Alastair G Sutcliffe, Brent Taylor, Jun Li, Simon Thornton, J Gedis Grudzinskas, Brian A Lieberman et al
Sutcliffe A G, Taylor B, Li J, Thornton S, Grudzinskas J G, Lieberman B A et al.
Children born after intracytoplasmic sperm injection: population control study
BMJ 1999; 318 :704
doi:10.1136/bmj.318.7185.704
What do we mean by control?
Suttcliffe et al (BMJ March 13, 1999) compare babies born through
intracytoplasmic sperm injection with controls, matched for social class,
maternal education, region and sex, who were born following unassisted
conception. In their table, they give means, standard deviations and 95%
confidence intervals for injection babies and controls separately. These
confidence intervals tell us about the estimation of the mean of the
population of which the babies are a random sample, which they are not.
They tell us nothing about the comparison between the cases and controls.
Now these controls have been chosen to be similar to the injection babies
in many ways, they are not a representative sample of babies. It is the
difference between the injection babies and the controls which is of
interest, and the confidence interval we should have is that for this
difference in mean. We cannot work this out from the data given, because
this is a paired study and we have unpaired standard deviations. All we
have for the comparison is the implication, not actually stated, that the
differences are not significant. What we should be given are confidence
intervals which tell us how different the cases and controls are.
Competing interests: No competing interests