In vitro techniques are widely used to predict the nutritive value of foods for farm animals. However, food evaluation systems are moving towards systems based on nutrient flows rather than on energy or protein. Nutrients are supplied via the digestive tract and presently available in vitro methods are inadequate to simulate the complicated and non-steady-state processes in this tract. This is partly due the lack of adequate in vivo reference values and partly due to a too-high degree of standardization and simplification. Future developments should therefore aim to develop in vitro systems which closely monitor the dynamics of the digestive processes. Adequate interpretation of the results may require sophisticated mathematical models.

Regressions relating in vivo digestibility to chemical composition of the food have residual standard deviations that are unacceptably high. The development of the two-stage in vitro technique inoculated with rumen liquor (Tilley and Terry, 1963) allowed dry-matter digestibility to be predicted with greater accuracy. This success was followed by a series of developments which replaced rumen liquor with inoculum produced from fresh or preserved faeces collected from sheep or cattle. Other methods used inoculum from a continuous fermentation containing rumen micro-organisms and enzymes produced by fungi. Another modification was to use gas production as a measure of in vitro digestion. The range of nutritional problems that could be measured by in vitro techniques was extended to include the estimation of voluntary food intake and protein degradation. All these in vitro techniques require standardization using food samples that have previously been analysed in vitro or offered as the sole diet to animals. The relative merits of these two calibration methods are discussed. Special facilities are required for storing and distributing these standard foods.

A common feature of nutritional models developed in the last three decades is that estimates of energy and protein are generally made independent of each other, even though interactions between them are recognized.

Most of the energy systems have adopted metabolizable energy (ME) or net energy (NE) as the unit for energy evaluation. Accurate measurement of the ME content of a food requires in vivo digestibility measurements using cattle or sheep, with methane losses either measured or predicted. This approach is clearly unsuitable for routine food evaluation and most systems predict the ME content from chemical analyses, using regression equations derived from calorimetric measurements.

The new generation of protein systems recognize the need to consider, separately, the energy and nitrogen requirements of the rumen micro-organisms and the protein requirements of the host animal. Despite this common base, systems differ markedly in estimates of energy supply to rumen bacteria, the degree to which dietary protein is degraded in the rumen, microbial protein yield and the digestibility of the microbial and undegraded dietary protein.

Although empirically based, current protein systems contain dynamic elements relating to rates of fermentation and protein degradation. In most cases, data have been derived using the in situ nylon bag technique, although this has major limitations. Recent developments in the gas production technique suggest that this approach has the potential for describing fermentation and degradation rates of individual foods.

To be effective, energy and protein systems need to describe foods in units that are both easily understood by non-specialists and can be measured using procedures that are rapid, reliable and cost effective. Near infra-red reflectance spectroscopy (NIRS) can be both rapid and cost effective and there is evidence that it may be able to predict not only proximate analyses but also dynamic characteristics of foods.

Several published reports on the nutritive value of whole-crop wheat (WCW) have been based on estimations from laboratory techniques, some of which were developed for grass silage. However, there is little information on the accuracy of such estimations. Therefore the aim of this study was to evaluate the suitability of predicting the in vivo digestibility of WCW from various less animal-dependent techniques.

Robust techniques that predict accurately and precisely the quality of forages are needed for developing improved feeding strategies and screening of large numbers of forage samples in breeding programmes. The release of gas in vitro (Menke et al., 1979) is attractive for this purpose because it does not require filtration and one single sample can be used to examine the kinetics of fermentation. The objective of this study was to determine the relationships between gas release in vitro and digestibility in vivo and intake of tropical forages.

The nylon bag technique (NBT) has been used widely to characterize the washable (W), insoluble potentially degradable (D) and insoluble non-degradable (U) fractions of grasses (e.g. van Vuuren, 1993) and concentrates (Tamminga et al., 1990). Recently the gas production technique (GPT) has been proposed as a method to evaluate the fermentation characteristics of different foods (Theodorou et al., 1994). Both techniques have shown good correlation with dry matter intake (Khazaal et al., 1993). One of the potential advantages of GPT over NBT is the ability to differentiate fermentation patterns (Groot et al., 1996) that might be related to changes in food composition. This trial was part of a larger grazing experiment and aimed to compare GPT and NBT when applied to samples of ryegrass at different ages. A specific objective was to determine whether or not GPT could extract more information concerning the fermentation patterns of different ryegrass fractions and how these patterns changed as a result of plant maturity.

Systems to measure gas production to study digestion kinetics have been developed at several locations. The system developed at Cornell University and the rationale behind its evolution are described with an emphasis on whether venting after each observation is necessary and on choice of sensors. Different non-linear-models used to fit gas production data are discussed with an emphasis on the dual-pool logistic model. The third section of the paper includes a theoretical discussion on how gas data can be integrated with data on passage to predict ruminal digestibility. The final section addresses the practical applications of these gas data and ways in which they can be used in models like the Cornell net carbohydrate and protein system. Also included are evaluations of ensiled and freeze-dried samples from the same source as an indication of how gas systems can be used to evaluate the soluble fractions of forages.

In vitro gas production techniques have become popular for characterizing the rate and extent of digestion of ruminant foods. In a typical gas production study, gas (predominantly carbon dioxide) is produced as particles of substrate are fermented by rumen micro-organisms in a bicarbonate buffered culture medium. Innovations in equipment design, including automated pressure recording systems and mathematical descriptions of the gas production profiles themselves, make the techniques both simple and precise and therefore good as laboratory procedure. The technique of measuring gas is of value in ruminant science because the kinetics of gas production and substrate degradation are very closely correlated. However, although it is relatively easy to measure gas volumes and to determine the kinetics of gas production, the underlying processes that give rise to the gas in the first place are complex and not well understood. Therefore, there is concern about what is being measured in gas production studies and how this relates to the digestion process in the ruminant animal. In this paper we review some of the fundamental properties of gases, describe their behaviour in liquids and consider some of the biological and chemical factors influencing gas production. From a knowledge of how gases behave in liquids and at gas-liquid interfaces, both at ambient and increased temperatures and pressures, it is possible to deduce what is happening in gas production studies. However, although the technique is invaluable for obtaining information about the digestion of particulate substrates in anaerobic ecosystems, we conclude that gas production should be used with caution in routine food evaluation studies.

At the DLO Institute for Animal Science and Health (ID-DLO) in Lelystad a fully automated time related gas production apparatus has been developed, using sensitive electronic pressure transducers in combination with electric micro-valves to release overpressure during the incubation. In this very sensitive system, no gas accumulation and pressure build up takes place. Each valve opening represents a known amount of gas, set at about 0-5 ml. By recording each valve opening the kinetics of degradation can be studied. Routine analyses are performed with rumen fluid from wether sheep, diluted with buffer (1 volume rumen fluid/2 volumes buffer). The gas production profiles are fitted with a multi-phasic model, which exists for most foods of three subcurves. It is shown that the first subcurve of gas production is caused by fermentation of the soluble fraction of a food, whereas the second subcurve is caused by fermentation of the non-soluble fraction. The third subcurve, starting after about 15-20 h of incubation, is not related to fermentation of the substrate, but is caused by gas production in the rumen fluid itself, as it is also observed in rumen fluid without substrate. It is discussed whether gas production profiles should be corrected for a blank or not. Gas production profiles of substrates with a high content of protein should be corrected for ammonia synthesis, as the ammonia inhibits the release of gas from the saturated carbonate buffer. The used gas production technique and the used three-phasic curve fit model show a high relationship with existing food evaluation techniques, like the Tilley and Terry technique and the nylon bag technique. Comparing the gas production technique with in vivo data has only been done with a very limited number of samples. At ID-DLO the gas production technique is used at present as a research instrument to study the fermentation characteristics of ruminant foods, the technique is used in breeding programmes of grass and maize, to search for optimal degradable genotypes, and in silage research. Moreover, the technique is used to study rumen physiology, with special emphasis on carbohydrate and protein fermentation and the balance between both. The gas production technique proved to be a cheap and accurate technique to determine differences in fermentation kinetics. The gas production equipment cannot only be used to study the fermentation in the rumen but can be used to study all kinds of fermentation, such as colon fermentation in all kind of animals and humans.

The automation of existing in vitro gas production methods (Beuvink et al., 1992) and the development of new ones (Theodorou et al., 1994) have created a need for suitable mathematical models to describe and interpret cumulative gas production profiles. Ideally a function is required which is capable of modelling a range of shapes with no inflexion point and also a range of sigmoidal shapes in which the inflexion point is variable. Several models have been proposed to describe gas production profiles (e.g. Blümmel and Ørskov, 1993; Beuvink and Kogut, 1993; Schofield et al ., 1994), sometimes blending empiricism with a more mechanistic view based on a compartmental scheme. The primary objectives of this paper are to present a unifying analysis of this area, pointing out compartmental interpretations of some of the candidate models and to link the gas production technique to animal performance by determining the extent of ruminal degradation for each model.

In order to understand and ultimately predict the voluntary intake and performance of ruminants, it is necessary to know the nutritional value of foods. Most recent systems for predicting nutrient supply are dynamic in nature and characterize foods in terms of the quantities of available nutrients and their potential rates of supply. The in vitro gas production system has been used to characterize the carbohydrate fraction of foods in this manner. For the technique to be able to do this, two assumptions must be satisfied. First, that the rate of fermentation is limited by characteristics of the food and secondly that the pattern of gas production correlates closely with the pattern of food fermentation.Low microbial activity within the system could invalidate both assumptions since it could (i) limit the rate of food fermentation, thus not allowing the potential rate determined by the physical and chemical nature of the food to be measured and (ii) result in partition of food carbohydrate into new microbial matter, thus reducing the amount of volatile fatty acids and hence gas produced per unit of food fermented.

The aims of this study were mathematically to simulate food fermentation within an in vitro system and to use this representation to investigate the potential effects of variation in microbial activity on the characterization of foods.

The separation of forages into a soluble fraction like cell contents and into an insoluble matrix like cell walls is an initial step in forage evaluation. Further analysis can be applied to study fermentation characteristics of soluble and insoluble fractions. Unfortunately, most laboratory in vitro techniques quantify incubation insoluble residues and are therefore inappropriate for the examination of food solubles. In vitro gas methods (for example Menke et al., 1979) reflect the production of short chain fatty acids (SCFA) and are therefore free of the disadvantages associated with gravimetric methods. However, it was recently pointed out that an inverse relationship may exist between SCFA production and microbial biomass yield (for review see Blümmel et al., 1997a). These authors suggested combining in vitro gas measurements with the quantification of substrate concomitantly degraded to calculate a partitioning factor (PF). This factor reflects the substrate-dependent variation in SCFA production and microbial biomass yield; the factor (PF) was shown to be valuable in voluntary food intake (DMI) predictions (Blümmel et al., 1997b). In the work presented here, the rate and extent of fermentation of whole roughages and extracted neutral-detergent fibres (NDF) of 54 roughages were examined in an in vitro gas test. The dry-matter degradability of extracted NDF was also quantified and the partitioning factor for NDF (PFNDF) was determined. The relevance of these findings for the analysis of extracted NDF and cell solubles is discussed and their implications for the prediction of DMI of roughages by gas tests are addressed.

When a food is ingested by a ruminant animal, the carbohydrate fraction of the food is fermented by the rumen micro-organisms to produce gas (predominantly carbon dioxide and methane) as well as volatile fatty acids (VFA). The gas production technique simulates this fermentation process and provides an estimate of both the rate and extent of fermentation. Comparing the gas production (GP) profiles of foods enables a comparison to be made of the fermentative characteristics of different foods. However, the technique uses a bicarbonate-based medium system with the rumen liquor. This complicates the GP profile because of the production of ‘indirect’ gas resulting from the reaction between the VFA and the bicarbonate ions.

Beuvink and Spoelstra (1992) measured the volume of gas produced from buffered rumen fluid when known amounts of VFA were added and observed that 20·8 ml gas were released per mmol VFA. However, there is variation between laboratories in terms of the composition of the medium that is used. Even when the same medium is used, significant differences have been observed in the GP profile when different types of apparatus were employed (Rymer and Givens, 1997). Media are gassed with carbon dioxide before they are added to the gas production system and it is possible that the concentration of carbon dioxide dissolved in the medium varies between experiments. The objective of this experiment, therefore, was to determine whether the volume of indirect gas produced was affected by the composition of the medium, the addition of carbon dioxide, and the technique employed to measure gas production.

Nitrogen-deficient fibrous crop residues are widely used as basal diets in less developed countries, particularly in dry seasons when alternative foods are often in short supply. One approach to improving animal performance on crop residue based diets is to include a supplement of improved quality food to provide fermentable protein and energy. There are no established in vitro methods for investigating interactions between foods but the in vitro gas production method shows promise in this regard (Prasad et al., 1994). This paper describes the interactions observed in vitro; an accompanying paper (Murray et al., 1998) describes in vivo responses to supplementation and relationships between in vitro and in vivo data.

The kinetics of gas production during ruminal fermentation may provide valuable information about foods that can be used to formulate diets and model animal responses. However, measurement of digestion kinetics is affected by methodology and techniques must be established that provide accurate and precise estimates of kinetic parameters. Because gas production measurements provide the opportunity to estimate the digestion kinetics of both soluble and insoluble matter in foods, it would be desirable to use this technique on a wide variety of forages, grains, supplements, and by-product foods. Applying an in vitro technique to such a wide variety of substrates raises questions about the type of inoculum that should be used. The objective of our study was to evaluate the effects of donor animal and its diet on the measurement of gas production kinetics using both forage and concentrate substrates.

New ruminant feeding systems being developed in Europe and North America have greatly increased the need for rapid measurement of rate and extent of ruminal digestion of complex carbohydrates and proteins in individual foods. Application of these systems is lagging because of a lack of reliable kinetic digestion data. This has increased interest in using cell-free enzymes to assess ruminal breakdown of carbohydrates and proteins. The simplest kinetic model describing ruminal digestion divides carbohydrates and proteins into three fractions: A (that already digested and/ or that soluble), B (that digestible) and C (that completely indigestible). Fractions A, B and C sum to 1·0; proportion of digestible fraction B that is degraded is given by: [kd/ (kd + kp], where kp and kd are, respectively, rates of ruminal passage and digestion. Some models divide fraction B into two or three subfractions while others include a digestion lag. Fully successful, cell-free enzyme systems would yield accurate estimates of digestion fractions and rates and, where appropriate, lag time. As a minimum, results with enzymes should be well correlated to in situ extents of digestion. In vivo total tract digestibility of organic matter (OM) and energy was predicted accurately when food samples were pre-treated with acid-pepsin or neutral-detergent solution, followed by treatment with cellulase from Trichoderma viride (T. reesei). Pepsin-cellulase and neutral detergent-cellulase methods have predicted in vivo digestibility more precisely than Tilley and Terry methods in some cases. Two-stage cellulase assays have given good results with many forages and, when including an amylase treatment, with mixed foods; OM digestibility of straws was not well predicted. Many different cell-free proteases have been tested to estimate ruminal protein degradability. Generally, effectiveness of proteases was assessed by correlating proportions of food nitrogen solubilized after specific incubation times with extents of in situ protein degradation. The broad specificity protease from Streptomyces griseus has been used most extensively for this purpose in incubations at pH 8·0. However, somewhat better correlations have been reported for ficin, bromelain, papain and neutral proteases of fungal and bacterial origin in incubations at pH more similar to the rumen. Prior treatment with amylases improved correlations for concentrate foodstuffs. As yet, cell-free proteases have not accurately predicted rates and extents of protein degradation observed in ruminal in vitro and in situ systems.

Intestinal digestibility of food protein is one of the important factors of most protein evaluation systems. Therefore it was determined in the experiments with duodenal infusions of foods (Hvelplund, 1985) or using mobile bags (Sauer et al., 1983; Hvelplund, 1985; de Boer et al., 1987; Antoniewicz et al., 1992). To avoid the use of cannulated animals, in vitro enzymatic methods have been developed (Antoniewicz et al., 1992). Incubation of food with pepsin and pancreatin correlates with the digestibility of food protein (Akeson and Stahmann, 1964) and is suitable (r = 0·95 to 0·99) to determine protein digestibility in pig foods (Babinszky et al., 1990). The aim of the study was to develop the best conditions of pepsin-pancreatin digestion and to evaluate the predictive ability of the method to determine protein disappearance measured using mobile bag technique.

Lucerne forage can provide significant amounts of both degraded and undegraded protein to ruminants. Changes in chemical composition taking place during advancement of maturity also affect ruminal degradability of protein. Therefore, forages harvested in a wide range of vegetation stages represent a good set of samples suitable to test and validate in vitro methods. The aim of the study was to elaborate the best conditions of using protease from Streptomyces griseus (PSG) to estimate ruminal degradability of dried lucerne forage and compare the results with those obtained when the same material was digested with ficin (FIC) or pancreatin (PNC).