r""-

,
"

--~-

f~- ~ "'"": -

Workshop Targets

Continuous Multiphase Aseptic Processing

of Foods
CONTENTS:
Continuous Multiphase Aseptic Processing of Foods .43
John W. Larkin

Measuring Residence Time and Modelling

the System Sudhir K. Sastry \ 44

I
,\

Biological Validation Joseph E. Marcy

48

i
I
,~

Statistical Design and Analysis ..................... 52

Michel Digeronimo, Wallace Garthright, and John Larkin

Issues Involved in Producing a Multiphase Food Product

Dominick Damiano

'56

Reprinted from Food Technology 51 (10) 43';62 @Institute of Food Technologists

!!

. 'I' .

1-

I
"II

I .
:1

5 PEe

C ~.T

Workshop Targets

'

Contiriuous l1ultiphase. AseptIc Processmg

Working groups discuss concerns that needed to be resolved to develop aseptic processes for foods containing particulates
JOHN W. L.ARKIN
The author. a Professional Member of 1FT,is Food Process Hazard Analysis Branch Chief, NaTIonal Center for Fcod Safety and TechnolOgy. Food and Drug AdministraTIon, 6502 S. Archer Rd, Summit-Argo, IL 60501 ,

f Foods
bution (the procedure used to determine the particle receiving the minimum thermal treatment), and validation of the model used to establish the process. What Pushed the Industry Forward? The continued interest in these types of processes, the limited number of filings, and the difficulty of documentCream of potato soup is ing the lethal treatment typical of the products delivered to each portion this workshop addressed. of the food product resulted in uncertainty in the food processing community as to what was needed to establish a process (Dal Porto, 1993). In 1995, in response to these concerns, the National Center for Food Safety and Technology (NCFST) and the Center for Aseptic Processing and Packaging Studies (CAPPS) organized a workshop for experts in aseptic processing of foods. The objective of the workshop was to establish a dialogue and a mechanism for exchange on issues surrounding the application and implementation of aseptic processing of multiphase foods, and to come to a consensus on these issues. Representatives of 11 universities, 5 government agencies, 22 food companies, and NFPA participated in the workshop, which was held on November 14-15, 1995, and March 12-13, 1996.The organizers wished to address the immediate con. . cerns that needed to be resolved to develop an aseptic multiphase food process. Th'e topic areas discussed by working groups were particle resi.. dence time, mathematical modeling, physical property measurement procedures, biological val'idation, and statistical design and analysis of data. The workshop also included a discussion of a case study developed by members of NFPA,
f'OODTECHNOLOGY 43

F
product.

or many years, food processors have tried to develop an aseptic process for

food products containing particles. The development of such a process has been hindered by the requirement to demonstrate an adequate thermal treatment for every portion of the

During the past 13 years, the Food and Drug Administration received only a few filings for aseptically processed low-acid canned foods containing particles. Two filings received in the midto-late 1980s generated many questions within FDA concerning process establishment methodology, test results, and the identification and control of certain critical factors. FDA had several discussions with each firm about its process and how it was established. For reasons unknown to FDA, both firms decided to discontinue their involvement in aseptic multi phase food processing projects and withdrew the filings from further consideration. In 1989, FDA identified several issues that it expected a food processor to address when developing a filing (Dignan et aI., 1989). What has proved to be the most difficult is the requirement to biologically validate the thermal treatment delivered by the system. The National Food Processors Association (NFPA) made a concerted effort to develop protocols for the establishment of an aseptic multiphase food process (Chandarana and Unverferth, 1996). In the early 1990s, a food industry consortium formed by NFPA commissioned two research projects to develop procedures that could be used to collect the necessary information to document an adequately scheduled process. Lengthy discussions between FDA and NFPA took place during the project. At the project's end, FDA was still unable to resolve several fundamental issues (Anonymous, 1995). In particular, questions. or concerns remained regarding biological validation (number of particles used and how this information would be used), residence time distri- .

VOL. 51, NO.10 -OCTOBER 1997

'If

S PEe

L. 5 E C T

AseRtic

Workshop Go N TIN U E 0

j'
I:

CAPPS, and FMC Corp. The case s udy was developed to try to bring the discussions of the workshop "down to eahh." The fictitious food product used in!lthe case study was cream of potato soup concentrate. The case study illustrated the types of issues that a processor bay need to consider when developing ~ scheduled thermal process for aseptic multiphase food products. A copy Clf the case study can be obtained from either NCFST or CAPPS. Not all of the issues surroundiri aseptic processing of multiphase foods were discussed at the workshop, but a major step forward was made in un~erstanding the important ones. The partic.ipants recognized that each process :rill be unique and that the results of the

workshop are not the official viewpoints or recommendations of FDA, the United States Army, NFPA, or any of the participating university or industrial organizations. They also recognized that it is the processor's responsibility to demonstrate the ability of his process to commercially sterilize every portion of the food product produced. What Happened Next? The progress made at the workshop was significant in helping to establish a direction for a food processor to pursue when establishing an aseptic multiphase food process. The following four articles in this issue of Food Technology present the results of several of the working group discussions. After the workshop, Tetra Pak, Inc., Buffalo Grove, Ill., used the case study as a guide and, with several new analytical procedures, developed a commercial filing for cream of potato soup. The filing was accepted by FDA in May 1997 (Palaniappan and Sizer, 1997). Future success-

ful fIlings will be dependent on the processor's use of scientifically sound principles, as outlined by the workshop participants, and documentation of the lethality delivered to every portion of the product processed.

REFERENCES
Anonymous. 1995. Food Chem. News 37(18): 26-27. Chandarana, DJ and Unverferth, JA 1996. Residence time distribution of particulate foods at aseptic processing temperatures. J. Food Eng. 28: 349-360. Dal Porto, A 1993. Aseptic advances tangled in bureau. cratic red tape. Food Proc. 54(10): 75-76. Dignan, D.M., Berry, M.R, Pflug, !.J., and Gardine, lD. 1989. Safety considerations in establishing aseptic processes for low-acid foods containing particulates. Food Techno!. 43(3): 118-121, 131. Palaniappan, S. and Sizer, GE. 1997. Aseptic process validation for food containing particulates. Food Techno!. 51 (8): 60-62, 64, 66, 68. Updated August 1997 from a paper presented during the forum, "NCFST and CAPPS Workshop on Aseptic Processing of Multiphase Foods, " at the Annual Meeting of the Institute of Food Technologists, New Orleans, La., June 22-26, 1996 .

,II

II'

Edited by Neil H. Mermelstein, senior Associate Editor.

Measuring Residence Time and Modeling the System

ll,
SUDHIR K. SASTRY
,i The author, a Professional Member of 1FT,is Professor, (Dept of Food, Agricultural and Biological II, Engineering, Ohio State University, 590 Woody Hayes Col~mbus, OH 43210.

q~.,

The National Center tor Food Safety and Technology and the Center for "I Aseptic Processing and Packaging Studies sponsored a workshop in November 1995 and March 1996 to: reach a consensus on concerns that: needed to be resolved to develop ani, aseptic muItiphase food process. This article presents the results of the working groups on (1) measuremen~ of residence time distribution (RTD) a~d (2) mathematical modeling and meaL surement of physical properties. The:~ scope covered aseptic processing usirtg conventional heating (surface, rather~ than internal generation). The RTD discussion is likely independent of mode of heating, but conclusions regarding system thermal behavior apply only to cbnventional aseptic processing. II,
'11

:i I

Measurement of RTD Reasons for measurement of RTD include determination of the fastest-moving particle, which is necessary for (a) designing a process via mathematical modeling to ensure commercial sterility, and (b) biological validation of a model. The biological test must contain sample particles that represent the fastest-moving particle. RTD measurement is needed because of the difficulty in noninvasive measurement of particle internal temperatures during continuous flow. If the cold-spot temperature of the slowest-heating particle could be measured at the end of the heater and the hold tube, RTD measurement would be unnecessary. Several methods for measurement of RTD are available: Optical Methods, such as particle tracking velocimetry or PTV (Zitoun, 1996) and other visualization tech-

niques. Magnetic Methods, which involve introduction of tagged particles containing small magnets (Chandarana and Unverferth 1996), the passage of which is detected by a voltage generated within coils at selected locations of the process equipment. Magnetic Resonance Imaging, used for flow visualization in food systems (Manavel et aI., 1993). History Methods, such as chemical markers and thermal memory cells, . which involve determining the effect of a process on a chemical reaction or diffusion process and back calculating processing parameters (Kim and Taub, 1993; Swartzel et al., 1991). Ultrasonic Methods, which involve detection of RTD by the Doppler scattering of ultrasound waves by the moving particles. Salt Tracer, a well-established approach used in the chemical engineering literature for pure fluids, in which RTD is detected by electrical conductivity measurements. The group opinion was that any of these methods could be used, within their respective applicable ranges.
OCTOBER 1997 VOL. 51, NO.1 0

..

44

FOODTECHNOLOGY

,"

Theoretical Velocity Profiles for Pure fluids While the solid-liquid.flow problem is too complex to be solved in detail theoretically, limiting cases may be understood for pure liquids. Practically all modifications of product and equipment "blunt" the velocity profile. For example, addition of solids decreases maximum velocity (Fig. 1); addition of macromolecules makes fluids pseudoplastic-the only reported dilatancy appears to involve ungelatinized starch suspensions (Dail and Steffe, 1990)-but this situation can be controlled by precooking the suspension; and bends and curvature in tubes create secondary flows that narrow RTD. Thus, the fastest velocity of a pure: fluid is expected to be greater than or equal to the, velocity of the fastest particle in a solid-liquid mixture. Therefore, if a velocity profile can be theoretically determined for a homogeneous carrier fluid, the fastest fluid velocity can be used as the fastest particle velocity in lieu of experimental measurements. For example, in holding tubes, it is sufficient to assume that the fastest particle moves at twice the mean fluid velocity, corresponding to the theoretical solution to steady, fully developed tube flow of a Newtonian fluid. Since the addition of ingredients and solids only blunts the velocity profile, the Newtonian case is considered conservative. Situations may exist where particles . may move faster than the local fluid. For example, a particle may be accelerated as it flows through a venturi and continue to flow faster than the local fluid downstream of the venturi as a result of inertia (Fig. 2). This phenomenon is not expected in steady, fully developed holding tube flows. Processing of Multiple Particle Types Commercial products may'contain more than one particle type, and it is necessary to assess which has the lowest residence time. Thus, RTD must be determined for samples for each particle type. This process may be expedited if it is determined which particle type best represented the characteristics of the system. Results may be product specific, and research in this area is needed. RTD Information from Separate Shifts One concern is the duration of the residence time experiments. If the mag-

.'

~.

netic particle approach is used, where the particles must be introduced one at a time and a large number (299) of particles need to be tested, one shift may not suffice for an experiment. However, Fig. l-"Blunting" effect on fluid velocity if a system is properly profile when solids are controlled, succeeding added to a carrier fluid. shifts would be expected (a) velocity profile of to have similar velocity profiles and RTDs. Thus, pure fluid; (b) velocity the use of data accumuprofile when solids and solutes are added lated over several shifts

(a)

C? _____ [2
(b)

_8
FOODTECHNOLOGY 45

VOL. 51, NO.1 0 OCTOBER 1997'

S PEe

Ii

A L 5 E C T

Residence Time

80

N TIN

U E D

(under such controlled conditions), would be an acceptable procedure. It is II assumed here that the processor understands and can control its shift-to-shift variations. Reuse of Sample Particles If a sample of transducer particles reasonably represent the actual product variations, they can be used for meJsureII ment purposes. RTD data need to be collected from independent samples from the population. If the same transduter 'I. particle is used repeatedly to deterl1}Ine RTD, it is not clear how it might represent variations in the sample population. An understanding of the impact of . 'I tracer's repeated use on th e Ind epen~ dence of a sample would be needed.: In particular, an insignificant effect of repeated tracer use on the sample RT~ would need to be shown.
II

behavior to the model. Also, the model assists in process control by identifying operating parameter adjustments needed to correct for process variations. Important process parameters can be selected for monitoring, and their measurement procedures and instrumentation decided using the model. Steps in Modeling It is desirable to mathematically predict all particle and fluid velocities, particle positions, orientations, and fluid and particle temperature distributions. However, this is not possible because of limited understanding of the basic physics, lack of precise knowledge of all initial conditions, and limitations in computational abilities. Thus, simplifying assumptions are necessary to make the problem tractable. It is only necessary to address the worstcase scenarios in arriving at an acceptable model. Most mathematical models Go+(e.g., Sastry, 1986; Chandarana and Gavin, 1989), involve the following steps: 1. Prediction of Fluid Temperature. Most models use an energy balance on the fluid phase to account for heat entering or leaving via system walls and heat exchange with particles to predict the fluid temperature. 2. Prediction of Representative Particle Temperature. The heat conduction equation is solved for a particle to be considered "representative" (defined below) of the system. 3. Iteration. Since the fluid temperature in step 1 is dependent on the particle temperature predicted in step 2 and vice versa, steps 1 and 2 are iterated until the temperature converges. This approach simulates the fluid temperature 4istribution throughout the heater. It is then possible to conduct a separate set of simulations to identify the worst-case scenario by solving the heat conduction equation for the slowestheating particle. For this purpose, the previously determined fluid temperature is used in a time-dependent convective , boundary condition (Sastry, 1986; Chandarana and Gavin, 1989), along with the physical properties, characteristics, and time scales of the slowest-heating particle, e.g., a particle of the largest size en-

countered during this process, and a time scale that would cause the particle to "experience" the fluid temperature conditions at a faster rate. Worst-case scenario simulations are needed for hold tube sizing. Minimum Acceptable Features Of a Model The following are the minimum desirable features of a mathematical model for aseptic multiphase processing. Heater Section. For thermally mixed systems (e.g., pure fluids in turbulent flow and solid-liquid mixtures in sweptsurface heat exchangers, where radial temperature variations are eliminated by mixing): For fluid temperature simulations, (1) an energy balance must be conducted on the fluid, combined with temperature

2-1l1ustration

of

eleration of a particle

0---

through a venturi, followed by inertial effect, whereby the particle is moving faster than the

Test Conditions , .'1 Data are needed from actu al pro~ess !l conditions, since the objective 1S presumed to be filing of a commercial process with the Food and Drug AdmiuJ.stration. II Reasons for Mathematical Modeling II There are various reasons for modeling a process: Since particle temperatures cannpt be measured during continuous flow, the mlJdel permits sizing the hold tube ~bquired for commercial sterility. Without modeling, hold tube sizing becomes a trial-and-error process, with potenti~lly disastrous consequences if the trials do not adequately simulate commercial practice. Certain product and process parameters have critical influences on the s~fety of the processed product, and their identification can be greatly simplified b~iuse of a model. By varying product and s,ystern parameters in simulation, it is pqssible to determine critical control poin:~s. Once we have a verified mathem~tical model, we can understand how a ,II, process behaves. Thus, any deviations or unusual behavior during processing may be detected by comparing real physical
II'

measurement at the inlet and outlet of the heater; (2) the value of the fluid-particle convective heat transfer coefficient h that yields the best agreement with rrieasured inlet and outlet fluid temperatures must be determined by trial-anderror or by some other numerical procedure; and (3) a "reality check" must be conducted to ensure that the estimate of hfpis realistic (e.g., Balasubramaniam and Sastry 1994a, b, 1996a, b; Cacace et al., 1994; Alhamdan, 1995; Gadonna et aI., 1996; and others). For these fluid temperature simulations, the assumption of a worst-case hrp(e.g., corresp~nding to a stagnant fluid) 1S not conservative, because such particles will not be a significant thermal "drag" on the fluid (Fig. 3). For worst-case simulations, the particle is considered to experience the fluid temperatures on a time scale corresponding to the fastest-moving particle. h fp, for these simulations must be conservatlve.

When the heater system is not thermally mixed (e.g., tubular heat exchangers), other considerations may apply. The following recommendations are made: (1) Although fluid temperatures approximate an exponential curve connectOCTOBER 1997 VOL. 51, NO.1 0

46

FOODTECHNOLOGY

I
=-"""""---~"".

PEe

I' A

SEC

o N
apparent specific heat C may either increase slightly or chang:' depending on phase transitions. For modeling purposes, the use of constant values is acceptable, if their variation is understood and accounted for by selection of the most conservative real value. The overall heat transfer coefficient was considered easy to measure and not. a major issue. An acceptable choice as a representaC tive particle is a particle of size and shape that yield reasonable fluid temperatures (with the same reality checks as for fluid temperature simulations).

ing inlet and outlet temperatures, the exact character of the "thermal trajectory" needs to be characterized. Thus, fluid temperature predictions along the tube axis need to be verified by checks of measured fluid temperatures at intermediate points along the heater. (2) It is desirable to quantify the extent of radial temperature variation. Monitoring at multiple points further complicates an already complex process. Monitoring of fluid inlet and outlet temperatures and their variations is necessary. Monitoring of at least one internal fluid temperature is considered desirable to establish the fluid thermal trajectory. (3) Fouling may alter the expected thermal trajectory away from the above (usually exponential) estimate. However, significant deviations due to fouling would likely be accompanied by changes in pressure drop and heating-medium temperature. to compensate for the fouling-hence, any such deviations should be detectable on-line. If fouling is part of the worst-case scenario, it should be tested. Hold Tube. The approach should be similar to that for'a tubular heater. Cooler. The approach should be similar to that for a tubular heater. However, for fluid temperature simulations, a low value of hfp is conservative. This assumption causes the fluid temperature to drop rapidly, with a low "thermal drag" from the particles. For worst-case simulations, however, a high value ofhfp is conservative, because the particle surface is considered to cool rapidly in response to the cold carrier fluid. Also, since the worstcase assumption that the particle achieves the carrier medium temperature is conservative, br~akup is not considered a major issue in modeling.

attached to it (Sastry et aI., 1990). 4. Liquid crystal methods, where a transducer particle is coated with liquid crystal, which changes color with temperature. The particle is videotaped during its passage through the process system, and the colors (P!ecalibrated against temperature) are analyzed to determine hrp (Stoforos et aI., 1989; Balasubramamam and Sastry, 1994b, 1996a). 5. Temperature pill, a temperature sensor which transmits temperature information to an external antenna (Balasubramaniam and Sastry, 1996b; Bhamidipati and Singh, 1995). 6. History indicators, such as chemical marker and thermal memory cell, detect an endpoint thermal conversion, and can be used to back calculate heat transfer over an entire heat-hold-eool process. 7. Melting-point indicator (Mwangi et al., 1993), where the melting of a poly-' mer indicator produces a color change. 8. Fluid temperature disturbance, where a large volume of cold particles is introduced into a flowing fluid and the fluid thermal disturbance determined and used for calculating hfp (Alhamdan, 1995). 9. Temperature monitoring by magnetic resonance imaging to determine the temperature proflle of an object (Schrader et aI., 1992). 10. Off-line sampling, which involves sampling particles at various points in the system, measuring internal temperature, and calculating hfp'

Periodic Verification by Monitoring

Periodic model verification is possible if flow rates, pressure drops, and inlet and outlet temperatures are monitored simultaneously and operational shifts could be detected. For example, an outlet fluid temperature which responds to changing inlet temperature in ways predicted by the model could be one measure of verification.

Software Documentation
Changes made in process modeling software should be properly documented, using procedures that are standard in the software industry.

Critical Control Points

The following are possible critical control points: 1. Product composition, including particle preparation steps (if appropriate), maximum particle size, particle loading, carrier formulation, preparation steps, viscosity, and product quality (as it affects heating rate). 2. Initial temperature, including preparation steps, agitation, and de- ;-' greeofmix. 3. Flow rate. 4. Process temperatures, including fluid temperature at the end of the heater, and fluid temperature at the end of the hold tube. 5. Heating-medium temperature. 6. Product pressure drop. 7. System pressure. 8. Swept-surface heat exchanger dasher speed (if appropriate);' '.

Thermophysical

Properties

In aseptic processing of foods, thermal conductivity k increases slightly and density decreases with temperature. The

Determination

of htp

To determine hfp' SUItable combinations of available methods could be used, including the following: 1. The assumption of Nu 2.0 based on classical solutions corresponding to heat conduction from an infinite medium to a sphere. 2. Stationary particle methods (Chandarana et aI., 1990; Chang and Toledo, 1989), where a stationary particle is held in a fluid stream, which is adjusted to a desired flow rate. 3. Moving particle method, where a particle is moved through atube at the same velocity as an unattached particle, while having a thermocouple

Distance along heater

3-1l1ustration of the effect of the .umed htp on fluid temperature prediction via energy balances in a heater section. Low hfpvalues do not yield conservative fluid temperature predictions

Guidelines for Process Filing

The above procedures are intended to serve as guidelines for processors interested in filing aseptic processes for multiphase foods. While the list is .

VOL. 51, NO.1 0 OCTOBER 1997

-FOODTECHNOLOGY

47

sip
Residence Time
Go
N TIN U E

ALSECT
M. Roco, pp. 127.140. Butterworth-Heinemann, Boston. Mwangi, J.M., Rizvi, S.S.H., and Datta, A.K. 1993. Heat transfer to a particle in shear flow: Application to aseptic processing. J Food Eng. 19: 55-74. Sastry, S.K. 1986. Mathematical evaluation of process schedules for aseptic processing of low-acid foods containing discrete particulates. J. Food Sci. 51: 1323-1328. Sastry, S.K., Lima. M, Brunn,T., Brim, J, and Heskitt, BF 1990. Liquid-to-particle heat transfer during continuous tube flow: Influence of flow rate and particle to tube diameter ratio. J. Food Proc. Eng. 13: 239-253. Schrader, GW, Litchfield. J.B., and Schmidt, S.J. 1992. Magnetic resonance imaging applications in the food industry. Food Techno!. 46(12): 77-83. Stoforos, NG, Park, KL., and Merson. R.L. 1989. Heat transfer in particulate foods during aseptic processing. Presented at Ann. Mtg., Inst. of Food Technologists, Chicago, June 25-29, 1989. Swartzel, K.R.. Ganesan, S.G, Kuehn, R.T., Hamaker, R.W., and Sadeghi, F. 1991. Thermal memory cell and thermal system evaluation U.S. patent 5,021 ,981. Zitoun, KB 1996. Continous flow of solid-liquid food mixtures during ohmic heating: Fluid interstitial velocities, solid area fraction, orientation and rotation. Ph.D. dissertation. Ohio State Univ.. Columbus.
Updated August 1997 from a paper presented during the forum. "NCFST and CAPPS Workshop on Aseptic Processing of Multiphase Foods, " at the Annual Meeting of the Institute of Food Technologists, New Orleans, La., June 22-26, 1996.

:I~

:l

not comprehensive, it represents a consensus among the academic, industrial, and governmental groups on the apJ proach to be used in process filing.

~I

REFERENCES
Alhamdan, AM. 1995. Particle residence time distiibution and bulk heat transfer coefficients of two-phase flow in scraped surface heat exchanger and holding tubes. Ph.D. dissertation, Ohio State Univ., Columbus. Balasubramaniam, V.M. and Sastry, S.K. 1994a. Liquidto-particle heat transfer in continuous flow through a horizontal scraped surface heat exchanger. FoodiBioprod. Proc.;Trans.IChemE, PartC, 72: 189-196. Balasubramaniam, V.M. and Sastry, S.K. 1994b. Convective heat transfer at particle-liquid interface in con,tinuous tube flow at elevated fluid temperatures. J Food Sci. 58: 675-681'1 Balasubramaniam, V.M., and Sastry, S.K.1996a. Liquidto-particle heat transfer in continuous tube flow: 80mparison between experimental techniques. Inti. J. 'Food Sci. TechnoL 31: 177-187. . Balasubramaniam, V.M. and Sastry, SK 1996b. Noninvasive estimation of convective heat transfer beMeen fluid and particle in continuous flow using a remole temperature sensor. J. Food Proc. Eng. 19: 223-240. Bhamidipati, S. and Singh, R.K. 1995. Determination of

fluid-particle convective heat transfer coefficient. Trans. ASAE 38: 857 -862. Cacace, D., Palmieri, L.: Pirone, G., Dipollina, G., and Masi, P. 1994. Biological validation of mathematical modeling of the thermal processing of particulate foods: The influence of heat transfer coefficient determination. J. Food Eng. 23: 51-68. Chandarana, 0.1. and Gavin, A ilL 1989. Establishing thermal processes for foods to be processed aseptically: A theoretical comparison of process development methods. J. Food Sci. 54: 198-204. Chandarana, D.L, Gavin,A III,and Wheaton, F.w. 1990. Particle/fluid interface heat transfer under UHT conditions at low particle/fluid relative velocities. J Food Proc. Eng. 13: 191-206. Chandarana, 0.1. and Unverferth, JA 1996. Residence time distribution of particulate foods at aseptic processing temperatures. J. Food Eng. 28: 349:360. Chang, S.Y.and Toledo, R.T. 1989. Heat transfer and simulated sterilization of particulate solids in a continuously flowing system. J. Food Sci. 54: 1017-1023. Dail, RV and Steffe, J.F. 1990. Rheological characterization of crosslinked waxy maize starch solutions under low acid aseptic processing conditions using tube viscometry techniques. J. Food Sci. 55: 1660-1665. Gadonna, J.P., Pain, JP, and Barigou, M 1996. Determination of the convective heat transfer coefficient between a free particle and a conveying fluid in a horizontal pipe. Food Bioprod. Proc.; Trans. IChemE. Part C, 74: 27 -39. Kim, H-J. and Taub, I.A. 1993. Intrinsic chemical markers for aseptic processing of particulate foods. Food TechnoL 47(1): 91-97, 99. Manavel, J. E., Powell, R.L., McCarthy, M.J., and McCarthy, K.L. 1993. Magnetic resonance imaging of multiphase systems. In "Particulate Two-Phase Flow," ed.

Edited by Neil H. Mennelstein, senior Associate Editor.

'iirr I

~~2!~~cal1alidation
.
.

Theauthor,a ProfessionalMemberof 1FT isAssociateProfessor,Dept.of FoodScienceandTechnology,VirginiaPolytechnicInstituteand State University,Blacksburg, , VA,24061.

The Food and Drug Administratiolhas suggested that the establishment of; each aseptic process for multiphase~ food products should consist of fou~ elements: identifying and selecting a sterilizing value F for the product; :: d d Ith 1f eve opmg a conservatIve rno e a" t reliably predicts the total lethality of the heat process; quantitatively

i .

'I

verifying the lethality delivered by II means of a bioindicator; and listing'llie critical factors of each aseptic proce!sl s
'I

and the procedures to be used for II controlling these factors (Dignan et al., 1989).

Prior Attempts at Validation Biological verification of an aseptic process for low-acid multiphase (particulate) foods requires some form of biological indicator that can be processed through the aseptic system, retrieved intact, and tested quantitatively for sterility. A variety of bioindicators using bacterial spores have been developed and used . In va'nous th er mal processes . A biological thermocouple system in which suspended spores are encapsulat~ ed in a carrier and do not come into contact with the product has been .used successfully with canned products (Pflug and Smith, 1977; Pflug et al., 1980; Jones et al., 1980). Similarly, Hersom and Shore (1981) suspended spores in glass beads which were then embedded within vegetable particles. Hunter (1972) embedded spores in polymethacrylate beads

and conveyed them through the thermal process, but embedding the spores is time consuming, recovery requires acetone which may be deleterious to the heat-injured spores, and the heat applied is dry rather than wet. Dallyn et al. (1977) suspended spores in alginate beads. This approach provided easy preparation of the particles, easy recovery of the spores, and wet rather than dry heating. This type ofbioindicator was also used by Bean et al. (1979) and Heppel (1985). The particle sizes used in these studies were quite small, however, and thus not completely applicable to larger particulate foods. A larger particle size was achieved by Brown et al. (1984), who embedded spores in cubes formed from alginate mixed with pureed potatoes, peas, or meat. However, particle shrinkage appar-

48

FOODTECHNOLOGY

OCTOBER 1997 VOL. 51, NO.1 0

PEe

T'

Biologic.al Validation

eo

N TIN

U E 0

ently resulted in significant differences between the calccl.ated and predicted lethality of the particles, and the particles were not shelf stable. Sastry et a1.(1988).also developed a bioindicator with a larger particle size. They suspended spores in an alginate solution, infused the spore suspension into mushrooms, and immobilized the spores by gelation of the alginate. These particles could then be freeze dried for shelf stability and rehydrated before use. A major disadvantage of this particle is the large inherent spore population present in mushrooms, which makes accurate quantitation of the number of infused spores difficult to determine. Biological methods are needed to validate mathematical models which are used to determine acceptable thermal processes for aseptically produced foods. Various biological methods are currently being used by food processing companies to validate models and to confirm the effectiveness of a specific aseptic thermal process, in much the same way that an inoculated pack is used to confirm lethality of thermal processes for canned foods (Segner et aI., 1989). An inoculated pack test assumes that the food will have the inoculum evenly distributed throughout the food. The test cans are placed at the slowest-heating position in the retort. This combination of factors guarantees that the inoculated pack will be measuring the worst-case process at the coldest point in the slowest-heating container. The conditions achieved in inoculated test packs cannot be replicated in biological challenge studies with multiphase aseptic foods. In a multiphase aseptic process, each particle of the food receives a different thermal process, with one particle receiving the least thermal process. Because it is not possible to identify which particle will re<:eive least therthe mal process, a statistically significant number of particles need to be tested. Developing a Protocol A working group of scientists from industry, trade associations, regulatory agencies, and academia was convened by the National Center for Food Safety and
50 FOODTECHNOLOGY

Technology (NCFST) and the Center for Aseptic Processing and Packaging Studies (CAPPS) in late 1995 and early 1996 to assess the status of biological validation of aseptic multiphase foods and produce a standard operating protocol for validation of aseptic processes. The working group addressed the following problem areas: . Biological validation test requirements. The working group agreed that microbiological challenge testing methods could be used (1) to validate a mathematical model developed to predict the lethality occurring in an aseptic rilul- tiphase food and (2) to serve as part of a~ process filing to confirm lethality of a specific operational system, . All microbiological testing for biological validation of a thermal process should be sufficient to achieve commercial sterility. Biological validation of multiphase aseptic foods is to demonstrate the effect of thermal processing on the particle which would receive the minimum thermal treatment. This' would be determined after considering which particle was the slowest heating and the fastest moving through the hold -tube. Selection of test microorganisms. The test microorganism should have stable thermal characteristics during an aseptic processing procedure and be appropriate for the intended use. Also, it must have appropriate thermal resistance (D value). If the D value is too high, "skips" (improbable results) occur in count reduction. If the D value is too low, large numbers of microorganisms are needed to ensure survivors. The D value and z value should be determined inthe same ~edia in which the incubation is done; this is often the food being processed. The D and z values should be determined experimentally for the test microorganisms at the temperature conditions of the aseptic process. Replications of the D and z value determinations are required to assess the variability of the microorganism's thermal resistance. Viability of the test microorganism must be established in the incubation media. Care should be taken to include the oxygen conditions (aerobic or anaerobic) which will exist in the package during incubation. The test microorganism may be either a mesophile or a thermophile, depending on the intended application. Because positive identification of the test microorganism is often needed, a

test microorganism with a unique identity is desirable. Inoculation procedures. The procedures for inoculation of particles for aseptic processing could be considered appropriate for either the count reduction method or the inoculated pack method. While there are distinct differences between the two procedures, both have appropriate uses. In the count reduction method, a known number of microorganisms are implanted into the appropriate location (usually the center) of a food particle,

Simulated potato cubes made of epoxy and embedded with magnets, as well as chickenalginate cubes, have been used to biologicallyvalidate a multiphase aseptic process.

and the particle is passed through the aseptic processing system, then recovered. The number of surviving microorganisms is determined, and the lethality of the aseptic process is established. This method is much more rapid than an inoculated pack method, which requires lengthy incubation periods. However, it is very costly because of the additional labor required to collect the inoculated particles and enumerate the survivors. Only a limited number of particles can be processed because of the difficulty of manufacturing and collecting the inoculatedparticles. Count reduction methods can be accomplished by inoculating particles with a series of microorganism levels such as 10" lOS, 106, etc., and recovering the inoculated particles for determination of survivors; this is usually difficult to accomplish. A more common approach is to inoculate all particles to the same level of microorganisms and then process them through the continuous-flow sterilization system at a series of appropriate process temperatures. Varying the length of the hold tube and varying the flow
OCTOBER 1997 VOL. 51, NO.1 0

'lip
,
,

E C

A L 5 E C T

o N
has been recovered, identification of the surviving microorganism is needed. Analysis of data. Count reduction data should be analyzed, based on the assumptions of the model used. If the lethality model is based on cold-point lethality, care should be taken not to use integrated sterilization values. Inoculated pack data need to be analyzed by the methods outlined by the' Statistical Design and Analysis Working Group (Digeronimo et aI., 1997).

Biological Validation .~ Go N TIN U E D

rate of the product are not acceptablJ procedures, since they easily could II change the residence time pattern of the particle in the hold tube. II The inoculated pack method re- II~ quires a relatively large number of particles to ensure that an inoculated particle will be placed in the containers to be incubated (Digeronimo et aI., 1997). Tlle physical characteristics of the critical:! heating food particle is determined a~d used to determine the characteristics Ibf the inoculated particles (Sastry, 1997t One proposed method of producing ~noculated particles is to use dehydrate~ particles and to rehydrate them in anl!inoculum suspension. The ability of the particle to retain !;heinoculum shoul~ be verified. This procedure can produce I: hundreds or even thousands of particles which can be used with virtually any III food product. Because a large number of inoculated particles are being used, is'~ sues such as clumping during commercial operations need to be addressed. II, Sample preparation. The inoculated food particle should have known and: reproducible physical and compositional characteristics. If a dehydrated food i~ II being used, care should be taken to ensure that the inoculated particle is fully hydrated. There is also a concern that: the physical characteristics of the food nbt have been altered in the dehydration4rehydration operation. The lethality rate should be no greater in the inoculated particle than in the food it represents! The inoculum level in the food partiJle should be appropriate for the test being conducted. Inoculum levels should be verified and the degree ofleaching determined. Placement of the inoculum i~ the particle should reflect the assumptions of the model being tested. Impact of thermal processing onlinoculated samples. Biological validati0n of an aseptic process requires that the food particle being used for testing behave in a manner identical to that of the food intended for processing. The tes~ particle should maintain its physical integrity throughout the aseptic process. The physical properties of the test partil cle should be known before and after 'II thermal processing to ensure that no:

changes have been made. Test particles should be evaluated to determine if the inoculum has been retained throughout the aseptic process. Many models for aseptic systems do not include the lethal effects of the cooling operation. Care should be taken to account for these effects, eitherby eliminating the heating effect during cooling or by deducting calculated lethality. Collection of samples and recovery of survivors. Collection of inoculated particles can be done in three ways: inoculated pack (growth/no growth), count reduction (number of survivors), and direct particle recovery (growth/no growth). In the inoculated pack method, the assumption is that any spoilage is caused by a single particle in the container, even if multiple particles could be present. Apparently sterile inoculated packs must have the absence of viable organisms confirmed by recovering a representative number of inoculated particles in an appropriate growth medium. Count reduction should be done with an appropriate growth medium and growth conditions. Since count reduction requires identification of the inoculated particle, a permanent method of distinguishing the inoculated particle is required. Once the inoculated particle

REFERENCES
Bean, PG, Dallyn, H, and Ranjith, H,MP, 1979, The use of alginate spore beads in the investigation of ultra. high temperature processing, In "Food Microbiology and Technology," ed, B, Jarvis, J.H,B, Christian, and H,D, Michener, pp, 281-294, Medicina Viva Seruizio Con. gressi S,r.l., Parma, Brown, K.L" Ayres, C.A., Gaze, J,E" and Newman, M,E, 1984, Thermal destruction of bacterial spores immobi. lized in food/alginate particles, Food Micro, 1: 187198, Dallyn, H, Falloon,WC., and Bean, P,G, 1977. Method for immobilization of bacterial spores in alginate gel. Lab, Pract. 26: 773-775, Digeronimo, M" Garthright, W, and Larkin, J,W 1997, Statistical design and analysis, Food Technol, 51 (10): 52-56, Dignan, D,M" Berry, M,R" Pflug, I.J" and Gardine, T.D, 1989, Safety considerations in establishing aseptic processes for low. acid foods containing particulates, Food Technol. 43(3): 118-121, Heppel, N,J. 1985, Measurements 'of the liquid.solid heat transfer coefficient during continuous sterilization of food stuffs containing particles, Presented at 4th IntI.

..........................

. . . . . ..

Statistical Design and Analysis

MICHEL DIGERONIMO: WALLACE GARTHRIGHT. AND JOHN W. LARKIN
Author Digeronimo is President, Dover Brook Associates, P,O, Box 177, Chester, NY 1091,8, Author Garthright is Mathematical Statistician, Food,and Drug Administration, 200 CSt., SW, Washington, DC 20204, And author Larkin, a Professional Member of 1FT, is Food Process Hazard Analysis Branch Chief, National Center tor Food Sately and Technology, Food and Drug Administration, 6502 S, Archer Rd" Summit-Argo, IL 60501, Send reprint requests to author Digeronimo,

Aseptic manufacturing of low-acid, single-phase foods such as milk and milk-based products, cheese sauces, and puddings is well established as a reliable and effective means of heat preservation. However, establishing a scheduled process to ensure commercial sterility of an aseptically manufactured food containing particulates is more complex.

Although several variables have to be considered when establishing a thermal process, one of the more critical ones is the particle's minimum residence time. Determining the length of time a particle is in the heating sections of the aseptic system, and especially the hold tube, is essential, because process lethality is based on both temperature and time. Introduction of particles into the fluid significantly complicates the flow conditions and consequently can affect residence time. The relevance and trustwor-

52' FOODTECHNOLOGY

OCTOBER 1997 ~ VOL. 51,NO. 10

Congo on Engineering and Foods, Edmonton, Alberta, Canada, July 7-10. Hersom, A.C. and Shore, OJ. 1981. Aseptic processing of foods COinprising sauce and solids. Food Technol. 35(5): 53-62. ,Hunter, G.M. 1972. Continuous sterilization of li'Quid media containing suspended particles. Food Technol. Austr. 24(4)158-159, 162, 164-165 , Jones, A.T, Pflug, I.J" and Blanchett, R. 1980. Performance of bacterial spores in a carrier system in measuring the F0 value delivered to cans of food heated in a Steritort J. Food Sci. 45: 940-945, Pflug, I.J. and Smith, G,M. 1977. The use of biological indicators for monitoring wet-heat sterilization processes. In "Sterilization of Medical Products," ed. E.R.L. Gaughran and K. Kereluk, p, 193-222 Multiscience, Montreal. Pflug, I.J., Smith, G" Holcomb, R., and Blanchett, R, 1980. Measuring sterilizing values in containers of food using thermocouples and biological indicator units. J. Food Protect: 43(2): 119-123, Sastry, S.K. 1997. Measuring residence time and modeling system. Food Technol. 51 (1 0) 52-56 Sastry, S,K., Li, SJ., Patel, P., Konanayakam, M" Bafna, P., Ooores, S., and Bulman, R.B. 1988. A bioindicator for verification of thermal processes for particulate foods. J. Food Sci.'53(5) 1528-1531, Segner, WP, Ragusa, T.J., Marcus, G.L" and Soutter, EA 1989, Biological evaluation of a heat transfer simulauon for sterilizing low-acid large paruculate foods for aseptic packaging. J. Food Proc, Preserv. 13: 257-274, Updated August 1997 from a paper presented during the NCFST and CAPPS Workshop on Aseptic Processing of Multiphase Foods at the Annual Meeting of the Institute of Food Technologists, New Orleans, La" June 22-26,

1996.
Edited by Neil H. Mermelstein, Senior Associate Editor.

thiness of residence time data depend on not only the experimental design but also the statistical technique used to analyze the results. Selection of the statistical method is crucial-incorrect use of a technique can result in an incorrect estimate of the sterilizing capacity of the heat process. Most of the time, statistical misuse is due to insufficient understanding of the concepts of the mathematicaltechniques involved and of the heating and flow characteristics of the product in the aseptic system. A firm comprehension of the underlying concepts of the statistical method and of the thermal process allows a proper evaluation of the legitimacy of the selected minimum process lethality (the Fo value of the scheduled process). Avoidance of a self-inflicted careless stalVOL.51,NO.10 I ,OCTOBER 1997

tistical analysis is essential. The National Center for Food Safety and Technology (NCFST) and the Center for Aseptic Processing and Packaging Studies (CAPPS) held a workshop on aseptic processing of multiphase foods at the 1996 Annual Meeting of the Institute of Food Technologists. A working group on statistical design and analysis met to discuss and develop appropriate statistical procedures for use in the establishment of a multiphase aseptic food pro-

cess. This article presents the group's findings. Particle Residence Times Several researchers have measured particle residence times through the ' heater(s), hold tube, or cooler(s) and attempted to fit the data to familiar statistical distributions. These have included a distribution-free method (Berry, 1989); a log-normal distribution model (Dutta and Sastry, 1990); aseries of superposi. FOODTECHNOLOGY 53

5 PEe

A L 5 E C T
dence times. The smallest such N is 299
(i.e., 0.95

o N
currence that prevents any reading of the result. Finally, all readable results from the 299 + f particles should be used as the basis for the validation. A processor cannot pick and choose which of the processed 299 + f particles it will use to demonstrate sterility. For example, if particle breakup is expected to be less than 10% and testing failures less than 1%, the test might be run with l.Ux 299 = 332 inoculated particles. If 330 survive intact and there are no testing failures, then all 330 must test satisfactorily, i.e., must show no growth of the target organisms. Because a processor cannot tell which particular particle is in the fastest percentile, every particle tested must be viewed as belonging to the fastest percentile and be processed to a total commercially sterile endpoint. This requirement to show adequate kill in every particle tested must not be confused with the combined overall kill measured by the quantal method. In the quantal method, the thermal kill of the process is determined from both the number of organisms used to inoculate each unit and the number of processed units that still demonstrate growth after the process (Halvorson and Ziegler, 1933). For example, if 100 units are inoculated with 1,000 microbes (3 logs), for a total of 100,000 organisms, and 2 of the 100 processed units show growth, an estimate of about 2-5 surviving organisms of the total 100,000 is made. This results in a claim of more than a 4-log reduction in the concentration of the inoculated organisms. This sort of calculation is not applicable for multiphase aseptic food processing, since each of 100 inoculated particles receives a differentthermal treatment. The quantal approach requires identical treatments. The residence time and thus the heating effects are too varied among the particles for them to be overlooked. Therefore, one either must show that all particles have been rid of viable test microbes or must recover the 299+ particles and enumerate the survivors for each particle. In the latter case, the log reduction for the process is determined on the particle demonstrating the least log reduction. The working group acknowledged that replications of a test only change the levelsof C and P used to determine the statistical reliability of the results, and that only one test of 299 + f intact particles is
OCTOBER1997 VOL. 51, NO.1 0

Statistical Design Go N TIN U E

<

1 - 0.99299).

D
I

. . tlOnedid' norma lstn'b'utlOns WIth tue normal distribution describing the fastest particles used to estimate the statistically fastest particle (Palmieri et al., 1992D; an autocatalytic logistic growth modell(Abdelrahim and Ramaswamy, 1993); the first, second, and third central moments of the normal distribution (Baptista et al., 1995); and the gamma and log-normal distribution functions (ChandJrana and Unverferth, 1996). After careful analysis of these approaches, the working group agreed that a distribution-free method was the bost appropriate method to determine r~liably the characteristic fastest particle of the system. Because the small tail of one side of the curve (i.e., the fastest pCU;lticle) is the area of the residence time distribution of primary interest and the topl percentile of the function's tail is difficult to accurately establish, this top percentile represents a characteristic fastest reSidence time for all of the product thlt will ever be processed in the system. A cautious approach to estimation of the top percentile was taken. The group wanted a 95% confidence (C) that the fastest time measured would be induded in the fastest percentile (P) of all residence times. This value of C = 0.95 was chosen because it represented the likelihood of collecting a characteristic parti'! cle within the fastest percentile in 19 out of 20 tests, a reasonable criterion for these types of tests. The fastest percentile was chosen as 1% of the particle's residence time population, which was Jonsidered consistent with the traditiorial use of 2-3 standard deviations of the mean to describe normal distributed variations in retort applications. The mathematical solution to tH.e probability of obtaining the fastest particle from a population of residence times is very straightforward. If we measure the residence times ofN particles, the ptoba,~, bility that none is a member of the fastest percentile is (1- P)I'. Therefore the probability that at least one of the residence times is from the fastest percentile i~ C = 1 - (1- P)N.Choosing a value ofNkuch that C > 0.95 will ensure that at leatt one particle residence tim~ will be in the top P percentile of the population of resi'

The collection of intact processed particles is necessary for an accurate assessment of the system. A fractured particle most likely flows through the system differently and is not characteristic of the population of particles that are being measured by the test. Statistics of Biological Validation A number of biological validation procedures have been developed for multiphase aseptic food processes, including an alginate-gel-bead procedure (Dallyn et al., 1977) and a knottedstring-enumeration method (Segner et al., 1989). Both groups of researchers mistakenly assumed that the statistical analysis procedures that had been used for a number of years in the canning industry (Halvorson and Ziegler, 1933) were appropriate for continuous multiphase aseptic food processing systems. Halvorson and Ziegler's underlying assumption of equal treatments for each particle cannot be satisfied for continuous multi phase aseptic systems because there is no way to gauge the residence times of the specific inoculated particles. The working group recommended inoculating a sufficient number of particles to ensure ,that at least 299 are recovered intact and enumerated successfully.This is a departure from traditional biological testing procedures. Each particle should be inoculated with the concentration of microbes that one wishes to demonstrate can be killed by the process. Because each particle may turn out to be in the fastest percentile, each one must get a sufficient kill. Because it is rarely desirable to enumerate separately the concentration of survivors for each particle, in practice this means that each particle must fail to show growth after processing. When 299 intact particles are recovered and show no growth, this gives a 95% confidence that a particle in the fastest 1% received a thermal treatment sufficient to destroy at least the concentration of microbes used to inoculate the particles. The processor should inoculate and treat enough particles to allow for particle breakup and for testing mishaps. The best practice is to inoculate and treat 299 + b particles to allow for at least 299 + f to be recovered intact and tested for sterility (or enumeration). A finding of surviving microbes is a process failure, not a testing failure. A testing failure is an oc-

54

f'OODTECHNOLOGY

5, PEe

A L S.E

C T

indicated by the workshop's residencetime, modeling, and biological-validation working groups-,-are employed. Every precaution' needs to be taken to ensure that the process developer has consulted the technical literature, available proprietary data, food engineering necessary to establish an understanding of theory, and comments from the other the lethal treatment delivered by the proworking groups when developing a concess. HO'Y'ever,he group agreed that it t tinuous multiphase aseptic food process. was critical for the processor to know During development of the processing ahead of time the type of trial-to-trial system, the developer will in alllikelifluctuations that may occur for the prohood need to conduct a sufficient numcess. The test run of 299 + f particles ber of experiments to determine and needs to be conducted on a system where measure the critical factors that are imthe process is known to deliver the miniportant to the process before actually atmum thermally treated particle. tempting to measure the characteristic Demonstration ofa lack of signififastest particle or biologically validate cant trial-to-trial fluctuations should inthe processing system. With this in clude a sufficient number of tests so that mind, the working group also agreed day-to-day variations, startup, fouling, that for design and optimization studies, personnel, and other factors are taken a smaller number of particles than 299 into account. A processor may want to can be used to determine failures or biologically validate the system, at the trends. scheduled process temperature, several It is the responsibility of the procestimes so as to include runs at the beginsor and the process authority to estabning and end of one process on one day lish a multiphase aseptic food process and another on a different d~y. that is adequate and verifiable. Particular care should be taken to ensure that all No Magic Formula . the requirements and assumptions assoThe working group strongly stipulatciated with a procedure used to measure ed that just performing an experiment an aspect of the process are satisfied. with 299+ particles is not going to be the The statistical procedures outlined by magic solution to the establishment of a the working group will ,only lend a meamultiphase aseptic food process. A prosure of I;onfidence to a procedure; they cessor needs to make sure that all of the , will not ~ake a bad proc~dure good. critical factors are defined and that appropriate measurement techniques-as

Statistical Design 80 N TIN U E D

REFERENCES
Abdelrahim, KA and Ramaswamy, H,S, 1993, Mathematical characterization of residence time distribution curves of carrot cubes in a pilot scale aseptic processing system, Lebensm, U.-TechnoI26: 498.504, Baptista, P,N, Oliveira, FAR" Cunha, L,M, ,and Oliveira, JC, 1995, Influence of large solid spherical particles on the residence time distribution of the fluid in twophase tubular flow, Inti J, Food Sci. Technol. 30: 625637 Berry, M.R. 1989, Predicting fastest particle residence time, In ':Proceedings of the First International Congress on Aseptic Processing Technologies," ed, J,V, Chambers, pp, 6-17, Food Science Dept., Purdue University, West Lafayette, Ind, Chandarana, 0.1. and Unverferth, JA 1996, Residence time distribution of particulate foods at aseptic processing temperatures J Food Eng, 28: 349-360, Dallyn, H, Falloon, w.e., and Bean, PG,1977. Method for the immobilization of bacterial spores in alginate gel. Lab, Pract. 26: 773-775, Dutta, B, and Sastry, S,K, 1990, Velocity distributions of food particle suspensions in holding tube flow: Distribution characteristics and fastest-particle velocities, J, Food Sci. 55: 1703-1710, Halvorson, H,O, and Ziegler, N,R, 1933, Application of , statistics to problems in bacteriology, I.A means of determining bacterial population by the dilution method, J, Bacterial 25(2): 101-121, - Palmieri, L, Cacace, D" Dipollian, G" and Dall'Aglio, G, 1992, Residence time distribution of food suspensions containing large particles when flowing in tubular ,systems, J, Food Eng, 17: 225.239 Segner, W.P,Ragusa, TJ" Marcus, e.L" and Soutter, EA 1989, BiologicaLevaluation of a heat transfer simulation for sterilizing low-acid large particulate foods for aseptic packaging, J Food Proc, Eng, 13: 257 -27 4, Updated August 1997 from a paper presented during the forum, "NCFST and CAPPS Workshop on Aseptic Processing of Multiphase Foods," at the Annual Meeting of the Institute of Food Technologists, New Orleans, La" June 22-26, 1996, Edited by Neil H. Mermelstein, Senior Associate Editor.

Issues Involved in Producing' a Multiphase Food Product

DOMINICK DAMIANO
"

The author is Senior ResearchScientist,NestleResearchand Development,Inc" 201 HousatonicAve" NewMilford,CT06776,

Thermal processing of multiphase food , products has received considerable attention in recent years. While there are continuous processes for low-acid particulate foods elsewhere in the world, there have been none in the United States. Many have attributed this to the difficulty and/or uncertainty in getting a

particulate low-acid continuous thermal process filing accepted by the appropriate regulatory agency. The Food and Drug Administration has, through scientific meetings and publications in journals,"stated its position over the past 10 yea~sor so as to what would be required in a successful filing (Dignan et al., 1989). This position has evolved in light of scientific publications over the same time span, but there still has been no ap~

parent commercial activity by industry in the U.S. To address this situation, a workshop 'on aseptic processing of multiphase foods was held in late 1995 and early 1996 (Larkin, 1997). The workshop brought together representatives of government, industry, and academia to address regulatory and other technical issues. The workshop went a long way in addressing these issues, but did not ad-

56 FOODTECHNOLOGY

OCTOBER1997

VOL. 51, NO. 10

'A L

T;

Issues Involved Go N TIN U E D II

dress other technical issues, as weIllas nontechnical and business issues. 11his article will broadly discuss these aJd other issues to provide a greater understanding of what is required for an\acceptable regulatory filing. I Working Group Issues Issues covered by the working iroups (Sastry, 1977; Marcy, 1997; Digerortimo et aI., 1997) are briefly described b~low: Critical Particle. The following, "theme issue" was discussed at the I orkshop: If there was a way to measure the temperature of the "critical part" of the "critical particle" continuously, then only a hold tube calculation would be rJ~ qui red, no different from what is needed for liquids. The "critical particle" is the one particle of all particles processJd which contains the discreet element which receives the least thermal tre~t. ment (or other treatment that contl:'ibutes to lethality). This can be becaJbe it is the fastest or it contains the discreet element which is the coolest, or a cOnl~ination of both. The key is to "prove" that the "critical" part of the critical pa~ticle received the target thermal treatment. Statistical Design and Analysis! The following issues were generally agrJed on as being resolved: (1) There is genefal dissatisfaction with existing models for predicting particle residence time d'istribution (RTD). These models are es~entially empirical and use all of the re~idence time data (fast, slow, and average), leading to possible errors in the preaiction of consequence-that of the fJstest particle(s). This issue was addressed by deciding that the fastest particle determination would require physical meaJurement. (2) For RTD measurements or for biological validation, a minimum sample size of299 different, intact, represeatative particles is required to concludb that a significantly fastest particle was collected (distribution free analysis). Addl'essing this issue has helped to put to r~st l' fears that very large numbers of pa~ticles would be required to collect a "fastest" particle. (3) More work is needed bffore most probable number (MPN) tecHniques are useful (i.e., the HalvorsohZeigler equation is inappropriate fo'r es,wI'
"

timating lethal treatment in aseptic systems). (4) A non-distribution-based experimental design should be used to assure that the fastest particle is collected. However, the following question remains: For a well-designed process (using nonparametric statistics), are repetitions of runs of 299 (replicate confirmations) particles necessary? A consensus on this point was not reached, so each processor will need to address this issue individually. This issue can have a substantial impact on time and cost of conducting RTD measurements and biological validation determinations. Mathematical Modeling arid Physical Properties. The following issues were generally agreed on as being resolved: (l) Biological validation should be used to verify a model, and the model used to determine the process. For a model, the fluid-to-particle surface heat transfer coefficient hfpmust be determined. One can claim a Nusselt number Nu = 2.0, which is the theoretical case of conduction through a film of stagnant liquid around a spherical particle. However, this represents the lowest theoretical value for hfpand is ultraconservative. (2) Methods to measure and determine physical property data for use in models are needed. (3) As an aid 'in model development, the fluid temperature can ale ways be directly measured anywhere in the system or conservatively estimated. (4) If credit is to be given for lethality during cooling, it must be demonstrated that particle breakup does not occur (tubular only) or that the particle never goes below the surrounding cooling liquid temperature. The following issues were generally agreed on as being unresolved: (1) Accuracy is needed in estimation ofhfp' (2) Heat exchanger fouling may lead to liquid temperatures different from those used in the model. (3) Developing a model was clearly outside the scope of the workshop. Fortunately, useful models do exist and are available from various sources. These models are quite complex, and any model used would need to be part of the regulatory filing. Residence Time Distribution. The following issues were generally agreed on as being resolved: (1) For situations where the fastest particles must be known, the RTD must be measured. Because a model is required to establish the thermal process and biological validation is required to verify the model, the , RTD data need to be as accurate as possi-

ble. (2) As paft of a flling, a processor can claim, for tubular flow, that no particle is faster than 2.0 times the mass average velocity or faster than the theoretical fastest liquid velocity proflle. However, except in unusual circumstances, these claims will probably not be used in any filings. (3) Various options exist for direct measurement of fastest particles; magnetic techniques seemed to be the easiest to apply in a commercial system. (4) If a product or process changes, the RTD needs to be reestablished; The following issues were generally . agreed on as being unresolved: (1) When using magnetic measurement techniques, what effect do particle density differences due to the magnets have on the data? (2) What is the cost of running a test, or series of tests, of 299 intact particles? For magnetic RTD measurement, the workshop estimated a particle reading every 2-3 minutes. With a margin of safety added, a minimum of 299 intact particles are required. On an industrial line, this would require 360 x 2-3 minutes, about 12-18 hours of running time with real product under real conditions-in addition to making the particles! Biological Validation. The following issues were generally agreed on as being resolved: (l) Guidelines for choosing and using the rest organism, and for preparing the inoculated particles were developed. (2) Guidelines for the collection of samples were given. One must measure and capture the "critical" particle. Inoculated pack was a favored method. (3) Methods for analysis of data and biological enumeration were presented. MPN techniques should be avoided. Data analysis is easier if spores are at the center, but this is more difficult to carry out properly. (4) Validation is not a single test, but a series of experiments whereby there are some survivors at certain experimental conditions (easiest is to vary temperature). (5) Particles can be collected aseptically into a growth medium. The following issues were generally agreed on as being unresolved: (l) Assessing the value of center point vs integrated sterilization value, and analysis of data when skips or tailing occurs. (2) The efforts required for maintenance of heat-resistant spore crops, and ensuring that sublethal spores grow in the biological validation tests. (3) Inoculum levels in particles must be verified, but how do you prove there is no leaching of spores? (4) How close can "inoculated particles"

II

60 FOODTECHNOlOGY

OCTOBER1997 VOL.51,NO.10

..-:;--"".

.
;:,

5 P C
be produced to the "actual" food product, considering physical properties? (5) How should the issue oflethality during cooling be addressed? Even if credit is not to be taken for this lethality as part of the filing, it still must be estimated as accurately as possible as part of any biological validation data analysis. (6) As for the RTD tests, what potentially are the time and cost of biological challenge tests? Other Issues There are other issues, technical and business: Technical. (1) The whole workshop centered on particles. The next question that should be addressed is,What is a "particle?" Clearly,it should not be defined as a dimension, but in terms which address heating rates. No doubt, a filing of a particulate-containing product that is handled as "nonparticulate" is much simpler. (2) Products and processes will always be changing for marketing and other business reasons. Under what circumstances will another filingbe necessary? (3) Software and associated models must be clearly understood by the user and the thermal process authority. (4) Real results of RTD tests and heating characteristics of actual particles vs formulated particles for biological validation must be taken into account. How easy will it be to make magnetic particles? Business. (1) Is there a business? What are the business risks?If a processor installs a new line for an aseptic particulate process and after some period of time the product doesn't sell,what will the processor do with the process line? How transferable is the equipment? Is the line or equipment so unique that it can't be used for anything else?In many firms there is much in the way of existing investments that are working fine and making money. They life not just going to be thrown away.The same goes for knowledge. (2) What is the justification for the process/ product? It has to be one or a combination of increased quality, innovative or less-expensive packaging, package size, and, most important, overall cost.

). L 5 C T

Case Study A separate working group was formed of participants from industry, CAPPS, NCFST; and the National Food Processors Association (NFPA) to complete a case study filing for an FDA-regulated product from a fictitious processor (Larkin, 1997). Much of the information and data in the case study were obtained from studies conducted by NFPA, but many unresolved issues from the working groups were still unresolved in the case study. Since this case study was based on a fictitious situation, contributors to the case study took liberties to address some problems. Things may not be so simple in real life. The following are issues I feel were not adequately resolved in the case study: (l) It presented data on alginate particles, but what are the differences in heating rate and RTD compared to the real product? Will it be possible to make particles with the same characteristics as the target particles? (2) The case study spore kinetics (z value) were assumed constant above 260F.We know that in reality the kinetics are not linear, but this is true for liquids as well. (3) How accurately is the contribution of lethality during cooling accounted for? (4) The cost and time of determining physical properties, h~, RTD tests, model simulations and vahdation, and bio~ challenge are still concerns. Although this can be a major undertaking, we now know essentially what is required, and processors are in a better situation to estimate the cost of a particulate filing. Using the caSestudy as a model, an industrial equipment supplier, Tetra Rex Packaging Systems, Buffalo Grove, Ill., recently had a filing for a particulate aseptic process accepted by FDA (Palaniappan and Sizer, 1997). Certainly, for this filing to be accepted, many of the issues discussed above had to be resolved and/or dealt with to FDA'ssatisfaction. Is It Worth Doing? Many but not all of the issues that have been considered hurdles from a

regulatory standpoint have been resolved. A greater understanding has been achieved of the necessary factors to be addressed for a successful regulatory filing. Every situation is unique, and there will probably never be a "cookbook" instruction to follow when a filing is prepared for particulates. The case study was compiled as an example only. It should not be implied that if a processor follows the example, it will result in an acceptance. The most important thing is to prove, using good science, that a process is safe. The workshop participants used the term "reasonably conservative" to describe how situations were to be treated. FDA'sacceptance of the recent aseptic particulate filing shows that it can be done. The cost and time of a filing may be a hurdle, but processors can now make better estimates of what the regulatory filing aspects of a project will cost in time and money. The case is no longer "How do we do it?" but "Is it worth do. mg.?" References
Digeronimo, M, Garthrighl, w., and Larkin, J. W. 1997. Statistical design and analysis.Food Technol. 51 (10): 52-56. Dignan, D. M., Berry, M. R, Pflug, I. J and Gardine,T. D. 1989. Safety considerations in establishing aseptic processes for low-acid foods containing particulates. Food Technol. 43(3): 118-121, 131. Larkin, J. W.1997. Workshop targets continuous multiphase aseptic processing of foods. Food Technol. 51(10) 43-44 Marcy, J. E., 1997. Biological validation. Food Technol. 51 (10): 48-53 Palaniappan,S. and Sizer, C.E 1997. Aseptic process validated for foods containing particulates. Food Technol.51 (8): 60-62, 64, 66, 68 Sastry, S. K. 1997. Measuring the residence time and modeling the system. Food Technol. 51 (10): 44-48
Updated August 1997 from a paper presented during the NCFST and CAPPS Workshop on Aseptic Processing of Multiphase Foods at the Annual Meeting of the Institute of Food Technologists, New Orleans, La., June 22-26, 1996. The author thanks Nestle Research and Development, Connecticut, for the support given him in participating in the workshop. Thanks also goes to all workshOp participants, especially those who participated in the case study

Edited by Neil H. Mennelstein, SeniorAssociate Editor.

I
62 FOODTECHNOLOGY OCTOBER 1997 VOL 51, NO. 10 ~

-----

---

II

~f
III

1~
'II

t'

-..

~I,
Ii~

...

:
II,

..

,-'

---

-~...