Can't resist some input here....  I'm Gail Hertz, pediatric resident who
wrote the little green breastfeeding book - My disclaimer is that I own
Pocket Publications.

It is important to consider what type of study is being done....
"statistics" can sound pretty impressive.

Things to ask when reading a study:
1. Who were the subjects and how were they chosen?  The gold standard is to
have the subjects randomly assigned to either group.  [This could be
problematic in some types of studies]  This will minimize  "selection bias"
- the probability that the two groups of subjects are not the same (for
example - there may have been more high risk pregnancies in the MD group)

2. What is the sample size?  That is, how many people were studied?  If the
sample size is too small the study may show effects that are actually due
to chance rather than actual effect of what is being studied.  The study
quoted had 3.9 million  An effect among 3.9 million people will be less
likely to be swayed by chance than if we were looking at a study with only
39 people.  Keep in mind that percentages change greatly with sample size -
 9 women with placenta previa  out of  39 is almost 25%  -  9  out of 3.9
million is about 0.000002 %.


3. Is the study prospective [going forward in time]  or retrospective
[going backward in time] ?  The problem with retrospective studies is that
again you can have unintentional selection bias.  An example of this would
be that say in 1991 only insurance company x would reimburse for midwife
deliveries.  That means that the people who would be likely to end up in
the midwife group would be those whose companies could afford insurance x
or those who could pay out of pocket for midwife care.  If somebody's
company only provided insurance y or z which would only reimburse for MD
deliveries they would be less likely to end up in the midwife group. --
These "confounding factors" [factors that confuse the issue by adding other
variables to the study]   can vastly change the end result of a study.  A
prospective study - one that goes forward in time allowing you to randomize
the subjects will help eliminate some of the selection bias.

4. What does the study tell me?  It is easy to read too much into a study,
especially if it's eloquently written.  Sometimes all you can say is that
these people got these results.  If the study was not well done because of
selection bias or other factors then it is not a good idea to generalize
the results to the rest of the world.

There.  That is just a start on how to "read" a study. Hope it's helpful.

Gail

Gail Hertz, MD
Pediatric Resident - PennState Geisinger Health System

Pocket Publications
40 Mapheliah Road
York PA 17402-8214