POTENTIAL BIASES IN COHORT STUDIES
1. Bias in the assessment of the outcome
2. Information bias
3. Biases from non-response and lost to follow-up
4. Analytic bias
BIAS IN THE ASSESSMENT OF THE OUTCOME
A person's judgment as to whether disease develop may be biased
Referring to the assessment of the disease or of the outcome
To minimize bias (i.e. case control study):
o we should have straight criteria
o there should be written and strict criteria to follow in order to determine whether the outcome
or the disease developed or did not develop
INFORMATION BIAS
The quality of information obtained should be comparable in both groups
Likely to occur in historical cohort studies
o because that will depend on the validity of the records that you are going to examine to establish
the exposure
o if these records are not that reliable, then we will have a problem okay of information bias
In a historical cohort study to determine the exposure you will need information obtained from past
records
Figure 1. Design of a cohort study. (Source: Celentano, D. D., Szklo, M., & Gordis, L. (2019). Gordis epidemiology
(6th edition.). Philadelphia, PA: Elsevier. Page 179.)
It is essential that the quality of the information obtained be comparable in both exposed and nonexposed
individuals.
Person’s judgment as to whether the disease developed may be biased if the person who decides whether
disease has developed in each subject also knows whether that subject was exposed: addressed by
masking
BIASES FROM NON-RESPONSE AND LOSS TO FOLLOW-UP
Non participation and non-response can complicate the interpretation of the findings
● In a cohort study, you will require the participants to report to you, not to your research center or to your
clinic or hospital, at several intervals or more than once and sometimes your participants will just get tired
of reporting to you every month or every six months or even every year
○ If they do get tired of reporting to you it is like they have dropped out from the study
○ If the number of participants that dropout is considerable or then we will have a problem in
the interpretation of findings
○ This is also referred to as attrition or just like in education when you say attrition rate. We are
referring to the dropout rate
● Those who dropped out from the study if they are many than that can affect the findings and complicate
the interpretation of the findings
● In connection with randomized trials, nonparticipation and nonresponse can introduce major biases that
can complicate the interpretation of the study findings
● The incidence rates calculated in the exposed and nonexposed groups will clearly be difficult to interpret
Analytic bias
Epidemiologists or the investigators themselves or these statisticians may have strong preconceptions that
may affect their interpretation of the data, but hopefully this will not happen
They may unintentionally introduce their biases into their data analyses and into their interpretation of the study
findings.
WHEN SHOULD WE CONDUCT A COHORT STUDY OR WHEN IS A COHORT STUDY WARRANTED?
So what would we expect if there is an association between the exposure and the disease or the exposure and
outcome? So if exposure is associated with the disease we would expect:
Figure 2. Expected findings if the exposure is associated with disease. Let us take a look at the size of the
rectangles in the outcome or in the disease. We will expect a greater number of those who are exposed to
develop the disease as compared to those who are not exposed. The comparison is between those who are
exposed and those who are not exposed. (Source: Celentano, D. D., Szklo, M., & Gordis, L. (2019). Gordis
epidemiology (6th edition.). Philadelphia, PA: Elsevier. Page 187.)
A cohort study is indicated when evidence suggests an association
● Studies indicate that when there is already evidence that suggests an association, you will not do a
cohort study as an initial study to investigate an association
○ Disadvantages of cohort studies: expensive and take a long time to conduct
● Cohort studies are usually done when there is already evidence of an association from clinical
observations, previous case control studies, or other studies
● You will not select a cohort study as the first study to conduct when investigating and Association
● To carry out a cohort study, we must have some idea of which exposures are suspected as possible causes
of a disease and are therefore worth investigating. Consequently, a cohort study is indicated when good
evidence suggests an association of a disease with a certain exposure or exposures
Cohort studies are easier to conduct when the interval between the exposure and the disease is short
● Because if the follow-up period is very low then there is a greater chance for attrition
● If the interval is short, this will minimize losses to follow up or attrition (dropping out, which we do not
want in a cohort study)
Figure 3. Design of a cohort study beginning with exposed and nonexposed groups. (Source: Celentano, D. D.,
Szklo, M., & Gordis, L. (2019). Gordis epidemiology (6th edition.). Philadelphia, PA: Elsevier. Page 181.)
● Example: Rubella infection in pregnancy and the development of congenital malformations
○ Your interval will be or your follow-up period will be at most 40-42 weeks
○ For instance, you want to determine the association between rubella infection in pregnancy and
the development of congenital malformations. Exposure: rubella infection. Outcome: congenital
malformations
● Because cohort studies often involve follow-up of populations over a long period, the cohort approach is
particularly attractive when we can minimize attrition (losses to follow-up) of the study population.
Consequently, such studies are generally easier to conduct when the interval between the exposure and
the development of disease is short.
SEVERAL CONSIDERATIONS CAN MAKE THE COHORT DESIGN IMPRACTICAL
Figure 4. Overview of the considerations in cohort study designs.
LACK OF A STRONG EVIDENCE OF A RISK FACTOR
● You should already have evidence of an association before you go into a cohort study because for all you
know, there is no association at all and many things will go to waste
INABILITY TO IDENTIFY A COHORT OF EXPOSED AND UNEXPOSED PERSONS
● exposed and unexposed persons
○ are needed in a cohort study
○ that is where a cohort study starts or that's where it begins
LACK OF APPROPRIATE PAST RECORDS OR SOURCES OF DATA
● In what type of cohort study is this a problem prospective or retrospective? So this is a problem for
retrospective or historical cohort studies
● Past records or sources: how you should establish that exposure
DISEASE OCCURS AT VERY LOW RATES OR IF THE DISEASE OR OUTCOME IS RARE
● Case-control study: best research design to use or the most appropriate research design to use for rare
outcomes or rare diseases
VERY LARGE COHORTS MUST BE ENROLLED TO ENSURE THAT ENOUGH CASES DEVELOP
● For cohort studies, you will be needing a very large sample size and when we say a very large sample size
you should have hundreds or maybe thousands of individuals participating in the study
● Large sample sizes are needed
CHECKPOINT
POTENTIAL BIASES IN COHORT STUDIES
A. Bias in the assessment of the outcome
B. Information bias
C. Biases from non-response and lost to follow-up
D. Analytic bias
1. Information obtained from past records
2. Number of participants that dropout is considerable
3. Experts unintentionally introduce their biases into their data analyses
4. Participants get tired of reporting to you every month
5. A need for written and strict criteria
6. Occurs historical cohort studies
7. Strong preconceptions of statisticians
8. (T/F) If exposure is associated with the disease, we would expect a greater number of those who are exposed
to develop the disease as compared to those who are not exposed
9. (T/F) Case-cohort study is done as an initial study to investigate an association
10. (T/F) Case-control studies are usually done when there is already evidence of an association from clinical
observations, previous case control studies, or other studies
11. (T/F) The cohort approach does not involve follow-up of populations and is particularly attractive when we can
minimize attrition (losses to follow-up) of the study population
12. (T/F) Case-cohort study is the best research design to use or the most appropriate research design to use for
rare outcomes or rare diseases
ANSWERS
1. B
2. C
3. D
4. C
5. A
6. B
7. D
8. T
9. F- Case-cohort study should not be done
10. F- Case-cohort studies
11. F-involves follow-up
12. F-Case-control study
