Food Allergen Recalls – What are the Root Causes and What Can Be Done to Mitigate the Risk? Joe Baumert, Ph.D. Associate Professor Co-Director, Food Allergy Research & Resource Program (FARRP) Department of Food Science & Technology University of Nebraska-Lincoln, USA Midwest AOAC Meeting Lincoln, NE June 6, 2018

Food Allergies Affect 3-4% of the U.S. population 4-8% of children 1-2% of adults Reactions can be life-threatening Reactions are preventable! Small amounts can cause reactions (mg or low ppm total protein) No current treatment: Must follow complete avoidance diets Burden on Consumers & Public Health Concern U.S. childhood food allergy: $24.8 billion annual economic impact © 2018

Reportable Food Registry A web-based portal for submitting a report whenever there is reasonable probability that the use of, or exposure to, an article of food will cause serious adverse health consequences or death to humans Report within 24 hrs of receipt of knowledge that a food product might be involved © 2018

2013-2014 FDA Reportable Food Registry Continued to see an increase in allergen-related entries (47%, up from 43.6% in 2012) Bakery (cookies, muffins & cakes) and chocolate/confections accounted for 37/95 entries (39%) © 2018

FDA Food Allergen Recall Incidents: 1988-2018 FSIS/ USDA Food Allergen Recalls: 1999-2018 * Includes FDA recalls & alerts © 2018

Food Allergen Recalls: Causes Review of FDA allergen recalls FY 2007- FY 2012 (Gendel and Zhu 2013) Identified 732 allergen recalls Gendel, S.M. and J. Zhu. Journal of Food Protection, 2013. 76(11): p. 1933-8. © 2018

FMI Foundation Grant FARRP obtained a small grant from the FMI Foundation to identify the root causes of labeling errors made in food allergen recalls and industry best-practice procedures Project duration: 2.3 months ending October 31, 2017 Obtained allergen recall publically available data from FDA and USDA-FSIS for 2015 and 2016 Surveyed companies involved in allergen recalls for information on root causes and preventive actions Surveyed FARRP member companies on best practices for allergen control © 2018

FMI Foundation Grant Root Cause Survey: 237 surveys emailed October 11, 2017 with an email reminder sent again after one week 90 personal emails 147 generic company emails 4 companies not allowed to fill out survey due to company policy 1 company permanently closed © 2018

FMI Foundation Grant FARRP Member Survey: 199 surveys emailed October 11, 2017 with an email reminder sent again after one week FARRP is a food industry-funded consortium with 93 member companies (at the time); primarily food manufacturers or ingredient suppliers Multiple surveys sent to some companies Some also received Roots survey due to recalls © 2018

2015-2016 FDA Food Allergen Recalls 1,593 total recalls involving 439 companies Unable to determine specific reason for over 700 FDA recalls (details not reported online) Peanut in cumin accounted for at least 323 recalls Peanut in wheat flour accounted for at least 142 recalls © 2018

Food Allergen Recalls: Trends Recent Recall Episodes Issues 2014-2015: Peanut in Cumin Imported, ground cumin with very high levels of peanut Potentially economically motivated adulteration in overseas supply chain Other cumin sources: potential low- level commodity co-mingling 2016: Peanut in Wheat Flour Wheat flour from mill in Georgia with moderate levels of peanut Potential root cause: cross-contact during transportation Supplier verification International supply chains Commodity transportation Commodity comingling © 2018

2015-2016 FDA Regulated Foods Accounting for Recalls © 2018

2015-2016 Reasons for FDA Regulated Food Allergen Recalls © 2018

2015-2016 USDA Food Allergen Recalls 95 total recalls involving 92 companies Unable to identify any true root cause or overall reason for error from USDA data Peanut in cumin accounted for at least 14 recalls © 2018

2015-2016 USDA Regulated Foods Accounting for Recalls © 2018

2015-2016 Reasons for USDA Regulated Food Allergen Recalls © 2018

Root Causes Survey 10 initial respondents, 7 finished entire survey 2 FARRP member companies finished Roots survey Need to continue to seek additional survey completions © 2018

Root Causes Survey – Corrective Actions QC made corrections on how new material is received, labeled and separated Reassess electronic labeling database Better identification of similar recipes Update packaging inspection SOP’s and training to ensure packaging matches allergens Corrective labeling Retrain employees © 2018

FARRP Member Company Survey 44/45 companies have an Allergen Management Program 38/39 companies provide training on allergen awareness to all employees other company provides training to select employees © 2018

Survey Summary Absence of good explanations of reasons for food allergen recalls in many cases in USDA and FDA recall reports Appears varying degrees of information provided by company and/or required by FDA or USDA on reasons for recalls Perhaps not all root causes of recalls can be identified but general reasoning helpful to learn from mistakes of other companies More survey respondents needed © 2018

Preliminary FDA Analysis of Recalls: 2017-2018 No allergen labeling Manufacturer failure to declare all the ingredients in a formulation Food mislabeled as wrong product Allergen cross-contact Jenny Scott (2018); US FDA presentation at Health Canada/FARRP Food Allergen Methodologies Conference, Toronto, Canada. © 2018

International Approaches to Use Thresholds & Risk Assessment © 2018

Current Status of Regulatory Thresholds Public health authorities have not established regulatory action levels for any of the allergenic foods With the exception of Japan (10 µg/g protein limit for labeling) Labeling laws/regulations in many countries impose a zero threshold for source labeling of ingredients Food industry and regulators are acutely aware of allergens How much allergenic residue is too much OR how clean is clean enough?? (Remember it is impossible to assure zero risk with anything in life) With little or no guidance on action levels/thresholds, extensive use of precautionary/advisory labeling (“may contain”) currently exists © 2018

Food Allergen Thresholds Clinical data exist on individual threshold doses of various allergenic foods from oral challenges conducted for diagnosis, threshold trials, and immunotherapy trials – published and unpublished Individual threshold data can be used to statistically model the population threshold distribution for allergenic sources *Note: data from food-allergic individuals rather than extrapolation from animal models as in classical toxicological approaches © 2018

ED10 = 3.8 mg peanut protein ©2018

The VITAL 2.0 ® base on ED01 or lower 95% confidence interval of the ED05 Expected to protect 97-99% of the allergic population ED01 = 0.2 mg peanut protein ©2018

VITAL® Reference Doses Allergen mg Protein Level Peanut 0.2 Milk 0.1 Egg 0.03 Hazelnut Soy 1.0 Wheat Sesame Crustacean shellfish 10.0 Mustard 0.05 Other Tree Nuts Based upon the ED01 for peanut, milk egg and hazelnut Based upon the 95% LCI of ED05 for the remaining 5 allergens Other tree nuts based upon the hazelnut reference dose ©2018

Current Status of Reference Doses Instituted by VITAL in 2012 Endorsed by ILSI-Europe in 2014 and iFAAM in 2017 Overall benchmark approach endorsed by US FDA (2006) and EuroPrevall (2007) Unofficially used by several public health agencies in several countries (UK FSA, CFIA, FDA on case-by-base basis) Similar approach adopted by different branches of the Dutch, Belgian or German authorities albeit each with their own different reference doses Continued questions by regulatory authorities has led to continued datamining and risk assessments for various inputs in the risk assessment ©2018

Quantitative (Probabilistic) Risk Assessment © 2018

Risk Assessment a function of the exposure dose (mg of protein from the allergenic source) compared to the threshold dose (mg of protein from the allergenic source) Quantitative risk assessment can evaluate the risk on an individual or population basis Exposure Dose < Threshold Dose = no predicted reaction Exposure Dose ≥ Threshold Dose = a predicted reaction ©2018

Key Components of a Risk Assessment: Primary Input Parameters: Understanding where UAP may occur Understand your vulnerabilities Tracking allergens Clinical threshold data from low-dose food challenges *Note: data from food-allergic individuals Exposure Assessment Food intake/consumption (g; quantity & frequency) Level of contamination (µg/g or ppm; & frequency) ©2018

Secondary Risk Assessment Input Parameters Prevalence of the Food Allergy Market Share for Specific Product in Category Number of Packages of Food Manufactured Propensity to Buy Advisory Labeled Products ©2018

Quantitative (Probabilistic) Risk Assessment ©2018

Quantitative (Probabilistic) Risk Assessment …… Iteration #5,000,000 Amount 6 g Concentration 1000 ppm Dose 6 mg Threshold 10 mg Iteration #1 Amount 40 g Concentration 3 ppm Dose 0.12 mg Threshold 0.5 mg Iteration #2 Amount 110 g Concentration 30 ppm Dose 3.3 mg Threshold 10 mg Iteration #3 Amount 260 g Concentration 300 ppm Dose 78 mg Threshold 25 mg Iteration #4 Amount 50 g Concentration 10 ppm Dose 0.5 mg Threshold 10 mg Calculate risk of predicted allergic reaction during a single eating occasion (%) ©2018

Exposure Assessment: Food Consumption Data Must ensure that the consumption estimates are reflective of the entire population of consumers portion size mean (average) consumption amount P75 of food consumption distribution maximum consumption amount (very conservative) ©2018

Exposure Assessment: Contamination Data The concentration of allergenic food residue (or protein from the allergenic source) can be determined either by calculation or by quantitative analysis Quantitative analysis commonly conducted on ingredients or finished food products that may contain an unintended allergenic residue Ideally the analytical method used to determine the concentration of the unintended allergic residue would detect proteins from the allergenic source (rather than DNA or ATP) There are important difference in target proteins that are detected and report units (ppm WHAT???) Commodity (e.g. NFDM) Total protein from the allergenic source (e.g. total milk protein) A certain protein fraction from the allergenic source (e.g. casein or whey) A specific allergen (e.g. α-casein or β-lactoglobulin) ©2018

Which ELISA Is More Sensitive? An ELISA with a sensitivity limit of 2.5 ppm NFDM? An ELISA with a sensitivity limit of 1.0 ppm BLG? © 2018

Which ELISA Is More Sensitive? BLG is approx. 10% of total milk protein So, 1 ppm BLG is equivalent to 10 ppm milk protein But milk protein is about 35% of NFDM So, 1 ppm BLG is equivalent to 28.6 ppm NFDM © 2018

Risk Assessment Example 1 RTE Popcorn Calculation of Milk Protein in Ready-To-Eat Popcorn Ingredient % Milk protein in flavor % Flavor in slurry % Milk protein in slurry Milk protein in slurry (ppm) % Slurry on popcorn % Protein in formula ppm allergen in formula Butter Flavor 0.000250% 5.00% 0.000012500% 0.13 16.00% 0.000002% 0.020000 Serving Size ( g) Allergen protein per serving (g) Allergen protein per serving (mg) VITAL action level for milk (mg) Amount of product achieving VITAL action level (g) 31 0.000001 0.00062 0.1 5,000.00 76.3 0.000002 0.001526 © 2018

Quantitative Risk Assessment Consumption Analysis Popcorn Consumption Estimates Using the 2003-2010 NHANES Dietary Surveys   Food Product Category # of Individuals Who Reported Consuming the Product Category Estimated % of U.S. Population that Consume the Product Category Consumption Estimates Per Eating Occasion (g) Average 90th Percentile 95th Percentile 99th Percentile Ready-To-Eat Popcorn 500 1.5% 30 77 83 85 © 2018

Quantitative (Probabilistic) Risk Assessment ©2018

Quantitative Risk Assessment Results ©2018

Quantitative Risk Assessment Results Quantitative Risk Assessment of the RTE Popcorn that Contains 0.02 ppm Milk Protein Residue. Product Allergen Analyzed in Risk Assessment ppm in Protein in Finished Product User Risk Milk Allergic Population Overall Population RTE Popcorn Milk ppm Milk Protein # of Reactions per # of Milk Allergic Users (%) # of Reactions per # of Milk Allergic Consumers (%) # of Reactions per # of Individuals in the U.S. (%) 0.02 ppm 8.0 per 10 million (0.00008%) 1.3 per 100 million (0.0000013%) 1.3 per 10 billion (0.000000013%) *User Risk: assumes that all individuals consuming RTE popcorn are milk-allergic which is a very conservative assumption and likely overestimates the true risk of allergic reaction occurring upon consumption of these products.   **Milk Allergic Population: assumes that 1.5% of the milk-allergic individuals consume RTE popcorn on any given eating occasion. ***Overall Population: assumes that 1.0% of the population is milk-allergic and 1.5% consume RTE popcorn on any given eating occasion. The most sensitive 1% of the milk allergic population would need to consume 5 kg of popcorn during a single eating occasion to reach their threshold dose. This is 58x the consumption of the 99 percentile consumers ©2018

Conclusions QRA and reference doses could be used advantageously by regulatory authorities to determine: determine which ingredients from allergenic sources need to be labelled by source (FALCPA notifications) curtail excessive use of precautionary labelling determine the degree of risk posed by undeclared allergens in recall situations assess the effectiveness of preventive allergen controls (FSMA) © 2018

Thank You for Your Attention Joe Baumert, Ph.D. Food Allergy Research & Resource Program Department of Food Science & Technology University of Nebraska jbaumert2@unl.edu farrp.org © 2018