Insect meal in feed : use of Near-Infrared Spectroscopy (NIRS) techniques to support the detection of authorized and unauthorized insect species

Since the bovine spongiform encephalopathy (BSE) crisis, many restrictive measures have been taken with regard to animal feed. Since 2017, the European authorities have allowed seven species of insects to be used in aquaculture1. More recently, in 2021, this authorisation was extended to pigs and poultry2, with the inclusion of an eighth insect species3. The correct characterisation and identification of insect meals and their detection are required to ensure the application of the legislation4. To date, the official control methods have shown some limitations, such as the need for expertise in light microscopy (LM) and Real-Time Polymerase Chain Reaction (RT-PCR) being a targeted method that does not allow the detection of several products at the same time. Furthermore, these different methods do not provide any insight into the chemical composition of the species studied.
In this context, various studies on insect meals have been carried out as part of the EU Farmyng project and the European Union Reference Laboratory for Animal Proteins in feedingstuffs (EURL-AP). Firstly, a sample bank of well-referenced and characterized insect meals has been built. This sample bank has been used as a starting point to develop NIRS models to predict chemical composition of insect meals. The first results gave a RMSEP of 2.21, 2.97 and 1.22 % for protein, fat and chitin contents respectively.
Secondly, Near-Infrared Microscopy (NIRM), combined with chemometrics, was investigated in order to develop a screening method to detect insect particles in feed without any chemical extraction step as is the case with current official methods. Different blends have been created from ruminant feed adulterated at levels of 1 %, 5 % and 10 % w/w with either H. illucens larvae meal or T. molitor larvae meal. Based on PLS-DA analysis, several spectra of insect particles in the blend were identified. The performance of the model was very good, with an accuracy for the respective detection of H. illucens and T. molitor particles always above 96% and 98%.