Efficiency of in vivo ear flap NIR scan in the detection of differences related to diet or pregnancy status in young rabbit does.

The aim was to determine whether in vivo spectroscopy of the ear flap was able to detect the diet-related differences and pregnancy status, with its prognostic in forward or backward perspective. Methods. Young grey Carmagnola rabbit does in groups of eight were fed diets enriched with: linseed at 5% (A); linseed at 5% and hazelnut skins at 1.5% (B); palm oil at 1.8% (C); and a standard diet (D), respectively. After the zero-point (P0) before A.I. a second point (P1) was fixed at d 21. A smart-NIR (SCIO, 740-1030 nm) was used. Chemometric analyses of the 331-point spectra were performed by using WinISI 1.5, without math pretreatment. Modified Partial Least Squares method was fitted to dummy binary variables. Outliers were identified once and excluded when t>2.0. R2 coefficient in cross-validation was retained as reference parameter. In parallel, for binary categorical discrimination, the proprietary software SCiO lab based on AKA (Also Known As) confusion matrix was used. Results. The NIR spectra were able to clearly perceive changes from a standard pre-experiment D diet to the A, B and C diets, with R2 around 0.60, while the value was lower for diet D (0.29). This first outcome is favorable for the repeatability of the method and for a discernment of nutritional factors. Prognostic of gestation in backward mode (scan at P1) was significant with mild NIR relationship (0.35), not suitable for practical uses, while in forward mode prediction (scan at P0) result was poor (0.20). The feeding treatments with different fatty acid profiles were instead more accurately perceived in the NIR spectra, especially for the C group, which was discriminated from the A and B groups with a R2 value of 0.71, which corresponds to a 75% reclassification as AKA. Interestingly, the antioxidant compound of the B diet was enhanced for a non-pregnant status (0.85 vs. 0.12 when pregnant) and this could indicate the presence of interactions between growth and pregnancy requirements, as assessed from skin variations. Conclusions. At the present state of knowledge, the miniaturized NIR device looks promising, mainly for the assessment of body fat composition. As far as the feeding experiment is concerned, it should be pointed out that it is rare to obtain model R2 values as high as those obtained during this preliminary work for a classic design, even multivariate ones, without extreme diets. The skin is an unbiased mirror of the health and nutritional status of rabbits.