Net energy systems in pig nutrition: what they measure and what they miss
Published 10 April 2026 · Updated 18 June 2026
For most of the twentieth century pig rations were formulated on digestible energy or, later, metabolisable energy. Both frameworks count the energy that disappears from the gut and enters the animal, but they count it at different doors and they disagree, sometimes badly, on the value of fibrous and protein-heavy ingredients. The shift to net energy, which is now the working standard in most of western Europe and a growing part of the Americas, was an attempt to close that gap by pushing the accounting one step further: past the gut, past urinary and gaseous losses, and past the heat the animal generates just to process the feed.
What net energy actually measures
A digestible energy value says how much of the gross energy in a feed ingredient survives passage through the gut. Metabolisable energy subtracts the energy that leaves as methane and urine. Net energy goes one step further still, subtracting the heat increment of feeding—the metabolic cost of digesting, absorbing and metabolising nutrients—so that what remains is the energy available for maintenance and for depositing protein and fat in the carcass.
The distinction matters most when two ingredients that look similar on an ME basis behave differently in the pig. Soybean hulls carry a fair amount of metabolisable energy, but a large fraction of it is fermented in the hindgut and released as heat; on a net energy basis the hulls are worth substantially less. Fat, on the other hand, has a low heat increment relative to its ME content, so its net energy value is proportionally higher than its ME value. A formulation built on ME overvalues the hulls and undervalues the fat. The pig, which has to deal with the heat either way, already knows this.
The French INRA system, published by Noblet and colleagues from the early 1990s onwards, is the reference that most European formulations lean on. It uses indirect calorimetry data—measurements from respiration chambers—to partition energy into its components, and it produces prediction equations that let a formulator estimate net energy from the chemical composition of an ingredient without putting every batch through a chamber. The Dutch CVB system and the Danish system run parallel sets of equations calibrated on their own data. The numbers do not always agree, which is worth knowing before mixing systems in the same spreadsheet.
Where the older systems go wrong
The practical case against DE and ME in pig nutrition is not that they are inaccurate—they measure what they claim to measure—but that they are incomplete. Two ingredients with identical ME values can produce different amounts of body gain, because the heat they generate in the process of being metabolised is not the same. The error is systematic rather than random: it tilts the ranking of ingredients in a predictable direction, penalising fats and favouring fibrous feeds.
That tilt runs straight into the least-cost formulation engine. If ME overstates the feeding value of beet pulp relative to tallow, the optimiser will put in more pulp and less tallow than the pig needs, and the growth response will fall short of the prediction. The formulator then compensates by oversupplying energy, which costs money and may also deposit more backfat than the market wants.
The same asymmetry shows up with amino acid supply. A high-protein ingredient carries a high heat increment because deaminating surplus amino acids costs energy; on an ME basis that cost is invisible, but it is very visible to the pig and to the carcass scanner. Net energy makes it visible to the formulator as well.
Where net energy itself falls short
Net energy is not the end of the story, and two of its limitations are structural rather than just gaps in the data.
The first is that the prediction equations assume a fixed partitioning of retained energy between protein and fat. In practice that partitioning changes with genotype, sex, body weight and ambient temperature. A lean modern genotype deposits proportionally more protein and less fat than the animals the original equations were calibrated on, and the energy cost of depositing a gram of protein is roughly twice the cost of depositing a gram of fat. If the equation assumes a fatter pig than the one being fed, it will overstate the net energy value of the ration.
The second is that the system is static. It gives a single value for a given ingredient at a given composition, but the pig's requirements change from day to day as it grows. Phase feeding and precision feeding both need a dynamic model underneath them, and a static NE value table does not supply one. The modelling sessions at symposia like the Granada meeting in 2022 were largely concerned with this: how to take a good static accounting system and make it move.
A third, more practical limitation: the underlying calorimetry data come from a relatively small number of experiments, most of them conducted in northern Europe on European genotypes. How well the prediction equations transfer to tropical breeds, to pigs raised in heat, or to ingredients that are regionally important but poorly represented in the calibration data is an open question, and one that the feed efficiency literature has not yet settled.
What this means for a formulator
Switching from ME to NE does not change the biology; it changes what the formulator can see. The pig was always penalised by a high heat increment; the only question is whether the spreadsheet knows about it. In the commercial feeds that dominate European production the shift happened years ago, and there is no serious argument for going back. In other markets the transition is slower, partly because the NE values for locally important ingredients are not always available, and partly because the existing ME-based systems have been tuned, empirically, to deliver acceptable performance—which means the gap between the model and the pig is being closed by experience rather than by physics.
Neither approach is wrong in a practical sense. One is more transparent about where the energy goes; the other compensates for what it cannot see. The argument for NE is that transparent is cheaper in the long run, because it does not need the safety margins that an opaque system requires.