Meta-analysis of effects of dietary vitamin E and post slaughter storage conditions on changes of redness (a*) of pork

A meta-analysis was carried out to quantify the effects of dietary vitamin E and storage conditions on colour changes of pork from M. longissimus dorsi . After standardisation procedures, redness of pork (CIE colour specification a*), one of the most important objective colour attributes, was used as an indicator for colour changes in this analysis. The analysis was based on results from five experiments, which met selection criteria. Analysis of changes of other objective colour attributes, lightness (L*) and yellowness (b*) was not possible due to lack of published data. The statistical analysis (using mixed models) found significant effects of tissue α-tocopherol concentration in M. longissimus dorsi , simplified supplemented vitamin E levels as well as storage time and storage light on redness of pork and its changes over time. The relationship between redness and α-tocopherol concentration was found to be linear, and between redness and storage time was non-linear (third degree polynomial) in one model. This model suggested that an increase of 1 μg of α-tocopherol in the muscle led to an expected increase a* value of 0.11. Another model identified significant interactions about 0.28 between α-tocopherol concentration and storage time in late storage periods. A third model found a significant difference of -0.48 between predicted a* values at lower (≤50 IU/kg feed) and higher supplemented vitamin E levels (≥100 IU/kg feed). The models predicted an initial increase for 3 days, a stable period for 5 days and then a decrease for a* values over storage time. The a* values were significantly lower by about 1.4 when samples were exposed to light in the models, the effect of light found to be constant over time. Further studies, carried out with standardized methods, are needed to increase the predictive power of the derived models and to validate the models for other muscles. der α-tocopherol Konzentration und der Lagerzeit. Ein drittes Modell fand einen signifikanten Unterschied von −0.48 in den a*-Modellwerten, entsprechend einer niedrigen (≤50 IU/kg feed) oder einer hohen (≥100 IU/kg feed) Dosierung von Vitamin E Zusatz. Die Modelle berechneten eine Erhöhung des a*-Wertes in den ersten drei Tagen nach der Schlachtung, gefolgt von einer stabilen Periode für etwa fünf Tage, und einer Verringerung über den restlichen Zeitraum. beeinflusst. Untersuchungen mit standardisierten Methoden wären von großem Nutzen für die Modell Validierung und die Schätzgenauigkeit der abgeleiteten Modelle erhöhen. diet (e.g. gross energy, crude protein, fat concentration), rearing conditions (e.g. indoor/outdoor, group size), slaughter processes (e.g. stunning method ), pork quality traits studied, sample size and treatment (e.g. storage temperature) as well as results from statistical analyses performed on the datasets from individual experiments.


Introduction
Visual appearance is an important characteristic of meat quality, as it strongly influences the consumer's purchase decision. From the various factors which contribute to the visual appearance of meat, colour is the main factor that affects the acceptability of meat products (FAUSTMAN and CASSENS 1990). The rate of discolouration of meat is related to several factors including ultimate pH, storage temperature, lipid oxidation and illumination (FAUSTMAN and CASSENS 1990). To maintain acceptable fresh meat colour over prolonged periods of time, it is necessary to delay pigment oxidation and/or enhance reduction of oxidized myoglobin (GRAY et al. 1996). The supplementation of feed with vitamin E, a well known antioxidant was found to significantly reduce the oxidation process (BUCKLEY et al. 1995). For beef, vitamin E supplementation resulted in an extension of retail display life from 1.6 to 5 days without compromising microbiological quality and similar trends have been reported for pork products (GRAY et al. 1996). There is strong evidence from various studies and review papers that dietary vitamin E supplementation decreases lipid oxidation in pork (BUCKLEY et al. 1995, LAHUCKY et al. 2000, PETTIGREW and ESNAOLA 2001, ROSENVOLD and ANDERSEN 2003, DUNSHEA et al. 2005, LAHUCKY et al. 2005, DIKEMAN 2007). However, results from individual studies on the effect of vitamin E on pork colour changes are ambiguous (ROSENVOLD and ANDERSEN 2003). While some papers have reported improved colour stability with vitamin E supplementation (e.g. ASGHAR et al. 1991, MONAHAN et al. 1992, several others found no significant effect (e.g. CANNON et al. 1996, JENSEN et al. 1997, PHILLIPS et al. 2001, KRSKA et al. 2001. To resolve this ambiguity and to determine quantitatively whether vitamin E influences colour changes in pork using a cross-experimental approach, meta-analysis may be used.
In animal science, meta-analysis has recently been described as a useful statistical tool to objectively integrate the results of individual experiments and to establish general response laws that are valid over a range of conditions (ST-PIERRE 2001, PHILLIPS 2005and SAUVANT et al. 2008. The aim of this paper was to use meta-analysis to quantify the effects of different dietary vitamin E levels and post slaughter storage conditions (e.g. storage time, storage light) on the changes of pork colour during storage. This paper focuses on the analysis of redness (a*) as an attribute of objective colour measurement of pork using the CIE L*,a*,b* colour system (CIE 1976).

Data sourcing and construction of meta-analysis datasets
The literature search used published information written in English from 1970 to 2009, from various scientific sources including journals, conference proceedings, book articles and abstracts from various electronic databases. Indices used in the search were combinations of the terms ›vitamin E‹, ›alpha-tocopherol‹, ›α-tocopherol‹, ›pork‹, ›dietary‹, ›pig‹, ›swine‹, ›hog‹ and plurals of these. In some cases, authors were contacted directly to complete missing information.
The literature research identified 60 initial references, related to the effect of dietary vitamin E supplementation on pork and pigs. Some studies reported aspects of dietary vitamin E utilization (e.g. blood α-tocopherol concentration), which were not relevant to the analysis and were therefore excluded. Since this study focused purely on the antioxidant effect of vitamin E, references using additional treatments which could possibly interfere with or modify the antioxidant effect of the vitamin E were also excluded. This was the case for experiments using diets with a fat to dry matter (DM) ratio of more than 5 %, as this may change saturated and unsaturated fatty acid ratio in pork making it more susceptible to oxidation (BUCKLEY et al. 1995). Experiments using diets containing in addition to vitamin E other antioxidants (e.g. vitamin C) or minerals (e.g. Se) above the normal maintenance level, which may interact with the vitamin E (OLDFIELD 2003), were also excluded. These selection criteria together with the standardisation criteria outlined below led to a drastic reduction of acceptable studies and the following references remained for inclusion in the meta-analysis (LANARI et al. 1995, DIRINCK et al. 1996, HOVING-BOLINK et al. 1998, PHILLIPS et al. 2001and VAN HEUGTEN et al. 2002. Table 1 summarizes experimental details of these studies. Experimental data were entered into a database that had been constructed according to the conceptual and structural requirements for meta-analysis outlined by VERNET and ORTIGUES-MARTY 2005. The database includes a detailed description of all relevant references (e.g. type, name and date of publication, title, author names), animals used in the experiments (e.g. gender, breed, beginning or slaughter weight), diet (e.g. gross energy, crude protein, fat concentration), rearing conditions (e.g. indoor/outdoor, group size), slaughter processes (e.g. stunning method ), pork quality traits studied, sample size and treatment (e.g. storage temperature) as well as results from statistical analyses performed on the datasets from individual experiments.

Dose definition and storage time standardiation
All experiments used in the analysis supplied vitamin E in form of all-rac-α-tocopheryl (DL-α-tocopheryl) acetate, which is an equimolar mixture of eight synthetic α-tocopherol stereoisomers in acetate form (LAURIDSEN et al. 2002). 1mg of this chemical compound has a bioactivity conversion factor of 1.0 and corresponds to 1 International Unit (IU) of α-tocopherol. For the three different terminologies referring to α-tocopherol that had been used in the studies to define vitamin E administration, i.e. vitamin E (e.g. HOVING-BOLINK et al. 1998) , α-tocopherol (e.g. DIRINCK et al. 1996 or analysed fed α-tocopherol (e.g. LANARI et al. 1995) the amount of supplementation could thus be expressed in IU of α-tocopherol.
In one study (PHILLIPS et al. 2001) different amounts of vitamin E had been administrated for different time periods throughout the supplementation time. We defined the administrated vitamin E (in IU) in this case by calculating weighted averages with the individual time periods as weighting factors.
Furthermore the different experiments varied in their definition of control or treatment vitamin E levels and in the accuracy of the provided vitamin E measures. To minimise the potential bias introduced by this variation, we rounded administrated vitamin E in units of 50 IU/kg feed. For example amounts of supplementation below 25 IU/kg feed were defined as vitamin E dose 0 IU/kg feed, and supplementations between 25 and 75 IU/kg feed were defined as vitamin E dose 50 IU/kg feed. Henceforth the rounded measure is denoted as vitamin E dose.

Standardization procedures related to colour of pork
Objective colour parameters of meat based on reflectance method can be given in different systems, e.g. HUNTERLAB (2008) or CIE (1976. All references used in the present study provided pork colour in term of the CIE colour system, which describes colour by three parameters: lightness (L*), redness (a*), and yellowness (b*) (CIE 1976). However only the redness (a*) parameter of this colour system was assessed in this paper, because of lack of sufficient data for the other two colour attributes. The value as a* value of this colour parameter is going to be equally referred as redness of pork in this paper. As storage conditions affect the rate of colour changes (FAUSTMAN and CASSENS 1990) redness data were considered from studies in which samples were: (i) stored at 4-7 °C; (ii) wrapped in oxygen permeable material (PVC, plastic foil); and (iii) stored under atmospheric conditions i.e. non oxygen/carbon-dioxide elevated or vacuumed storage environment. If samples were exposed to illumination, redness data received the attribute ›light‹ and if samples were kept either in dark or where the description of storage conditions demonstrated that samples were not exposed to light, they received the attribute ›no-light‹ in the metaanalysis database.
In addition, it was necessary to standardize the time scales provided in different experiments to a uniform scale. The time of slaughter was considered as the starting point of storage (day 0) and expressed the storage time in units of days post slaughter.

Statistical analysis
The aim of this meta-analysis was to derive statistical models which describe how redness of pork changes over time, and how the trend is affected by factors e.g. dietary vitamin E dose, storage conditions.
The inputs for the meta-analysis were the results of the statistical models used in the original publications in the form of least square (LS) means or means (if LS means were not provided) and the corresponding standard errors of LS means (SEM) or standard deviations (SD) where SEM were not provided.
Meta-analysis was performed using the PROC MIXED procedure of SAS (SAS 2000) where the effects of different covariates and fixed effects as well as individual study effects and autocorrelation between repeated measurements could be taken into account (VERBEKE andMOLENBERGHS 1997, WHITEHEAD 2002). The fitted models contained α-tocopherol concentration of M. longissimus dorsi as covariate, vitamin E dose, vitamin E supplementation time and storage light as fixed effects, as well as interaction terms between covariates and fixed effects. Effect of storage time was examined separately in the models either as covariate or as fixed effect. Including storage time in the model as a continuous covariate provides a description of the time trend of redness, whereas inclusion as a fixed effect lends itself better to the statistical comparison of redness associated with different stages during storage and of interactions between storage time and other factors influencing it. The random experimental effect was included as a »random intercept«. For one study (DIRINCK et al. 1996) that did not report muscle α-tocopherol concentrations, LS means predicted from a nonlinear model (TREFAN et al. 2010) were used. Autocorrelations between repeated measurements were examined using the repeat statement of PROC MIXED, where vitamin E dose was nested within experiment identifier and various residual covariance matrix structures (SAS 2000) were explored.
To compare the models, the following model fitting criteria were used: correlation between the predicted and observed values through Pearson correlation R; Akaike's information criteria (AIC) (SAS 2000); RSD (residual standard deviation) where ε ij is the residual value of i-th experiment at j-th dose level, n is the number of observations and p is the number of parameters in the model; CD (coefficient of determination) where ε ij is the same as in RSD and Y ij is the predicted value i-th experiment at j-th dose level (BÜNGER and HERRENDöRFER 1994). In addition, visual inspection of predicted values, residuals and normality tests of residuals were carried out (SAUVANT et al. 2008).
Since research designs and accuracy varied across studies, weighting factors based on the published experimental errors were included in the models. The weighting factor was the inverse of the square of these errors divided by their mean value (ST-PIERRE 2001). In one study (DIRINCK et al. 1996) neither SEM nor SD were published, and error terms were estimated based on the coefficient of variations of the other experiments. Weighting of models was performed using the WEIGHT statement of PROC MIXED procedure of SAS (SAS 2000).

Results
After stepwise removal of statistically non-significant factors influencing redness and comparison of different assumptions for storage time dependence (i.e. linear, quadratic, cubic), the statistical model (1) provided the best fit out of all models where storage time was a continuous variable (AIC: 227.8, R: 0.85, RSD: 0.96, CD: 0.99). According to this model accumulated α-tocopherol concentration in pork (at slaughter), storage time and storage light have significant effects on redness of pork: where a* ijkl is the a* value of i-th experiment of j-th dose level at the k-th storage time point under l-th storage light condition; TOC_cont ij is the α-tocopherol concentration in M. longissimus dorsi of the i-th experiment of j-th dose level and ß 0 its slope; Time ijk, ,Time 2 ijk , Time 3 ijk are the storage time, square and cube of k-th storage time point of the i-th experiment of j-th dose level, respectively, ß 1, ß 2 and ß 3 their corresponding coefficients; StoreLight l is the effect of storage light (l=0 refers to ›no light‹ and l=1 refers to ›light‹); b i is the random effect (»random intercept«) of the i-th experiment and e ijkl denotes the residuals. There were no significant interactions between the different factors.
The slope ß 0 for muscle α-tocopherol concentration and the coefficients ß 1, ß 2 and ß 3 for linear, quadratic and cubic storage time terms as continuous variables were significantly different from zero (P<0.0309, P<0.0001, P<0.0001 and P<0.0011, respectively). Muscle α-tocopherol concentration was found to have a positive effect on redness with a slope value ß 0 =0.11 (SE: 0.05). The linear, quadratic and cubic storage time coefficients ß 1 , ß 2 and ß 3 , respectively were ß 1 =1.04 (0.21), ß 2 =−0.13 (0.03) and ß 3 =0.004 (0.001). A significant difference of 1.35 (SEM: 0.23) (P<0.0001) was found for the predicted redness with lower a* values when light was present during storage.         Confidence intervals of the individual predicted a* values ranged from 0.34 to 0.92. Distribution of the residuals was uniform over storage time. The predicted values are presented according to whether pork samples were stored in dark ( ) or exposed to illumination ( ).
times showed that the interaction was significant at 1, 6 and the last 3 days post slaughter storage times and their LS means were −0.70 (0.25), 0.39 (0.19), 0.30 (0.12), 0.28 (0.12), 0.28 (0.12), respectively, implying that α-tocopherol concentrations at slaughter affect the change in redness at rather later stages of the oxidation process. This coincides with the period in which redness changes most as LS means of a* values estimated from model (2) increased from 7.58 (0.54) to 9.30 (0.63) over the first 3 days post slaughter, after which values remained approximately stable until day 8 after which they gradually decreased to 5.05 (0.37). Multiple comparisons of LS means of redness between the different storage days generally showed no significant differences between successive storage days, but significant differences were found between estimates corresponding to short (i.e. four days or less) and long (above a week) storage. As in the case of model (1) a significant difference of 1.43 (0.19) (P<0.0001) was found for predicted redness with lower a* values when illumination was present during storage. Confidence intervals of the individual predicted redness varied between 0.47 and 1.47. Distribution of the residuals was uniform over storage time.
In the derivation of the statistical models vitamin E dose as a multi-level fixed effect as well as its accumulation in the form of α-tocopherol were introduced into the models. However, α-tocopherol concentration was found to be the better predictor for redness, and the effect of vitamin E was not significant at the 95 % confidence limit and hence not included in these models. In order to answer the question whether supplementation of vitamin E has a direct effect on the change of redness of pork, a further model was tested in which vitamin E dose was attributed to one of two dose categories corresponding to doses either equal to or less than 50 IU/kg feed (control) or greater than or equal to 100 IU/kg feed (supplementation), respectively. The statistical model applied to the data contained the two simplified dose categories, storage time and storage light as fixed effects. Although the fit of the model did not improve (AIC: 232.7, R: 0.88, RSD: 0.97, CD: 0.99), the model was: a* rstv = simpDose rs + StoreTime t + StoreLight v + b r + e rstv where a* rstv is the a* value of r-th experiment of s-th simplified dose level at the t-th storage time point under v-th type storage light condition; simpDose rs effect of s-th evel of simplified vitamin E dose of r-th experiment; StoreTime t is the effect of time at t-th storage time point; StoreLight v is the effect of storage light (v=0 refers to ›no light‹ and v=1 refers to ›light‹); b r is the random effect, as random intercept of the r-th experiment and e rstv denotes the residuals. All fixed effects were found to be significant, and interactions between fixed effects were found to be non-significant at the 95 % confidence limit. Model (3) gave similar estimation for changes of LS means of redness over time as model (2) e.g. estimated a* values increased from 7.64 (0.60) to 9.20 (0.60) on the first 3 days post slaughter, after which values remained approximately stable until day 8 after which they gradually decreased to 5.10 (0.42). Furthermore multiple comparisons of estimated redness between the different storage days showed the same patterns as in the case of model (2) and a significant difference of 1.47 (0.22) (P<0.0001) with the same magnitude of model (1) & (2) was found between light and dark storage conditions. In particular a significant difference of -0.48 (0.19) (P<0.0162) was found between predicted redness at lower (≤50 IU/kg feed) and higher vitamin E doses (≥100 IU/kg feed). Confidence intervals of the individual predicted a*values varied between 0.55 and 1.20. Distribution of the residuals was uniform over storage time.
Random intercepts were found to be not significant in either model, implying that experimental conditions of the individual studies did not contribute to the modelled values. Weighting did not improve the fitting of the models, nor did it change general tendencies in results and only marginally modified predicted values. Hence, weighting factors were excluded from the final models.
The present meta-analysis was based on the period between days 0 and 17 post slaughter. However, data between days 11 and 17 originated from a single reference (LANARI et al. 1995). To test for potential bias in the statistical results caused by this single reference for later storage times, the meta-analysis was also carried out using only data from the period between days 0 and 11 post slaughter. Removal of the data referring to the later storage times was however found to have only negligible influence on the statistical model results.

Discussion
In this study a meta-analysis was performed to determine the effects of dietary vitamin E as an antioxidant on the changes of redness of pork during storage.
The meta-analysis combined results from five independent studies to establish quantitative relationships between the changes in pork redness with time and influencing factors of significant importance. To the best of the authors' knowledge this study is the first to quantitatively describe the time dependence of the discolouration process of pork when related to dietary vitamin E and storage in different light conditions. Model (1) revealed a significant influence of accumulated α-tocopherol concentration in M. longissimus dorsi, as well as storage time and storage light on redness and its change over time. The relationship between redness and α-tocopherol concentration was linear, whereas the change of redness with storage time was best described by a third order polynomial. The moderate but positive slope value associated with α-tocopherol concentration in model (1) suggested that an increase of 1 μg of α-tocopherol in the muscle as a result of dietary vitamin E led to an expected increase of 0.11 in redness across all storage times.
A second model was derived using storage time as a fixed effect. Model (2) found an interaction between α-tocopherol concentration of M. longissimus dorsi and storage time, indicating that muscle α-tocopherol concentration (and hence vitamin E supplementation) affects the evolution of redness over time. This influence was more frequent after 6 days of storage. Model (2) further shows that redness decreases only slowly over time, and that differences become significant only after about a week of storage.
A third model was developed to establish quantitative relationship between supplemented vitamin E dose, storage conditions and changes in redness. In model (3) definition of dose categories was required to quantify the effect of dose. This simplification may have been due to the relatively low number of data points (80) from 5 references compared to the available number of vitamin E dose levels (7), however this model predicted the changes of redness over time as model (2). Considering redness changes over time the initial increase in redness during storage is coherent with findings of earlier studies unrelated to the effects of vitamin E (LINDAHL et al. 2006, LI et al. 2009). While these two studies found that after the initial increase, redness remained stable, other studies found a decrease in redness over several days of storage (JUNCHER et al. 2001, TIKK et al. 2008, which is in accordance with the results of this metaanalysis.
To correctly predict redness, all models needed to take into account whether samples were exposed to illumination. This result is a good agreement with the existing literature (FAUSTMAN andCASSENS 1990, HONIKEL 1998) which shows that redness of pork and pork products decrease during exposure to light (CALKINS et al. 1986), (ANDERSEN et al. 1988. Storage light was found to have a significant influence on redness of pork with a difference of about 1.4 between light and dark storage conditions and this effect was found constant over all storage time by this meta-analysis. Several other factors that may influence the relationship between vitamin E and colour changes could not be assessed in the present analysis. The length of time supplementary vitamin E was fed was identified as a non-significant factor in the models. Earlier studies established that blood α-tocopherol concentration of pigs stabilise after more than 2 weeks of vitamin E supplementation and a linear relationship was found between blood-and tissue α-tocopherol concentration at slaughter (HOPPE et al. 1993). The duration of vitamin E supplementation was considerably longer than 14 days pre slaughter in all references used for the present meta-analysis (Table 1). Thus tissue α-tocopherol concentrations may have reached equilibrium with vitamin E supply, which might explain why supplementation times did not significantly contribute to the models. Furthermore, although studies included in the meta-analysis gave information on the breeds used, mainly modern crossbreds (Table  1), any breed effect could not be assessed. The reason is that pigment (myoglobin) content of muscle fibres is primarily responsible for meat colour (HEDRICK et al. 1994, FAUSTMAN andCASSENS 1990) and no marked difference in muscle fibre numbers and size of the longissimus muscle were apparent between modern meat-type pig breeds and crosses (PAS 2004). Similarly, the present meta-analysis concerns only redness of a specific muscle, M. longissimus dorsi. Among the references used in this analysis only two (HOVING-BOLINK et al. 1998, PHILLIPS et al. 2001) published changes of a* values of M. psoas major over time. This dataset was too small to carry out a meta-analysis for this muscle. Finally among the references, which satisfied our selection criteria, only 2 references (DIRINCK et al. 1996, VAN HEUGTEN et al. 2002 contained information on time dependent changes in lightness (L*) and only one ( VAN HEUGTEN et al. 2002) on time dependent changes in yellowness (b*) following the different amounts of vitamin E administration. In summary the present study used metaanalysis based on a small number of carefully selected experiments to establish quantitative relationships for the effect of dietary vitamin E supplementation on change of redness of pork under different storage conditions.
In conclusion, the three statistical models together suggest that vitamin E supplementation affects redness of pork, but only when supplementation of dose exceeds 100 IU/kg and after 6 days post slaughter.
The number of experiments included in this meta-analysis was small, as the aim was to produce statistically robust estimates rather than less reliable estimates based on a larger number of studies but inconsistent experimental conditions. Further studies would be necessary to confirm the results, to establish and validate the found relationships for other muscles and colour attributes.