Abstract:
63 Strategies for the assessment of livestock feed ingredient quality R.J. van Barneveld Barneveld Nutrition Pty Ltd and the BECAN Consulting Group, PO Box 42, Lyndoch SA 5351 Summary Potential exists to improve current strategies and methods employed to assess livestock feed ingredient quality. Of paramount importance to any assessment procedure is representative sampling of the test ingredient. Sampling must be sufficient to facilitate the most sensitive form of analysis, such as that conducted for chemical residues. Arguably, the greatest proportion of time and effort directed towards the assessment of livestock feed ingredient quality should be focussed on the sampling process. Recent research has resulted in the development of rapid methods for the direct assessment of the nutritional quality of feed ingredients for pigs, poultry and ruminants and significant potential exists to improve the use of this technology in commercial animal and feed production systems . Opportunities also exist for the development of multi screen ELISA assays for chemical contaminants, while a number of ELISA based test kits are in existence for specific mycotoxins. Quantitative analysis of weed seeds may be achieved through the use of image analysis, but there is an urgent need for rapid methods for the assessment of natural plant toxins such as pyrrolizidine alkaloids. The cost benefits of assessing nutritional quality can be clearly demonstrated, but when compared against the risks, analysis of ingredients for contaminants requires more strategic thought. This process may provide the enduser with some comfort that a quality assurance protocol is in place, however a procedure of this nature can only be considered as a very rough guide to ingredient quality. It will provide no data on the presence or absence of chemical residues or mycotoxins, weed seeds may be identified but the level of toxin contributed will not be quantified, and the chemical components measured correlate poorly with nutrient contributions to the target animal. Furthermore, if more detailed analysis of a sample is pursued, the length of time taken to complete the analysis is often too long to be useful. For these reasons, a review of practical strategies for the assessment of livestock feed ingredient quality is long overdue. Research over recent years has improved our ability to define the nutritional requirements of livestock, largely through the use of simulation models (e.g. Black et al. 1986). This advance necessitates more refined definition of the nutritional value of feed ingredients if the nutritional requirements of the target species are to be met as accurately as possible. Some progress has been made in the rapid assessment of the nutritional quality of feed ingredients (Morgan 1995; van Barneveld et al. 1999; BECAN Consulting Group 1999; Wrigley 1999), yet these techniques are still to be adopted widely. There is a need to increase awareness of these methods and of the potential benefits that could be derived from their routine adoption. A recent survey of Australian stockfeed manufacturers (BECAN Consulting Group 1999) suggests a high awareness of potential grain contaminants and the associated risks but a limited capacity to identify these contaminants and a limited knowledge of the costeffectiveness of analysis. It is clear that with the advent of more rapid methods of analysis, some enduser education is required in relation to grain sampling, the frequency of use of rapid tests and the course of action in the event that a contaminated sample of grain is identified. To provide some practical guidelines for the assessment of livestock feed ingredient quality prior to Introduction Current routine procedures for the assessment of livestock feed ingredient quality prior to inclusion in compound feeds are far from adequate. Our ability to rapidly and accurately assess the nutritional quality and subsequent nutritional value of a feed ingredient is limited and restricts our capacity to optimise overall livestock production efficiency . For example, a typical testing procedure for a grain received at a feedmill might involve collection of a sample from the truck during unloading, a visual inspection of the grain for obvious signs of grain damage or contamination, and analysis of the sample for moisture and crude protein content. Recent Advances in Animal Nutrition in Australia, Volume 12 (1999) 64 van Barneveld, R.J. inclusion in mixed feeds, this paper will: Detail adequate sampling procedures for grains and other feed ingredients Outline rapid, accurate methods to predict nutrient contributions from feed ingredients Discuss the potential for analysis of grain contaminants Provide an estimate of the costeffectiveness of sample analysis Sampling bulk ingredients Samples of bulk ingredients are most commonly collected using a spear sampler at various points in the load. While accepted by many, a more representative sample may be obtained from a grain stream falling from the end gate of the truck as follows: Use an appropriate sampling device that will collect grain from the entire stream without overflowing. The grain stream may need to be controlled to allow the sampler to pass through the stream without overflowing Make sure the sampling container does not contain seeds or other material before sampling Samples should not be taken from the first or last portions of a load since these areas do not provide a representative cross section Samples should be collected from the grain stream where the stream is established, approximately 30 cm below the endgate of the truck The entire stream must be cut (sampled) with a side to side sweep of the sampler, cutting the full thickness of the streamfront to back. The sampler should be held in a horizontal position while passing it through the grain stream to facilitate even filling of the sampler At least two, preferably more, samples should be collected at regular intervals, with intervals selected so that the entire lot of grain is represented. For example, three samples could be taken in the middle of each third of the load. Each sample should be approximately 2 kg and bulked for the entire load The quantity of sample sent for analysis (5001000 g) will be significantly less than the quantity collected during sampling. The portion used for analysis must be divided so that it is representative of the whole sample. Hand mixing and subsampling tends to cause fines to settle to the bottom of the container. Segregation occurs in the receiving container when pouring from Ingredient sampling Ingredient sampling is the most critical step in the assessment of livestock feed ingredient quality. A sample that is not representative can result in misleading analysis with the consequences being unnecessary rejection of an ingredient, or acceptance of an ingredient that will lead to losses in livestock production efficiency. In this instance it is fair to say that the wrong answer arising from incorrect sampling is worse than no answer. The degree of accuracy required when taking an ingredient sample will differ depending on the type of analysis planned. If only one sample is taken from a load, it will need to be adequate for the most sensitive form of analysis. Sampling for nutritive value A survey of variability in grain protein content conducted by the Canadian Grain Commission (1999) demonstrates the need for representative sampling when assessing nutrient contributions from an ingredient (Table 1). Given that significant variation in the nutrient content of grains can occur even within a single plant, sampling of bulk loads must be performed methodically and meticulously if a representative sample is to be obtained . The following protocols derived from Hellevang et al. (1992), Richardson (1995) and the Canadian Grain Commission (1999) are likely to result in a representative sample from bulk loads and bagged ingredients, respectively. Table 1 Variation in grain protein content (g/kg) at different sampling levels (Canadian Grain Commission 1999). Source of variation Range in protein content (g/kg) 40 20 30 60 50 60 80 110 Kernels within a head Heads from one plant Plants within one row (0.55 metres) Rows within one apparently uniform field Farms delivering to one elevator point Farms within one crop district Farms within a province Farms within Western Canada Strategies for the assessment of livestock feed ingredient quality 65 container to container, so grain used for analysis needs to come from a cross section of the sample as it is being poured. Place a container on the unloading grate or tarpaulin, then with a sweeping motion pour the grain from the sample while allowing the stream to move across the stationary container. The grain should be poured at a rate so that after several passes the sample container is empty at the time the desired quantity for analysis is collected. The procedure should allow collection of a subsample for analysis and a subsample for storage and future reference if required The sample must be handled in a manner that does not allow the condition of the sample to alter from the time it was collected. Samples must be labelled with the date collected, the source, the type of sample, and any other relevant information Duplicate determinations are recommended for all variables measured disease vectors Moulds and mycotoxins can be distributed unevenly in a sample of grain or feed In addition, our ability to test for the presence of moulds and/or mycotoxins can be limited by the accuracy of the analytical methods employed. For example, sample size and sampling method have limited effect on the variability of vomitoxin or deoxynivalenol test results for barley (Anon 1998). Increasing sample size did not significantly decrease variability of test results, sample selection method did not appear to cause greater variability and no single factor was identified that will reduce variability of these measurements in an easy and costeffective manner. It was concluded that the bulk of the variation observed in this study was due to the analytical methods employed. When variability in analytical methods can be minimised, the influence of sampling method on the accuracy of analysis for the presence or absence of moulds and mycotoxins can be significant. Howell et al. (1984) used a 12.5 kg sample of whole grain maize and an 11 kg sample of soybean meal, both shown to be contaminated with approximately 50 �g/kg of zearalenone, to demonstrate the distribution of the mycotoxin in a sample. Each sample was divided into 100 subsamples and the level of zearalenone in each was measured. The distribution of zearalenone in 40 x 110 g samples of soybean meal was shown to be normal at a mean value of 50 mg/kg with a variance of 77. The distribution of zearalenone in 100 x 125 g samples of maize was best described by a loge normal distribution when the mean value was 52 mg/kg with a variance of 14126. It is concluded that for soybeans the proposed sampling technique of 3 x 4 kg aggregate samples each composed of 20 x 200 g incremental samples was adequate, but for maize the aggregate samples should be bulked together to give a 12 kg sample composed of 60 x 200 g incremental samples. For maximum accuracy more than one 12 kg aggregate sample should be used. Given the low incidence of moulds and mycotoxins in Australia and the difficulties associated with accurate analysis, sampling for analysis is unlikely to be routine, but when completed it must be completed meticulously. Sampling bagged ingredients A slotted grain trier or spear should be used for bagged ingredients and approximately 400 g of sample should be collected from each bag sampled. For lots of one to ten bags, all bags should be sampled. For lots of eleven bags or more, at least ten bags should be sampled. Samples can be bulked and a subsample analysed in duplicate, or at least three separate samples can be analysed and the results averaged. Sampling for moulds and mycotoxins The difficulties associated with sampling and analysing for moulds and mycotoxins was summarised by van Barneveld (1999a) and include: There is a wide range of moulds and mycotoxins that can potentially occur in grains and grain by products Not all moulds that produce mycotoxins are visible The presence of a mycotoxin producing mould does not necessarily mean the mycotoxin is active (for example, Chelkowski et al. (1983) tested 636 samples of wheat, oats, barley and maize in 197981 to find only three had zearalenone at levels of 200, 700 and 2000 mg/kg, respectively. Fungi forming zearalenone were present in 28% of samples of cereal grain and about 48% of Fusarium fungi tested were able to synthesise the mycotoxin The antinutritional effects of moulds and mycotoxins are not always easy to detect in the target species Analytical methods are usually specific for a particular mycotoxin Many of the clinical symptoms associated with the presence of mycotoxins can be confused with other Sampling for chemical contaminants Recommendations for the sampling of feed ingredients for chemical contaminants were made by the BECAN Consulting Group (1999). Sampling procedures were as described above with the following qualifications for sample size and sampling frequency: Sampling is undertaken by taking at least three, preferably five, primary samples at random from the commodity as presented Each primary sample must be the same size (e.g. weight or volume). The primary samples are mixed 66 van Barneveld, R.J. to form the sample for analysis. Mixing should be complete to ensure the sample analysed is uniform Equipment used in the collection and transfer of samples must be clean and of an inert structure. The equipment must not be exposed to pesticides from any source (other than through the grain being sampled) during storage or use. Details of sample size and other sample requirements for various commodities are provided in Table 2. term storage of grains, insect identification and modelling the effects of storage and holding conditions on insect damage. Information is available on sampling equipment (Figure 1) and the source of this equipment. The sampling module can be used to develop an insect sampling program for specific storage situations. Based on the number of samples taken, it will calculate 95% confidence intervals for insect trap and probe samples. Estimations can also be made of the probability of detecting insects in grain based on insect density and the number of samples taken. Sampling for insects Tools such as the Stored Grain Advisor (Version 3.02; Instructional Media Centre, Kansas State University, Manhattan, Kansas) can be used for information on how to sample stored grains (particularly wheat) for insects in addition to providing strategies for the long Table 2 Sampling for weed seeds When assessing the consequences of weed seed contamination from a livestock perspective, the biggest consideration is the natural toxin that the weed seed contains. If a weed seed does not contain a toxin, its Sample sizes and other sample details for analysis of chemical contaminants in various commodities (BECAN Consulting Group 1999). Commodity Minimum size* (kg) of primary sample (5 or more primary samples taken) 0.8 1.5 1.5 0.8 1.5 1.3 1.3 Minimum size* (kg) of primary sample (3 or 4 primary samples taken) 0.5 1.0 1.0 0.5 1.0 1.5 1.5 1.0 1.0 Minimum size* (kg) of sample despatched to laboratory Cereal grains Other grains Forage/Fodder Hay/Straw Silage 0.7 1.3 1.3 0.7 1.3 Wet feeds (e.g. Pomace, Brewers waste) Other (e.g. cane tops, pineapple tops, etc.) *The minimum sample size assumes primary samples will be taken and the remainder of each primary sample will be retained for further testing if required. A sub_sample of each primary sample will be mixed together to form the laboratory sample. Figure 1 Example of the information provided by the `Stored Grain Advisor ' (Version 3.02; Instructional Media Centre, Kansas State University, Manhattan, Kansas). Strategies for the assessment of livestock feed ingredient quality 67 presence in a grain sample may be of limited concern, except for the effect it may have on our ability to define the supply of amino acids and energy that could be expected from the grain sample in question (van Barneveld 1999). To quantify the presence of plant toxins, sampling procedures similar to those employed for chemical contaminants should be employed. Enzymebased test kits use specific enzymes to hydrolyse a target compound, usually a complex carbohydrate. As these methods are well suited to identifying classes of polysaccharides that can influence nutritional quality, they have reasonable potential as a possible method for the prediction of available energy. Unfortunately, these methods are less than rapid. Near infrared spectroscopy (NIRS) analysis depends on detecting the presence and intensities of spectral bands from overlapping overtones corresponding to various chemical bonds, particularly CH, OH and N H. NIRS represents one of the truly rapid forms of analysis and requires limited sample preparation. The success of the technique relies heavily on the composition of the base sample set used to derive calibrations. There is potential to apply NIRS in two ways to predict the available energy content of feed ingredients either by quantifying factors that may correlate with available energy content, or by simultaneously accounting for those factors that influence available energy content (such as starch, nonstarch polysaccharide and protein content). As NIRS is already widely in use at grain receival points for the assessment of moisture and energy, and given its unrivalled combination of speed, accuracy and simplicity (Osborne et al. 1993), it holds the greatest potential in any strategy for the assessment of livestock feed ingredient quality. Rapid assessment of nutritional quality Having collected a representative sample of a feed ingredient, it is desirable to assess that sample for nutritional quality prior to inclusion in a mixed diet. To accomplish this, rapid and accurate methods of analysis for nutritional parameters such as available energy and available amino acids are required at the point of receival. To date, rapid assessment of nutritional components of feed ingredients is restricted to crude protein and moisture content. For monogastrics, these measures correlate poorly with available energy and available amino acids (van Barneveld et al.1999b). The studies completed to date (e.g. Batterham et al. 1980; Morgan and Whittemore 1982; Yin 1993) all conclude that neutraldetergent fibre is the best predictor of the nutritive value of feed grains for pigs. It is likely that more specific nonstarch polysaccharide components will give more accurate correlates with nutritional value, but further research is still required to define these relationships. Current methods for accurately measuring the available energy content of feed grains for livestock involve detailed animal experiments. Hence they are no use when it comes to assessing samples as they are delivered to grain receival points or sites of stockfeed manufacture. To achieve this, an objective analytical method is required that meets the criteria of being rapid, accurate, inexpensive, safe, easy to use and environmentally friendly. Candidates for this type of analysis for the measurement of the nutritive values of feed grains were suggested by Wrigley (1999) and include: Singlekernel characterisation system determines the mass, thickness, moisture content and hardness of individual grain kernels. The only potential for this type of system to predict available energy content would be through the addition of near infrared spectroscopy equipment. Rapid visco analysis based on the principles of the Falling Number test aims to determine the extent of starch breakdown due to the activity of the starch hydrolysing enzymes of germination . As rapid visco analysis provides indications of the pasting properties of starch in grain, there is potential for correlation with starch digestibility and hence energy contributions. This technique does, however, involve wet chemistry which can limit the speed of testing. Use of NIRS for the assessment of feed ingredients for pigs Recent research completed by van Barneveld et al. (1999) has demonstrated that NIRS can predict the digestible energy (DE) content of cereals for pigs within 0.38 MJ/kg, which is comparable to the accuracy achieved when using in vivo reference methods. Examination of difference spectra for the samples within one grain type (barley) having the highest and lowest DE values reveal that there are a number of factors (hemicellulose/cellulose, starch, lipid, protein etc.) influencing the DE predictions. Validation studies completed to date demonstrate that this calibration can successfully predict the DE content of wheat samples ranging in DE content from 13.814.7 MJ/kg and barley samples from 10.713.2 MJ/kg. Further, this analysis can be completed with equal degrees of accuracy on both whole and milled samples of grain. Use of NIRS for the assessment of feed ingredients for poultry The development of NIRS calibrations to predict apparent metabolisable energy in feed grains fed to poultry has met with variable levels of success (van Barneveld 1998a). A large factor contributing to this may be that apparent metabolisable energy 68 van Barneveld, R.J. measurements in poultry can be highly variable due to inherent individual bird variation. The extent of the influence of individual bird variation on apparent metabolisable energy measurements has been demonstrated by Hughes and van Barneveld (unpublished data). When a single diet containing barley was fed to 96 individually housed broiler chickens (17 days of age) simultaneously under identical conditions, the apparent metabolisable energy measurements ranged from 12.25 to 13.5 MJ/kg DM. Despite this, Windham et al. (1994) and Leeson and Valdes (1996) demonstrated close agreement between predicted and in vivo apparent metabolisable energy content of selected ingredients. Use of NIRS for the assessment of feed ingredients for ruminants The application of NIRS to predict the available energy content of feed grains for ruminants requires additional considerations. A confounding factor in addition to the inherent variability within feed ingredients used in ruminant rations is that the energy and amino acid availability of any feed fed to ruminants is heavily influenced by the other components of the ration. Many rumen variables such as flow rate, rumen pH and nitrogen levels will affect the nutritive value of a diet ingredient. For example, the relative nutritive value of a cereal grain depends on its proportion in the total diet, its degree of processing, other dietary constituents and the level of productivity of the animal . Consequently, variability in response to grain feeding has a lot to do with other dietary factors rather than simply the variability in grain nutritive value. For these reasons, the use of NIRS to predict the nutritional quality of feed ingredients for ruminants is more difficult due to the interactions that take place between different dietary components within the rumen. NIRS may still have a role, however, when combined with rumen function and amino acid and energy computer simulated growth models. Rumen function models will evolve with the capacity to account for a variety of rumen inputs and predict the subsequent outputs . NIR could be effectively used to measure the inputs required by such simulation models (e.g. protein, moisture, nonstarch polysaccharide and starch content of ingredients), and the subsequent use of computer simulation models would facilitate the accurate prediction of nutritional quality regardless of the feeding regimen (Figure 2) . Flinn and Downes (1996) have focussed on the development of calibrations for functional properties of ruminant feeds such as in vivo dry matter digestibility, voluntary feed intake and animal production. Dry matter digestibility has potential for use in the prediction of energy availability in grains for ruminants, however, this measurement will not account for interactions that occur between grains and forages in the rumen environment. Based on the above descriptions, it is clear that significant advances have been made in our ability to directly assess the nutrient contributions from various feed ingredients for pigs, poultry and ruminants. Further research is underway to refine these methodologies, but they hold significant potential to improve livestock production efficiency in their current form and warrant more attention from nutritionists and stockfeed manufacturers when routinely assessing feed ingredient quality. Analysis of grain contaminants To fully define nutritional quality, an assessment of chemical and physical contaminants in addition to potential nutrient contributions is required. To date, our ability to complete some of these analyses is limited, however, recent and current research will assist completion of these assessments in the future. Analysis of chemical contaminants The incidence of chemical contamination of grains used in livestock feeds is comparatively low, but the level of enduser concern is high. Grain and milling offals have a higher potential for chemical contamination, and hence closer scrutiny of these products is required (van Barneveld 1999a). Current methods of analysis for chemical contaminants of grains are very slow and prohibitively expensive. There is a need for a rapid method of screening chemical contaminants of grains . The BECAN Consulting Group (1999) developed the Feed grains NIR (Protein, fibre, measurements etc.) Rumen function model + Postrumen amino acid energy model ME Forage Figure 2 Schematic diagram outlining the possible role of NIRS in the prediction of nutritional quality (e.g. metabolisable energy) of feed grains for ruminants. Strategies for the assessment of livestock feed ingredient quality 69 following priorities for chemicals to be included in residue management strategies based on a risk ranking for individual chemicals (SP, synthetic pyrethroid; OC, organochlorine; OP, organophosphate; C, carbamate): Highest priority: Bifenthrin (SP), Chlorpyrifosmethyl (OP), Deltamethrin (SP), Endosulfan (OC), Fenitrothion (OP). Second highest priority: Beta cyfluthrin (SP), Chlorpyrifos (OP), Cypermethrin (SP). Third highest priority: Bioresmethrin (SP), Carbaryl (C), Esfenvalerate (SP), Maldison (OP), Methoprene (N/A), S Methoprene (N/A). Potential exists for the development of multi screen ELISA assays for synthetic pyrethroids, organochlorines, carbamates and methoprenes in addition to persistent organochlorines such as DDT and DDE. Biological testing techniques are required for the rapid analysis of organophosphates. rarely widespread, with many accounts in the literature referring to a small number of animals on individual enterprises. The National Agricultural Commodity Marketers Association of Australia (NACMA 1994) have listed the presence of moulds and mycotoxins at a nil acceptance level in feed grains . That is, grains contaminated with a mould or mycotoxin should not be purchased. On this basis, only a qualitative measure of these compounds is required. Some test kits available for this type of analysis are presented in Table 3 (J. Skerritt, CSIRO Plant Industry, unpublished data). Analysis of weed seeds and natural toxins Contamination of feed grains with weed seeds presents a major hazard to the Australian livestock industry, and yet the procedures for rapid screening and identification of weed seeds and their toxins are limited. Two levels of measurement are required for weed contaminants of feed grains. Qualitative measurements are required for weed seeds such as heliotrope, crotolaria, mustards (Sinapsis sp.), common tares, Senecio sp. (ragwort, fireweed) and Amsinckia sp. which prevent the acceptance of feed grains for use in livestock diets when they are present (van Barneveld 1999a). Quantitative measurements are required for weeds like saffron, variegated thistle, Australian and hoary cress, bindweed, double gee, mexican poppy, wild radish, thornapple and burrs. The presence of these seeds is acceptable, but the levels of contamination will dictate the potential for use of the feed grain. Image analysis, a computerised form of visual examination without subjectivity, holds Analysis of moulds and mycotoxins The volume of research and literature available on the effects of mould and mycotoxin contamination of grain on livestock production, and techniques available for the measurement of moulds and mycotoxins, far exceeds the potential risk these contaminants pose to the livestock industries in Australia (van Barneveld 1999a). In addition, the effects of moulds and mycotoxins are Table 3 A selection of commercial field tests available for the analysis of mycotoxins (J. Skerritt, CSIRO Plant Industry, unpublished data). Analytes Aflatoxins Aflatoxins Zearalanone Aflatoxins Ochratoxin Affinity column* Fluorometry 20 min Format Microwell ELISA Cup ELISA Read out Colour change Colour change Test time 10 min 10 min Kit name Neogen `Agriscreen' International Diagnostic Systems `AccuCup' Vicam Rhone Poulenc `Aflatest' Aflatoxins Affinity column Fluorescent card or fluorometry 15 min 10 min diAGnostix Inc `EZ_Screen' Aflatoxins Ochratoxins T_2 toxin Zearalanone Card ELISA Visual (colour change) IDEXX `Cite' *Concentrates toxin Aflatoxins Cup ELISA Visual (colour change) 10 min 70 van Barneveld, R.J. potential for the quantitative identification of weed seeds in grain samples. In addition to measurements for the presence of weed seeds per se, quantitative measures are required for toxic compounds such as pyrrolizidine alkaloids, glucosinolates, canavanine and vicine. Pyrrolizidine alkaloids are of particular concern as the levels of this toxin are poorly correlated with the number of weed seeds present and vice versa. Cargill and Slade (1996) reported that visual counts for seeds such as heliotrope which contain pyrrolizidine alkaloids tend to underestimate the level of contamination, while immunoassay methods (e.g. ELISA) to test for the presence of the alkaloids themselves may overestimate the degree of contamination. This statement is supported by a study involving the addition of 25 heliotrope seeds to a clean sample of wheat which was then analysed for pyrrolizidine alkaloids. The estimated number of seeds was 33.8 based on the pyrrolizidine alkaloid content. However, when the number of seeds present in a contaminated grain sample were counted visually and then estimated by analysis, the results were 23 seeds by visual count and 11.7 seeds by estimation from chemical analysis. Australian industry is not expected to change. Assuming the consuming animal industries bear the cost of the tests, it was also suggested that there will be a reduction of 2%, 1.1% and 0.3% in Australian grown pig, poultry and beef meat, respectively. A sensitivity analysis on the basic assumptions indicated that the price of tests was strongly influenced by industry risk, delays caused by the tests at the delivery terminals, the incidence of grain rejection and particularly the average size of delivery loads. Given that it is unlikely that a suite of tests for moulds, mycotoxins, chemicals, insects and weed seed will be able to be developed within this cost constraint, strategic selection o