effect of chilling on prevalence of Salmonella spp. on pig - - PowerPoint PPT Presentation

A meta-analysis study of the effect of chilling on prevalence

• f

Salmonella spp.

• n

pig carcasses

Ursula Gonzales Barron, Donal Bergin and Francis Butler

UCD School of Agriculture, Food Science and Veterinary Medicine

INTRODUCTION

 Meta-analysis

refers to ‘the statistical analysis of a large collection of results from individual studies, such as experimental studies, opinion surveys and causal models, for the purpose of integrating the findings’.

 The primary aim of meta-analysis is to

produce a more precise estimate of the effect

• f

a particular intervention

• r

treatment, with an increased statistical power, than is possible using only a single study.

…INTRODUCTION

 There is a need for conducting meta-

analysis in the field of food safety, to start identifying, appraising and summarizing the results of otherwise unmanageable quantities of research, so that policy- makers and decision-makers can access trustworthy and concise information on effectiveness of interventions to control and prevent food-borne illnesses in humans.

OBJECTIVE

 The

• verall
• bjective:

introduce a traditional parametric approach of meta- analysis with the purpose of synthesizing findings of prevalence studies of pathogens within the food processing chain.

 Specific

• bjective:

investigate whether there is support in the sampled population

• f studies for the causal inference that the

chilling stage within pork production had a statistically-significant decreasing effect on Salmonella prevalence of pig carcasses

METHODOLOGY

 A

meta-analysis begins with the formulation of a focused study question: population, intervention or treatment and

• utcome.

Problem statement: Estimation of the overall effect of chilling

• n the Salmonella prevalence of pig carcasses during pork

production. Population: Eviscerated pig carcasses post-meat inspection in slaughterhouses. Intervention

• r

treatment: Chilling stage during pork processing, which includes cooling and posterior cold storage (18-24 hours) at ~5 C. Measured outcome: Presence of Salmonella spp. on the pig carcass surface.

…METHODOLOGY

 The effect size (θ) refers to the degree to

which the hypothetical phenomenon (i.e, decrease in Salmonella prevalence due to chilling) is present in the population (i.e, pig carcasses during processing at slaughterhouses).

 For

studies to be compatible, meta- analysis converts the effect size into a ‘parameter’

• r

common metric that permits direct comparison and summation

• f the independent studies.

…METHODOLOGY

 Effect

size parameterisation (θ) chosen  Relative risk is defined as the probability of the outcome in the treatment group relative to the probability in the control group.

…METHODOLOGY



Table 1. Occurrence of Salmonella-contaminated pig carcasses before and after chilling as detected in individual studies

Study number Study reference Pre-chilling (control) Post-chilling (treatment) sC

1

fC nC sT fT nT

1 2 3 4 5 6 7 8 9 Oosterom et al. (1985) Saide et al. (1995) Davies et al. (1999) UCD study (2000)2 Quirke et al. (2001) Booteldoorn et al. (2003) Bouvet et al. (2003) Lima et al. (2004) Prendergast et al. (2008) 27 3 7 3 6 138 7 5 18 183 267 18 160 413 232 113 25 153 210 270 25 163 419 370 120 30 171 12 1 3 1 1 12 3 4 5 198 269 22 162 418 63 117 26 156 210 270 25 163 419 75 120 30 161

…METHODOLOGY

 Effect size parameterisation (θ) chosen 

Relative risk

 The

se(θi) gives an indication

• f

the degree of precision to which each study estimates the effect size: a small se(θi) indicates a precise estimate, usually from a large study.

C C T T i

n s n s RR ln ln

5 .

) (ln

C C C T T T i

n s f n s f RR se se

…METHODOLOGY

 Since different studies estimate the

true effect size with varying degrees

• f precision, a weighted average is

used to combine individual study estimates.

 A

common method

• f

weighting individual estimates is by means of their inverse variances

2

1

i i

se

…METHODOLOGY

 A fixed-effect meta-analysis makes

the fundamental assumption that each study is estimating the same underlying effect size, with a random error that stems only from a chance factor associated with subject-level sampling error.

i i

…METHODOLOGY

 The global null hypothesis that the

effect size in all studies is equal to zero is tested by comparing the statistic with the chi-squared distribution with one degree of freedom.

i i i

U

2

…METHODOLOGY

 Assuming that there is a common

effect size in all studies, the overall fixed effect θ and its standard error se(θ) are estimated by

i i i

i

se 1

…METHODOLOGY

 Because

meta-analyses are

• ften

performed retrospectively, in many situations it might be expected that differences in the study protocols will produce heterogeneity. A large-sample test for heterogeneity in effect size parameter across studies exists, and it is based on the Q statistic:

U Q

i i i i 2 2

…METHODOLOGY

 If the hypothesis of homogeneity

across studies is rejected, then there must be additional sources of variability other than carcass-level sampling error. Under such a condition, a random-effects model can be assumed.

i i i

RESULTS AND DISCUSSION

 The effect size parameterization of ln-RR led to

a highly-significant U statistic (p<0.001) providing strong evidence of the reduction due to chilling on Salmonella prevalence on pig carcasses.

 All

individual studies presented a negative estimate of ln-RR, which shows their agreement

• n the beneficial effect of chilling.

 The Q statistic for the ln-RR parameterization

was not statistically significant (p=0.96), indicating that there was no strong evidence of heterogeneity among studies  Fixed-effects approach more suitable.

…RESULTS AND DISCUSSION



Table 2. Fixed-effects meta-analysis for the effect size parameterization

• f

‘ln- relative risk’

• f

Salmonella presence on pig carcass after chilling in relation to control (before chilling) Study number Effect size (θi) Standard error (se(θi)) Relative weight (ωi) 1 2 3 4 5 6 7 8 9

• 0.811
• 1.098
• 0.847
• 1.098
• 1.792
• 0.846
• 0.847
• 0.223
• 1.211

0.333 1.151 0.629 1.149 1.078 0.273 0.678 0.619 0.493 9.021 0.754 2.524 0.757 0.861 13.42 2.176 2.609 4.107 U = 27.304; (1 df) p<0.001 Q = 2.448; (8 df) p=0.964 Overall effect θ: -0.868 Standard error of overall effect se(θ): 0.166

…RESULTS AND DISCUSSION

 Forest plots use point

estimates

• f

the individual studies along with their confidence intervals and may help to reveal discernable patterns in the data among studies.

 The

marker size illustrates the contribution

• f

each study to the

• verall

effect estimate

Study 1 Study 2 Study 3 Study 4 Study 5 Study 6 Study 7 Study 8 Study 9 Fixed

• 5
• 4
• 3
• 2
• 1

1 2

Log relative risk (log pT/pC)

…RESULTS AND DISCUSSION

 Proportion of Salmonella-positive carcass

after chilling relative to before chilling was: 0.4197 with a 95% CI of 0.303 – 0.581.

 The meta-analysis of the studies identified,

indicated that chilling, on average, would be expected to reduce the number

• f

carcasses with detectable Salmonella by a factor of 2.38 (1/0.4197) (95% CI: 1.720- 3.299).

CONCLUSIONS

 Because

• f

the systematic approach

• f

meta- analysis and its reliance on actual data, the effect size outcome distribution can be used instead of, or in addition to, expert judgment in quantitative risk assessment models.

 Hence, it is expected that the normal distribution of

the effect size of chilling on Salmonella prevalence from the RR meta-analysis (Prevalence after chilling/Prevalence before chilling) ~ e N(-0.868, 0.166) will provide a more precise and realistic input distribution of the chilling stage for risk assessment models of this pathogen along the pork production process.

…CONCLUSIONS

 This meta-analysis has confirmed that chilling can

be an effective control for pathogenic Salmonella

• spp. when the operation is properly undertaken,

as it may reduce by ~2.4 the detected incidence

• f this pathogen on pig carcass surfaces.

 Finally, this meta-analysis has helped identify:

(i) data gaps in the existent literature with regards to sensitivities for the whole range of Salmonella detection protocols, (ii) a common methodological flaw in the available research, which is the lack of standardization for Salmonella detection in swab samples

• f

pig carcasses.

ACKNOWLEDGMENTS

SafeFood and the Irish Department

• f

Agriculture, Fisheries and Food.