Validating Thermal Process Lethality in Low Moisture Food Approaches to Modeling Lisa A. Lucore, Ph.D. The Kellogg Company, Battle Creek, MI 49017 GMA Science Forum - April 4, 2012 1
Topics in Brief Why is lethality in Low Aw foods needed? Variables that influence lethality in Low Aw foods? Creating a robust predictive model Framework Significant variables Choices for underlying calculation process 2
Why Lethality Models for Low Aw Food? Count reduction using surrogate or target organism is simpler in principle, but may be challenging to execute. Changes to process parameters may require new test. May be cost prohibitive for many. Development of model to calculate lethality within process allows changes to the process to be analyzed and predicted. Robust if designed effectively and validated. Food forms of comparable composition may be able to utilize similar models for specified pathogens. 3
Limited Information Available Few published D & z values relevant to low Aw foods: TDT tubes in water Animal feed Liu, et. al, 1969 TDT tubes in oil Chocolate syrup Sumner, et. al, 1991 Mixer cup in oil bath Milk chocolate Goepfert and Biggie, 1968 Swept surface kettle Milk chocolate Barille & Cone, 1970. Open dishes Flour Archer, et. al, 1998 May have test parameters different from desired production environment, different test substrate, etc. Studies underway for some commodities: Almonds, Wheat Kernels, Dates, Powders & Pastes Marks, et. al, 2011 Peanut Paste & Oil GMA, 2011 Individual companies work on data collection but may not have clear avenue to share findings. 4
Format: Model Structure: Overview Collect TDT (D & z) values in sealed system. Prevents loss of moisture in test; which may skew results of TDT Use general method calculation such as American Meat Institute Model (AMI, 2010), but adjust D&z values based on Aw changes in thermal process. This transitions the model from calculating F-value to Log reduction Primary Variables in Model: Internal Food Temperature, Dwell Time, Aw Model accuracy requires precise measurement of primary variables, understanding of heat penetration of the food and temperature mapping of the thermal system. If non-homogeneous system, determine worst case. (coldest food temp, shortest time, lowest Aw) 5
D & z Values Defined - D-value: the time required to reduce by 90% (1 log cycle) the population of the target microorganism at a reference temperature - z-value: the number of degrees of temperature required for the TDT curve to traverse one log cycle z=(t 2 -T 1 )/(logd 1 -D 2 ) Example: (D 185 F = 10.0 minutes, z = 10 F ). 6
General Method Calculation Used in conventional thermal processing since 1920 Relies on heat-penetration data and TDT data 1. Lethal rate: 2. Lethality: 3. Cumulative log reduction: (T Tref)/z L = 10 Lethality = L Δt Σ [( 10 (T Tref)/z )Δt]/D or Σ (L Δt)/D D = # min required for 1-log (90%) reduction of the target organism t = time (min) T = actual temperature in the heat penetration study Tref = reference temperature of the TDT study z = # of degrees (C or F ) for TDT results to move 1 log cycle. 7
Express TDT in Weibull or Linear? Linear: dn/dt=-kt Weibull: Log S = -bt n N= # living cells after exposure at time t k= rate constant b= scale, n= shape (when n=1, constant rate linear) Weibull solved for n determines if shape significantly different from linear. Linear distribution is special case of Weibull where n=1. n>1 = underestimate time and temp needed. n=1 or n<1 a linear distribution is most conservative estimate of TDT. 8
Log reduction Linear is Best Fit for TDTs short processing times Little to no curve Linear fit is less sensitive at ends/tails 0.85 Aw 0.57 Aw 0.21 Aw Time 9
Thermal Death Time Test Results Of multiple strains tested, S. Tennessee slightly more heat resistant than others, thus S. Tennessee selected for model (Lucore et al, 2011). TDT within Cereal Food Form (Flour, Solute, Water) Strain Aw z-value ( C) D Value ( C) Salmonella Tennessee 0.96 6.9 10^(-0.1444T*+20.763) Salmonella Tennessee 0.85 7.4 10^(-0.1346T*+10.543) Salmonella Tennessee 0.57 10.8 10^(-0.0921T*+8.0944) Salmonella Tennessee 0.21 11.5 10^(-0.0873T*+9.0586) * T = Temp in C (e.g. D 93C (200F) @.21 Aw = 8.1min) 10
Another way of viewing TDT Results D-Values at given Aw Temp (C) Temp (F) 0.21 0.57 0.85 0.96 27 80 5377742 434911 8988075 1613119966 38 100 576235 41216 287152 2083425 49 120 61745 3906 9174 2691 60 140 6616 370 293 3.48 71 160 709 35 9 0.00 82 180 76 3.32 0.30 0.00 93 200 8.14 0.32 0.01 0.00 104 220 0.87 0.03 0.00 0.00 116 240 0.09 0.00 0.00 0.00 11
D value (minutes) Thermal Death Time Studies 2.5 D values of S. Tennessee in Cereal Food Forms at Different Aw (NFL, 2010) 2 1.5 1 0.5 2.13 Milled cereal Finished cereal Cereal dough At 100C(212F), 2000 times longer to kill the same population when comparing.21 &.85 Aw food. 0 0.075 0.001 0 0.2 0.4 0.6 0.8 1 Aw
1 10 20 30 40 50 60 70 80 90 100 110 120 F-value (min) 1 10 20 30 40 50 60 70 80 90 100 110 120 Temperature (F) Date: PROCESS LETHALITY DETERMINATION Organism: _Salmonella Product name: Example EXAMPLE: Lethality Data from Literature Microbial Heat Tolerance T ref z D User Must: Organism Product ( ºF) ( ºF) (min) 1. Identify organism and product of concern Salmonella Meat Patty (Scott and Weddig, 1998) 150 10 0.172 2. Provide at least 20 time/temp data points Gr. Beef (25% fat) (Juneja, 2003) 140 46.51 19.31 E. coli O157:H7 Lean Gr. Beef (2% fat) (Line et al., 1991) 145 8.3 0.30 Instructions: Gr. Beef (25% fat) (Juneja, 2003) 140 43.39 20.90 1. Select the organism and product of concern Lean Gr. Turkey (Juneja and Marmer, 1999) 149 43.7 1.45 and identify corresponding T ref, z, and D Lean Gr. Lamb (Juneja and Marmer, 1999) 149 44.4 1.90 values in the table. These values should be Lean Gr. Pork (Juneja and Marmer, 1999) 149 43.7 1.60 obtained from your own companies challenge Listeria Lean Gr. Beef (2% fat) (Fain, et al., 1991) 145 9.3 0.6 study data, from scientific literature, or other monocytogenes Gr. Beef (25% fat) (Juneja, 2003) 140 42.98 27.7 reliable sources. These values need to be Hot Dog Batter (30% fat) (Mazzotta and Gombas, 2001) 160 11 0.5 relevant and appropriate for the type of product Note: This model is a tool for calculating F-values. To ensure correct results, the proper z, T-ref, and and the organism of concern. D-values for each product and organism must be used. 2. Enter the T ref, z, and D values into the appropriate labeled cells below the table that z = 10 ºF T ref= 150 ºF contains the lethality data from literature. 3. Clear and enter at least 20 time/temp data D = 0.172 min points into the data table. 4. Once the table is completed, a cumulative F value will be given as the very last number in the right hand column of the data table. This Model Structure: Modified Method Example of AMI* format: Example of Modified format: Log Reduction of Process = 116577.95 Data Table number adds up the lethality values for each Time (min) Core Temp (ºF) F-value (min) time interval and calculates an approximation 0 40 0.000 of the area under the lethal rate curve. 5 49 0.000 5. After the data is entered, a core 10 58 0.000 200 150 100 50 0 Core temperature 0 50 100 150 Time (min) Time (s) Aw Temp (ºc) D Log red 1 0.8 29.40 32366.97 0.00 10 0.76 43.30 2399.55 0.00 20 0.74 68.30 14.47 0.01 30 0.69 87.80 0.36 0.24 40 0.63 98.80 0.06 1.68 50 0.58 101.60 0.05 3.40 60 0.53 104.40 0.04 5.46 70 0.47 105.70 0.05 7.10 80 0.42 105.60 0.08 8.16 90 0.35 105.70 0.14 8.77 100 0.31 104.20 0.26 9.08 110 0.25 104.10 0.44 9.27 120 0.2 82.20 68.16 9.27 10.00 8.00 6.00 4.00 2.00 0.00 Cumulative Log Lethality Across Thermal Process temperature and lethality curve are produced. 15 67 0.000 6. The total log reduction of the process is 20 76 0.000 automatically determined by dividing the 25 85 0.000 cumulative F value by the D value that was 30 94 0.000 entered into the appropriate labeled cell. The 35 103 0.000 resulting value equals the total log reduction 40 112 0.001 of the process. 45 121 0.004 7. By using these estimates, you or a process 50 130 0.032 authority should determine if the process 55 139 0.256 meets regulatory requirements as safe. 60 148 2.032 Additional documents, such as Appendix A, 65 157 16.139 which discuss desired log reductions should 70 166 128.195 also be considered when evaluating a lethality 75 175 1018.292 process. 80 184 8088.577 85 180 16868.293 90 170 19618.293 95 165 19947.350 100 160 20051.407 30000.000 20000.000 10000.000 0.000 Lethality 0 100 200 Time (min) 120.00 100.00 80.00 60.00 40.00 20.00 0.00 Change in Food Temp (C) Across Thermal 13
Step 1 (wet) Using D-values 1. Step-by-Step = holding D value constant through each Aw. - Limitation - Likely to overestimate results. HEAT STEP A B DWELL FOOD TIME TEMP ( F/ C) (MIN) C WATER ACTIVITY IN D WATER ACTIVITY OUT E Data Used for D value calculations Dwell time at Aw F Calculated Log D (Time in minutes to = 1 log kill) G D reduction (# log reduced based on dwell time) COOK IN 80/27 45 0.58 aw =.58 Tr = 72 Dr = 45 z = 30.3 22.5 1375.2 0.0 COOK OUT 225/107 45 0.91 T r = 65.5 aw = 0.89 D r = 4.8 z = 6.9 22.5 0.0 10431531.8 A The product/food temperature reported from the facility B The dwell time within the heat step divided by 2 to split dwell time between entering and exiting water activity C Entering water activity of food D Exiting water activity of food E Reference Information Collected from Archer et al., 1998 and Sumner et al. 1991, respectively F Time in minutes to enable 1 log kill at temperature and water activity reported G # of Logs of kill achieved based on reported dwell time within the reported termperature and water activity Log D = Log (D R )+(T R -T A )/Z R 14
Log reduction Using D-values 2. Regression = use TDT results to create prediction formula and smooth the results across the thermal process. Using regression from results of individual TDTs: Change in D-value at a given temperature as Aw changes. 0.85 Aw 0.57 Aw 0.21 Aw Time 15
2. Regression Using D-values - Limitation curve may extrapolate incorrectly at temperatures that exceed the parameters of the TDT tests. 212F/100C 5 log Time.8 Aw.5 Aw.2 Aw 16
Effect of Different D-Value Options * Representative Time & Temp Time (s) Aw Core Temp D Regression 10 0.80 85F (29.4C) 33175.11 0.000 20 0.74 110F (43.3C) 2771.42 0.000 30 0.69 155F (68.3C) 22.18 0.008 40 0.63 190F (87.8C) 0.57 0.298 50 0.58 210F (98.8C) 0.09 2.233 60 0.53 215F (101.6C) 0.08 4.452 70 0.47 220F (104.4C) 0.07 6.998 80 0.42 220F (104.4C) 0.10 8.622 Time (s) Aw Core Temp D Step-by-step 10 0.80 85F (29.4C) 559113.53 0.000 20 0.74 110F (43.3C) 20183.66 0.000 30 0.69 155F (68.3C) 87.10 0.002 40 0.63 190F (87.8C) 0.81 0.207 50 0.58 210F (98.8C) 0.06 3.171 Step-by-Step determination of D-Value 60 0.53 215F (101.6C) 0.03 8.950 70 0.47 220F (104.4C) 0.01 20.218 80 0.42 220F (104.4C) 0.01 31.486 90 0.36 215F (101.6C) 0.29 9.199 Regression determination of D-Value 100 0.31 210F (98.9C) 0.81 9.404 110 0.25 200F (93.3C) 4.12 9.444 120 0.20 180F (82.2C) 67.84 9.447 90 0.36 215F (101.6C) 0.03 37.265 100 0.31 210F (98.9C) 0.06 40.229 110 0.25 200F (93.3C) 0.21 41.008 120 0.20 180F (82.2C) 3.09 41.062 * Regression is considered best case as it calculates a more conservative result across changing Aw and temperature. 17
0.80 0.69 0.58 0.47 0.36 Log Kill Suggested Best Practice for Use of D-value 3. Controlled Regression = prediction formula holds values constant beyond parameters of lab TDTs to prevent over-estimation of lethality. - Additionally, if results exceed 10 log, report as >10 as is difficult to accurately quantify higher results. 12.000 10.000 8.000 6.000 Regression of all linear TDT results to address changing Aw across process time 4.000 2.000 0.000 Temp (C) 98.89 Aw 87.78 18
Effect to Model of Precise Aw Data Linear reduction between points of measurement provides most conservative result. 1 0.9 0.8 0.7 0.6 0.5 0.4 Food A Food B Food C 0.3 0.2 0.1 0 0 min 2 min 4 min 6 min 8 min 19
Effect to Model of Precise Aw Data Additional data points collected improves precision. Time (s) Aw Core Temp Lethality Time (s) Aw Core Temp Lethality 10 0.80 85F (29.4C) 0.000 20 0.80 110F (43.3C) 0.000 30 0.79 155F (68.3C) 0.018 40 0.75 190F (87.8C) 0.781 50 0.71 210F (98.8C) 6.500 60 0.68 215F (101.6C) 14.511 70 0.57 220F (104.4C) 20.288 80 0.46 220F (104.4C) 22.605 90 0.36 215F (101.6C) 23.182 100 0.31 210F (98.9C) 23.387 110 0.25 200F (93.3C) 23.427 Linear Decline between measured Aw Points 10 0.80 85F (29.4C) 0.000 20 0.74 110F (43.3C) 0.000 30 0.69 155F (68.3C) 0.008 40 0.63 190F (87.8C) 0.298 50 0.58 210F (98.8C) 2.233 60 0.53 215F (101.6C) 4.452 70 0.47 220F (104.4C) 6.998 80 0.42 220F (104.4C) 8.622 90 0.36 215F (101.6C) 9.199 100 0.31 210F (98.9C) 9.404 110 0.25 200F (93.3C) 9.444 120 0.20 180F (82.2C) 23.430 120 0.20 180F (82.2C) 9.447 20
Example of a Log-Reduction Calculation Baking process is divided into 4 segments. TDT study data is applied within each segment. D-values increase through the process. Lethality during cooling is not calculated. 21
Mathematical Modeling Segment Time (min) Aw Temp (ºC/ºF) D - Value Log red 1 1 0.75 65/150 26.73 0.04 2 1 0.74 85/175 0.43 2.35 3 1 0.35 100/195 0.46 2.18 4 1 0.21 110/215 0.18 5.64 Total 10.20 Log reduction: 2.4 2.2 5.6 22
Summary of Model Recommendations Test multiple strains of target organism Best choice should be selected from those found within food form or ingredients within food form. Maintain stable Aw during TDT tests within appropriate food form. Provides multiple snapshots of thermal effect at each Aw sampled in process. Calculate lethality using Controlled Regression to create most conservative result. Linear reduction of Aw between each measured Aw point of process. Worst case process variables selected for calculating lethality in thermal system 23
References American Meat Institute Foundation. 2010. AMI Foundation Process Lethality Determination Spreadsheet Aug 2010 http://amif.org/ht/d/sp/i/26870/pid/26870/ Anderson, D. G. and L. A. Lucore. 2012. Validating the Reduction of Salmonella and Other Pathogens in Heat Processed Low-Moisture Foods. Alliance for Innovation & Operational Excellence, Alexandria, VA. Published online at http://community.pmmi.org/alliance/home/. Accessed [scheduled for April, 2012]. Archer J., E. T. Jervis, J. Bird, and J. E. Gaze, 1998. Heat resistance of Salmonella weltevreden in low-moisture environments. J. Food Prot. 61: 969-973. GMA, 2011. Summary of GMA Scientific and Regulatory Affairs Projects 2010, "Thermal Inactivation and Survival of Salmonella in Food as a Function of Water Activity and Fat Level". p. 8. Grocery Manufacturers Association, 1350 I Street, NW, Suite 300, Washington, DC 20005. www.gmaonline.org 24
References Goepfert, J. M., and R. A. Biggie. 1968. Heat resistance of Salmonella typhimurium and Salmonella senftenberg 775W in milk chocolate. Applied Microbiology 16:1939-1940. Liu, T. S., G. H. Snoeyenbos, and V. L. Carlson. 1969. Thermal resistance of Salmonella senftenberg 775W in dry animal feeds. Avian Diseases 13:611-631. Lucore, L. A; M. A. Moorman, B. L. S. Jackson, 2011. Lethality Validation of Thoroughly Cooked Products: A Dry Foods Toolbox. IAFP annual meeting, August 2, 2011. Milwaukee, WI. Marks, B.P., J. Tang, E. T. Ryser, S. Wang and S. Jeong, project directors. 2011. Improving process validation methods for multiple pasteurization technologies applied to low-moisture foods. USDA project number MICL05056. Biosystems & Agricultural Engineering, Michigan State University, East Lansing, MI 48824. Sumner, S., T. M. Sandros, M. C. Harmon, V. N. Scott, and D. T. Bernard. 1991. Heat resistance of Salmonella typhimurium and Listeria monocytogenes in sucrose solutions of various water activities. J. Food Sci. 6:1741-1743. 25
Thank You Lisa Lucore, Ph.D. Food Safety Scientist The Kellogg Company 26