Sunday, February 14, 2016

How long does it take breast milk to regenerate after a feeding?


My wife has been breastfeeding and pumping since our son's birth. Starting at his two-month birthday, she switched to pumping 15 minutes per session and we started tracking how much she produced with more criteria than before. The two things we tracked very consistently that ended up being interesting were the time since the last feeding or pumping, and the time of day.

How long does it take to regenerate?

The plot at the beginning has the raw data. There is a pretty clear trend towards greater production with greater time since feeding. Aggregating the results provides a slightly clearer picture:

The error bars here are 95% confidence intervals. Are these columns significantly different? Running a one-way ANOVA yields a p value of 4.4097e-09. Assuming p < 0.05 or p < 0.01 implies statistical significance, then yes...we can reject the hypothesis that all of these results are from the same distribution. Which columns are actually different though?

There are a massive number of ways to test this. A common one is Tukey's Test, and the results are: 

  • 240 minutes compared with 30, 60, 90, 120, 150, and 180 minutes yields p < 0.01
  • 210 minutes compared with 30 minutes yields p < 0.01 and compared with 60 minutes yields p < 0.05
  • 180 minutes compared with 30 minutes yields p<0.05


She gets significantly less milk within an hour of feeding than if she waits 3-4 hours. On average, she gets more than 2.5x as much milk at four hours than at 30 minutes.

Time of Day

We got no significant results when we factored in time of day, but that could very well be due to a lack of data (we have very little data in the evening, night, and early morning). It's difficult to display the time of day results because the time since feeding dominates the overall results. The clearest way I can think of displaying it is to take each bin of time since feeding data and subtract the mean of the bin from each result in the bin and then plot that vs time. This attempts to normalize data from all bins and plot them on the same scale. The results are:

It looks like the period before lunch has better production than the other periods. Separating the day into four hour bins, you get that, on average:
  • midnight to 6 am is 9.8% below average
  • 6 am to noon is 16.9% above average
  • noon to 6 pm is 8.7% below average
  • 6 pm to midnight is 14.7% below average

1 comment :