Clinical Biostatistics: breast pump trial exercise

This website is for students following the M.Sc. in Evidence Based Practice at the University of York.

Paul et al. (1996) reported that 14 women were recruited to a cross-over trial comparing a breast pump with manual breast milk expression. Each women had three 15 minute sessions on each method, three sessions on the fourth and three on the fifth postnatal day. The order of application of the two methods was alternate.

They produced the following table:

Volume of breast milk expressed by manual and pump methods on postnatal days 4 & 5 in 15 minutes sessions in 14 subjects
  Pump       Manual  
Milk expression sessions 42 42
Mean output per session (ml)   46.8 31.2
SD 26.3 15.5
Pump vs Manual t = 3.29, P<0.01

A two sample t test comparing two groups of 42 observations yielded t = 3.31, d.f.= 82, 2 tailed probability P = 0.0014. This is almost the same as what they report, so this appears to be what they did, given that there would be rounding errors.


1. There are at least three things wrong with this test. What are they?

Check suggested answer 1

2. What effect might each of these problems have on the t test?

Check suggested answer 2

3. What better analysis could the authors have done?

Check suggested answer 3


Paul VK, Singh M., Deorari AK, PachecoJ, Taneja U. (1996) Manual and pump methods of expression of breast milk. Indian Journal of Pediatrics 63, 87-92.

Thanks to Felicia McCormick for this example.

Back to Clinical Biostatistics index.

To Martin Bland's home page.

This page maintained by Martin Bland.
Last updated: 9 July, 2008.

Back to top.