Evidence-Based Medicine: Types of Study Design

Study designs immediately tells you

1) Basic structure of the study
2) What you can/can't get from study
3) potential problems you should look for
Many ays of classifying eg
!"bser#ational #s$ %&perimental/ 'nter#entional( Based on hat )*% '+,%S)'-.)"/S do$
"bser#ational don0t pro#ide any diagnostic or therapeutic inter#ention specific to the research1 +ote tat
patient2s) may still recei#e inter#entions outside the study
! /etrospecti#e #s$ 3ross!Sectional #s$ 4rospecti#e( based on ho the in#estigators collect data in time
Individual types:
5/emember that any study ithout a control group cannot make causations/associations$ )hey can only be
descripti#e as they simply describe a simple group$
1) 3ross sectional 2.lays obser#ational)( 6escribes the pre#alance orf disease and looks at
associations1 but 6"%S +") tell you hether one is causing the other
2) 7ollo up 2ithout a control)( 4rospecti#e obser#ational study1 descriptpi#e only$ Since no control
group1 they simply describe a single group$ 3onclusions regarding association "/ causation can't be made
3) 3ase report/3ase series( /eport is one anecdotal e#idence1 as a series is a bunch of ppl$ )his type of
study is only helpful in pro#iding initial clues to recogni8ing a clinical problem
9) 3ase control( 3ases 2 disease) compared to 3"+)/":S 2/o disease)$ ;sually retrospecti#e
4ros( useful study to do hen the onset of disease takes a long time$ .lso useful hen rare disease$
!useful for multiple risk factors
3ons( recall bias 2systematic error) or ha#ing the disease e#en before e&posure to risk factor 2random
error)$ RISK can't be deterined t!erefore use "dds ratio#
$$%& &o $$%& Total
'luo(etine in )rd triester 19 < =
&o fluo(etine in )rd triester 3<3 >3? =
Total 3@@ >3< =
"dds Ratio$
"dds of infants *it! 44*+ ha#ing been e&posed to fluo&etine= 19/3<3$
What are the "dds of infants *it!out 44*+ ha#ing been e&posed to fluo&etine= </>3?$
)herefore1 the "dds ratio A 19/3<3 A B$39$
</>3?
"dds that a case as e&posed ./3 .6
"dds ratio A A A
"dds that a control as e&posed B/6 B3
case control needs to be #erified ith prospecti#e 3ohort study$
)hus1 the usual order of studies is as follos(
1$ 3ase report/ case series
2$ "bser#ational studies
3$ 3ase control studies
9$ 3ohort studies
B$ ;ncontrolled1 clinical trials
<$ /andomi8ed1 controlled clinical trials
6ef'n( Incidence( 4roportion of subCects ho -%) the disease o#er a period of time 2ie$ only ne cases)
5+ote that the denominator ill only be indi#iduals ho are suscpetible to the disease$ )hose
ho already had the disease before the study ill be e&cluded from denominator
$revalence( 4roportion of subCects ho ha#e the disease 2ie$ .:: cases1 old and ne)
6ef'n( +onfounder( ,ariable associated ith the risk and outcome1 leading to false conclusion$
6ef'ns(
eg) Risk of myocardial infarction in a high risk population is 10% for subjects who do not smoke and 25% for
subjects who smoke
/iskA D.bsolute /iskD A ,-.
.ttributal /iskA risk difference$$$ smokers #s$ non!smokers /0-.1
/elati#e /iskA /isk /atio A2risk ith smoking / risk as non smoking)A ,#-
55cohort study can sometimes be done Eretrospecti#ely0 by identifying from medical records subCects ith or
ithout the putati#e risk factor and no finding and e#aluating them to see if the outcome has de#eloped$ T!us2
t!e 3ey difference bet*een case control and co!ort studies is t!at in case control *e start *it! sub4ects
*it! and *it!out t!e outcoe2 *!ile in co!ort studies *e start *it! sub4ects *it! and *it!out t!e ris3
factor#
B# E(periental /Interventional1 Studies
inter#entionalA researchers pro#ide some type of inter#ention 2t&1 education etc)
;ncontrolledA no control group$ 3an't really conclude that the outcome is 6;% to the inter#ention
3ontrol group( can be a placebo "/ an acti#e medicine
2