of hygiene indicators, particularly Enterobacteriaceae, in kefir. Since the microbiota of the kefir grains is usually uncharacterised, these results may be generated by the starter culture and further work on this food type will help to determine an appropriate test profile and interpretation.
Cheese
Amongst all the 984 cheeses tested as part of routine monitoring, 80% were of satisfactory microbiological quality, 5% were borderline, 10% were unsatisfactory and 5% unsatisfactory/potentially injurious to health. Goat milk cheeses were of poorer microbiological quality than those prepared from milk of other species (Table 5). The 47 cheese were categorised as unsatisfactory potentially injurious to health because of high levels of L. monocytogenes or CPS, or the isolation of Salmonella, E. coli O157 or STEC (Table 8) and a summary of the results from these samples is shown in Table 9. Apart from two possible cases of salmonellosis with indistinguishable Salmonella Newport isolated from a hard cow's milk cheese, analysis of national surveillance databases did not provide any other evidence for disease associated with the consumption of these products, or any other of the cheeses sampled here. The samples of cow's, goat's or and sheep milk cheeses collected at the point of production and categorised as unsatisfactory potentially injurious to health were collected from 15 (15%) of the total of 98 dairies sampled. Amongst 41 samples of goat's milk cheese from a single dairy, unsatisfactory levels pathogens were detected in 10 samples: three with unsatisfactory levels of L. monocytogenes, five with unsatisfactory levels of CPS and two with unsatisfactory levels of both L. monocytogenes and CPS. Although these results may reflect bias from resampling, the decision to sample will have been taken on the recognition of risk within this manufacturer's premises. Therefore within the data presented here, adverse microbiological results were more common in a subgroup of manufacturers where efforts to improve hygiene should be concentrated.
Table 9. Summary of microbiological results from testing 47 cheese samples where an interpretation of unsatisfactory potentially injurious to health was obtained
Based on product descriptions, amongst all the 984 cheeses tested as part of routine monitoring, 34 (3%) were classified as fresh, 355 (36%) as hard, 35 (4%) as semi-hard, 57 (6%) as blue, 108 (11%) as semi-soft and 145 (15%) as soft: the remaining 250 (25%) could not be classified. The proportion of semi-soft and soft cheeses with unsatisfactory or borderline levels of CPS (25/102 (23%) for semi-soft and 34/121 (28%) for soft) was significantly higher than for hard and semi-hard cheeses (54/364 (15%): Fisher's exact test; P = 0.025 for semi-soft and P = 0.002 for soft cheese). Levels of E. coli were determined for 576 cheeses that could be classified into different types (hard, soft, etc.), and the distribution of these two parameters is shown in Table 10. The proportion with unsatisfactory levels of E. coli (>102 cfu/g) varied from 6% of the hard cheeses, 27% of the semi-hard, to 33–36% of the blue, semi-soft, soft and fresh. The highest levels of E. coli occurred in the semi-hard, blue, semi-soft and soft cheese types and corresponded to the types with the lowest proportion of a satisfactory or borderline interpretation.
Table 10. Levels of E. coli detected as part of routine monitoring of different categories of cheese prepared from unpasteurised milk
It is well recognised that there is a relationship between E. coli levels and cheese-type with high levels of E. coli being more common in soft compared to hard cheese at the end of production or at retail [17, Reference Donnelly29]. To achieve process control during manufacture and provides assurance that food safety management plans are operational, cheese makers routinely monitor the microbiological (as well as the physicochemical) quality of the cheese [17, Reference Donnelly29]: an important component of this monitoring utilises testing for levels of generic E. coli. The UK Specialist Cheesemakers Association has provided criteria for levels of E. coli, with a distinction made between hard cheese (<100 cfu/g considered satisfactory) and soft or semi-soft cheese (<104 cfu/g) (http://www.specialistcheesemakers.co.uk/). The Guidance for raw cheese enforcement from the Scottish Food Enforcement Group (https://www.foodstandards.gov.scot/downloads/Guidance_for_Local_Authorities_–_Cheese_made_from_Unpasteurised_Milk_-_May_2019.pdf) recommended that a target level of <100 cfu/g is achievable for some cheese types, and where this is exceeded, further evidence should be provided to verify food safety. The results presented here show an association between higher levels of E. coli and the presence of unsatisfactory levels of CPS (Table 11). Therefore, the data indicate that the application of more lenient E. coli criteria for unpasteurised cheeses, or soft cheeses, for example, cannot be justified in terms of public health, and it is more appropriate to apply the criteria set out in the HPA guidelines [14] to all ready-to-eat foods, including those made from unpasteurised milk. Both E. coli and CPS are likely to reduce as cheeses mature, although the biological risk associated with staphylococcal enterotoxins will remain even in the absence of viable S. aureus [Reference Seo, Bohach, Doyle and Beuchat30]. The presence of CPS just after manufacture as well as being an indicator of public health risk on its own, may also be a useful indicator of STEC. Furthermore, this should prompt investigation of the use of E. coli as an indicator for other raw foods such as raw milk rather than the use of coliforms and ACC.
Table 11. Unsatisfactory detection of pathogens at different E. coli levels detected as part of routine monitoring of cheese prepared from unpasteurised milk
STEC, Shiga toxin-producing E. coli; stx, shiga toxin gene; CPS, coagulase positive staphylococci.
There have been major advances in understanding microbiological hazards in dairy products over the past decades, particularly in relation to STEC [31, 32]. Not only is STEC O157 recognised as a major hazard, but this also extends to other STEC serogroups. The methods to detect and isolate STEC from food matrices are technically demanding and we report here on routine monitoring results obtained from testing 142 cheese prepared unpasteurised milk of which STEC was isolated from four samples: one was STEC O157, and the remaining three were either STEC O2 or O6 (Table 9). Although much effort has been directed towards controlling STEC O157, the isolation of any STEC is unsatisfactory/potentially injurious to health and this is consistent with advice from the Food and Agriculture Organization of the United Nations [18] stated: ‘It is not prudent to regard any STEC strain as being non-pathogenic or not posing a health risk, as all STEC strains probably have the potential to cause diarrhoea and be of risk, especially to susceptible individuals’. Although this presents difficulties to food regulators (as well as food manufacturers), Food Standards Scotland issued a policy statement in 2019 [18] stating ‘the presence of any STEC in a ready-to-eat food is a potential risk to health, and could cause food poisoning’. For this, amongst other reasons, further surveillance is ongoing on the microbiological quality of unpasteurised milk cheeses, including the presence of STEC.
In the three samples where STEC was isolated, stx genes were initially detected by polymerase chain reaction. There were a further 10 cheese samples where stx genes were detected but STEC was not isolated. The relationship between different levels of E. coli and the occurrence of unsatisfactory and borderline levels of L. monocytogenes and CPS and the isolation of STEC and detection of stx genes are shown in Table 11. The proportion of samples with borderline and unsatisfactory levels of L. monocytogenes was similar across the different levels of E. coli (Spearman's rank correlation, rho = 0.03, P = 0.4). However elevated levels of CPS (Spearman's rank correlation, rho = 0.21, P = <0.001) and stx detection (Spearman's rank correlation, rho = 0.19, P = 0.02) was significantly more likely to occur in the presence of higher levels of E. coli. The proportions of samples where STEC was isolated was significantly reduced in the presence of higher levels of E. coli (Fisher's exact test P = 0.04). The significantly reduced isolation of STEC in the presence of higher levels of generic E. coli may reflect the greater technical difficulty in isolating STEC in the presence of a greater competing microbiota (especially non-STEC E. coli). The significant association between stx detection and higher generic E. coli is intriguing but we are unclear how to interpret these results. While the most likely reason for non-isolation is the background microbiota, this may also be as a result of free stx-containing phage. Recent studies suggest that the cheesemaking process triggers the production of stx containing prophages, potentially interfering with the analysis of STEC in the finished products [Reference Bonanno33]. These studies demonstrated that oxidative (aeration and exposure to oxygen) and salt stress, which are both likely to occur during cheesemaking, had the ability to induce stx phages in vitro. Additionally, production of stx phages was also observed during cheesemaking when milk was inoculated with a strain of STEC O26.
Summary
We reviewed here the results of microbiological testing of over 2500 raw drinking milk and dairy products made with unpasteurised milk examined in England between 2013 and 2019. Samples were collected as part of incidents as well as part of routine monitoring and were tested for a range of bacterial indicators and pathogens. This review provides baseline data and interpretation for subsequent monitoring of raw cow's drinking milk and reports a similar level of adverse microbiological results to that reported previously for samples tested between 2014 and 2016: there is no evidence to support an improvement in microbiological quality of this product despite the efforts by the Food Standards Agency. While results presented here, and previously, indicate that the statutory hygiene indicator tests for raw drinking milk do not correlate well with the presence of pathogens. Analysis of data from cheese demonstrated an association between increasing levels of indicator E. coli with elevated levels of CPS and detection of stx genes. The isolation of STEC was significantly associated with lower levels of indicator E. coli. These data provide evidence for setting criteria for E. coli in cheeses made from unpasteurised milk. This group of products is a concern for public health, and there is a need for continued surveillance and implementation of controls during production and throughout the food chain.
Acknowledgements
The authors would like to thank colleagues for their contributions to this study including colleagues within PHE (particularly C.F.L. Amar, A. Charlett, C. Jenkins and L. Larkin for reference laboratory analysis, surveillance data and statistical support), M. James from Food Standards Scotland for insightful critical comments and staff in environmental health departments throughout England.
Financial support
This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors.
Conflict of interest
None.
Data availability statements
The datasets used or analysed during the current study are available from the corresponding author on reasonable request.
Open access
