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