Dear all:

I'd like to make some comments about sample size because there is a view out there that more is 
always better or if we don't like a conclusion, maybe the sample size was not big enough.  Even 
after a lot of epidemiology when I did my MHS, I still didn't quite get the difference between a 
large sample size and an adequate sample size.  My doctoral advisor would get quite annoyed with 
us when we didn't think through all the implications of sample size.

The one that I originally didn't get is that if a study detects a statistically significant difference, 
regardless of how small a sample size was used in the study, then the sample size WAS adequate.  
The statistical power is contained in the p-value statements.  You cannot state that the sample 
size was too small when the result is statistical significance.  To give the recent example, a 
response of lowered intake in 11 out of 12 subjects could very well be statistically significant and 
that would mean that the sample was big enough. 
  
That does not mean that you have established causality, however.  You need to ask several other 
questions, such as:

1) Was the difference that was detected relevant (i.e. is the difference physiologically important)?
 
So, in the case of the 12 women and the infants suckling for longer, but had a lower intake, I can 
think of a few question that should be considered sucvh as:
a) Is the reduction in intake at that feeding enough to make a difference to the overall intake of 
that infant or is it so minimal a difference as to be physiologically irrelevant?  E.g.  If it was 0.1 oz 
lower versus 2.0 oz lower at a feeding you might have different conclusions about the physiologic 
relevance.
b) Do infants compensate by increasing their intake after a feeding when they suckle longer and 
take in less?  If you measured the infants total intake over the course of a 24 hour period and 
discovered there was no difference you might feel differently than if they had a deficit that was not 
compensated for.
c) What sort of dose response occurs in the infant's intake if the mom were to increase the amount 
she drinks before a particular feeding or if she were to increase the frequency of drinking? (This 
one would probably never get past human ethics review boards).  You might feel differently about 
a baby that had a lower intake a one feeding a week than a baby that had a lower intake every day.

2) Is the sample selected for the study reflective of the specific population that the study is meant 
to represent?  (Was it biased?)
a) Were the drinking habits of these women similar to the drinking habits of the general 
population?  Would the infants of women who drank more or less on a routine basis than the 
women selected for this study respond differently? 
b) Were the infants in this study reflective of the general population?  Would particular subsets of 
infants (premies, SGA, etc) respond differently to the ingestion of alcohol by their mothers?  You 
might feel differently about the consequences of  a lower intake in a preterm 3 lb baby than a 
fullterm 8 lb baby.

I'm sure all of you can come up with a lot more questions of this nature.  When you see a positive 
result and you want to explore all the angles (and you should) you don't really need to question 
sample size, you do need to look at representivity, bias, & relevance.


For negative studies, however, you need to look very very carefully
at the adequacy of the sample size, particularly when serious but rare side effects or illnesses
are concerned. This is where the study comparing NEC rates in premies between those taking 
formula and those taking human donor milk (plus formula) clearly did not have an adequate 
sample size.  Even though the initial calculations of sample size were statistically valid, the 
assumptions were wrong.  The sample size was on an expected prevalence of NEC that was much 
higher than the current rates in the study population.  As a condition becomes more rare, you 
need to have larger sample sizes to detect a difference.  The results showed that NEC was twice as 
high in the formula group, but because the sample size was too low, what may have been an 
extremely important physiologic difference was not statistically significant.  A larger sample would 
have answered the question.


This is also why the studies on new drugs are tricky.  If the sample sizes are small, they may not
pick up rare but very serious side effects.  If the time of observation is too short, they may not
pick up on very important problems that develop gradually over time.  If a conclusion about safety 
is reached on the basis of one trial, it is not the same as the same conclusion being drawn by 
various different researchers on a consistent basis.  For something like NEC or SIDS which are 
serious, you may have a much lower tolerance for what is an acceptable increase in risk than an 
outcome like improper palate development.  For an immediate outcome like NEC, you might have 
a different tolerance than for something increased risk of cancer as an adult.

This is why mothers who evaluate the risk of formula based on the small sample size of their own 
children may reach the conclusion that their children are fine on formula if they get lucky and
have a children that don't have serious consequences. Or their children may experience the 
consequences such as ear infections when their children are in nursery school and weaned off of 
formula and not make the connection that this is due to the fact that she stopped breastfeeding at 
2 weeks of age.

When it comes to the brain of a developing newborn, I'd rather err on the side of safety with 
alcohol and suggest a moderate approach such as the Institute of Medicine recommendations - 
not too restrictive, but still protective.  Yesterday, Rachel Myr just gave us one more reason to 
suggest sensible moderation in terms of the mother's ability to care for her infant.

Best, Susan Burger, MHS PhD, IBCLC

             ***********************************************

To temporarily stop your subscription: set lactnet nomail
To start it again: set lactnet mail (or digest)
To unsubscribe: unsubscribe lactnet
All commands go to [log in to unmask]

The LACTNET mailing list is powered by L-Soft's renowned
LISTSERV(R) list management software together with L-Soft's LSMTP(R)
mailer for lightning fast mail delivery. For more information, go to:
http://www.lsoft.com/LISTSERV-powered.html