Field and laboratory methods in human milk research



This article is corrected by:

  1. Errata: Erratum to Field and laboratory methods in human milk research Volume 25, Issue 3, 442, Article first published online: 20 April 2013


Human milk is a complex and variable fluid of increasing interest to human biologists who study nutrition and health. The collection and analysis of human milk poses many practical and ethical challenges to field workers, who must balance both appropriate methodology with the needs of participating mothers and infants and logistical challenges to collection and analysis. In this review, we address various collection methods, volume measurements, and ethical considerations and make recommendations for field researchers. We also review frequently used methods for the analysis of fat, protein, sugars/lactose, and specific biomarkers in human milk. Finally, we address new technologies in human milk research, the MIRIS Human Milk Analyzer and dried milk spots, which will improve the ability of human biologists and anthropologists to study human milk in field settings. Am. J. Hum. Biol., 2013. © 2012 Wiley Periodicals, Inc.


Human milk is a complex and highly variable fluid with a fundamental role in infant health, nutrition, and development (see Table 1 for elements of milk composition). Long the purview of nutritionists and clinicians, investigations of milk composition are increasingly conducted by anthropologists and human biologists. These studies forgo the assumption that milk is passively transferred from mother to infants (Hall,1979), instead reframing human milk as a dynamic evolutionary and cross-cultural phenomenon that links mothers and infants. The nuances of the direct physiological exchange—from potential developmental signaling to the life history trade-offs of mothers and infants—have considerable potential for understanding population and individual biological variability for a host of phenotypes. Mother's milk, including its composition and volume, is part of the dynamic lactation strategy of mothers and nutritional strategy of infants that extend beyond observational measurements of nursing behavior.

Table 1. Notable constituents of mature human milk
Per dL milkaReferences
  • a

    g/dL can also be interpreted as % of total milk volume.

Total energy(60–88 kcal)Jenness,1979; Mandel et al.,2005; Michaelson et al., 1994; Nommsen et al.,1991
Water (86.0–88.0 g)Ogra et al.,2006; Vaughan et al.,1979
Fat (2.5–6.0 g)Brown et al.,1986; Mitoulas et al.,2002; Nommsen et al.,1991; Prentice et al.,1981a; WHO, 1985
 Myristic acid (14:0)
 Palmitic acid (16:0)
 Linoleic acid (18:2n-6)
 Alpha-linolenic acid (18:3n-3)
 Arachidonic acid (20:4n-6)
 Docosahexaenoic acid (22:6n-3)
Protein (0.83–1.30 g)Brown et al.,1986; Mitoulas et al.,2002; Nommsen et al.,1991; Prentice et al.,1981a; WHO, 1985
 Whey protein
 Casein protein
  Insulin-like growth factor-1 (IGF-1) 
  Insulin-like growth factor-2 (IGF-2) 
 Immune factors 
  Secretory IgA (sIgA) 
  Transforming growth factor beta (TGF-β) 
  Interleukin 1 (IL-1) 
  Interleukin 6 (IL-6) 
  Interleukin 10 (IL-10) 
  Tumor necrosis factor alpha (TNF-α) 
Carbohydrates (6.3–8.1 g)Brown et al.,1986; Mitoulas et al.,2002; Nommsen et al.,1991; Prentice et al.,1981a; WHO, 1985
Ash (0.2 g)Jenness,1979
 Vitamin A 
 Vitamin D 

Human milk is the first food for the majority of infants, evolutionarily, historically, and cross-culturally today. In exclusively breastfed infants, milk alone provides the resources necessary for the first six months of postnatal growth. Although milk is not unique to humans—the synthesis of milk in mammary glands is the defining characteristic of mammals—understanding its central role during critical growth and developmental windows in infancy makes milk an important topic of research for anthropologists in general and human biologists in particular. Researchers have only begun to define normal milk variation across populations and how it influences the postnatal period and beyond (Neville et al.,2012). Many of the emerging questions in human biology are linked to lactation, from unique aspects of primate (Hinde et al.,2009; Milligan and Bazinet,2008) and human life histories (Fujita et al.,2011), parental investment (Fujita et al.,2012; Hinde,2009; Powe et al.,2010), and developmental programming (de Moura et al.,2008; Hinde and Capitano,2010; Miralles et al.,2006; Newburg et al.,2010; Palou et al.,2009; Pico et al.,2007; Prentice,2005; Quinn,2011; Quinn et al.,2012; Savino et al.,2009; Stocker and Cawthorne,2008; Weyerman et al., 2007). In addition, lactation is implicated in human evolutionary biology, including the evolution of large brains and body fat (Kuzawa,1998; Martin,1981), childhood (Bogin,1999; Konner,2010; Sellen,2007), reproductive timing (Al-Sahab et al.,2011) and the developmental origins of adult metabolism (Kuzawa and Quinn,2009; Wells, 2003).

Human milk has likely been under considerable evolutionary selection (Hinde and Milligan,2011; Martin,1981), reflecting its central role in infant nutrition. Historical documents suggest that infant mortality rates were very high in the absence of maternal milk, illustrating the selective pressure on milk (Hinde and Milligan, 2011). Milk, then, likely communicates critical ecological information to the infant, such as maternal condition or nutritional availability in the environment (Fujita et al.,2011; Hinde and Capitanio,2010; Hinde,2013), pathogen burdens (Miller and McConnell,2011; Prentice et al.,1983), and potentially long-term environmental information (Quinn,2011).

Milk is a dynamic process between mother and infant, and as with most areas of physiology, sampling methodologies can only approximate the true biological exchange. Collection of any biological fluids is inherently challenging, and this is particularly true for milk. As best practice, we advocate an informed, ethical approach sensitive to the research question, the appropriate milk constituent of interest, and the population studied. The bulk of this review will focus on currently available, validated methodologies for collecting human milk under variable conditions, and analyzing milk for constituents that may be of interest to human biologists. The ethical and informed practices of milk collection for the population must be resolved by the individual researcher after consideration of research question, available resources in the field and the laboratory, and the well-being of subjects. Our own narratives of human milk collection in the field are highly varied (Miller, Fujita, Aiello, and Quinn) and illustrate how population specific some constraints can be.


The physiological processes by which mammary glands synthesize breast milk necessarily reflect, in part, maternal characteristics at the time of lactation, or even in the months, and years before lactation (Hinde and Milligan,2011; Quinn et al.,2012). For these reasons, milk can vary between populations, across lactation, among women within populations, and even within mother during a single nursing bout and throughout the day. Micronutrients (e.g., minerals and vitamins), macronutrients (e.g., fatty acids), energy density (kcal/g), and volume have all been reported to vary, to some extent, among non-Western populations (Jelliffe and Jelliffe,1978; Martin et al.,2012; Prentice and Prentice,1995). Micronutrients and fatty acids may be wholly or partially derived from current maternal diets (Francois et al.,1998; Innis,2007; Milligan,2013; Stuetz et al.,2012; Yakes et al.,2011), although macronutrients appear moderately buffered from short-term nutritional fluctuations because mothers can mobilize body reserves for milk synthesis during lactation (Prentice et al.,1981b; Villalpando and del Prado,1999). Unlike milk macronutrient composition, milk volume may be more sensitive to changes in maternal condition (Ettyang et al.,2005; Prentice et al.,1981b; Villalpando et al.,1992; but see also Pérez-Escamilla et al.,1995). Hormones in milk, such as glucocorticoids and adipokines, are correlated with concentrations in maternal circulation (Bronsky et al.,2011; Savino et al., ,2009; Sullivan et al.,2011; Uysal et al.,2002). Selection of appropriate collection techniques, volume of sample required, and assay methods should be determined by the parameter(s) of interest with sensitivity to the culture, nutritional ecology, and health of the subject population.

Human milk composition and volume change across the lactation period. For example, colostrum, the first milk produced in low volume after delivery, is low in fat but high in protein and specifically immune factors (Ogra et al.,2006). Subsequent to colostrum, mothers produce transitional milk. This transitional milk is characterized by a gradual increase in volume and fat concentration reflecting physiological changes within the milk-producing mammary epithelial/lactocyte cells in the mammary gland and the infant's capacity to digest more complex milk after the initial establishment of the intestinal microbiome (Martin and Sela,2012). At ∼3–4 weeks postpartum, transitional milk is replaced by mature milk with continued volume increase and further physiological changes in the mammary gland (Ogra et al.,2006). More detailed discussions of the process of secretory activation and milk synthesis are available elsewhere (Jensen,1995; McClellan et al.,2008; Neville et al.,1984, 2002). Age-related changes persist in mature milk for many but not all constituents. Prior analyses have primarily investigated macronutrients and energy (Mitoulas et al.,2002; Neville et al.,1984; Picciano,1984; Prentice et al.,1981a), although there is a growing body of research for other milk constituents such as hormones (Bronsky et al.,2011), vitamins, (Allen,1994; Haskell and Brown,1999), and immune factors (Weaver et al.,1998). Infant age, as a measure of the duration of maternal lactation, must be considered during subject selection and as a covariate in statistical models, as age may affect constituent concentration.

Individual milk volume varies considerably among women, with typical production ranging from 550 to 850 mL per day (Neville et al.,1988). Milk supply is determined by the flow of milk through the mammary gland (Akers,2002). Milk throughput is the result of behavioral negotiation between the mother and infant through nursing intensity, frequency, and duration. When the infant is permitted to nurse to satiety, maternal supply is primarily, but not totally, linked to infant appetite (Daly et al.,1992; Wilde et al.,1995). However, prior research has shown extensive variation in maternal behaviors that determine infant access to milk (Gray,1995; Vitzthum,1994).

In addition to inter-individual variation, there is considerable variation in milk production within an individual woman, and even within a single feed (Daly et al.,1993a; Kent et al.,2006). Foremilk, the first milk expressed during a feeding, is usually relatively dilute and low in fat compared with hindmilk (Daly et al.,1993a), the last milk consumed during in a feeding bout. Milk synthesis and composition is related both to the total volume of milk removed at a feeding (Prentice et al.,1981a) and the interval between feedings (Daly et al.,1993a; Jackson et al.,1988; Lai et al.,2010). Longer inter-nursing intervals are associated with higher milk fat in hindmilk, reflecting downregulation of lactose synthesis in relation to milk stasis in the mammary gland (Akers,2002; Hinde et al.,2009; Wilde et al.,1995). Differences in time intervals between feedings can be problematic for crossstudy comparisons, especially those that rely on single samples. These factors make selecting an appropriate milk collection methodology paramount to minimize the potential impact of sampling on composition. Similarly, recording the inter-nursing interval and degree of milk evacuation during the previous feed before collection are important for interpreting the concentration of constituents in milk and volume collected (Hood et al.,2009).

Finally, there is diurnal variation in some milk components, particularly fat. In one study of milk fat content, Kent et al. (2006) found that milk fat concentration was highest during the day and evening and was lowest at night and in the morning in Australian women. Similarly, Garza and Butte (1986) found that total milk energy was lowest in the morning in the United States, which they attributed to daily variation in feeding schedules. By contrast, Prentice et al. (1981a) reported higher levels of milk fat in the morning compared with the afternoon in rural Africa, possibly attributed to a greater degree of night feeding in this non-Western population. Standardizing milk collection to a particular time of day, preferably in the morning, is preferred for milk collection protocols (Garza and Butte,1986; Ruel et al.,1997).


Obtaining a representative milk sample for research purposes has numerous logistical and ethical challenges, which are often amplified for researchers working in remote conditions or with nutritionally stressed populations. The methods described in the following sections detail current protocols for collecting milk. These represent common practices co-opted from clinical and field-based studies by nutritionists and lactation biologists that are currently used by human biologists. However, the application of any of the following protocols should reflect best practice scientific methodology while also accommodating population specific limitations.

Milk expression

Milk may be extracted using manual expression or a breast pump. To manually express milk, mothers use their hands to stimulate the mammary gland and direct the flow of milk. Manual expression is an effective and extremely low-tech methodology, which can be easily demonstrated to women. Alternatively, either hand or electric powered pumps can be used to express milk. Electric pumps require power and may be cost prohibitive; by comparison, hand pumps are cheaper, easier to operate in populations with limited knowledge of pumps, pose less risk of contamination due to having fewer parts, are disposable, and can be used in any field setting (Boo et al.,2001). Several reliable pumps are manufactured by Kaneson (Osaka, Japan) or Medela (McHenry, IL). There is a growing body of evidence, however, that suggests pumps alter milk composition by evaporating water content and artificially changing milk composition. Moreover, the use of the pump can affect milk synthesis during the subsequent feed (De Curtis and Rigo,2012; Morton et al., 2012). Additionally, pumps require sterilization between uses and are not recommended for multiperson use, as there is a risk of spreading infection between mothers. Finally, our field observations (Fujita, Miller, and Quinn) indicate that pumps may be uncomfortable or unfamiliar to mothers and are generally discouraged for field applications. Regardless of the expression method, handling of milk requires strict compliance to biohazard safety measures. Milk can contain viruses such as HIV and Hepatitis B virus (de Oliveira et al.,2009; Kline,1999; Oxtoby,1988).

Milk sampling

Currently, there exists no single sampling protocol that is universally applicable for every constituent of potential interest in human milk. This reflects the variability in milk macronutrient composition across a single feeding, and the as yet largely unknown variation of microconstituents in human milk. In particular, the changes in fat and fat-soluble components of milk can be problematic for researchers attempting to collect a representative milk sample (Garza and Butte,1986). Herein, we review several sampling methods that may be of interest to researchers (Butte et al.,1983; Neville,1995; Neville et al.,1984, 2002; Picciano,1984; Prentice et al.,1981a; Ruel et al.,1997) but should note that this list is primarily targeted toward human biologists and will not contain all methodologies currently available to clinicians. Ultimately, the sampling protocol should be the “best fit” for the milk constituent of interest while also considering possible limitations from the population or the field conditions (Neville,1995). Collection protocols must also be sensitive to possible diurnal variation in milk composition (Ruel et al.,1997), longitudinal changes associated with postpartum duration (Mitoulas et al.,2002), possible seasonal effects (Prentice et al.,1981b), time since last feed (Daly et al.,1993a; Kent et al.,2006), volume of milk consumed at the prior feed (Daly et al.,1993a), and maternal physiological let down (Neville et al.,1984). We recommend standardizing collection as much as possible and recruiting as many participants as the research can support to help statistically control for these variables.

Full expression

Many researchers consider the collection of all the milk from a full mammary gland to be the optimal method of determining representative lipid content in milk (Neville et al.,1984; Villalpando and del Prado,1999). One or both mammary glands are fully evacuated of all accumulated milk, then homogenized and aliquoted until assaying. Surplus milk, beyond the needs of the target assays, may be returned to the infant, may be discarded (see section on ethical considerations below), or may be archived for future assays.

Expressed alternate breast

The expressed alternate breast protocol creates an amalgam of milk using samples collected over a 24-h period. Milk is removed from alternating breasts while the infant feeds on the opposite breast which triggers the physiological milk letdown response in the mother but does not disturb the infant by removing him from the breast (Butte et al.,1984). At the end of each collection, the volume is recorded, and a small subsample collected. The remainder can be returned to the infant or discarded. This procedure is repeated for every subsequent feed over 24 h, and the subsamples are then mixed together in proportion to the volume collected at each feed (Neville,1995). This method is time consuming for participants, provides a no more accurate sample than other methods (Neville,1995), and the nutritional and behavioral disruptions described previously remain a concern.

Mid-feed sampling

A mid-feed sample is taken 3–5 min after let-down from the mammary gland. This method has been shown to give reliable measures of milk fat at the population level (Allen et al.,1991; Prentice et al.,1981b; Xiang et al., 2005), although it may not give an adequate estimation at the individual level (see Neville et al.,1984). This collection method has relatively low nutritional impact on infants and may be a good fit for researchers working with undernourished populations. In particular, coupling this sampling strategy with feed-specific test weighing (described below) may be a useful strategy for those working with marginally nourished populations when full mammary expression is not possible.

Fore and hind sampling

Fore and hind sampling requires relatively small amounts of sample (Neville,1995). In this method, milk (4–5 mL) is collected from the mother immediately before the baby is put to the breast and immediately after the baby has finished suckling. The samples are then placed into separate 10 mL polypropylene tubes (Jackson et al.,1988), analyzed separately for fat, and a formula (Y = 0.61XFORE + 0.58XHIND − 0.273) derived from full breast expression is used to determine average lipid content (Saint et al.,1984). This collection method has a relatively low nutritional impact on infants and may be a good fit for researchers working with undernourished populations (Jackson et al.,1988; Prentice et al.,1981a).

Fore (or hind) milk sampling

Fore (or hind) milk samples may provide an acceptable surrogate for the above methods if ethical, logistical, or practical considerations prohibit the collection of a more representative milk sample and where the use of a single fore or hind sample has been accepted as a viable alternative to collecting a more representative milk sample. This approach may be useful for constituents where prior work has demonstrated that fore or hind milk is an adequate representation of interest: for example, vitamin A concentrations per gram of milk fat at the population level (Stoltzfus and Underwood,1995) and calcium (Prentice,2000) or if the constituent of interest demonstrates minimal variation across a breastfeeding bout (such as many proteins; Neville et al.1984). However, this approach is problematic for macronutrients and many hormones, such as leptin (Daly et al.,1993b; Karatas et al.,2011).

The strategies described above work best when coupled with feed-specific test weighing (described below), which provides a measurement of milk volume. Encouraging the infant to nurse on the opposite breast during collection promotes let down and ease of collection and may minimize disruption to normal nursing behaviors.

Sample preparation and storage

If possible, samples should be chilled immediately to reduce the likelihood of bacterial growth and deterioration of milk constituents (Neville,1995). Bacteria in milk will consume sugar and reproduce, artificially increasing “protein” content and decreasing sugar content (Oftedal, pers. comm.). Many, but not all, milk constituents are stable for several days at refrigerator temperatures (Slutzah et al.,2010), although stability under refrigeration is unknown for many hormones and cytokines. For analysis of immunoproteins or cortisol, sodium azide or protease inhibitors can be added to samples to prevent the growth of bacteria; however, these substances can destabilize other metabolic hormones (such as adipokines) and should be used cautiously. For storage, the samples should be aliquoted into smaller containers, filled nearly to the top to minimize oxidation, and frozen at −70°C, although they can be stored at −20°C for up to 12 months with minimal degradation (Jensen,1995; Neville,1995).

Some constituents may require stricter sample handling and preparation procedures to promote the measurement validity. For example, vitamin A is susceptible to decay on light or oxygen exposure and requires amber tubes with air tight caps or equivalent (Stoltzfus and Underwood,1995). Measuring hormones may require collection in polyprolene vials instead of polystyrene and a well-maintained cold chain to maintain sample integrity. For fat soluble constituents, it is important that the milk be homogenized with gentle swirls when preparing aliquots for analysis of the constituent and fat. Considerable validation work remains necessary for many of the constituents in milk, and pilot work and validations should be conducted before data collection. When minimal information is available on the constituent, using methods previously validated for other biological fluids, such as saliva, urine, or plasma, provides a reasonable starting point.

Measuring volume and infant intake

The volume of milk ingested by the infant represents another important piece of information for the study of human milk and milk synthesis. Infant intake increases as the infant ages until human milk alone cannot sustain the infant's metabolic requirements and complimentary feeding becomes necessary, at around 6 months (WHO,2003; Sellen,2007), after which volume of intake stabilizes before a gradual decline as nursing frequency decreases.

Milk intake varies considerably between infants and is determined by maternal-infant compromise on conflict between milk supply and milk demand (Daly et al.,1993b; Hinde and Milligan,2011). Because the amount of milk mothers are capable of producing may not accurately represent the amount of milk infants ingest, determining infant nutritional intake is more complicated than determining milk macronutrient concentrations (Butte et al.,1991; Daly et al.,1992; Hinde and Milligan,2011; Wilde et al.,1995). In addition, because milk macronutrients vary throughout the day, especially fat, assumptions of energy transfer are difficult to determine. It is also important to note that the following methods are primarily used to estimate infant nutritional intake and do not reflect the intake of other constituents in human milk.

There are two primary methods for estimating volume transfer to the infant. These are (1) indirect test weighing of the mother or infant; and (2) use of doubly labeled water ingested by the mother and measured over 10–14 days in the infant's urine. Behavioral observation of nursing, while cheaper and requiring less technology than other methods, is largely inaccurate and should not be used as a measure of infant milk intake (Furman and Minich,2006; Scanlon et al.,2002).

Indirect test weighing

Indirect test weighing relies on weighing the infant before and after each feed on a scale (accurate to the gram) and calculating the weight change between the measurements divided by the specific gravity of human milk (1.032 g/mL; Neville et al.,1988). Measurements are collected each time the infant nurses over a 24-h period. Drawbacks to this method include the interruption of usual nursing behavior, particularly overnight, and an inconsistent relationship between infant milk intake and changes in infant weight (Brown et al.,1986; Neville et al.,2002; Scanlon et al.,2002). Indirect test weighing must be adjusted for water loss due to infant respiration (Woolridge et al.,1987), sometimes called “insensible water loss.” There is likely between-population variation in insensible water loss due to ecological aridity and temperature that may confound the interpretation of results, although the extent of this variation is currently unknown. Indirect test weighing can also be modified to the weight-test-weigh method, where the infant is weighed before and after a single feeding, rather than over a 24-h period (Meier and Engstrom,2007). However, the use of one feed may be difficult to standardize without knowing individual variation in “typical” nursing bouts.

Doubly labeled water

Isotope dilution is also used for measuring milk energy intake. Mothers are given deuterium labeled water, which is incorporated in breast milk. Urine is collected from the infant repeatedly during an observation window (up to 14 days), and a decay function is used to calculate infant intake (Butte et al.,1991). The cost of analyzing samples is considerable—often several hundred dollars per participant. In addition, doubly labeled water may also under- or over-estimate intake, with ranges (15% over to 5% under) similar to those seen in test weighing protocols (Haisma et al.,2003; Reilly et al.,2005; Scanlon et al.,2002; Wells et al.,1996).

Ethical and practical issues

Collecting human milk is not without ethical considerations. Research sampling of milk relies on taking a primary foodstuff from an infant. Great care should be taken to minimize the disruption (either nutritional or behavioral) to the mother and the infant. Based on our experiences with nutritionally marginal populations, we recommend either mid-feed sampling or collection of fore and hind milk samples coupled with weigh-test-weigh of the infant. In some populations and field settings, fore or hind milk samples can be collected provided researchers justify the use of a small sample for their constituent of interest. It has been estimated that a small volume of around 5 mL is well within the estimated excess production of the mammary each day (Jensen,1995). The caloric equivalent of such a small volume of milk is less than 7 kcal (based on roughly 70 cal per 100 mL), which is an acceptable compromise for infants whose average milk intake ranges between 400 and 800 kcal/day during the first year of life (Butte et al.,1984; Neville et al.,1988).

However, these conservative collection protocols cannot determine daily milk intake or production, which may be necessary for understanding relative differences in milk production and composition. We have previously described two detailed methods of estimating infant milk intake, each of which has drawbacks (Butte et al.,1991; Scanlon et al.,2002). One possible compromise may be the expressed alternate breast method described above. However, this method involves potential issues surrounding the disposal of the remaining milk after sample collection. The milk can theoretically be fed to the infant, but this requires either the introduction of a bottle or the use of a LACT-AID or similar system. The LACT-AID system is essentially a bag and tube system that is attached the nipple and allows the infant to receive expressed milk while nursing at the breast. As with a breast pump, the LACT-AID system requires appropriate sterilization of equipment between mothers. The alternative, a bottle, is an ethical conundrum: there is a possibility that a researcher-provided bottle may promote its use, including the use of breast milk substitutes (for discussion see Palmer,2009). This can result in adverse health consequences for infants especially where clean water is not available and sterilization of the bottle cannot be ensured. Researchers must also consider the possibility of unsafe storage of the milk in households without refrigeration.

Simply disposing of the excess milk is itself problematic: if done on site, it may devalue the milk in the eyes of the mother, reflecting the power imbalance between the lactating woman and the researcher in addition to the loss of nutrition for the infant. One solution we propose is to collect the entire milk sample and dispose of excess elsewhere or to provide the remainder (after a subsample is collected) via cup but not bottle for immediate consumption. A similar solution has been used by Martin et al. (2012) during milk sampling among the Tsimane in Bolivia without incident. This solution preserves the relationship between researcher and participant without promoting bottle use and helps avoid nutritional disruptions that may accompany human milk research.

Overcoming the real and perceived loss of nutrition is paramount for researchers working in nutritionally stressed populations. Maternal perception of the volume of loss compared with supply may negatively influence willingness of women to participate in the study. This maternal perception is based in fact: collection does take energy from both the baby and the mother. Under normal nutritional situations, such loss is relatively minor and likely compensated by the mammary gland that day or at a minimum over the next several days (Geddes et al., 2008; Kent et al.,2006). However, in limited nutritional contexts, the resulting deficit may be more severe, especially if repeated measures or a full mammary expression are used. One strategy may be nutritionally provisioning women with supplemental calories following the milk collection interview, provided this is within social norms. The potential impact on the infant should be relatively minor, but full mammary evacuation should not be used for closely spaced, repeated measures such as daily or weekly samples. By comparison, the fore and hind, fore or hind, or mid samples have a smaller nutritional impact and could be used several times a day or week for fine-grained analysis of diurnal variation. However, the overall nutritional impact of such a frequent collection may approach that of a single full mammary expression, and care should be made to ensure that even modest additional energy requirements (10–20 kcal/day) are compensated. It should be noted that these nutritional needs will change with infant age, and disruptions will likely be less severe in older infants than in younger infants.

The perception of milk loss can go beyond consideration of the needs of mothers and infants, affecting other family members or members of the community. One example is in Kenya, where some husbands believed rumors that researchers were draining their wives of milk and depriving their infants of nutrition (Fujita,2008). This unfounded rumor was corrected by explaining to the men that women manually express their own milk assisted by a local nurse and that only a few “spoonfuls” of milk were taken. In studies requiring a larger sample volume, community and family involvement may be a barrier to women's participation in research.

Although the wide variety of alternatives for milk sampling can help facilitate field research in diverse populations, it poses a problem for crosspopulation comparisons and study interpretation. The frequently used method of full breast expression for measuring milk synthesis and volume can create ethical, social, and logistical barriers to research. On the other hand, alternative milk sampling strategies introduce measurement variation to results that can obscure the true ranges of variation in human milk across the world. One solution to this dilemma is to carefully pilot test sampling strategies before use in field settings, and to accurately and thoroughly describe methodology in publications so that the limitations of crossstudy comparisons can be clearly defined and strategies can be developed to overcome such constraints.

We do not propose to have a single solution for the above issues, nor do we suggest that there is a universal answer. We encourage researchers to carefully balance the nutritional, health and safety, and social needs of the study population, logistical challenges of the field setting, and the desired milk sampling method to reach the ideal protocol for their research needs following best practice. We also believe that it is imperative that researchers strive for clear and thorough communication with participants and appropriate parties on the sampling protocol and to explain why such sampling is necessary.


Milk samples can be analyzed for a variety of nutritive, hormonal, and immunological components. Before laboratory analysis of human milk for macronutrients and other biomarkers, researchers should consult detailed methodological references such as Filteau (2009), Hood et al. (2009), or Jackson et al. (1999) for discussion of analytical methods. In addition, researchers should consider how they report their results: common units found in the lactation literature include density (mass/volume of sample), mass per unit protein, mass per kcal of milk energy, total mass from a given expressed volume, and others. There is no consensus on the reporting of results, and researchers should consider reporting more than one measure to facilitate comparison across studies. Table 2 provides a summary of methods for analyzing macronutrients, their cost and requirements, and companies providing assay reagents.

Table 2. Methods for analyzing macronutrients in human milk
 Approximate cost and requirementsCompany/institutionReferences
  1. Prices are valid as of publication.

Total fat
 Micro-Rose-Gottlieb (gravimetric)∼300 μL whole milkVWR, Sigma-AldrichAOAC,1975; ISO,2001; Oftedal,1984
∼$100–200 for reagents
Wet lab with explosion-proof fume hood, sensitive scale
 Creamatocrit∼75μL whole milkEBay (hematocrit centrifuge), Medela, VWRLucas et al., 1978; Wang et al.,1999
∼$150–$1500 for centrifuge
$30 for capillary tubes and sealer
Total protein
 Bicinchoninic acid (BCA)∼5 μL whole milkPierce, FisherSciKeller and Neville,1986
∼$160 for kit
∼$150–250 for 96-well plates
Wet lab and spectrophotometer
 BioRad/Bradford/Coomassie∼10–50 μL whole milkBioRad, FisherSciKeller and Neville,1986
∼$120 for kit
∼$150–250 for 96-well plates
Wet lab and spectrophotometer
 CHN elemental analysis for total nitrogen∼20 mL whole milkPerkin-ElmerPower et al.,2002
Call for price
 Kjeldahl total nitrogen∼10 mL whole milkSigma-Aldrich, VWROftedal and Iverson,1995
∼$50 for reagents
$various for digestion, distillation, and titration equipment
Total sugar (Lactose)
 Dahlquist enzyme immunoassay∼50 μL whole milkAbnovaDahlqvist,1964
∼$360 for kit
Wet lab and spectrophotometer
 Phenol-sulfuric∼5 μL whole milkSigma-Aldrich, VWRDuBois et al.,1956
∼$100 for reagents and supplies
Wet lab with explosion-proof fume hood, personal protective gear

Although this laboratory review focuses on the macronutrient content of human milk, there are a wide variety of other factors that are of interest to human biologists. In particular, immunoproteins, hormones, micronutrients, oligosaccharides, and fatty acids are all biologically active compounds that play an important role in infant nutrition, health, and physiology. It is beyond the scope of this review to weigh the merits of all possible techniques; we recommend that researchers further explore the milk literature and pilot test their assay of interest before a population analysis for a given constituent.

Analysis of proteins

Milk proteins provide nutritional, immunological, and hormonal support to the developing infant. Milk proteins can be broadly classified as casein or whey proteins, with whey proteins generally representing 60% of total protein at midlactation (in contrast to cow's milk, which is primarily casein protein). The primary specific proteins in milk, such as α-lactalbumin, immunoglobulins, lactoferrin, and lysozyme, are whey proteins (Conti et al.,2007). Proteins are generally assayed for total protein and then for specific hormones and immune factors; if desired, analyses on casein and whey fractions can be performed.

Total protein can be measured by several inexpensive colorimetric assay techniques. The two most suited for human milk are bicinchoninic acid (BCA; Pierce) or BioRad/Bradford (BioRad), with comparisons suggesting Pierce BCA as the best suited for human milk (Keller and Neville,1986). The Pierce BCA kit protocol should be modified for a higher dilution and reduced incubation time to adjust for the levels of protein in human milk (Miller,2011). Although reagents are inexpensive, analysis requires a wet laboratory and a spectrophotometer. There is some concern that these colorimetric assays may overestimate total protein (Keller and Neville,1986; Lonnerdal et al.,1987) and may be sensitive to the choice of standard.

Alternatively, elemental analysis of nitrogen bonds is an option for determining total protein for researchers looking to outsource to a specialized laboratory. Elemental analysis of carbon, hydrogen, and nitrogen (CHN) determines the total nitrogen bonds in a sample of milk by combustion and measurement of waste NO2 and is a direct measure of total protein (Power et al.,2002). Finally, researchers may wish to choose the “gold standard” Kjeldahl method for determining total nitrogen, which uses large volumes of sample (10 mL; Oftedal and Iverson,1995). For more information on the Kjeldahl method, interested readers are referred to Hood et al. (2009) or Atkinson and Lonnerdal (1995).

There are a wide variety of specific proteins of interest to human biologists in milk (Table 1) including immunoglobulins A, G, and M, lysozyme, lactoferrin, hormones, growth factors, and cytokines (Conti et al.,2007). Generally, these biomarkers are quantified using immunoassay techniques. Immunoassay methods are based on the ability of an antibody to bind to a specific molecule and of a label to produce a specific signal in response to this binding. Two major immunoassay labels are enzyme immunoassays (EIA) and radioimmunoassay (RIA; Skelley et al.,1973; Wisdom,1976). EIAs are ubiquitous, and commercial kits are available to measure hundreds of different potential biomarkers. Kits can be somewhat expensive and may require modification for use with human milk. Investigators can save money but invest more time developing their own EIA using commercially available antibodies (See Crowther,2009 for a detailed handbook on ELISA development). EIAs require a wet laboratory with standard equipment, including a spectrophotometer. RIAs use radioactively tagged antibodies to measure concentrations of unknown sample. RIAs are less often used by smaller laboratories due to the risks associated with handling radioactive materials.

Analysis of milk fat

There are several well-described methodologies for analyzing the fat content of human milk. The two best known for work with human milk are the creamatocrit method and the micro-Roese-Gottlieb (alternatively, micro Rose-Gottlieb) method.


Using the same equipment as for measuring hematocrit, one can measure the fat layer in milk, known as the creamatocrit. Approximately 75 μL of milk is drawn by capillary action into standardized glass capillary tubes (75 × 1.5 mm2) and capped on one end. Tubes are then spun for 15 min at 12,000g, immediately removed and placed upright with the cream layer oriented up. The thickness of the cream and total column length later can be read using Vernier calipers in 0.05 mm increments. The ratio of the two can then be converted to grams of fat per liter using standardized equations (Lucas et al.,1978; Wang et al.,1999). For field purposes, we recommend creamatocrit methods as the equipment and space constraints tend to be less than those for the micro-Roese Gottlieb method.


The micro-Roese-Gottlieb method analyzes samples in duplicate using ammonium hydroxide, ethyl alcohol, ethyl ether, and petroleum ether in combination to extract the soluble lipids from milk samples (AOAC,1975; ISO,2001; Oftedal,1984). As ether is lighter than milk, soluble lipids from milk will be dissolved into the ether fraction. This fraction will sit on top of the rest of the sample and can be removed, placed into test plates for initial weighing, then slowly removed using a specific washing technique. The change in weight, once corrected for the initial weight of the sample, allows for determination of milk fat as a percentage of total weight. For human milk, a modified micromethod only requires a volume of 125 μL in duplicate (250–300 μL total). This method requires an explosion-proof hood, which may limit use.

Analysis of milk sugar/lactose

As lactose is the predominant sugar in milk, lactose concentrations can be measured directly or indirectly as the total sugar content of milk. Measuring the carbohydrate portion of milk can either be done as an exclusive measure of lactose content or as total sugar (including oligosaccharides). Total sugar will overestimate the amount of lactose in milk; therefore, two terms cannot be used interchangeably.

One primary method used for measuring milk lactose is a colorimetric technique based on a modified Dahlquist protocol (Dahlqvist,1964). Briefly, a small volume (∼50 μL) of milk is heated to 37°C and is sequentially treated with Ba(OH)2 at 0.3 M and Zn(SO4) at 5%, before vortexing and centrifuging at 4°C/3000 g for 15 min. The clear supernatant is added to sodium phosphate buffer with β-galactosidase, and incubated for 1 h. A total of 500 μL of this solution are transferred to fresh tubes with 2 mL Glox Solution (Sigma Chemicals) and 0.5 mL of dH2O added. After vortexing, tubes are read at 450 nm on a linear curve with standards prepared from commercially available lactose in the range of 2–10 mg/mL. This assay is available as a commercial kit, which can be modified to run on a standard spectrophotometer (Abnova Lactose Assay Kit, KA167).

Total sugars can be measured using a phenol-sulfuric acid method (DuBois et al.,1956; Marier and Boulet,1959). A total of 5 μL of milk is diluted in 20 mL of dH20. Samples are measured in triplicate, using 1.6 mL each of the initial dilution. One milliliter of phenol (at 11% concentration) is added to each tube (3 per sample) and vortexed before the addition of sulfuric acid. After 10 min, each tube is transferred into a water bath to stop the chemical reaction. Samples are read on an ultraviolet visible spectrophotometer, using a standard curve of known quantities of lactose in the range of 0–50 μg lactose/mL. The readings from the spectrophotometer are then multiplied by the total weight of the dilution and divided by the initial sample weight to calculate the percentage of sugars in the milk sample (for more information, see Hood et al.,2009; Oftedal,1984; Oftedal and Iverson,1995).

Total milk energy

Although bomb calorimetry is the gold standard for measuring the total energy content of milk, its applications to human milk have been fairly limited. A large volume of sample is usually required for combustion. While small scale bomb calorimetry used to be available, the machines are no longer manufactured (Hood et al.,2009). The calculation of milk energy from proximate assays described above has been validated with bomb calorimetry previously in non-human primate milk (Hinde et al.,2009). Energy density in human breast milk can be calculated from the relative concentrations of fat, carbohydrates, and protein using the formula 9.25(F) + 5.65(C) + 3.95(P) where fat, protein, and carbohydrates are measured in grams per 100 mL of milk (Garza et al., 1985).

New field-friendly techniques

Recent work has established field-friendly methodologies for milk collection and analysis. Dried milk spots (DMS) provide an opportunity for the assessment of specific biomarkers within human milk, while the MIRIS Human Milk Analyzer (HMA) is a point-of-care machine for the determination of milk macronutrients. Currently, neither technology has addressed milk sampling and volume concerns raised by this review, but we hope that future work will establish best practices for these technologies.

DRIED Milk Spots

In addition to on-site analysis of macronutrient composition, recent techniques have explored dried milk spots (DMS) on filter paper as a field-friendly method for storage, transport, and analysis of biomarkers in human milk. This work was pioneered by Brown et al. (1982), who matched whole and DMS for anti-rotavirus and anti-enterotoxin IgA. Titers of IgA were detectable in dried spots after storage at room temperature; however, comparisons were not made with whole milk samples. Recently, Miller and McConnell (2011) matched IgA levels in whole milk and DMS samples using ELISA. Nearly 70% of IgA was recovered from the DMS, and after adjusting for time at ambient temperature, the R2 between whole and dried samples was 0.68. There was a small but statistically significant decline of IgA with time at ambient temperature; this did not bias the relationship between whole and matched pairs. In the field, DMS collection requires a single-channel pipette that handles volumes between 50 and 150 μL, tips, Whatman 903 filter paper, desiccant, and plastic bags. It is recommended that researcher's pilot test this technique for their biomarker(s) of interest before using in the field. Future work will expand the repertoire of available biomarkers in DMS, including lactoferrin, lysozyme, cytokines, growth factors, and metabolic hormones.

MIRIS Human Milk Analyzer

The Human Milk Analyzer (HMA), produced by MIRIS, represents new opportunities for the study of human milk composition (García-Lara et al., 2012; Menjo et al.,2009). The machine uses infrared transmission spectroscopy to measure the macronutrients in milk against a known reference library included with the machine. Sample measurement requires ∼1 mL per test (MIRIS recommends analyses in triplicate) with each sample taking less than 1 min for analysis. The HMA performs best when coupled with a sonicator for analyses.

Measured quantities of fat, protein, and lactose are used to calculate milk energy, with the units of g/100 mL for macronutrients and solids and kilocalories per 100 mL for milk energy. Casadio et al. (2010) compared paired milk samples from 20 lactating women with HMA analysis and commonly used analytic techniques, as described above. Overall, they identified significant but modest differences in milk macronutrients between specific laboratory assays and HMA results. Mean fat was 0.3 (±0.4) g/dL higher, protein was 0.2 (±0.2) g/dL higher but lactose was underestimated by 0.4 (±0.5) g/dL) compared with standard analytical methods. These differences reflect known issues with the comparison assays and HMA reference standards. Although some of these differences in milk macronutrients did reach statistical significance, the overall effect on energy calculations was 1.7 kcal/100 mL, a nonsignificant difference in milk energy.

Unlike other currently available, common analytical methods, the HMA is not dependent on specialized laboratory facilities. Instead, it is a standalone unit weighing ∼3 kg which runs on battery power and produces no waste besides the initial milk sample, making it ideal for field conditions with limited electricity and transportation limitations. Although sample volumes are higher than those for most other techniques, the portability removes many current barriers to successful field collection. At present, the MIRIS HMA is available in Europe and Australia but is not currently available in the United States.

The HMA has not been rigorously tested under field conditions or used in field based studies of human milk to date. Its applications have been primarily clinical, where it is used for direct bedside adjustment of milk composition for preterm infants (Menjo et al.,2009) and similar applications, although it has been used for research purposes in Australia (Casadio et al.,2010). In particular, one of the benefits of the machine is much like current point of care devices in use by human biologists (such as the Hemocue Hb201+), the HMA machine is essentially a point of care device for analyzing human milk and can be used to provide rapid feedback to participants on milk composition. This immediate feedback, when coupled with milk volume intakes, may be useful to mothers in determining infant energy intakes.


The study of human milk is at a critical point in its development (Neville et al.,2012). While prior clinical and nutritional researchers have thoroughly described the composition of milk, human biologists have the opportunity to develop hypotheses filtered through the lens of human variation (Fujita et al.,2011; Miller,2011; Quinn,2011), evolution (Hinde and German,2012; Hinde and Milligan,2011), and culture (Miller,2011) to further understand this integral experience of mothers and infants. As human biologists, we are well placed to incorporate a global perspective into the study of human milk. Although we as researchers must pay considerable attention to balancing appropriate collection and analytical practices with the needs of our study participants, doing so will greatly improve our ability to understand human milk as an important aspect of human biological variation and health. We hope that human biology as a discipline can promote the sharing of collection and laboratory protocols, data, and field experiences to those researchers who are interested in incorporating milk studies into their research. With the arrival of new field-friendly analysis techniques, the study of human milk can expand into previously under-studied populations.