[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Definitions of study designs



RuthS and all,

Correction accepted with embarassment.

Cohort studies are not the same as case-control studies.

NSWS  -- not same as Iowa Radon study.

In simplifying I became imprecise (but perhaps not inaccurate, because true in a broader sense).



Groups do not = individuals. I make the same argument that [type A] physicians may not be good controls for radiologists [usually typeB]



However, the lively arguments here further impel db placebo control studies, so that, "The controls should be matched as closely as possible" (Ruth S)



Howard Long



Ruth Sponsler wrote:



> Hello all -

>

> A bit of clarification is needed here.  It seems that

> there is a bit of confusion regarding the definition

> of a case-control study vs. an occupational cohort

> study.

>

> At the bottom, I have provided a couple of links that

> explain some epidemiological study designs, and may

> help to clarify these designs a bit further.

>

> The most basic, "fact-finding" type of epidemiological

> study is a correlation analysis.  My class project

> last semester (radon vs. life expectancy in the U.S.)

> was a simple correlation analysis.  The results of a

> correlation analysis may "point to" a phenomenon that

> should be further investigated via a more tightly

> designed epidemiological study.

>

> The NSWS was not a case-control study, but rather an

> occupational cohort study.  A cohort study is a

> comparative analysis of mortality of non-exposed vs.

> exposed persons.  The NSWS stratified by exposure and

> by age and compared exposed vs. non-exposed shipyard

> workers.  The strength of its design, as opposed to

> many other cohort studies in occupational epidemiology

> was that the study compared workers (exposed) vs.

> workers (not exposed), rather than workers vs. the

> general population.  The general population contains

> more individuals disabled by serious illnesses or by

> drug/alcohol problems than do most worker populations.

>  The NSWS design eliminated this 'healthy worker

> selection criterion' of being 'on-the-job' by

> comparing two groups, both on-the-job.  The NSWS study

> of 8 shipyards was initiated because a small cohort

> study of one shipyard indicated an excess leukemia

> risk.  The NSWS involved many more workers at 8

> shipyards and showed this result from the small study

> to be probably due to random chance.

>

> The goal of an occupational cohort study is to examine

> a cohort of workers in an industry vs.an unexposed

> cohort, matched for age and gender,  to see if the

> exposed workers are experiencing an elevated risk of

> any disease from a factor or factors at work.  Most

> occupational cohort studies involve analysis of death

> certificates, and some also link these certificates

> with hospital records.   A non-nuclear example of a

> controlled occupational cohort study would be an

> examination of a cohort of butchers at supermarkets,

> compared vs. a cohort of supermarket personnel who did

> not work with meat i.e. checkout clerks, bakery

> employees etc.  [I didn't make this up.  Studies have

> actually examined this].  This design is relatively

> similar to the NSWS because workers in the same

> facility are compared with one another, the main

> difference between them being the level of exposure to

> the agent under consideration.  A somewhat less

> controlled cohort study, where one would expect a

> 'healthy worker effect,' would be a comparison of oil

> refinery employees vs. the general population.

>

> A case-control study design is entirely different from

> a cohort study, and has a different goal.  In a case

> control study, the researcher is trying to identify

> unique aspects of the disease case's life that may

> have put them at risk for the disease.  In a

> case-control study, the researcher evaluates actual

> cases of a disease, along with controls that are

> individually matched to the cases.  The controls

> should be matched as closely as possible with the

> cases with respect to factors such as age, gender,

> weight or body mass index, smoking status, alcohol

> consumption, parity (especially for studies involving

> breast cancer, not the subject here), and a number of

> other factors that may influence health status.

> Case-control studies have been discussed on the list

> extensively, as the Iowa and Missouri lung cancer

> studies are examples of this design.

>

> Perhaps a non-radiation example will clarify what a

> case-control study is.  If I were to go to a

> developing country, and do a case-control study of

> diarrhea among children, then my controls would be

> kids who did not have diarrhea recently, and my cases

> would be kids with the 'runs.'  I would examine the

> food, water, and living conditions of the cases vs.

> the controls.  Most likely, I would find that the

> diarrhea cases had worse water supply or were perhaps

> eating food that was not prepared in a sanitary

> fashion while controls had access to cleaner water

> and/or food.

>

> The other type of study that I will briefly mention is

> the randomized controlled trial.  This is an actual

> experiment.  Certain types of randomized controlled

> trials are unethical, especially if the something

> being evaluated is harmful.  For example, if I

> suspected that Giardia (a parasite) causes intestinal

> illness, it would be unethical for me to separate some

> nice friendly student volunteers into two groups and

> give one group a glass of water with Giardia in it and

> the other a glass of clean water.  However, in certain

> instances with minor illnesses (the common cold,

> caused by rhinoviruses), randomized controlled trials,

> with consent from the subjects, have been performed to

> evaluate efficiency of transmission of the

> rhinoviruses between different settings i.e.

> temperature, how long the cold virus has 'sat' on a

> surface before the person touches his nose, etc.

>

> Randomized controlled trials are often performed in a

> double-blind fashion, where neither the patients nor

> the caregivers know which treatment is assigned to

> whom.  The most common application of this study

> design is evaluation of medicines.  Only the

> experiment designer knows whether the person is

> getting [silly example!] Viagra or a placebo.  The

> health care providers, who do not know who is getting

> what, record the data and provide the information to

> the researchers, who then determine if there is any

> difference between the Viagra group and the placebo

> group.

>

> In all the types of studies mentioned here, statistics

> improve with improvements in matching between subject

> groups and with numbers of members in the groups.  The

> key constraints are time and money.

>

> Below are links to a couple of pages that explain the

> various designs a bit further.

>

> http://cebm.jr2.ox.ac.uk/docs/studies.html

> http://bmj.com/epidem/epid.html

>

> ~Ruth 2 aka Ruth Sponsler

> >

> > Furthermore, this rejected null hypothesis and the

> > many animal and human

> > confirmations - like the 4,000 case NSWS case

> > control study - suggest

>

> > Howard Long

>

>

>

> __________________________________________________

> Do You Yahoo!?

> Send FREE Valentine eCards with Yahoo! Greetings!

> http://greetings.yahoo.com

> ************************************************************************

> You are currently subscribed to the Radsafe mailing list. To unsubscribe,

> send an e-mail to Majordomo@list.vanderbilt.edu  Put the text "unsubscribe

> radsafe" (no quote marks) in the body of the e-mail, with no subject line. You can view the Radsafe archives at http://www.vanderbilt.edu/radsafe/



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

You are currently subscribed to the Radsafe mailing list. To unsubscribe,

send an e-mail to Majordomo@list.vanderbilt.edu  Put the text "unsubscribe

radsafe" (no quote marks) in the body of the e-mail, with no subject line. You can view the Radsafe archives at http://www.vanderbilt.edu/radsafe/