Assessment of human milk composition using mid-infrared analyzers requires calibration adjustment

Nutrient composition of human milk (HM) is highly variable. Targeted HM fortification has been proposed to address these variations and reduce the cumulative nutritional deficit in preterm infants. Near-infrared analysis is used to measure the protein and fat content in HM; however, the reliability of this technique has not been evaluated. The objective of this study is to evaluate the reproducibility and accuracy of two generations of HM analyzers (HMA1 and HMA2) in estimating protein and lipid contents.

Reproducibility was assessed by analyzing in duplicate 146 and 128 HM samples with HMA1 and HMA2 (Miris), respectively. To evaluate the accuracy, lipid and protein concentrations were assessed in 31 and 39 samples using HMA1 or HMA2, respectively. Values were compared with measurements obtained using reference methods and correction equations were calculated. After applying the correction equations on 12 HM samples, the performance of the two devices were compared and the equation was validated according to the reference methods.

The coefficients of variation for protein and lipid assessments were below 3% for both HMA1 and HMA2. Protein concentrations were significantly underestimated by HMA2 (-0.53±0.23 g dl^-1). Lipid content was significantly overestimated by both devices, but the error was greater with HMA1 (0.76±0.48 g dl^-1) than with HMA2 (0.36±0.33 g dl^-1). Correction equations were specific for each generation of HMA. Finally, after correction, both instruments provided similar and accurate results.

HMAs require calibration adjustment before their use in clinical practice, to avoid inappropriate HM fortification.