Validation of Correction Algorithms for Near-IR Analysis of Human Milk in an Independent Sample Set-Effect of Pasteurization

Abstract

Commercial infrared (IR) milk analyzers are being increasingly used in research settings for the macronutrient measurement of breast milk (BM) prior to its target fortification. These devices, however, may not provide reliable measurement if not properly calibrated. In the current study, we tested a correction algorithm for a Near-IR milk analyzer (Unity SpectraStar, Brookfield, CT, USA) for fat and protein measurements, and examined the effect of pasteurization on the IR matrix and the stability of fat, protein, and lactose. Measurement values generated through Near-IR analysis were compared against those obtained through chemical reference methods to test the correction algorithm for the Near-IR milk analyzer. Macronutrient levels were compared between unpasteurized and pasteurized milk samples to determine the effect of pasteurization on macronutrient stability. The correction algorithm generated for our device was found to be valid for unpasteurized and pasteurized BM. Pasteurization had no effect on the macronutrient levels and the IR matrix of BM. These results show that fat and protein content can be accurately measured and monitored for unpasteurized and pasteurized BM. Of additional importance is the implication that donated human milk, generally low in protein content, has the potential to be target fortified.

Keywords: analyzer; donor milk; fat; lactose; milk; nutrition; preterm; protein; term.

Figure 1

Study Design: To validate the Near-IR human milk analyzer in unpasteurized breast milk, group B samples were compared to the corresponding group A samples. The correction algorithm was also tested for pasteurized breast milk, through a comparison of all group D samples with the corresponding samples in group C. The effect of pasteurization on macronutrient levels was determined through two separate comparisons: one between groups A and C, and another between all group D samples and the corresponding samples in group B.

Figure 2

Fat (A) (n = 20) and protein (B) (n = 20) values as measured through chemical reference methods correlated against corrected Near-IR human milk analyzer values. Bland–Altman plots for fat (C) and protein (D) indicating the difference obtained by the reference method (x-axis) and the corrected Near-IR human milk analyzer values represented in y-axis.

Figure 3

Fat (A) (n = 10) and protein (B) (n = 10) values in pasteurized breast milk as measured through chemical reference methods correlated against corrected Near-IR human milk analyzer values; Bland–Altman plots for fat (C) and protein (D) indicating the difference in pasteurized breast milk obtained by the reference method (x-axis) and the corrected Near-IR human milk analyzer values represented in y-axis.

Figure 4

(A) shows the correlation between samples of pasteurized and unpasteurized breast milk (n = 50) as measured through Near-IR milk analyzers and subsequently adjusted with the correction algorithm; (B) shows a similar correlation between pasteurized and unpasteurized samples (n = 10) as measured through validated chemical reference methods.