Go Figure: Equations For Estimating Infant Human Milk Intake

Written by: Lauren Milligan Newmark, Ph.D. | Issue # 123 | 2024

“One size fits all” is acceptable for rain ponchos and mittens but little else. Humans are just too variable in body size and shape—and most other biological traits—for one product to work for everyone. It seems surprising, then, that researchers studying human infant nutrition would apply a “one size fits all” approach to estimate infant milk intake when this trait is known to vary across infants [1]. Assuming all infants consume the same amount of milk each day has the potential to over- or underestimate individual human infant intake of nutrients, immune factors, hormones, and other bioactive milk components. “One size” approaches are often used, however, because methods for accurately measuring daily milk intake can be costly and cumbersome [1, 2]. 

The gold standard for measuring milk intake—called the dose-to-mother stable isotope technique—requires mothers to consume a labeled (stable) isotope, followed by multiple urine collections from mother and infant pairs over several weeks. The quantity of isotope in each urine sample (which is used to calculate the volume of human milk transferred from mother to infant) must then be measured using specialized laboratory equipment. Less technical but equally impractical for most studies is the test weighing method, where infants are weighed before and after a feed. The difference in weight is assumed to be equal to the volume of milk consumed but is only accurate when using highly sensitive scales [3]. Researchers must either have participants complete feedings at the study site or provide all participants with highly sensitive scales to use at their homes. 

A new study [1] from a team of New Zealander nutrition and health researchers offers a clever solution to these methodological problems. The research team hypothesized that data commonly collected in surveys and studies of infant health or nutrition—such as infant age, infant weight, maternal age, number of feedings per day, or types of complementary foods— could be used to reliably predict the mean daily human milk intake from the study population [1]. 

Their goal was to create two predictive equations to estimate human milk intake in infants aged 7 – 10 months [1]. The first equation would include data usually provided by questionnaires and anthropometrics and the second would also add in data from dietary assessments. Study participants were part of the First Foods New Zealand (FFNZ) study, a cross-sectional study aimed at understanding health and nutrition of infants aged 7 – 10 months. From the larger study population, equations were derived from data collected from 157 mother-infant pairs.

The researchers first needed to establish accurate daily milk intake data for each infant. They employed the dose-to-mother stable isotope technique to determine the volume of milk consumed each day over a 14-day assessment period for each infant. Then, they turned to the scientific literature to identify potential predictor variables. Variables previously shown to influence human infant milk intake included continuous traits such as infant length and weight, number of feedings per day, and maternal BMI, as well as discrete traits such as whether the mother is employed outside the home and highest level of maternal education. Most of these variables were assessed by a questionnaire sent out to all participants of the FFNZ study. Additional data on duration and number of breastfeeds, amount of formula intake, and any information on complementary foods came from two 24-hour dietary recalls taken across a two-week period. 

The team now had a long list of variables that could influence human infant feeding behavior as well as the corresponding data on these variables from their study population—but which variables were most likely to predict milk volume? To weed out the least helpful variables, researchers turned to statistical tests, including linear regression. This test determines how well the value of the variable on the x axis (such as infant weight or feedings per day) predicts the value of the variable on the y axis (milk volume). 

After running all the statistical analyses, the researchers found that the strongest predictors of human milk intakes were infant age, infant BMI, number of breastfeeds per day, and infant formula consumption (as a yes/no question) [1]. These variables and 14 others with high predictive value were then put through a more rigorous type of regression analysis, called LASSO. This type of statistical modeling further narrowed the list of predictive variables while also defining the relationship between these variables and milk intake in an equation. Once again, infant age, infant BMI, number of breastfeeds per day, and infant formula use came out as the strongest predictive variables [1]. The LASSO regression also identified maternal employment as a strong predictor, likely because it influences the number of human milk feedings in a day.  

The two resulting equations, called Human Milk Intake Level Calculations (or HuMILC), demonstrated “near perfect” agreement [2] with actual values of milk intake [1]. The mean value for human milk intake from the study’s dose-to-mother method was 762 milliliters per day (ml/d) and both equations provided mean estimates within 0.5 ml/d of this value. However, the predictive equations were less accurate on an individual level. Although determining individual infant intakes would be optimal, accurate study population means are still more valuable than using a “one volume fits all” model as they can highlight variation across populations. This data from infants 7 – 10 months old are lacking but are important for understanding how human milk provides nutritional support for infants who are also consuming complementary foods [2]. 

This study is a perfect example of what it means to work smarter, not harder. The equations may be lengthy (and require you to remember what PEMDOS stands for), but with just a calculator researchers can transform their study data into reliable estimates of milk volume. It is not yet known whether these equations are applicable to populations outside of New Zealand, but the authors encourage external validation using milk data from other populations. Larger data sets will enable future researchers to improve the predictive capability of these equations and expand our understanding the role of human milk in infant nutrition. 

References

1. Haszard JJ, Heath AL, Taylor RW, Bruckner B, Katiforis I, McLean NH, Cox AM, Brown KJ, Casale M, Jupiterwala R, Diana A. Equations to estimate human milk intake in infants aged 7 to 10 months: prediction models from a cross-sectional study. The American Journal of Clinical Nutrition. 2024 Jul 14: 102-110
2. Bandyopadhyay S, Schulze KJ. Predicting human milk intake: a step forward for infant nutritional assessment. The American Journal of Clinical Nutrition. 2024 Jul 14: P5-6
3. Savenije OE, Brand PL. Accuracy and precision of test weighing to assess milk intake in newborn infants. Archives of Disease in Childhood-Fetal and Neonatal Edition. 2006 Sep 1;91(5): F330-2.

Back to SPLASH!^® Home