Development of a Salmonella source-attribution model for evaluating targets in the turkey meat production

Report (0.7 Mb)

Summary

Project developed on the procurement project NP/EFSA/BIOHAZ/2011/04

Following a request from the European Commission, the Scientific Panel on Biological Hazards (BIOHAZ) was asked to assess the relative public health impact if a new target for reduction of Salmonella is set in fattening turkey flocks being 1 % or less remaining positive for all Salmonella serovars with public health significance, compared to (1) the theoretical prevalence at the end of the transitional period (1 % or less flocks remaining positive for Salmonella Enteritidis and/or Salmonella Typhimurium), and (2) the real prevalence in 2010 reported by the Member States (MSs). This external scientific report describes the work conducted in order to support the BIOHAZ Panel in answering this request.

A turkey-target source attribution model (TT-SAM) was developed to provide estimates for the quantitative contribution of turkeys and other major animal-food sources to the estimated true burden of human salmonellosis in the EU. The mathematical model was based on the so-called microbial subtyping approach, which allows for distinguishing between the different Salmonella serovars. The basic principle is to compare the serovar distributions observed in different animal-food sources with the serovar distribution found in humans. A similar model has previously been applied to answer an equivalent question for Salmonella targets in the broiler production.

The TT-SAM model employed the following data: (i) the results from the harmonized EU monitoring in turkey, broiler and laying-hen flocks in 2010, (ii) the results from the EU-wide Salmonella Baseline Surveys on slaughter pigs, (iii) the reported cases of human salmonellosis in EU in 2010 as provided by the European Centre for Disease Prevention and Control (ECDC), and (iv) the amount of each food source available for consumption by MS as estimated from different data sources on production, import and export. The model included data from 25 MSs, four animal-food sources (turkeys, broilers, laying hens and pigs) and 23 individual serovars. To take account for differences in underreporting of human salmonellosis cases, MS-specific underreporting factors were applied in the model. Some sources of Salmonella (e.g. cattle/beef) were not included in the model due to lack of data. The possible influence of this is discussed.

First a baseline model applying reported prevalence data from the harmonized monitoring in turkey flocks in 2010 was developed. Then in order to answer the Terms of Reference, seven different scenarios, where the combined prevalences of specific serovars were changed, were developed and the results compared to the results of the baseline model.

The results of the baseline model indicated that 2.6 % (95 % CI: 1.2-5.2) of all human salmonellosis cases (i.e. estimated true number of cases when accounting for underreporting) in the EU were attributed to the turkey reservoir. This corresponds to 135 100 (95 % CI: 60 790-293 600 human cases in 2010. Around 63 % of the turkey-associated human salmonellosis cases were caused by serovars other than the currently regulated serovars S. Enteritidis and S. Typhimurium. Four serovars (S. Kentucky, S. Saintpaul, S. Senftenberg and S. Kottbus) had turkeys as the most important reservoir for human infections.

For the other animal-food sources included in the model, the attribution estimates were that 56.8 % (95 % CI: 48.2-65.8), 10.6 % (95 % CI: 5.1-18.3) and 17.0 % (95 % CI: 11.3-24.0) of the estimated number of human salmonellosis cases could be attributed to the pig, broiler and laying-hen reservoir, respectively. However, when looking at the relative risk between turkey meat and the other three sources weighted by the tonne of meat/eggs available for consumption, this picture changes, indicating that the risk of infection for the individual consumer is highest when consuming shell eggs closely followed by the consumption of pig meat, whereas the risk is lower for turkey and broiler meat.

For the scenario analyses, the largest reduction was found for the scenario, where the overall prevalence (i.e. the combined prevalence of all serovars) in turkey flocks per MSs is reduced to 1 %. Here a reduction in the number of turkey-associated human cases of 83.2 % (95 % CI: 79.0-87.4) compared to the baseline model was estimated. In absolute numbers, this corresponds to a reduction of 112 300 (95 % CI: 50 410-243 400) human salmonellosis cases. Overall, this scenario was estimated to reduce the percentage of human turkey-associated cases from 2.6 % to 0.4 %.

A combined prevalence of the top-6 serovars in turkeys that contribute most to human cases is reduced to 1 % or less in turkey flocks per MSs gave the next largest reduction in the number of turkey-associated human cases of 37.2 % (95 % CI: 19.2-54.0) compared to the baseline model. In absolute numbers, this corresponds to a reduction of 48 110 (95 % CI: 22 580-100 500) human salmonellosis cases. Overall, this scenario was estimated to reduce the percentage of human turkey-associated cases from 2.6 % to 1.7 %.

The least reduction was obtained in the scenario, where the achievement of the current target of the EU control programme of Salmonella in turkey flocks would be met. This analysis resulted in an estimated reduction in the number of turkey-associated human salmonellosis cases of only 0.4 % (95 % CI: 0.1-1.3) compared to the baseline model. In absolute numbers, this corresponds to an estimated reduction of 594 (95 % CI: 121-1 901) human cases. Since, all MSs except one have already met the transitional target, this result is not unexpected.

Several assumptions and factors contributing to the uncertainty and validity of the results are discussed. These include the variability in the human surveillance systems in place in the countries as well as the different details with which serovar information is reported in both humans and animal-food sources. Such uncertainties cannot be statistically quantified, but should be kept in mind when interpreting the results.

The lower attribution estimate obtained for the laying-hen reservoir (i.e. shell eggs) by the TT-SAM model as compared to previous models is supported by data, since both the reported number of cases in EU (particularly S. Enteritidis cases) and the prevalence of Salmonella (particularly S. Enteritidis) in laying hen flocks have been decreasing from 2008 to 2010. The improved surveillance and control of S. Enteritidis in laying hens in many MSs is assessed to be responsible for a major part of this reduction.

The conclusions also emphasise that the reduction of the overall burden of human salmonellosis must be expected to change the attribution estimates, particular the relative estimates, following the logic that if one or more sources contribute significantly less to the overall burden other sources will contribute relative more. The high relative attribution estimate obtained for pig meat by the TT-SAM model, is believed to be partly explained by this.

Despite data limitations and the resulting uncertainty in the results, the source attribution estimates are considered to reflect the best current knowledge about which sources are most important for human salmonellosis in the EU, and highlight differences in the contribution of different food-animal sources for disease and on the efficiency of surveillance systems in place in EU MSs. The results are expected to be useful for the delineation of risk management strategies.

The report concludes with a number of recommendations, one of them being that based on the model results, pig meat is likely to be the most important source in a majority of MSs. Harmonised monitoring and control of Salmonella in pigs and pig meat should therefore be considered.