Risk-based Sampling
Perspective from Different EU and Non-EU Member States
Jenny Scott: Senior Advisor to the Director of the Office of Food Safety at FDA’s Center for Food Safety and Applied Nutrition.
1
Annemarie Pielaat: Senior researcher at the Dutch Institute for Public Health and the Environment (RIVM)
Akos Jozwiak: Vice-director at the Hungarian National Food Chain Safety Office
Risk Based Sampling | May 2016
The presenters ● Akos Jozwiak:
– Head of the coordination working group on planning of sampling plans, supervising different sampling programs from soil and plant protection to plant and animal health, feed and food domains. – Presentation: theoretical background of the planning process at central authority level with practical and organizational examples, problems identified and solutions provided recently.
● Jenny Scott: – Technical lead of the team that developed the new Preventive Controls for Human Food rule, with a risk-based approach to control of food hazards. – Presentation: risk-based microbial sampling by industry as part of a food safety system. FDA’s approach to its risk-based sampling including data use. (prepared in conjunction with Dr. Yuhuan Chen (RA at FDA).
● Annemarie Pielaat: – Project manager for different research projects in food safety RA. – Presentation: Development of a risk based sampling program for the Netherlands Food and Consumer Product Safety Authority (NVWA) 2
Risk Based Sampling | May 2016
Risk Based Sampling A public health perspective from The Netherlands Annemarie Pielaat
Jurgen Chardon, Lucas Wijnands, Ellen Delfgou, Angelina Kuijpers, Aarieke de Jong, Eric Evers Risk Based Sampling | May 2016
Introduction ● Netherlands Food and Consumer Product Safety Authority (NVWA) ● Monitoring program occurrence microorganisms on food products ● Insight in public health risk caused by foodborne pathogens Optimization of program through risk based sampling
4
Risk Based Sampling | May 2016
Approach ● Optimization of monitoring foodborne pathogens for public health risk estimates ● Input variables: – Prevalence, concentration microorganisms – Consumption figures food products & food handling by consumers – Disease burden – Sampling costs ● Definition of Risk based sampling
Distribute the NVWA sampling capacity over retail products proportional to their contribution to the DALY’s* *DALY:
Disability Adjusted Life Year (Havelaar et. al, 2012) is a consistent, quantitative measure to compare the disease burden of different foodborne pathogens 5
Risk Based Sampling | May 2016
Distribute the NVWA sampling capacity over retail products proportional to their contribution to the DALY’s
Methodology A. Exposure assessment
B. Attribution to DALY’s
Pathogen prevalence and concentration at time of purchase, food preparation and amount consumed
Distribute DALY’s on pathogen-product group over products proportional to exposure
C. Sampling
D. Costs
Initial prevalence to assess the number of samples needed for each product
Personnel and analysis costs
Optimizing criterion: (C * D) / B(A) 6
expressed as costs per DALY
Risk Based Sampling | May 2016
Case studies Initial pathogen product combinations of interest to NVWA ● ● ● ●
7
Salmonella on pork meat Campylobacter on pork meat and on poultry meat Toxoplasma on pork meat Shiga toxin-producing Escherichia coli (STEC) O157 on beef
Risk Based Sampling | May 2016
Exposure Assessment ● Based on swift QMRA tool with deterministic calculations (Chardon and Evers, 2016)
● Changes in numbers of micro-organisms on food products Retail Storage Preparation Exposure
Chardon and Evers 8
Risk Based Sampling | May 2016
Food product
- Consumption data - prev/conc pathogens
Pre - retail processing
-reduction due to smoking/salting/drying/cooking/heating
Consumer storage
Cross cont amination
Heating
- fractions room/fridge/freezer - storage times - growth/inactivation
Using fit-for-purpose models from sQMRA •
Time / temp dynamics
•
D/z inactivation models
•
…
- CC fractions - CC transfer rate
-Raw consumed products - Heating Model - fraction meat balls/ raw / medium / done - D/z inactivation data - heating time and temp
Exposure 9
No. contaminated portions at consumption and total no. of micro-organisms ingested (pop.n year-1) Risk Based Sampling | May 2016
Attribute DALY’s to Exposure
Food product
Exposure*
DALYs
Pork, chops, rib
1.6E+03
6.1E+01
Ham, raw
1.1E-01
4.3E-03
Bacon, smoked
1.9E-02
3.7E-03
Sausage, BBQ
1.8E+02
6.9E+00
…
…
…
*No. 10
contaminated portions at consumption and total number of micro-organisms ingested (pop.n year-1 ) Risk Based Sampling | May 2016
Sampling ● NVWA does not monitor DALYs but prevalences for pathogen-product combinations Mean pathogen prevalences (%) in the Netherlands in retail 2010-2013. Pathogen / food animal Fresh Meat Source combination meat product STEC beef&veal 1,31 3,88 A STEC mutton/lamb 6,69 3,44 A Salmonella pork 0,97 1,29 A B Campylobacter pork 0,37 0,16 A C C Toxoplasma pork 0,90 0,90 Zomer et al. (2014) Campylobacter chicken 27,38 11,60 Zomer et al. (2014) B A: data from the NVWA. : no measurements available, calculated using ratio for chicken (0,37*11,6/27,38); C: based on serological prevalence in slaughter house.
11
Risk Based Sampling | May 2016
● Link Sampling with Prevalence
Sampling PATH.-PRODUCT Campy-pork Campy-pork Campy-pork Campy-pork Campy-pork Campy-pork Campy-poultry Campy-poultry Campy-poultry Salm-pork Salm-pork Salm-pork STEC-beef STEC-beef STEC-beef Toxopl-pork Toxopl-pork Toxopl-pork Toxopl-pork … 12
Exposure RETAIL PRODUCT NO PORTIONS DALYs PREVALENCE Pork, spare rib (SH 786) 2,5E+06 0,0 0,37% Pork, fricandeau 1,1E+07 0,0 0,37% Pork, silverside 5,7E+07 0,1 0,37% "rotterdammertje" 7,5E+05 0,1 0,16% Salami 2,3E+08 29,7 0,16% "Worst, met" 5,9E+07 8,1 0,16% Chicken n.s. 5,0E+07 15,3 27,38% Chicken, wing 2,7E+07 16,2 27,38% Chicken, leg 4,6E+07 46,4 27,38% Dried sausages n.s. 4,8E+07 5,3 1,29% Cervelaat 2,7E+08 13,2 1,29% "rotterdammertje" 7,5E+05 0,1 1,29% Salami 2,3E+08 0,9 3,88% Roasted meat, cold n.s. 1,3E+07 0,0 3,88% Filet americain 3,2E+08 12,8 3,88% "Worst, paling" 4,8E+07 0,0 0,90% Dried sausages n.s. 4,8E+07 76,4 0,90% Cervelaat 2,7E+08 185,0 0,90% "Worst, chorizo" 2,0E+07 28,3 0,90% … … … …
● How many samples are needed to estimate prevalence within defined uncertainty ● Basic principle: Numbers based on one positive sample Number of samples with number of positives Uncertainty about prevalence estimate slowly
Risk Based Sampling | May 2016
Sampling
PATH.-PRODUCT Campy-pork Campy-pork Campy-pork Campy-pork Campy-pork Campy-pork Campy-poultry Campy-poultry Campy-poultry Salm-pork Salm-pork Salm-pork STEC-beef STEC-beef STEC-beef Toxopl-pork Toxopl-pork Toxopl-pork Toxopl-pork …
13
k
RETAIL PRODUCT NO PORTIONS DALYs PREVALENCE SAMPLES Pork, spare rib (SH 786) 2,5E+06 0,0 0,37% 540 Pork, fricandeau 1,1E+07 0,0 0,37% 540 Pork, silverside 5,7E+07 0,1 0,37% 540 "rotterdammertje" 7,5E+05 0,1 0,16% 1250 Salami 2,3E+08 29,7 0,16% 1250 "Worst, met" 5,9E+07 8,1 0,16% 1250 Chicken n.s. 5,0E+07 15,3 27,38% 6 Chicken, wing 2,7E+07 16,2 27,38% 6 Chicken, leg 4,6E+07 46,4 27,38% 6 Dried sausages n.s. 4,8E+07 5,3 1,29% 153 Cervelaat 2,7E+08 13,2 1,29% 153 "rotterdammertje" 7,5E+05 0,1 1,29% 153 Salami 2,3E+08 0,9 3,88% 50 Roasted meat, cold n.s. 1,3E+07 0,0 3,88% 50 Filet americain 3,2E+08 12,8 3,88% 50 "Worst, paling" 4,8E+07 0,0 0,90% 220 Dried sausages n.s. 4,8E+07 76,4 0,90% 220 Cervelaat 2,7E+08 185,0 0,90% 220 "Worst, chorizo" 2,0E+07 28,3 0,90% 220 … … … … …
n
1 1 1 1 1 1 1 1 1 1 1
Risk Based Sampling | May 2016
d
1250 540 220 205 153 151 56 50 28 16 6
0,16% 0,37% 0,90% 0,97% 1,29% 1,31% 3,45% 3,85% 6,67% 11,11% 25,00%
Pathogen/ Product
Costs
What is the best sampling strategy from a public health perspective within this budget?
Fresh meat
Pork
17,-
Fresh meat
Poultry
50,-
Meat product
Pork
15,-
Meat product
Poultry
30,-
Fresh meat
Beef/Veal
25,-
Meat product
Beef/Veal
37,-
Fresh meat
Mutton/Lamb
52,-
Meat product
Mutton/Lamb
35,-
Fresh meat
Pork
20,-
Meat product
Pork
21,-
Pork
100,-
STEC
● Catch maximum number of DALYs for minimum budget
● Optimization criterium:
Salmonella
Toxoplasma Fresh meat
#Samples • Costs DALYs 14
Costs*
Campylobacter
● NVWA budget: e.g. €225,000.-
● Sort on Costs per DALY (min
Source
* € Per sample, incl admin
max) Risk Based Sampling | May 2016
Results Top 26 pathogen-product combinations for sample analysis
TOP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
maximised for DALYs as measure for public health risk
15
ID 149 146 147 155 151 150 145 144 148 157 156 154 500 497 502 208 265 377 501 496 153 262 159 443 267 266
foodnum 0863 0860 0861 0870 0865 0864 0859 0857 0862 2667 0871 0868 0970 0966 0972 0989 0970 1012 0971 0965 0867 0966 0869 0989 0972 0971
pathogen/matrix 1. Campy-kip 1. Campy-kip 1. Campy-kip 1. Campy-kip 1. Campy-kip 1. Campy-kip 1. Campy-kip 1. Campy-kip 1. Campy-kip 1. Campy-kip 1. Campy-kip 1. Campy-kip 2.Toxopl-varken 2.Toxopl-varken 2.Toxopl-varken 2. Salm-varken 2. Salm-varken 2. STEC-rund 2.Toxopl-varken 2.Toxopl-varken 1. Campy-kip 2. Salm-varken 2. Campy-kip 2.Toxopl-varken 2. Salm-varken 2. Salm-varken
Aantal monsters 6 6 6 6 6 6 6 6 6 6 6 6 220 220 220 153 153 50 220 220 6 153 16 220 153 153
Product VCP-product DALY's Prevalentie Kip filet 579,28 27,38% Kip poot 46,41 27,38% Kip drumstick 41,69 27,38% Kip braadstuk 39,50 27,38% Kip half 21,56 27,38% Kip heel 17,22 27,38% Kip vleugel 16,25 27,38% Kip n.s. 15,26 27,38% Kip borst met bot 13,34 27,38% Kip reepjes 11,54 27,38% Kip karbonade 8,57 27,38% Kip TV stick 6,48 27,38% Salami * 374,59 0,90% Cervelaat 185,02 0,90% Worst snij 145,08 0,90% Ham coburger 28,92 1,29% Salami 26,74 1,29% Filet americain 12,78 3,88% Worst met 102,26 0,90% Worst gedroogd n.s. 76,41 0,90% Kip soep 1,12 27,38% Cervelaat 13,21 1,29% Kip rollade 1,85 11,60% Ham coburger 55,56 0,90% Worst snij 10,25 1,29% Worst met 7,30 1,29% *Net *
Totale kosten € 505 € 505 € 505 € 505 € 505 € 505 € 505 € 505 € 505 € 505 € 505 € 505 € 30.283 € 30.283 € 30.283 € 6.209 € 6.209 € 3.255 € 30.283 € 30.283 € 505 € 6.209 € 906 € 30.283 € 6.209 € 6.209
Kosten/ DALY 8,72E-01 1,09E+01 1,21E+01 1,28E+01 2,34E+01 2,93E+01 3,11E+01 3,31E+01 3,79E+01 4,37E+01 5,89E+01 7,79E+01 8,08E+01 1,64E+02 2,09E+02 2,15E+02 2,32E+02 2,55E+02 2,96E+02 3,96E+02 4,50E+02 4,70E+02 4,89E+02 5,45E+02 6,06E+02 8,51E+02
Cumulatieve Cumulatieve aantal kosten monsters € 505,0 6 € 1.010,1 12 € 1.515,1 18 € 2.020,1 24 € 2.525,2 30 € 3.030,2 36 € 3.535,2 42 € 4.040,3 48 € 4.545,3 54 € 5.050,3 60 € 5.555,3 66 € 6.060,4 72 € 36.343,4 292 € 66.626,4 512 € 96.909,4 732 € 103.118,6 885 € 109.327,9 1038 € 112.583,1 1088 € 142.866,1 1308 € 173.149,1 1528 € 173.654,2 1534 € 179.863,4 1687 € 180.769,8 1703 € 211.052,8 1923 € 217.262,0 2076 € 223.471,3 2229
number of samples: 1894
Risk Based Sampling | May 2016
Trend analysis ● Minimum no. samples in two consecutive sampling events (n1=n2) to identify a significant difference in prevalence (d1 and d2) with certain probability (power of the test) H0: d1 = d2 Ha: d1 ≠ d2 ● Power: P(reject H0 in favour of Ha when Ha is true) depends on: – Actual difference between d1 and d2 – Significance level, α and – Number of samples (n)
16
Risk Based Sampling | May 2016
Trend analysis d1
d1 d2
d2
d1
d1 d2
17
d2
Risk Based Sampling | May 2016
Conclusion The Netherlands: ● If €225,000.● n ~2000 samples over ~20 different meat products ● Top 12: Campylobacter on different poultry products – Chicken filet no. 1 product ● Not sufficient for trend analysis in two consecutive years – Power of the test ≤ 55% ● Consistent iterative implementation of risk based sampling plan – Monitor pathogen prevalence based on public health risk – Adapt intervention programs based on public health risk – Trend analysis 18
Risk Based Sampling | May 2016
Discussion ● Extend for other pathogen-product combinations – Next important DALY
● Update prevalence estimates – Use of molecular methods for monitoring? (reduce costs) ● Improve exposure assessment parameter estimates – Pathogen survival on meat (STEC on beef)
19
Risk Based Sampling | May 2016