Interventions to reduce Salmonella and Campylobacter during chilling and post-chilling stages of poultry processing: a systematic review and meta-analysis

Abstract

Salmonella and Campylobacter are common bacterial hazards causing foodborne illnesses worldwide. A large proportion of Salmonella and Campylobacter illnesses are attributed to contaminated poultry products that are mishandled or under cooked. Processing interventions such as chilling and post-chill dip are critical to reducing microbial contamination of poultry. A comprehensive search of the literature published between 2000 and 2021 was conducted in the databases Web of Science, Academic Search Complete, and Academic OneFile. Studies were included if they were in English and investigated the effects of interventions against Salmonella and/or Campylobacter on whole carcasses and/or parts during the chilling or post-chill stages of poultry processing. Random-effects meta-analyses were performed using the “meta” package in the R programming language. Subgroup analyses were assessed according to outcome measure reported, microorganism tested, processing stage assessed, and chemical treatment used. The results included 41 eligible studies. Eighteen studies reported results of 28 separate interventions against Salmonella and 31 reported results of 50 separate interventions against Campylobacter. No significant difference (P> 0.05) was observed when comparing the combined mean difference of all interventions targeting Salmonella to the combined mean difference of all interventions targeting Campylobacter or when comparing chilling times within each pathogen subgroup. For analyses examining antimicrobial additives, peroxyacetic acid (PAA) had the largest reduction against Salmonella population regardless of chilling time (P< 0.05). PAA also had the largest reduction against Campylobacter population and prevalence during primary chilling (P< 0.01). Air chilling showed a lower reduction for Campylobacter than any immersion chilling intervention (P< 0.05). Chilling time and antimicrobial used during poultry processing had varying effects depending on the pathogen and outcome measure investigated (concentration or prevalence). High heterogeneity and low sample numbers in most analyses suggest that more high-quality research that is well-designed and has transparent reporting of methodology and results is needed to corroborate the results.

INTRODUCTION

Foodborne illnesses have a major impact on health around the world. The World Health Organization (WHO) estimates 600 million illness and 420,000 deaths from foodborne illnesses per year globally (WHO, 2022). The U.S. Centers for Disease Control and Prevention (CDC) estimates that foodborne diseases cause 48 million illnesses and 3,000 deaths in the United States every year (CDC, 2018). Of these foodborne diseases, Campylobacter and Salmonella are among the most common bacterial hazards causing an estimated 1.5 million and 1.35 million illnesses, respectively, in the United States each year (CDC, 2021, 2022). A large portion of Salmonella and Campylobacter illnesses are attributed to contaminated chicken and chicken products that are mishandled or under cooked. A meta-analysis by Golden and Mishra (2020) reported that, at retail, 55 to 59% and 19 to 23% of broiler chicken products from alternative and conventional production systems were contaminated with Campylobacter and Salmonella, respectively. The Interagency Food Safety Analytics Collaboration (IFSAC) between the CDC, the U.S. Food and Drug administration (FDA), and the U.S. Department of Agriculture (USDA) reported that between 1998 and 2019, 65% of the Campylobacter outbreaks and 17% of the Salmonella outbreaks investigated could be attributed to chicken (IFSAC, 2021).

The CDC's Foodborne Diseases Active Surveillance Network (FoodNet) data suggests that progress in controlling the 8 pathogens monitored by the Network, including Campylobacter and Salmonella, has plateaued because incidence has remained stable or even increased in recent years (Tack et al., 2020). Measures for controlling pathogens such as Salmonella and Campylobacter in poultry can be applied both pre-harvest on the farm and post-harvest through interventions at processing facilities. While control strategies to prevent and reduce colonization of these pathogens in poultry are desirable, none have proven adequate to eliminate the pathogens from poultry products (Robyn et al., 2015; Backert, 2021). Therefore, processing interventions are critical to reducing the contamination of poultry products before reaching retail establishments.

Most modern poultry processing plants follow similar general steps: shackling, stunning, killing, bleeding, scalding, de-feathering, evisceration, inside-outside washing, chilling, and post-chilling (Rasschaert et al., 2020; Backert, 2021). If managed properly, the chilling step is thought to account for the largest reduction of pathogens from carcasses in both population and prevalence (Russell, 2012). The goal of the chilling step is to reduce the temperature of poultry carcasses to 4°C or less in order to minimize bacterial growth (European Parliament and the Council of the European Union, 2004; USDA-FSIS, 2014). In the United States, immersion chilling is the most common technique, whereas in Europe, air chilling is the preferred method (Sams and McKee, 2010; Lu et al., 2019). Several factors are used to enhance efficacy during immersion chilling including the use of 1) several discrete tanks with subsequently decreasing temperatures to gradually cool the carcasses over multiple steps; 2) paddles or an auger system that slowly pushes carcasses through the tanks; 3) counter-current flow where carcasses and water flow in opposite directions, so carcasses encounter increasingly cold and clean water as they move through the chill tank; 4) air injected into the chill tank for agitation of the water to prevent a layer of warmer water forming at the surface of the carcasses; and 5) chemical treatments (Sams and McKee, 2010). During air chilling, carcasses move through large chilled rooms with circulating cold air until they reach optimal deep muscle temperature (<4.4°C) (Davis et al., 2010). Controlling the humidity of the room and spraying, misting, or dipping carcasses with water can facilitate rapid cooling and reduce moisture loss (Sams and McKee, 2010). There is conflicting evidence on which method is better at reducing pathogen levels on carcasses, and the use of either practice may be more related to the regulations of the countries that use them (Huezo et al., 2007; Berrang et al., 2008; Demirok et al., 2013).

A post-chilling step immediately following the primary chiller may be used as a final intervention to reduce pathogens on carcasses before packaging or further processing. This can include another immersion treatment or a spray application. Although post-chill interventions have a relatively short treatment time (≤30 s) compared to the primary chilling step, they can have enhanced efficacy because the treatment solution used has a lower organic load and a higher antimicrobial concentration (McKee, 2011; Nagel et al., 2013).

The aim of this systematic literature review and meta-analysis was to determine the impact of chilling and post-chilling interventions used in poultry processing on the concentration and prevalence of Salmonella and Campylobacter. To our knowledge, this is the first meta-analysis to evaluate post-chilling interventions on Salmonella and Campylobacter in poultry processing.

METHODS

Research Question and Search Strategy

A systematic review and meta-analysis was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guide (Page et al., 2021) as well as guidelines and recommendations from Sargeant et al. (2006). The research question of interest was: “Do the chilling and post-chilling stages of processing reduce the concentration and/or prevalence of Campylobacter and/or Salmonella on poultry?” An initial search was conducted on December 13, 2018 using the terms (poultry OR chicken OR turkey) AND (process* OR product* OR slaughter*) AND (chill* OR "post-chill*") AND (Salmonella OR Campylobacter) AND (control* OR limit* OR intervention OR reduc* OR decontaminat* OR treatment OR prevent* OR inhibit* OR antimicrobial OR sanitiz* OR disinfect* OR inactivat*) and the databases Web of Science, Academic Search Complete, and Academic OneFile. The search was limited to peer-reviewed journal articles published from 2000 to 2018 to assess modern poultry processing practices. Results from this search were imported into the online reference management service RefWorks (ProQuest LLC, Ann Arbor, MI). The “find duplicates” function in RefWorks indicated that “no duplicates were found,” so duplicates were removed manually before screening. An updated search was conducted on January 27, 2022 using the same search terms and databases to find articles published from January 2019 to January 2022. For this search, the EndNote reference management tool (EndNote 20, Clarivate Analytics, Philadelphia, PA) was used to manage the search results and remove duplicates.

Relevance Screening

The title and abstract of each article were reviewed using the following eligibility criteria: 1) English language; 2) peer-reviewed journal articles; 3) interventions focused on chilling or post-chilling stages of poultry processing; 4) interventions tested on whole carcasses or parts; and 5) interventions tested against Salmonella or Campylobacter. After screening abstracts, full text copies of all potentially eligible studies were obtained to perform an in-depth review. Along with the previously mentioned screening criteria, additional eligibility requirements used during full-text review included 1) reporting sufficient detail on the interventions tested (e.g., chilling time, chemical used) and 2) reporting sufficient data to perform a meta-analysis (i.e., sample size, mean, and standard deviation for concentration outcomes and sample size and number of positives for prevalence outcomes for both the control and experimental groups). Initially, articles testing all types of poultry were included. However, articles that examined poultry types other than broiler chickens (e.g., turkeys or ducks) were removed at this screening stage since a limited number of articles were found for these products. Additionally, studies with inconsistencies in reporting of results between text, tables, and figures were excluded. Reasons for exclusion of screened articles were recorded in all cases. The reference lists of included studies and other relevant review articles were manually searched to locate additional published studies that may have been missed in the initial searches.

Data Extraction

Qualitative and quantitative data were extracted manually from all relevant articles that passed full-text screening and were stored in Microsoft Excel 2016 spreadsheets (Microsoft Corporation, Redmond, WA). Collected qualitative data included country where the research was conducted, study design, study setting, treatment method, treatment chemicals and concentrations, exposure time, pathogen tested, matrix tested, and sampling method used. Quantitative data included sample size, mean outcome, and standard deviations of bacterial concentration or prevalence for both control and treatment groups. For articles that only presented data in graphs, the Plot Digitizer tool (Plot Digitizer 2.6.9, http://plotdigitizer.sourceforge.net/index.html) was used to digitize the graphical results.

Data were grouped according to outcome measures reported (concentration or prevalence), microorganism tested (Salmonella or Campylobacter), processing stage assessed (chilling or post-chilling), and chemical treatment used. A number of studies used treatment times that fell between the standard definitions of chilling (45 min to 1.5 h) and post-chilling (≤30 s) (Nagel et al., 2013; Vaddu et al., 2021a), so we settled on grouping the data into full chill (≥30 min), short chill (2–15 min), and post-chill (≤30 s). No assumptions were made according to classification of processing stages, so only studies that explicitly reported data immediately before and after each processing stage were included (e.g., post-washing was not considered the same as pre-chilling). Subgroups needed to contain at least 2 independent studies to be included in the meta-analysis.

Quality Assessment of Included Studies

Systematic reviews and meta-analyses commonly include an assessment to determine the quality of the evidence presented by each study, such as GRADE (Guyatt et al., 2011). However, previous research has shown that these types of quality scores can influence the interpretation of meta-analyses and may create their own selection bias, so quality scores for the included studies were not determined (Jüni et al., 1999; Herbison et al., 2006; Stone et al., 2019).

Data Analysis

Quantitative results were synthesized by meta-analytic approach using the “meta” package, version 5.0-0 (Schwarzer et al., 2015), in the R programming language, version 4.1.2 (R Core Team, 2021). Two summary measures were calculated to address changes in either concentration or prevalence. For studies reporting concentration changes (continuous outcome), the mean, standard deviation, and sample size were extracted for the treatment and control groups of each study. The extracted data were used to calculate raw mean differences rather than standardized mean differences because the reported raw units of measure (log₁₀ CFU/ml, log₁₀ CFU/g, log₁₀ CFU/cm², or log₁₀ CFU/carcass) are inherently meaningful when discussing bacterial concentrations (Borenstein et al., 2009). Since the outcome of interest for the meta-analysis was mean difference (treatment mean minus control mean), no attempt was made to convert the units to a standardized form (e.g., CFU/carcass) before meta-analysis.

For studies reporting prevalence changes (binary outcome), the number of positive samples and the sample size were extracted for the treatment and control groups of each study. From these data, odds ratios (OR) were calculated using the formula:

$$\text{OR}=\frac{\text{Odds}\text{in}\text{the}\text{treatment}\text{group}}{\text{Odds}\text{in}\text{the}\text{control}\text{group}}$$ (1)

where the odds in the treatment group was calculated using the formula:

$$\text{Odds}\text{in}\text{the}\text{treatment}\text{group}=\frac{\text{Number}\text{of}\text{positive}\text{samples}\text{in}\text{the}\text{treatment}\text{group}}{\text{Number}\text{of}\text{negative}\text{samples}\text{in}\text{the}\text{treatment}\text{group}}$$ (2)

and odds in the control group was calculated using the formula:

$$\text{Odds}\text{in}\text{the}\text{control}\text{group}=\frac{\text{Number}\text{of}\text{positive}\text{samples}\text{in}\text{the}\text{control}\text{group}}{\text{Number}\text{of}\text{negative}\text{samples}\text{in}\text{the}\text{control}\text{group}}$$ (3)

Prevalence studies were pooled using inverse variance weighting. When zero-cell counts were present in either the treatment or control groups, a continuity correction of 0.5 was added to all cells of the 2 × 2 table where the problems occurred (Sweeting et al., 2004; Higgins et al., 2022).

Random-effects meta-analyses were performed with subgroup analyses of data based on the groups mentioned above (i.e., outcome measure reported, microorganism tested, processing stage assessed, and chemical treatment used) when 2 or more studies were available. The between-study variance (τ²) was estimated using the DerSimonian and Laird method (DerSimonian and Kacker, 2007; Schwarzer et al., 2015). The effect of heterogeneity was quantified on a relative scale using the I² value with thresholds for interpretation as follows: 0 to 40% heterogeneity might not be important, 30 to 60% moderate heterogeneity, 50 to 90% substantial heterogeneity, and 75 to 100% considerable heterogeneity (Higgins et al., 2003; Schwarzer et al., 2015; Higgins et al., 2022).

Test for Publication Bias Due to Small-Study Effects

Ioannidis and Trikalinos (2007) emphasized the limitations of statistical tests to quantify asymmetry in funnel plots and recommended these tests only be conducted when there is a sufficient number of studies (n ≥ 10) and low heterogeneity (I² < 50%). Unfortunately, none of the meta-analyses examined in this study met these criteria, so publication bias could not be appropriately assessed.

RESULTS

Study Selection

An overview of the systematic review process is provided in Figure 1. Our initial search in 2018 yielded 649 articles while the updated search in 2022 yielded 218 records. After removing duplicates and screening titles and abstracts, 151 and 41 full-text articles were reviewed for eligibility in 2018 and 2022, respectively. After full-text review, 147 articles were excluded. The reasons for exclusion included: 1) not assessing processing (n = 1); 2) not testing chicken products (n = 1); 3) not evaluating chilling or post-chilling stages (n = 31); 4) not testing Salmonella or Campylobacter (n = 9); 5) not testing whole carcasses or parts (n = 14); 6) insufficient reporting of intervention methods (n = 12); 7) insufficient reporting of data to calculate reductions (n = 55); 8) assessing outcomes other than enumeration or prevalence (n = 9); 9) not being in English (n = 12); and 10) only describing modeling parameters (n = 3). Manually searching the reference lists of eligible articles and reviews resulted in 10 additional articles for a total of 55 eligible studies. However, after data extraction, an additional 14 articles were excluded for a total of 41 articles included in the meta-analyses.

Figure 1.

PRISMA flow chart of the systematic review process.

Characteristics of Included Studies

Characteristics of the 41 studies included in the meta-analyses are presented in Tables 1 and 2. Eighteen studies reported results of 28 separate interventions against Salmonella. Of those, there were 12 and 6 studies that reported concentration and prevalence outcomes, respectively. Of the studies reporting concentration outcomes, 6 tested full chill times and 9 tested post-chill. Of the studies reporting prevalence, 6 tested full chill times and none tested post-chill. Full chill interventions included immersion chilling in water (n = 3), chlorine (n = 7), or peroxyacetic acid (PAA, n = 5). Post-chill interventions included immersion in PAA (n = 7) or spray with water (n = 2), polylysine with acidified sodium chlorite (n = 2), or lauric arginate with acidified sodium chlorite (n = 2).

Table 1.

Summary of included studies testing interventions against Salmonella.

Reference	Country	Chill type	Study design	Study setting	Treatment application	Treatment chemical	Matrix	Sampling method	Outcome measures
Benli et al. (2011)	US	PC	CT	Laboratory	Spray	Water	Carcass	Rinse	C
						Polylysine & ACS	Carcass	Rinse	C
						Lauric Arginate & ACS	Carcass	Rinse	C
Benli et al. (2015)	US	PC	CT	Laboratory	Spray	Water	Carcass	Rinse	C
						Polylysine & ACS	Carcass	Rinse	C
						Lauric Arginate & ACS	Carcass	Rinse	C
Boubendir et al. (2021)	CA	FC	B&A, NC	Processing	Immersion	PAA	Carcass	Rinse	P
Fabrizio et al. (2002)	US	FC	CT	Laboratory	Immersion	Water	Carcass	Rinse	C
						Chlorine	Carcass	Rinse	C
Huezo et al. (2007)	US	FC	B&A, C	Pilot plant	Immersion	Water	Carcass	Rinse	C
Kataria et al. (2020)	US	FC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
		PC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
Kumar et al. (2020)	US	PC	CT	Laboratory	Immersion	PAA	Breast Meat	Rinse	C
Lemonakis et al. (2017)	US	PC	CT	Laboratory	Immersion	PAA	Wings & Carcass	Rinse	C
Nagel et al. (2013)	US	PC	CT	Pilot plant	Immersion	PAA	Carcass	Rinse	C
Northcutt et al. (2003)	US	FC	B&A, NC	Pilot plant	Immersion	Chlorine	Carcass	Rinse	P
Northcutt et al. (2008a)	US	FC	B&A, NC	Processing	Immersion	Chlorine	Carcass	Rinse	P
Ramírez-Hernández et al. (2017)	CO	FC	B&A, NC	Processing	Immersion	Chlorine	Carcass	Rinse	P
Russell and Axtell (2005)	US	FC	CT	Pilot plant	Immersion	Water	Carcass	Rinse	C
						Chlorine	Carcass	Rinse	C
Schambach et al. (2014)	US	FC	CT	Pilot plant	Immersion	Chlorine	Carcass	Rinse	P
Scott et al. (2015)	US	PC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
Vaddu et al. (2021a)	US	FC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
		PC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
Vaddu et al. (2021b)	US	FC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
		PC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
Wideman et al. (2016)	US	FC	B&A, NC	Processing	Immersion	Chlorine	Carcass	Rinse	P
						PAA	Carcass	Rinse	P

Abbreviations: ACS, acidic calcium sulfate; B&A, C, before and after, challenged; B&A, NC, before and after, no challenge; C, concentration; CA, Canada; CO, Columbia; CT, challenge trial; FC, full chill; P, prevalence; PAA, peroxyacetic acid; PC, post-chill; US, United States.

Table 2.

Summary of included studies testing interventions against Campylobacter.

Reference	Country	Chill type	Study design	Study setting	Treatment application	Treatment chemical	Matrix	Sampling method	Outcome measures
Berrang and Dickens (2000)	US	FC	B&A, NC	Processing	Immersion	Chlorine	Carcass	Rinse	C & P
Bourassa et al. (2019)	US	FC	B&A, NC	Processing	Immersion	PAA	Carcass	Rinse	P
		PC	B&A, NC	Processing	Immersion	PAA	Carcass	Rinse	P
Boysen and Rosenquist (2009)	DK	FC	B&A, NC	Processing	Air	None	Carcass	Rinse	C
Chantarapanont et al. (2004)	US	SC	CT	Laboratory	Immersion	Chlorine	Skin	Rinse	C
Dickens and Ingram (2001)	US	FC	B&A, NC	Pilot plant	Immersion	Water	Carcass	Rinse	C
Dickens et al. (2000)	US	FC	B&A, NC	Pilot plant	Immersion	Water	Carcass	Rinse	C
Duffy et al. (2014)	AU	FC	B&A, NC	Processing	Immersion	Chlorine	Carcass	Rinse	C
Huezo et al. (2007)	US	FC	B&A, C	Pilot plant	Immersion	Water	Carcass	Rinse	C
					Air	None	Carcass	Rinse	C
Kameyama et al. (2012)	JP	FC	B&A, NC	Processing	Immersion	Chlorine	Breast Meat	Swab	C
Kataria et al. (2020)	US	FC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
		PC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
Kim et al. (2017)	US	PC	B&A, NC	Pilot plant	Immersion	PAA	Carcass	Rinse	C
Kumar et al. (2020)	US	PC	CT	Laboratory	Immersion	PAA	Breast Meat	Rinse	C
					Spray	PAA	Breast Meat	Rinse	C
Li et al. (2017)	US	PC	CT	Pilot plant	Immersion	PAA	Carcass	Rinse	C
						Lactic Acid	Carcass	Rinse	C
Meredith et al. (2013)	IE	PC	CT	Laboratory	Immersion	Water	Skin	Rinse	C
						PAA	Skin	Rinse	C
						ASC	Skin	Rinse	C
						Lactic Acid	Skin	Rinse	C
					Spray	PAA	Skin	Rinse	C
Nagel et al. (2013)	US	PC	CT	Pilot plant	Immersion	Water	Carcass	Rinse	C
						Chlorine	Carcass	Rinse	C
						PAA	Carcass	Rinse	C
Northcutt et al. (2003)	US	FC	B&A, NC	Pilot plant	Immersion	Chlorine	Carcass	Rinse	P
Northcutt et al. (2006)	US	FC	B&A, C	Pilot plant	Immersion	Water	Carcass	Rinse	C
Northcutt et al. (2008a)	US	FC	B&A, NC	Processing	Immersion	Chlorine	Carcass	Rinse	C
Northcutt et al. (2008b)	US	FC	B&A, NC	Pilot plant	Immersion	Water	Carcass	Rinse	C
Osiriphun et al. (2011)	TH	FC	B&A, NC	Processing	Immersion	Chlorine	Carcass	Rinse	C
Oyarzabal et al. (2004)	US	PC	B&A, NC	Processing	Immersion	ASC	Carcass	Rinse	C
Park et al. (2002)	US	SC	CT	Laboratory	Immersion	Chlorine	Wings	Rinse	C
						Water	Wings	Rinse	C
		FC	CT	Laboratory	Immersion	Chlorine	Wings	Rinse	C
						Water	Wings	Rinse	C
Reich et al. (2008)	DE	FC	B&A, NC	Processing	Air	None	Carcass	Rinse	C & P
Sarjit and Dykes (2015)	MY	SC	CT	Laboratory	Immersion	Water	Skin & Meat	Rinse	C
Schambach et al. (2014)	US	FC	CT	Pilot plant	Immersion	Chlorine	Carcass	Rinse	P
Stella et al. (2021)	IT	FC	B&A, NC	Processing	Air	None	Skin	Rinse	C
Vaddu et al. (2021a)	US	FC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
		PC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
Vaddu et al. (2021b)	US	FC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
		PC	CT	Laboratory	Immersion	PAA	Wings	Rinse	C
Wagle et al. (2021)	US	FC	CT	Laboratory	Immersion	PAA	Skin	Rinse	C
Wideman et al. (2016)	US	FC	B&A, NC	Processing	Immersion	Chlorine	Carcass	Rinse	C & P
						PAA	Carcass	Rinse	C & P
		PC	B&A, NC	Processing	Immersion	Chlorine	Carcass	Rinse	C
						PAA	Carcass	Rinse	C & P
Zweifel et al. (2015)	CH	FC	B&A, NC	Processing	Air	None	Skin	Rinse	C & P

Abbreviations: ASC, acidified sodium chlorite; AU, Australia; B&A, C, before and after, challenged; B&A, NC, before and after, no challenge; C, concentration; CH, Switzerland; CT, challenge trial; DE, Germany; DK, Denmark; FC, full chill; IE, Ireland; IT, Italy; JP, Japan; MY, Malaysia; P, prevalence; PAA, peroxyacetic acid; PC, post-chill; SC, short chill; TH, Thailand; US, United States.

Thirty-one studies reported results of 50 separate interventions against Campylobacter. Of those, concentration and prevalence outcomes were reported by 28 and 7 studies, respectively. Of the studies reporting concentration outcomes, 20 tested full chill times, 3 tested short chill, and 10 tested post-chill. Of the studies reporting prevalence, 7 tested full chill and 2 tested post-chill. Full chill interventions included air chilling (n = 5) and immersion chilling with water (n = 6), chlorine (n = 9), or PAA (n = 6). Interventions examining short chill times included immersion chilling with water (n = 2) or chlorine (n = 2). Post-chill interventions included immersion in water (n = 2), chlorine (n = 2), PAA (n = 10), lactic acid (n = 2), or acidified sodium chlorite (n = 2) as well as spray with PAA (n = 2).

Meta-Analyses Comparing Interventions Against Salmonella and Campylobacter

When the combined mean difference for all intervention types targeting Salmonella populations (−1.17 log₁₀ CFU, 95% CI: −1.45 to −0.88) was compared to the combined mean difference for all intervention types targeting Campylobacter populations (−1.20 log₁₀ CFU, 95% CI: −1.42 to −0.97), the test for subgroup differences had a p-value greater than 0.05 indicating no significant difference was observed (Figure 2A). However, significant heterogeneity existed in the analyses of both subgroups (I² = 99%, P< 0.01). Similar results were observed when comparing change in prevalence for both pathogens (Figure 2B). The odds ratio was 0.29 (95% CI: 0.13–0.68) for Salmonella and 0.40 (95% CI: 0.21–0.77) for Campylobacter with no significant difference in the test for subgroup analysis (P> 0.05). Heterogeneity was significant for both Salmonella studies (I² = 59%, P< 0.05) and Campylobacter studies (I² = 76%, P< 0.01) indicating that results should be interpreted with care.

Figure 2.

Forest plot of subgroup analysis for changes in population (A) and prevalence (B) for Salmonella or Campylobacter due to included interventions.

Meta-Analyses of Studies Reporting Data for Salmonella

When comparing the change in population of the chilling and post-chilling steps against Salmonella (Figure 3), the full chill subgroup had an overall mean difference of −1.22 log₁₀ CFU (95% CI: −1.87 to −0.56) while the post-chill subgroup had a mean difference of −1.26 log₁₀ CFU (95% CI: −1.57 to −0.95). However, the test for subgroup differences was not significant (P> 0.05), and there was significant heterogeneity between studies (I² = 98%, P< 0.01). A comparison of the chilling and post-chilling steps against Salmonella using prevalence data could not be made since there was only one study that had sufficient data to be included in the post-chill subgroup.

Figure 3.

Forest plot of subgroup analysis for population of Salmonella by chilling type.

For the analysis comparing change in population due to different full chill interventions against Salmonella (Figure 4A), 3 intervention groups were assessed: immersion in water, chlorine, or PAA. The subgroup that tested plain water immersion showed the lowest reduction with an overall mean difference of −0.59 log₁₀ CFU (95% CI: −1.57 to 0.39). Of the subgroups that assessed immersion with chemical treatments, the PAA subgroup showed a larger mean difference (−1.73 log₁₀ CFU, 95% CI: −1.90 to −1.56) than the chlorine subgroup (−0.87 log₁₀ CFU, 95% CI: −2.66 to 0.91). The test for subgroup differences showed that there was a significant difference between the 3 interventions tested (P= 0.05), however, there was significant heterogeneity in the chlorine (I² = 96%, P< 0.01) and water (I² = 97%, P< 0.01) subgroups. Only the PAA subgroup did not have significant heterogeneity (I² = 51%, P> 0.05).

Figure 4.

Forest plot of full chill subgroup analysis for population (A) and prevalence (B) of Salmonella by treatment.

For the analysis comparing the change in prevalence due to different full chill interventions against Salmonella (Figure 4B), only 2 intervention groups were assessed: immersion in chlorine or PAA. The odds ratio for the chlorine immersion subgroup was 0.45 (95% CI: 0.17–1.19) while the odds ratio for the PAA subgroup was 0.09 (95% CI: 0.02–0.36) indicating a reduction in prevalence regardless of the antimicrobial used. The test for subgroup differences showed no significant difference between the 2 antimicrobials (P< 0.05). Heterogeneity was significant for the chlorine subgroup (I² = 66%, P< 0.05) but not the PAA subgroup (I² = 0%, P> 0.05).

When assessing the change in population due to different post-chill interventions against Salmonella (Figure 5), only the immersion in PAA group could be accurately analyzed. Two studies assessed spray with plain water, polylysine/acidic calcium sulfate, and lauric arginate/acidic calcium sulfate, but one of the studies reported a standard deviation of zero for the control group (Benli et al., 2015), so no summary statistics could be calculated for these subgroups. The PAA immersion group had a fairly large mean reduction of −1.27 log₁₀ CFU (95% CI: −1.62 to −0.92), but heterogeneity was significant (I² = 98%, P< 0.01). A comparison of post-chill interventions against Salmonella using prevalence data could not be made since there was only one study that had sufficient data to be included.

Figure 5.

Forest plot of post-chill subgroup analysis for population of Salmonella by treatment.

Meta-Analyses of Studies Reporting Data for Campylobacter

For the comparison of population changes for varying chilling times against Campylobacter (Figure 6), there were 3 different chill types assessed: full chill (≥30 min), short chill (2–15 min), and post-chill (≤30 s). Each subgroup showed a reduction in population after treatment with the mean difference for full chill being −1.39 log₁₀ CFU (95% CI: −1.78 to −0.99), short chill being −0.97 log₁₀ CFU (95% CI: −1.50 to −0.44), and post-chill being −1.14 log₁₀ CFU (95% CI: −1.54 to −0.73). The test for subgroup differences found no significant difference between the various chill times (P> 0.05), and there was significant heterogeneity (I² = 99%, P< 0.01). Only the full chill and post-chill groups could be assessed using prevalence data (Figure 7) with the full chill group showing a reduction in prevalence after treatment (OR: 0.30, 95% CI: 0.13–0.68). The post-chill group showed an overall increase in prevalence after treatment (OR: 1.11, 95% CI: 0.37–3.33). The test for subgroup differences was not significant (P> 0.05). Heterogeneity was significant for the full chill group (I² = 77%, P< 0.01) but not for the post-chill group (I² = 56%, P> 0.05).

Figure 6.

Forest plot of subgroup analysis for population of Campylobacter by chilling type.

Figure 7.

Forest plot of subgroup analysis for prevalence of Campylobacter by chilling type.

For the analyses comparing the change in population due to different full chill interventions against Campylobacter (Figure 8A), 4 intervention groups were evaluated: immersion chilling in chlorine, PAA, or plain water as well as air chilling. The test for subgroup analysis found a significant difference between the interventions assessed (P< 0.01). While all of the groups showed a reduction in population after treatment, air chilling had the smallest mean difference (−0.50 log₁₀ CFU, 95% CI: −0.96 to −0.04), and chlorine immersion had the largest mean difference (−1.96 log₁₀ CFU, 95% CI: −2.69 to −1.24). Heterogeneity was significant for all subgroups (I² = 96–99%, P< 0.01).

Figure 8.

Forest plot of full chill subgroup analysis for population (A) and prevalence (B) of Campylobacter by treatment.

Prevalence data comparing various full chill interventions against Campylobacter included immersion chilling in chlorine or PAA as well as air chilling (Figure 8B). Similar to the population analysis, all of the subgroups showed reductions in prevalence after treatment, and the test for subgroup differences expressed significant differences between each intervention (P< 0.01). The air chilling group had the least effect on prevalence (OR: 0.91, 95% CI: 0.60–1.38) while immersion in PAA showed a greater reduction in prevalence after treatment (OR: 0.09, 95% CI: 0.02–0.47) than immersion in chlorine (OR: 0.27, 95% CI: 0.10–0.75). Heterogeneity for the subgroups in this analysis was lower than most other analyses with no group having significant heterogeneity (I² = 0–68%, P> 0.05).

Only 2 groups were able to be analyzed for changes in population due to short chill interventions against Campylobacter (Figure 9): immersion in chlorine or plain water. Immersion in chlorine showed a larger reduction (−1.64 log₁₀ CFU, 95% CI: −3.61 to 0.33) than immersion in plain water (−0.89 log₁₀ CFU, 95% CI: −1.60 to −0.18), but the test for subgroup differences was not significant (P> 0.05). Heterogeneity was significant for chlorine immersion (I² = 96%, P< 0.01) but not water immersion (I² = 63%, P> 0.05). A comparison of short chill interventions against Campylobacter using prevalence data could not be made since there were no studies that had sufficient data to be included.

Figure 9.

Forest plot of short chill subgroup analysis for population of Campylobacter by treatment.

Six different interventions against Campylobacter were analyzed for change in population after post-chill treatment (Figure 10A): immersion in chlorine, PAA, acidified sodium chlorite, lactic acid, or plain water as well as spray with PAA. All interventions showed a reduction in population after treatment with the smallest mean difference as a result of immersion in plain water (−0.42 log₁₀ CFU, 95% CI: −0.97 to 0.12) and the largest mean difference from PAA immersion (−1.26 log₁₀ CFU, 95% CI: −1.75 to −0.76). However, the test for subgroup differences was not significant (P> 0.05). Heterogeneity was significant for chlorine immersion (I² = 98%, P< 0.01), PAA immersion (I² = 99%, P< 0.01), water immersion (I² = 96%, P< 0.01), and PAA spray (I² = 94%, P< 0.01).

Figure 10.

Forest plot of post-chill subgroup analysis for population (A) and prevalence (B) of Campylobacter by treatment.

Prevalence data comparing post-chill interventions against Campylobacter (Figure 10B) only included immersion in PAA which had an odds ratio of 1.11 (95% CI: 0.37–3.33) indicating an increase in prevalence after treatment. Heterogeneity was not significant between the 2 studies assessed (I² = 56%, P> 0.05).

DISCUSSION

The chilling and post-chilling stages of poultry processing are important for reducing pathogens since they precede final packaging and further processing operations. No significant differences were found between different chilling types (i.e., full chill, short chill, and post-chill) against Salmonella or Campylobacter. In fact, chilling type had varying effects depending on the pathogen and outcome measure investigated. When comparing treatment times alone, the longer treatment time used for full chill (≥30 min) would be expected to have a greater effect on the reduction of pathogens on poultry than the shorter treatment time used for post-chill (≤30 s) (Nagel et al., 2013; Vaddu et al., 2021a). However, post-chill treatments use higher concentrations of antimicrobials than the primary chiller. For example, PAA is generally used at concentrations up to 220 ppm in the primary chiller and up to 2,000 ppm in post-chill immersion tanks (USDA-FSIS, 2021). These differences in treatment time and concentration may account for the variable reductions seen between chilling types.

Some research suggests that the “washing” effect of immersion chilling in water alone may be enough to reduce pathogen levels on poultry products (Park et al., 2002; Russell and Axtell, 2005; Northcutt et al., 2006). The meta-analyses conducted in this study showed that immersion chilling in water alone did reduce pathogen populations regardless of chilling type, and full chill times seemed to reduce Campylobacter populations more than short chill or post-chill times. However, treatment with water alone had the lowest reduction of all the immersion interventions. These findings are in agreement with a meta-analysis by Dogan et al. (2021) which observed that immersion treatment without antimicrobial additives during primary chilling reduced Campylobacter, but immersion chilling with antimicrobial additives resulted in larger reductions. On the other hand, a meta-analysis by Bucher et al. (2012) that evaluated the effect of primary chilling on Salmonella reported varying results for immersion chilling in water alone. The authors attributed these variations mostly to differences in study design emphasizing the fact that challenge trials tend to report larger reductions than those without a challenge (e.g., before and after trials). These results suggest that the addition of antimicrobials during immersion chilling could be beneficial to reduce pathogens on poultry carcasses.

Historically, chlorine has been used as the main antimicrobial during poultry chilling in the United States with acceptable concentrations up to 50 ppm of free available chlorine (Singh et al., 2019; USDA-FSIS, 2021). However, several factors can influence the effectiveness of chlorine during poultry processing. High loads of organic and inorganic material in the chilling water reduce free chlorine within minutes, even when chlorine is continuously added to the chiller, which decreases the ability of chlorine to act against microorganisms (Tsai et al., 1992; Yang et al., 2002; Bartenfeld et al., 2014). High pH can also decrease the effectiveness of chlorine against microorganisms. As pH increases, available hypochlorous acid (HOCl) is converted to hypochlorite ions (ClO^–) which have lower antimicrobial activity than HOCl (Bashor et al., 2004). Finally, chlorine treatments tend to be more effective when longer contact times are used (Tsai et al., 1992). Our meta-analyses found that immersion chilling with chlorine showed a reduction in both population and prevalence of Salmonella and Campylobacter with slightly greater reductions for Campylobacter seen during full chill and short chill compared to post-chill. Similar results were found by Bucher et al. (2012) as well as Dogan et al. (2021) when assessing primary immersion chilling with chlorine against Salmonella and Campylobacter, respectively. However, our study found that chlorine immersion showed lower reductions than most other antimicrobial additives while Dogan et al. (2021) found that chlorine immersion had a higher reduction in Campylobacter population than other additives. The other additive tested in the meta-analysis by Dogan et al. (2021) was an herbal extract which may have lower effectiveness than the traditional chemical antimicrobials tested in our study.

In recent years, peroxyacetic acid (PAA) has replaced chlorine as the antimicrobial additive of choice in poultry processing operations for both primary and post-chill immersion (McKee, 2011; Ebel et al., 2019). PAA is an equilibrium solution of peroxyacetic acid, acetic acid, hydrogen peroxide, and water (Kim and Huang, 2021). The U.S. Department of Agriculture's Food Safety and Inspection Service (USDA-FSIS) has approved the use of PAA at concentrations of up to 2,000 ppm, but it is generally used at concentrations up to 220 ppm in the primary chiller and up to 2,000 ppm in the post-chiller (Kataria et al., 2020; USDA-FSIS, 2021). Some reasons why PAA is not used at higher concentrations during primary chilling include potential yield losses from the carcasses, potential “graying” of the carcasses, greater hazard to plant employees, and greater cost associated with maintaining the higher concentration for longer periods of time (Bourassa, 2017). The greater effectiveness of PAA seen in this meta-analysis compared with chlorine could be due to several factors. First, changes in pH have less of an influence on the effectiveness of PAA than on chlorine (Kataria et al., 2020; Vaddu et al., 2021a,b). This is useful because poultry processors prefer using higher pH values in the primary chiller (≥8.0 pH) to prevent moisture loss from carcasses and to improve carcass yields (Vaddu et al., 2021a). Also, the effectiveness of PAA seems to be less affected by organic matter than chlorine (Briñez et al., 2006). However, studies examining the specific impact of organic matter in the poultry chiller on the effectiveness of PAA are lacking and further research is needed. Finally, while PAA shows a greater reduction of pathogens with longer contact times (e.g., 60 min), short contact times such as those used in post-chill interventions (e.g., <30 s) still have the ability to reduce pathogen levels on poultry by 1–2 log₁₀ CFU (Nagel et al., 2013; Kataria et al., 2020; Laranja et al., 2021; Vaddu et al., 2021a). In our meta-analyses, PAA immersion showed the largest reduction in population and prevalence for all Salmonella analyses. However, the results for Campylobacter were more varied with chlorine showing the greatest reduction for full chill population data, PAA showing the greatest reduction for full chill prevalence and post-chill population data, and PAA showing an increase rather than a reduction for post-chill prevalence data. Similar published meta-analyses by Bucher et al. (2012), Bucher et al. (2015), and Dogan et al. (2021) did not specifically assess PAA during chilling against Salmonella or Campylobacter.

While immersion chilling is common in the United States and Australia, air chilling is more common in Europe, Brazil, and Canada (Chen et al., 2020; Belk et al., 2021). During air chilling, poultry carcasses are cooled using circulating cold air in a humidity controlled environment (Davis et al., 2010; Sams and McKee, 2010). There is conflicting evidence on whether immersion or air chilling is more effective. On the one hand, the physical removal of microorganisms during immersion chilling (e.g., the washing effect) may be an advantage over air chilling (Northcutt et al., 2006; Buncic and Sofos, 2012). However, air chilling seems less likely to cause cross-contamination than immersion chilling (Stella et al., 2021). One way to prevent cross-contamination during immersion chilling, as well as reduce microorganisms present on carcasses, is the use of antimicrobial additives, such as chlorine or PAA, in the chilling water (Xiao et al., 2019), but these antimicrobials are not generally used during air chilling (Buncic and Sofos, 2012; Chen et al., 2020). In our meta-analyses, we were only able to assess articles that examined the effectiveness of air chilling on Campylobacter during primary chilling. The results found that air chilling had the lowest reduction in population and prevalence for Campylobacter compared to all other full chill interventions. Similar results were seen in other meta-analyses examining primary chilling of Salmonella and Campylobacter (Bucher et al., 2012; Dogan et al., 2021).

While both primary chilling and post-chilling interventions reduce microorganisms on chicken carcasses during processing, neither are effective at completely eliminating pathogens by themselves. This is one reason poultry processors rely on the multiple hurdle approach, i.e., the use of multiple-sequential interventions to augment microbial reductions as products move through the processing line (Leistner, 1995). The studies assessed in this meta-analysis evaluated chilling and post-chilling interventions independently to limit confounding factors during analysis. However, information on whether each sequential intervention used during processing could have an additive effect are necessary to determine the overall impact of processing on poultry products as well as the most efficient use of resources for processors.

Limitations

Some limitations exist in the meta-analyses presented here. First is the number of studies that were excluded during screening due to inadequate reporting of methodology and results. A number of studies did not include enough information on the intervention tested, such as whether an antimicrobial was used or which antimicrobial was used. Also, this systematic analysis made no assumptions according to classification of processing stages, so a number of studies were excluded because they did not report taking samples immediately before and after chilling or post-chilling. For example, studies that took samples post-washing and considered that the same as pre-chilling were not included in our analyses. Finally, several studies did not report sufficient data to calculate reductions in bacterial populations and were excluded.

Most of the meta-analyses conducted in this study had significant levels of heterogeneity, so care should be taken when interpreting the results. Summary effect estimates were still presented regardless of heterogeneity because heterogeneity is common in meta-analyses, and we wanted to provide a foundation for future research. Much of the heterogeneity is likely explained by the wide range of experimental conditions of the included studies. For example, there were differences in the study design (e.g., challenge or no challenge), study setting (e.g., laboratory, pilot plant, or commercial plant), treatment time, matrix (e.g., whole carcass, skin, or meat), sampling type (e.g., rinse or swab), antimicrobial composition, and antimicrobial concentrations. Other meta-analyses have grouped interventions by study design (e.g., before and after versus challenge) or study setting (e.g., laboratory and pilot plant versus commercial plant) in order to reduce heterogeneity. However, we did not do this because we wanted to present a greater number of analyses as a starting point for future research. The large number of meta-analyses with significant levels of heterogeneity limits the generalizability of our results.

Finally, due to the high heterogeneity and small number of studies included in each meta-analysis, we were unable to assess and report publication bias. Publication bias is the increased chance of small studies to be published if they have a stronger effect (Schwarzer et al., 2015). Various methods exist for determining publication bias, but research has shown that these statistical tests have limitations. Therefore, it is recommended that publication bias only be conducted when there are more than 10 studies and low heterogeneity in a meta-analysis (Ioannidis and Trikalinos, 2007).

CONCLUSIONS

No significant difference was seen when all interventions against Salmonella were compared to all interventions against Campylobacter, but high heterogeneity limited the interpretation of these results. Similarly, there was no significant difference in chill times for Salmonella or Campylobacter, and heterogeneity was high in all subgroups except the Campylobacter post-chill prevalence group. For analyses examining antimicrobial additives, PAA had the largest reduction against Salmonella population and prevalence regardless of chill time. Results were more varied for Campylobacter with PAA immersion showing the largest reduction for full chill and post-chill populations only. Finally, air chilling showed lower reductions for Campylobacter than any immersion chilling intervention. The high heterogeneity and low number of samples in most analyses means that more high-quality research that is well-designed and has transparent reporting of methodology and results is needed to clarify the results presented here.

DISCLOSURES

The authors declare no conflicts of interest.