Friday, May 13, 2011

Gourdis : Chapter 9 : Cohort Studies


 

General

-          determine whether there is an association between factor / characteristic and development of disease

-          derive appropriate inferences regarding possible causal relationship from patterns of association

 

Design of a Cohort Study

 

-          investigator selects group of

o   exposed individuals

o   non exposed individuals

-          incidence : a/ (a+b)

-          example

o   coronary heart disease CHD

§  smokers à exposed  / non smoker à not exposed

§  both groups followed for development of CHD

§  incidence in both groups compared.

-          identify new cases

o   can determine new cases à can determine whether a temporal relationship exists between exposure and disease

o   temporal relationship must be established if we are to consider exposure possible cause

 

Comparing Cohort Studies with Randomized Trials

 

-          difference

o   presence / absence of randomization

-          without randomization

o   when we observe an association

o   left with uncertainity

§  association may be result of fact that people were not randomized to exposure

§  not the exposure, but rather factors that led people to be exposed that are associated with diseae

§  example

·         increased risk of disease in workers at factory

·         most workers live in certain area

·         exposure

o   place of work or residence

 

Selection of Study Populations

-          two ways

o   create study population by selecting groups for inclusion on bais of whether or not exposed.

o   select defined population before any members become exposed (or before exposure identified)

§  select based on factor not related to exposure

·         community

 

 

 

Types of Cohort Studies

 

-          prospective cohort study

o   also               

§  concurrent cohort / longitudinal study

o   issues

§  time involved

-          retrospective cohort

o   also

§  historical cohort study

§  non current prospective study

o   use historical data to ascertain exposure

o   outcome (disease / no disease) is known at time study is begun.

 

Examples

 

-          The Framingham Study

o   cardiovascular disease

o   many exposures were defined.

o   second method of selection

§  selected on residence

o   cohort study that begins with exposed / non exposed group -> focuses on specific exposure

o   cohort study that begins with defined population à can explore role of many exposures

-          Incidence of breast cancer and progesterone deficiency

 

 

Cohort Studies for Investigating Childhood Health and Disease

-          exposure dose was calibrated on bais how far person was from point of bomb drop

 

 

-          at what point should individuals in cohort be first identified

-          sample from one centre / several clusters / nationally

-          how long should cohort be followed  / changes in knowledge etc

 

Potential Biases in Cohort Study

 

-          bias in assessment of outcome

o   can be fixed by masking

-          information bias

o   if quality and extent of info obtained is different for exposed and non exposed persons

-          bias from non response and loss to followup

-          analytic bias

 

 

When is a cohort study warranted

 

-          when we have good idea of which exposures are possible causes of disease

-          need to minimise attrition

-          historical records not available to do retrospective cohort study à case control

-          disease has low rate of occurrence.

-           

 

 

 

 

No comments: