|
|
|
Given the frequency with
which studies are posted on a.s.m. and because so many press releases and
news reports begin with "a study released today," or "studies suggest,"
it is important to know how to evaluate a study and examine its conclusions.
This brief outline attempts to list some of the principles needed to do
such evaluations.
(A mnemonic: People say fleas on every dog really seem unhealthy) 1. Examine
population..
1. Examine the population
of the study from two different angles.
b) Are controls and subjects evenly matched on all relevant variables? If not does the study point out the variables which may have affected the results but which they could not control. 2. Look at the
size of the study population.
3. Who is financing/supporting
the study?
4. How are the results
being obtained?
b) Is there a distinction made between in vivo (i.e. results in living human beings) and in vitro (i.e. results in test tubes and petri dishes). Very frequently results in test tubes and petri dishes do not translate into results in living human beings. Similarly results on animals are a step removed from results on humans and frequently will not translate. 5.
What is being studied?
6.
How is improvement or prevention defined?
7. Examine the
raw numbers.
8. Does the body of the
study actually support the conclusion?
9. Finally even if the study
itself seems logical and all seems to be in order, is it being used
appropriately.
Or is is being used to justify a course of action or to prove a theory
which does not logically follow.
I have quoted the following with the permission of Dr. William. Rich who has authored a web site on gyn cancers. URL is http://www.gyncancer.com. I highly recommend Dr. Rich's web site. He has given permission for us to use it in any way which we feel will help the members of this group. "Numbers don't lie, but what is inferred from them is almost always a distortion. For instance, what is the probability that the next time you fly someone will have brought a bomb onboard to blow up the plane? This can be estimated and will be a very small number. Assume that it is one in one hundred thousand flights (1:100,000). What is the probability that there will be two people with bombs on your flight? This will be an exceedingly small number, and is calculated by multiplying 1:100,000, byWhen you look at the statistics of a study, make sure no one has invalidated the equation in advance by bringing his/her own bomb aboard. If you don't know what's wrong with the mathematical conclusion here, email me. Tetje |
The
articles below appear in the British Medical Journal. They are eminently
readable.
How to read a paper : the Medline Database How to read a paper : Getting your bearings (deciding what the paper is about) Extract:The science of "trashing" papers It usually comes as a surprise to students to learn that some (perhaps most) published articles belong in the bin, and should certainly not be used to inform practice. Below are some common reasons why papers are rejected by peer reviewed journals.
How to read a paper : Statistics for the non-statistician How to read a paper : Statistics for the non-statistician. II: "Significant" relations and their pitfalls How to read a paper : Papers that report drug trials How to read a paper : Papers that report diagnostic or screening tests Study Design This page gives a brief comparison of the advantages and disadvantages of the different types of study.
Case-Control Study
|