Section 12: Cohort Studies

Front 1

Relative risk

Back 1

Value of 1 = rate of disease in exposed and unexposed are equal (no association)

Value greater than 1 = increased risk among the exposed (harmful exposure effect)

Value less than 1 = decrease risk among the exposed (protective effect)

Front 2

Relative risk (RR) you can detect can move further away from the value of 1 as:

Back 2

Power increases. Level of significance decreases. Expected number of cases decreases.

Front 3

Attributable Risk (AR)

Back 3

Quantifies the risk of disease in the exposed group that can be considered attributable to the exposure by removing the risk of disease that would have occurred anyways from other causes

value 0 indicates rates of exposed and unexposed are equal, therefore no association

value >0 increased risk among the exposed
value <0 decreased risk among exposed (protective effect)

Front 4

Attributable risk percent (AR%) also referred to as:

Back 4

Attributable rate percent, attributable proportion, etiologic fraction

Front 5

Interpretation of attributable risk percent (AR%)

Back 5

AR% value of 87.8%: 87.8% of cancer deaths among smokers may be attributable to their smoking

If 2023 lung cancer deaths occurred among smokers during 4 year period, how many would not have occurred if the smokers had not smoked? 2023 x .877 = 1744 deaths. 1744 deaths could have been averted during the period if smokers didn't smoke.

Front 6

Population attributable risk percent (PAR%)

Back 6

The percentage of disease or death in the population that is attributable to an exposure. PAR% is affected by the prevalence of exposure in a population.

Front 7

If P=1 (the entire study population is exposed) then PAR% equals

Back 7

AR%

Front 8

Relative risk (RR)

Back 8

Strength of association btw exposure and disease. Provides info that can be used to judge whether a valid observed association is likely to be causal.

Front 9

Attributable risk (AR)

Back 9

Provide the public health impact of an exposure. Assumes the exposure is CAUSAL

Front 10

Relative risk is used more by ______.

Back 10

Epidemiologists

Front 11

Attributable risk is used more by ______.

Back 11

public health officials and policy makers

Front 12

Magnitude of relative risk (DOES/DOES NOT) predict the magnitude of attributable risk.

Back 12

DOES NOT!

Front 13

Attributable risk percent and Population attributable risk percent

Back 13

Attributable risk percent: measures the percentage of disease or death due to a selected cause among exposed.

Population Attributable risk percent: measures the percentage of disease or death due to a selected cause IN THE POPULATION

Front 14

Seroepidemiology

Back 14

measure serum antibody status - most infectious agents can be measured this way to develop a cohort

Front 15

Seroepidemiology cannot be done when...

Back 15

Reinfection is common in the presence of antibody. Disease results from reactivation of a latent infection.

Front 16

No good indicator of susceptibility is available for chronic diseases. So remove people without organ at risk and people who have known suspected history of disease.

Back 16

*

Front 17

Stool DNA Test

Back 17

Not yet cost effective. tests being developed.

Front 18

Choice of cohort

Back 18

general population. limited population subgroups: people working in a particular industry or occupation (school etc). People with prepaid health plans

Front 19

When cohorts not strictly representative of the general population:

Back 19

Caution when attempting to generalize to the general population

Front 20

Cohorts require large numbers of people

Back 20

chorort studies often conducted among groups NOT due to their exposure status, but rather are they able to provide relevant info

Front 21

Sources of data

Back 21

Major consideration is the availability of the data, its accuracy, and its completeness.

Preexisting records. (PROS: available for high proportion, inexpensive, unbiased)
(CONS: exposure info may be insufficient, freq do not contain data on potential confounding variables)

Front 22

Exposure info

Back 22

Rather than medical records, interviews and questionnaires completed by cohort. (PROS: allows data collection on exposures that are not routinely recorded) (CONS: Potential for recall bias, time consuming, and expensive)

Front 23

Exposure info

Back 23

Inaddition to med records and questionnaires, Direct measurements may also be taken. (PROS: more accurate) (CONS: ensure reliability, pertains only to point in time taken, time consuming, and expensive)

Front 24

Death certificates

Back 24

good for diseases that are highly fatal. completely acceptable when endpoint is death. far less reliable for cause-specific mortality: depends on the disease under study, potential bias when certificate is vague or incomplete.

Front 25

Internal comparison group

Back 25

a study group is chosen and divided by exposure assessment into exposed and non-exposed groups. usually decreases the likelihood of bias

Front 26

External comparison group

Back 26

often consists of the general population from the area from which the exposed group is obtained. NOT IDEAL for evaluating risks from occupational exposures because: employed ppl are average healthier than non-employed, they also have better economic circumstances and access to med care.

Front 27

External comparison group (CONT)

Back 27

Often consists of a group believed to be susceptible to the same selective forces as the exposed group, provided this group does not have exposures that might also increase the risk for diseases of interest.

Front 28

Sometimes BOTH internal and external groups are used!

Back 28

Increases confidence of the validity of the findings. if comp groups have dif results, investigator must find out why

Front 29

Diagnostic criteria

Back 29

MUST BE DEFINED BEFORE the study begins - to ensure the absence or presence of disease be made in uniform manner for exposed and unexposed cohort.

The person assigning the outcome should be blinded to the exposure status of the cohort members

Front 30

Potential for selection bias is MORE/LESS of a concern in cohort than in case-control studies

Back 30

LESS

Front 31

Non-differential misclassification (random misclassification)

Back 31

Inaccuracies exist in the categorization of subjects by exposure or disease status but these inaccuracies occur in SIMILAR PROPORTIONS in each of the study groups. Effect is to dilute or underestimate the true risk because the random misclassification increases the similarity between the study groups.

Front 32

Differential or nonrandom misclassification

Back 32

Differential misclassification can result in a biased risk estimate that is either an underestimate, and overestimate, or miraculously, the same as the true measure of association

Front 33

__________ is affected by differences between participants and non-participants.

Back 33

External validity (generalizability)

Internal validity is usually not affected by differences between participants and nonparticipants.

Front 34

Losses to follow-up produce the same sorts of bias in risk estimates as do losses to non participation. MINIMIZE losses through intensive efforts to locate each cohort member.

Back 34

Procedures to facilitate follow-up:
Ask participant to prov a name/number of someone who will know of their whereabouts. DOT. Voter's list. Doctor's office. etc

Front 35

Selection of groups for study.

Back 35

Must be a group in which a large number of people have been exposed to the agent of interest. Excess incidence for the disease under investigation is likely to be detected. Reasonably accurate data for the cohort is available. Exposed cohort must also have an appropriate comparison group.

Front 36

Measurement of exposure

Back 36

When exposure happened many years before, may be difficult to get an accurate measure of a crude classification.

Errors of measurement for exposure or disease are likely to bias the association.

Need a refined classification to determine whether an association exists.

Front 37

Once a cohort has been established for a retrospective study, valuable info can easily be obtained inexpensively by adding a prospective component.

Back 37

*

Front 38

Nested Case-control study within a cohort

Back 38

Controls ARE OFTEN matched to cases on calendar time, length of follow-up, age, sex, and other variables

Front 39

Case-cohort study within cohort

Back 39

Controls not matched to cases - possible to study different diseases using the same sub cohort

Front 40

Advantages of Case-control studies WITHIN a cohort

Back 40

If questionnaire date obtained before the development of the disease, recall bias is minimized. Cost savings. Greater comparability between cases and controls.

Front 41

Disadvantages of Case-control studies WITHIN a cohort

Back 41

Case populations cannot be selected until disease status of cohort is known.

Front 42

Difference between NESTED CASE-CONTROL STUDY and CASE-COHORT STUDY

Back 42

NESTED: Controls are matched. CASE-COHORT: subcohort is not matched (possibly to study another disease)

Front 43

STRENGTHS of Cohort studies

Back 43

Efficient for RARE exposures. Info on MULTIPLE EXPOSURES. Yield info on MULTIPLE OUTCOMES. Clear TEMPORAL RELATIONSHIP between exposure and disease. MINIMIZES bias. DIRECT MEASUREMENT of disease incidence in exposed and non-exposed groups. Strongest observational design for establishing CAUSE-AND-EFFECT relationships.

Front 44

LIMITATIONS of Cohort studies

Back 44

TIME CONSUMING and expensive - especially if prospective. If retrospective, requires availability of adequate records. LARGE sample size. Losses to followup can hurt the VALIDITY of the study. TIME IN DIAGNOSIS methods may bias results.