The efficacy and safety of high-pressure processing of food

Abstract

High-pressure processing (HPP) is a non-thermal treatment in which, for microbial inactivation, foods are subjected to isostatic pressures (P) of 400-600 MPa with common holding times (t) from 1.5 to 6 min. The main factors that influence the efficacy (log₁₀ reduction of vegetative microorganisms) of HPP when applied to foodstuffs are intrinsic (e.g. water activity and pH), extrinsic (P and t) and microorganism-related (type, taxonomic unit, strain and physiological state). It was concluded that HPP of food will not present any additional microbial or chemical food safety concerns when compared to other routinely applied treatments (e.g. pasteurisation). Pathogen reductions in milk/colostrum caused by the current HPP conditions applied by the industry are lower than those achieved by the legal requirements for thermal pasteurisation. However, HPP minimum requirements (P/t combinations) could be identified to achieve specific log₁₀ reductions of relevant hazards based on performance criteria (PC) proposed by international standard agencies (5-8 log₁₀ reductions). The most stringent HPP conditions used industrially (600 MPa, 6 min) would achieve the above-mentioned PC, except for Staphylococcus aureus. Alkaline phosphatase (ALP), the endogenous milk enzyme that is widely used to verify adequate thermal pasteurisation of cows' milk, is relatively pressure resistant and its use would be limited to that of an overprocessing indicator. Current data are not robust enough to support the proposal of an appropriate indicator to verify the efficacy of HPP under the current HPP conditions applied by the industry. Minimum HPP requirements to reduce Listeria monocytogenes levels by specific log₁₀ reductions could be identified when HPP is applied to ready-to-eat (RTE) cooked meat products, but not for other types of RTE foods. These identified minimum requirements would result in the inactivation of other relevant pathogens (Salmonella and Escherichia coli) in these RTE foods to a similar or higher extent.

Keywords: High‐pressure processing; food; microbial inactivation; milk; ready‐to‐eat products; safety concern.

Figure 1

Representation of temperature and pressure profile of a food when exposed to a HPP treatment

Figure 2

Diagram representing a commercial HPP treatment equipment

Figure 3

Observed (points) and predicted (response surface) log₁₀ reductions of Listeria monocytogenes in response to pressure (P, MPa) and holding time (min), in various milk types (excluding UHT milk)

1. Heat map bars represent magnitudes of log₁₀ reductions. The two figures (a and b) represent two different angles of the same 3D graph. Source of data: Patterson et al. (1995); Dogan and Erkmen (2004); Gao et al. (2006); Hayman et al. (2008); Viazis et al. (2008); Xu et al. (2009); Amina et al. (2010); Mishra et al. (2013); Huang et al. (2015); Ramos et al. (2015); Allison et al. (2018); Misiou et al. (2018); Stratakos et al. (2019); Komora et al. (2020).

Figure 4

Predicted vs. observed log₁₀ reductions of Listeria monocytogenes by HPP in different milk types, as predicted by the global linear model fitting

1. ‘Milk’ refers to data from Xu et al. (2009), where the milk type was not specified; ‘Milk buffer’ refers to a buffered solution that contains equivalent amounts of mineral and lactose as the whey from rennet casein (Gao et al., 2006).

Figure 5

Observed (points) and predicted (response surface) log₁₀ reductions of Staphylococcus aureus in response to pressure (P, MPa) and holding time (min), in various milk types (excluding UHT milk)

1. Heat‐map bars represent magnitudes of log₁₀ reductions. The two figures (a and b) represent two different angles of the same 3D graph. Source of data: Records included in text and used for data extraction: Patterson et al. (1995); Patterson and Kilpatrick (1998); Gervilla et al. (1999); Viazis et al. (2008); Tabla et al. (2012); Ramos et al. (2015); Windyga et al. (2015).

Figure 6

Predicted vs. observed log₁₀ reductions of Staphylococcus aureus by HPP in different milk types, as predicted by the global linear model fitting

1. ‘Milk’ refers to experiments human milk from Windyga et al. (2015). ‘PastRawMilk’ refers to raw milk that was centrifuged and standardised to different fat contents (6 or 50%) and pasteurised without prior homogenisation (Gervilla et al., 1999, 2000).

Figure 7

Observed (points) and predicted (response surface) log₁₀ reductions of Escherichia coli (including E. coli O157:H7 and non‐pathogenic E. coli) in response to pressure (P, MPa) and holding time (min), in various milk types (excluding UHT milk)

1. Heat‐map bars represent magnitudes of log₁₀ reductions. The two figures (a and b) represent two different angles of the same 3D graph. Source of data: Patterson et al. (1995); Patterson and Kilpatrick (1998); Garcia‐Graells et al. (2000); Gervilla et al. (2000); Dogan and Erkmen (2003); Pandey et al. (2003); Buzrul et al. (2008); Buzrul et al. (2009); Foster et al. (2016); Bernedo‐Navarro et al. (2016); Machado et al. (2019); Stratakos et al. (2019); Viazis et al. (2008).

Figure 8

Predicted vs. observed log₁₀ reductions of Escherichia coli by HPP in different milk types, as predicted by the global linear model fitting

1. ‘PastRawMilk’ in the legend, refers to raw milk that was centrifuged and standardised to different fat contents (6%) and pasteurised without prior homogenisation (Gervilla et al., 2000).

Figure 9

Observed (points) and predicted (response surface) log₁₀ reductions of Salmonella spp. in response to pressure (P, MPa) and holding time (min), in various milk types

1. Heat‐map bars represent magnitudes of log₁₀ reductions. The two figures (a and b) represent two different angles of the same 3D graph.Source of data: Erkmen (2009); Xu et al. (2009); Erkmen (2011); Foster et al. (2016); Stratakos et al. (2019).

Figure 10

Predicted vs. observed log₁₀ reductions of Salmonella spp. by HPP in different milk types, as predicted by the global biphasic model fitting, including (a) or excluding (b) the records relevant to colostrum

1. ‘Milk’ in the legend, refers to data from Xu et al. (2009), where the milk type was not specified.

Figure 11

Observed (points) and predicted (response surface) log₁₀ reductions of Campylobacter jejuni in response to pressure (P, MPa) and holding time (min), in various milk types

1. Heat‐map bars represent magnitudes of log₁₀ reductions.Source of data: Martinez‐Rodriguez and Mackey (2005), Solomon and Hoover (2004).

Figure 12

Predicted vs. observed log₁₀ reductions of Campylobacter jejuni by HPP in UHT milk, as predicted by the global linear model fitting

Figure 13

Observed (a) and predicted (b) log₁₀ reductions of Mycobacterium bovis (using Mycobacterium avium subsp. paratuberculosis as surrogate) in response to pressure (P, MPa) and holding time (min), in UHT milk and colostrum

1. Heat‐map bars represent magnitudes of log₁₀ reductions.Source of data: Donaghy et al. (2007), Foster et al. (2016), López‐Pedemonte et al. (2006).

Figure 14

Predicted vs. observed log₁₀ reductions of Mycobacterium bovis (using Mycobacterium avium subsp. paratuberculosis as surrogate) by HPP in UHT milk and colostrum, as predicted by the global linear model fitting

Figure 15

Dependence of D _p ‐values on pressure, based on D _ref and z _p ‐values estimated by global fitting of log‐linear (single‐phase inactivation) or biphasic primary models shown in Table 5

1. For Listeria monocytogenes not including UHT milk and colostrum; for Staphylococcus aureus not including UHT milk and colostrum; for STEC not including UHT milk and colostrum with the studies used to fit the model used only O157:H7 and non‐pathogenic Escherichia coli; for Salmonella spp. (including colostrum)– D1 considering the first rapid inactivation phase; for Salmonella spp. – D2 considering the second slower death phase; for Campylobacter jejuni only including UHT milk; for Mycobacterium bovis not including UHT milk and colostrum and using Mycobacterium avium subsp. paratuberculosis (MAP) as surrogate.

Figure 16

Isoreduction curves of HPP conditions (pressure/holding time combinations) needed to achieve the four targeted performance criteria (log₁₀ reductions), according to the global model parameters shown in Table 5 for all six relevant pathogens in milk

Figure 17

Isoreduction plot of HPP conditions (pressure/holding time combinations) that cause 90% reduction of the activity (enzymes) or denaturation (whey proteins) of some milk inherent compounds compared with the same HPP conditions that cause 5 log₁₀ reductions of Staphylococcus aureus in raw milk (see Section 3.2.3.4)

1. ALP: Alkaline phosphatase; GGT: γ‐Glutamil Transferase; Xox: Xanthine oxidase; LF: Lactoferrin; β‐Lg: β‐Lactoglobuline. Sources: (1) Mussa and Ramaswamy (1997), estimated from z_p/D_p‐values; (2) Ludikhuyze et al. (2000), estimated from z_p/D_p‐values; (3) Rademacher and Hinrichs (2006), estimated from C_f/C_o‐values; (4) Pandey and Ramaswamy (2004), estimated from z/D values; (5) Olsen et al. (2004), estimated from C_f/C_o‐values; (6) Mazri et al. (2012b), estimated from C_f/C_o values; (6) Mazri et al. (2012a), estimated from C_f/C_o‐values (see Section 2.3.2).

Figure 18

Estimates of the pathogen‐specific log₁₀ reductions in milk achieved by HPP at three different pressure‐time combinations (Minimum or HPPMin using 450 MPa for 5 min, Intermediate or HPPInt using 600 MPa for 3 min, Maximum or HPPMax using 600 MPa for 6 min) as reported to be applied by the industry

1. The minimum and maximum reference values to be achieved by thermal pasteurisation (i.e. 5 log₁₀ reductions or TT5 and 8 log₁₀ reductions or TT8) are shown as well as the assumed 12 log₁₀ reductions for UHT milk. Mycobacterium avium subsp. paratuberculosis (MAP) was used as surrogate for Mycobacterium bovis while Shiga toxin‐producing E. coli (STEC) considered E. coli O157:H7 and non‐pathogenic E. coli.

Figure 19

Probability of at least 1 CFU being present per serving of 250 mL (a) and location parameters (i.e. 5th, 50th and 95th percentiles) of the Poisson distributions describing the variability in the level of S. aureus in a 250 mL serving of milk (N₁, in log₁₀ CFU in 250 mL) (b)

1. Results are presented disaggregated by initial contamination level (N _o ).

Figure 20

Probability of at least 1 CFU being present per serving of 250 mL (a) and location parameters (i.e. 5th, 50th and 95th percentiles) of the Poisson distributions describing the variability in the level of L. monocytogenes in a 250 mL serving of milk (N₁, in log₁₀ CFU in 250 mL) (b)

1. Results are presented disaggregated by initial contamination level (N ₀ ).

Figure 21

Probability of at least 1 CFU being present per serving of 250 mL (a) and location parameters (i.e. 5th, 50th and 95th percentiles) of the Poisson distributions describing the variability in the level of Salmonella spp. in a 250 mL serving of milk (N₁, in log₁₀ CFU in 250 mL) (b)

1. Results are presented disaggregated by initial contamination level (N ₀ ).

Figure 22

Probability of at least 1 CFU being present per serving of 250 mL (a) and location parameters (i.e. 5th, 50th and 95th percentiles) of the Poisson distributions describing the variability in the level of E. coli in a 250 mL serving of milk ((N₁, in log₁₀ CFU in 250 mL) (b)

1. Results are presented disaggregated by initial contamination level (N₀).

Figure 23

Probability of at least 1 CFU being present per serving of 250 mL (a) and location parameters (i.e. 5th, 50th and 95th percentiles) of the Poisson distributions describing the variability in the level of M. bovis in a 250 mL serving of milk (N ₁ , in log₁₀ CFU in 250 mL) (b)

1. Results are presented disaggregated by initial contamination level (N ₀ ).

Figure 24

High‐pressure inactivation of different strains of L. monocytogenes inoculated on cooked chicken (a; enumeration done on non‐selective Tryptone Soya Yeast Extract (TSAYE) agar and selective Oxford Listeria Selective Agar (OLSA), Simpson and Gilmour, 1997) and sliced cooked ham (b; enumeration done in chromogenic agar, Serra‐Castelló et al., 2021b)

1. Strains: Lm1 (NCTC 11994, isolate from cheese); Lm2 (isolate from poultry meat of a local processor); CTC1011 and CTC1034 (isolates from RTE meat products); Scott A (clinical isolate).

Figure 25

Listeria monocytogenes inactivation (log (N/N ₀ )) when treating three RTE food categories (cooked meat products, soft cheese and smoked fish) using various pressure (in MPa) and holding time (in min) combinations

1. Source of data for cooked meat products: Bover‐Cid et al. (2019); Chen (2007); Hayman et al. (2004); Hereu et al. (2012a); Hereu et al. (2014); Jofré et al. (2007); Jofré et al. (2008); Lavieri et al. (2014); Lucore et al. (2000); Marcos et al. (2008b); Montiel et al. (2015); Myers et al. (2013); Pavli et al. (2019); Serra‐Castelló et al. (2021b); Simpson and Gilmour (1997); Stratakos et al. (2015); Teixeira et al. (2016); Teixeira et al. (2018). Source of data for smoked and gravid fish: Amanatidou et al. (2000a); Ekonomou et al. (2020); Lakshmanan and Dalgaard (2004); Medina et al. (2009); Mengden et al. (2015); Misiou et al. (2018); Montero et al. (2007); Montiel et al. (2012); Montiel et al. (2014). Source of data for soft or semi‐soft and fresh cheese: Argues et al. (2005); Batty et al. (2019); Carminati et al. (2004); Evert‐Arriagada et al. (2018); Evrendilek et al. (2008); Goncalves et al. (2021); López‐Pedemonte et al. (2007); Martinez‐Rodriguez et al. (2012); Misiou et al. (2018); Morales et al. (2006); Opkala et al. (2010); Shao et al. (2006); Tomasula et al. (2014).

Figure 26

Log D_p‐values(dots) collected from HPP inactivation kinetic reported in the scientific literature as a function of pressure (MPa) for cooked meat products (a) and cheese (b)

1. Black line represents the linear fit, light grey line shows the 95% confidence interval and dark grey line show the 95% prediction interval.
Source of data for cooked meat products: Fonberg-Broczek et al. (2005); Hereu et al. (2012a); Lucore et al. (2000); Serra-Castelló et al. (2021b); Simpson and Gilmour (1997); Teixeira et al. (2016). Source of data for soft or semi-soft and fresh cheese: Carminati et al. (2004); Morales et al. (2006); Shao et al. (2006); Tomasula et al. (2014).

Figure 27

Log D_p‐values (dots) collected from HPP inactivation kinetic reported in the scientific literature as a function of pressure (MPa) for cooked meat products (red circles), cheese (orange diamonds) and smoked fish (blue squares)

1. Black line represents the linear fit, light grey line shows the 95% confidence interval and dark grey line shows the 95% prediction interval.
Source of data for cooked meat products: Fonberg-Broczek et al. (2005); Hereu et al. (2012a); Lucore et al. (2000); Serra-Castelló et al. (2021b); Simpson and Gilmour (1997); Teixeira et al. (2016). Source of data for soft or semi-soft and fresh cheese: Carminati et al. (2004); Morales et al. (2006); Shao et al. (2006); Tomasula et al. (2014). Source of data for smoked and gravid fish: Amanatidou et al. (2000a); Medina et al. (2009).

Figure 28

Observed log₁₀ reduction data (dots) for cooked meat products and the simulation of HPP inactivation (lines) according to the model by Santillana Farakos and Zwietering (2011)

Figure 29

Observed log₁₀ reduction data (dots) for cooked meat products and the simulation of HPP inactivation (lines) according to the model developed by Hereu et al. (2012b)

Figure 30

Isoreduction (log₁₀ units) diagrams showing the combination of pressure and holding time for HPP causing a given number of log₁₀ reductions for L. monocytogenes as predicted by the predictive model developed in Santillana Farakos and Zwietering (2011)

Figure 31

Dispersion of the log₁₀ reduction data (symbols) of Escherichia coli and Salmonella vs. times grouped by pressure intensity, and simulated L. monocytogenes inactivation curves (solid lines) by the models of Santillana Farakos and Zwietering (2011) and Hereu et al. (2012b), including + 1 log₁₀ (dotted lines)

1. Source of data for E. coli in cooked meat products: Chung and Yousef (2010); Garriga et al. (2002); Waite‐Cusic and Yousef (2011); Salmonella in cooked meat products: Al‐Nehlawi et al. (2014); Garriga et al. (2002); Jofré et al. (2008); Kruk et al. (2011); Porto‐Fett et al. (2010); Stiebing et al. (2000); E. coli in smoked and gravid fish: Mengden et al. (2015), Salmonella in smoked and gravid fish: Stollewerk et al. (2014), E. coli in soft or semi‐soft and fresh cheese: Capellas et al. (1996); De Lamo‐Castellvı et al. (2006); Goncalves et al. (2021); O'Reilly et al. (2000); Rodriguez et al. (2005); Shao et al. (2007), Salmonella in soft or semi‐soft and fresh cheese: De Lamo‐Castellvı et al. (2007).