Monday, August 11, 2014

Swirled or shaken? Does shaking actually damage milk - the scientific evidence


Perhaps one of the most widespread pieces of advice women expressing milk will hear is about the best way to remix milk after expression. Human milk separates after expression (Figure 1) and needs to be remixed before feeding. 

Figure 1: Milk samples (1.5 mL) from 3 different mothers allowed to separate to show the variation in milk fat. Photo: EA Quinn/Biomarkers & Milk.

Many, many websites and books have strict recommendations for the remixing: swirl, never shake. 

As an anthropologist and a bench scientist, I am always interested in the natural history of advice, Where did this advice to swirl, never shake, come from? Upon investigation, I found 3 primary reasons given for why expressed milk should be swirled, never shaken:
1)      Shaking denatures proteins
2)      Swirling helps to remove fat globules stuck to the side of the container
3)      Shaking damages cells.

But, like many before me, I can’t find any actual scientific evidence. I started with PubMed, the national, searchable database of scientific literature ( Figure 2).   

Figure 2: Screenshot of my PubMed search for shaking breast milk. Stirring breast milk looked similar, but with less hits. None were relevant. Image: me.
Here is what I found – and how I went about trying to solve this issue.
Let’s start with #1: shaking denatures proteins. There are many, many different types of proteins in human milk and these are highly variable in size. In addition to size variations, there are also going to be major differences in the way in which proteins are folded – with denaturing being the unfolding of these proteins. 

There are no published papers on this topic. Since the literature was not an option, I turned instead, to math and physics. The idea that shaking denatures proteins is based on the shear force the proteins would be exposed to during shaking.  We need two pieces of information here: what level of force is generated by shaking and what level of force denatures proteins. 

Several reference values for the shear force necessary to denature proteins were available in the literature. Most data however, were based on experimental models of the protein in isolation, when micro-tweezers could be used to literally rip the protein apart. This model is not valid here – what we need is a measure of the shear force necessary to denature a protein in a liquid medium.  Again, we don’t have any studies in human milk, so we will have to substitute water as a medium – and given the composition of human milk, this is a reasonable substitute. In a highly viscous medium, similar to milk, α-amylase (a protein involved in starch digestion found in breast milk), requires a force of 3 x 10^4 Pa to denature the protein.
Figure 3: Alpha-amylase, of pancreatic origin. Image from:
Proteins with beta folds, it is estimated, would be much more resistant to shear force. The predicted force (in a highly viscous medium) necessary to shear a beta protein would be 2 x 10^5 to 10^7 Pa. 

So how much force can a human arm generate? Again, there is no direct measurement for a human shaking a highly viscous medium (but there is plenty of data on ketchup).  If you’ve goggled this (or seen Mythbusters) you know an elite boxer can punch with 5000 pounds of force, or more than 22,000 Newtons. 
Figure 3: The action of boxing, as demonstrated by Manny Pacquiao, is very, very different than the action of shaking breast milk in a container. Image:

But boxing, pitching, and shaking are very different actions – and this causes some interesting differences in the way in which force must be calculated.
When you pitch or punch, the entire body is involved in the action. Punching involves rotation at the waist, shoulder, and elbow. Pitching involves the same rotation, plus the fingers. But shaking is typically done with a stationary shoulder and body and the primary point of movement at the elbow. This is going to limit the force the arm is generating – and the forces extended to the container.  The best analogy in the literature for shaking a container is, remarkably, swinging a hammer, as the hammer swing comes mostly from the elbow. Even a hammer swing is probably an over-estimation, as the shoulder may be involved. 

The average speed for swinging a hammer is 4 meters per second, with maximum times closer to 10 meters per second. The average hammer weights about 3 pounds – the average container of breast milk will weigh a little bit more than 4 ounces.  Now, one thing about a liquid medium is that the forces within the fluid may vary considerably – but it is still unlikely that the human arm will generate enough force through shaking to damage the proteins.  Earlier studies (Thomas and Dunnill 1979) reported that proteins were often stable under shear forces exceeding 9000 s-1 for more than 15 hours. 

One additional factor serves to protect the proteins in human milk, particularly those proteins that are hormones or immune factors rather than more nutritional proteins. We know for example, that many of the hormone proteins are bioactive infant circulation, and thus survive digestion in the infant stomach. Many of these protein hormones are found in a glycosylated form – that is, with the protein has added sugars attached to it that protect the protein structure and serve to reduce the risk of denaturing.  Other proteins may be packaged within the membrane bound fat globules, which will further act to protect the proteins from damage. 

Skipping ahead to #3 – shaking damages cells – the math from above remains important. Again, it is unlikely that the human arm is capable of generating enough force to damage the cells in the milk. Most of the research looking at shear forces and cell damage uses a platelet cell model (Christi 2001). Platelets are not found in human milk, and are also more prone to cell damage and death than many of the other cells commonly found in human milk. Again, human milk specific data are not available – except for spinning in a centrifuge – and we are substituting a leukocyte model for the reference cell. Moazzam et al., (1997), in a study of leukocytes exposed to shear forces in a rat model, found that leukocytes incurred very little damage from shear forces.  Breast milk cells are likely exposed to high shear force at multiple points in their normal life course – from milk ejection to swallowing to digestion, and may be more resistant to cell damage (Papoutsakis 1991). 

Concern #2: Swirling helps remove the fat stuck to the side of the contained.
Again, there are no available data. However, in a study of ultrasonic mixing versus stirring, Garcia-Lara et al., (2013) found that samples mixed by ultrasonic waves had higher fat, suggesting that the ultrasonic mixing was better at removing fat adhering to the sides of the container compared to manual mixing. Current research protocols for measuring milk fat in samples have used multiple inversion techniques to mix milk to ensure adequate mixing – and inversion is a lot closer to shaking than swirling.

So what is the final verdict? There is no published evidence to support that shaking actually damages breast milk when compared to swirling. Many of the issues identified with shaking are better described as myths, and simply do not hold up when the actual shear forces are calculated. Certainly, it would be awesome if we could do an in depth study of this – have women swirl and shake milk with sensors on the hand and in the milk cup and actually measure the acceleration of the hand and then analyze the milk. I suspect however, that we wouldn’t find much damage. 

Sarah and I were discussing the origins of this myth while I was working on this post over the last several days.  She made a really excellent point about this myth – “Really I think it's just one more way to make breastfeeding seem super hard and easy to mess up.” And it seems to be one piece of advice that while well meaning, may contribute to the persistent idea that human milk is fragile, easily damaged, and requires a high degree of care. It serves as one more perceived “threat” mothers (and fathers and caregivers) pose to human milk – the “if you aren’t careful, you’ll damage it and you can’t damage formula*” underlying subtext that serves to undermine breastfeeding mothers.

*see all the recalls and allowable insect parts

Bee JS, Stevenson JL, Mehta B, Svitel J, Pollastrini J, Platz R, Freund E, Carpenter JF, Randolph TW. Response of a concentrated monoclonal antibody formulation to high shear. Biotechnol Bioeng. 2009 Aug 1;103(5):936-43. doi: 10.1002/bit.22336.

Yusuf Chisti. Hydrodynamic Damage to Animal Cells Critical Reviews in Biotechnology, 21(2):67–110 (2001).

García-Lara NR, Escuder-Vieco D, García-Algar O, De la Cruz J, Lora D, Pallás-Alonso C. Effect of freezing time on macronutrients and energy content of breastmilk. Breastfeed Med. 2012 Aug;7:295-301. doi: 10.1089/bfm.2011.0079.

Jaspe J, Hagen SJ. Do protein molecules unfold in a simple shear flow? Biophysical Journal. 2006;91(9):3415–3424.

Moazzam F1, DeLano FA, Zweifach BW, Schmid-Schönbein GW. The leukocyte response to fluid stress. Proc Natl Acad Sci U S A. 1997 May 13;94(10):5338-43.

Papoutsakis ET. Fluid-mechanical damage of animal cells in bioreactors. Trends Biotechnol. 1991 Dec;9(12):427-37.

Physics@ UNWA. Smashing bricks and the ballistic pendulum: more collision examples. URL: Accessed: 8/9/14.

Thomas CR, Dunnill P. Action of Shear on Enzymes - Studies with Catalase and Urease. Biotechnology and Bioengineering. 1979;21(12):2279–2302.

Thomas CR, Greer D. Effects of shear on proteins in solution. Biotechnology Letters 2010; 33(3) 443-456. DOI : 10.1007/s10529-010-0469-4.

van der Veen ME, van Iersel DG, van der Goot AJ, Boom RM. Shear-induced inactivation of alpha-amylase in a plain shear field. Biotechnology Progress. 2004;20(4):1140–1145.

Wednesday, June 11, 2014

Might the gut explain colic?

Infant colic is a common problem among infants, with between 10-30% of US infants identified as colicky infants. Despite the common experience that is colic, colic is in fact poorly understood. Medically, colic is defined using Wessel’s criteria, perhaps better known as “The Rule of 3s”: crying that last more than 3 hours a day, for more than 3 days a week, for over 3 weeks, although most parents do not seek medical advice for colic.  There are also stereotypical behaviors associated with a bout of colic – the infant will pull their legs up as if in pain, there is increased abdominal bloating, passing of gas, face flushing, and a specific, high pitched cry.  Colic usually resolves by 4 months, making it difficult to determine if treatments worked, or the colic naturally resolved on its own. 

Historically, any number of factors have thought to be play a role in the development of colic ranging from the tired old trope of the refrigerator mother (cold, uncaring mother), to reflux, protein sensitivity, or issues with lactose digestion. 

Emerging evidence does suggest that some of the cases of colic may be caused by infant reflux, GERD, or protein sensitivity. However, in a large number of infants, these causative factors can be ruled out. New evidence however, suggests that the factors underlying colic may be quite complex – and definitely non-human. Colic, it seems, may be a behavioral response by infants to differences in the microorganisms living in their GI tracts.

Three studies have investigated the association between infant gastrointestinal microbiome and infant colic, with some very interesting results.
For the first study, Savino et al., (2004) collected fecal samples from 71 infants aged 15-60 days with no prior use of antibiotics or probiotics. There were no differences in the amount of several common aerobic GI bacteria. However, infants with colic were much less likely to have bacteria from the genus Lactobacillus. Those colicky infants that did have Lactobacillus had much smaller colony forming units (a way of quantifying the amount of bacteria in the infant’s GI tract) 1.26 cfu per gram, compared to 2.89 cfu/gram.

Figure 1: Lactobacillus acidophilus at high resolution.  This is one of the more common forms of Lactobacillus, and is frequently used in the production of yogurt. Image:
In the follow-up study, Savino et al., (2005) investigated these microflora differences in 30 colicky and 26 unaffected controls; all exclusively breastfed. Colic was defined by physician assessment, and fecal samples were collected from each infant.  Fecal samples were analyzed for bacterial type – down to the species level.  Unlike the prior study, Savino et al., reported no differences in total Lactobacillus colony forming units between the colicky and non-colicky infants. However, at the species level, there were some striking differences.

None of the colicky infants had Lactobacillus acidophilus. No non-colicky (control) infants had Lactobacillus brevis or Lactobacillus lactis lactis. These differences may be tremendously important in influencing infant health and GI function. First, while all Lactobacillus are anaerobic bacteria (they do not require oxygen), a few species are capable of glucose fermentation, producing as byproducts carbon dioxide and ethyl alcohol. The species that produce carbon dioxide and ethyl alcohol? You guessed it: Lactobacillus brevis and Lactobacillus lactis lactis – the very bacteria ONLY present in colicky infants.

Lactobacillus acidophilus (Figure 1), the bacteria not found in colicky infants, is an important contributor to immune function in the GI system, and likely promotes immune activity and the development of oral tolerance to food antigens (reduces risk of reactions to foods). Presence and absence here becomes a perfect storm: the increase in gas and ethyl alcohol producing species and the loss of protective species may increase the risk of GI infection in colicky infants, and may contribute to the gas and distress commonly associated with colic. This may also explain why fecal calprotectin, a hormonal marker of inflammation commonly used as a measure of GI damage, is increased in the stools of infants with colic (Rhoads et al., 2009).

However, these three studies are somewhat limited in their study design – primarily by the use of a single sample per infant. Infant microflora may differ between colicky and non-colicky infants – but how does this process occur?

deWeerth et al., (2013) have some answers. Piggy-backing on an existing study, they identified 12 infants with colic and 12 infant without colic from a larger sample of infants. The infants had similar ages, birth weights, and current weights, but differed in the amount of crying reported by the parents. They then analyzed nine fecal samples collected from day 2 to 4 months postpartum, using DNA measurements to determine the types of bacteria in gut microflora. As found in the earlier studies, there were significant differences in the types of bacteria, especially Lactobacillus. These differences were present in the samples collected at 2 weeks postpartum - colicky babies already had less Bacteroidetes, and more E. coli and Enterobacteria, while non-colicky babies had more Bifidobacteria and Lactobacillus gasseri. Most striking however, was the decreased diversity in bacterial species found on days 14 and 28 in the feces of the future colicky infants - and remember, these samples were collected before the colic emerged. 

Figure 2: Image from deWeerth et al., 2013. The distribution of bacterial groupings in the fecal samples collected from infants at day 14 (before the onset of colic) classified by whether or not the infant developed colic. Each Circle with a letter is a non-colicky infant, each red square with a letter is a colicky infants. There is little overlap in the bacterial groupings of the infants. The paper is open access and you can read it here
By 4 months postpartum, there were no differences in the microflora between colicky and non-colicky infants. This is about the time that colic usually resolves, and deWeerth et al., speculate that this shift in the microbiome may be one of the potential mechanisms.

Here is what I want to know : if there are known differences in the GI microflora between colicky and non-colicky infants, what factors contribute to these differences? deWeerth speculates genetics or chance encounters may contribute to these differences. But what is if is something else? What is the differences start with milk? Is it possible that differences in the oligosaccharides in milk may promote the growth of different types of bacteria? Alternatively, may the cfu units in milk differ between women? If the later, could maternal probiotics be a treatment for colic? No one really knows – as far as I can tell, no one has investigated milk composition and microflora differences in the milk or guts (and maybe vaginas?) of mothers who have infants with or without colic. Or, perhaps following the elegant study design of deWeerth, a longitudinal study utilizing maternal and infant microflora measurements of control and colicky infants recruited into the study at birth and followed over the first four months of life. In any event, there are a lot of questions and missing pieces remaining – and milk may be an important one!

Next month: the milk microbiome: or there are bacteria in human milk and that is a good thing.


Rhoads JM1, Fatheree NY, Norori J, Liu Y, Lucke JF, Tyson JE, Ferris MJ. Altered fecal microflora and increased fecal calprotectin in infants with colic. J Pediatr. 2009 Dec;155(6):823-828.e1. doi: 10.1016/j.jpeds.2009.05.012.

Savino F, Bailo E, Oggero R, Tullio V, Roana J, Carlone N, Cuffini AM, Silvestro L. Bacterial counts of intestinal Lactobacillus species in infants with colic. Pediatr Allergy 
Immunol. 2005 Feb;16(1):72-5.

Savino F, Cresi F, Pautasso S, Palumeri E, Tullio V, Roana J, Silvestro L, Oggero R. Intestinal microflora in breastfed colicky and non-colicky infants. Acta Paediatr. 2004 Jun;93(6):825-9.

de Weerth C, Fuentes S, Puylaert P, de Vos WM. Intestinal microbiota of infants with colic: development and specific signatures. Pediatrics. 2013 Feb;131(2):e550-8. doi: 10.1542/peds.2012-1449.