Principle of cohort studies Flashcards
What is a cohort study?
- longitudinal, prospective study
following groups with different exposures forward over time, providing disease incidence
What is the design of a cohort study?
exposed subject vs unexposed subjects
do they develop disease or not?
What is the definition of a cohort study?
Hennekens et al, 1987
A prospective, cohort study is one in which a
group of people with different exposures is
followed over time to see if they acquire a
disease/outcome
What are the strengths of a cohort study?
- strong research design (exposure measured before disease)
- provide info on several parameters (RR, incidence and attributable risk)
- several expos and several disease outcomes in one study
BUT not quick or cheap option
What are the classic cohort studies in CHD?
Global:
- Framingham study (5000 people)
- Seven countries study (12000 people)
UK based:
- Whitehall (12000)
- British Regional heart study (8000)
- Scottish heart and health study (10000)
What are the key pitfalls in cohort study design?
- hypothesis, confounding factors
- size and statistical power
- selection bias
- follow up methods
- approach to analysis
What is a confounding factor?
factor associated both with the exposure and with the outcome of interest
What are the important considerations for study size and statistical power?
- strength of association (RR)
- is expo common?
- incidence rate in unexposed group
- what p value will be statistically significant
- what are the chances of detecting an association if present (study size)?
What is important to consider when selecting a cohort study population?
- range of exposures low to high
- general population would be helpful: representative
2 options:
- geographically define
- well-define groups *(e.g. occupation)
How can cohort studies inform on exposures that are uncommon in general population?
e.g. occupational hazards - asbestos
highly exposed occupational group vs low expo comparison group" (+ve/-ve control) (internal comparisons in same factory w/o expo and external group of similar exposure)
How can you define exposure as accurately as possible in a cohort study?
- questionnaire, examination, blood, medical records
- make OBJECTIVE measurement
- repeat assessment if possible: more accurate associations
Where does the ‘outcome data’ for cohort studies come from?
N.B multiple outcomes may be measures
- routine surveillance/registers
- medical records
- questionnaires
- physical examination
How is risk presented from a cohort study?
usually expressed as number of outcomes per 1000 recruited in a defined period (expo vs. non-expo)
relative risk or attributable risk
What is relative risk (RR)?
RATIO
measured as risk in exposed/ risk in unexposed
What is attributable risk (AR)?
EXCESS risk (difference)
risk in exposed MINUS risk in unexposed
What are the ways that cohort data may be analysed?
- logistic regression
- Cox proportional hazards modelling
What is logistic regression?
takes account of WHETHER an event has occurred during follow up ( 0 or 1)
What is Cox proportional hazards modelling?
takes account of WHETHER and WHEN and event has occurred
Can account for graded expos and for confounding
MORE PRECISE
What criteria are used to assess whether an association is causal?
Bradford-Hill criteria
What are the Bradford Hill criteria?
- Strong
- Independent
- Dose response
- Temporal sequence (exposure precedes outcome)
- Consistent
- Specific
- Reversible
- Biological plausibility (biological mechanisms)
- Analogy (similar to other studies)
What factor lead to high total and LDL-cholesterol levels in the population?
- high dietary saturated fat intake
- high dietary sat fat: polyunsaturated fat ratio
- obesity
- genetics
What is familial hypercholesterolaemia?
genetic predisposition to elevated total and LDL cholesterol
autosomal dominant
1:500 hets
What are the determinants of high HDL cholesterol levels?
- high levels of exercise
- higher EtOH intake
- lower adiposity
- non-smoker
(potentially protective factors)
What are the important causes of CHD identified in cohort studies?
- high blood cholesterol
- high BP
- smoking
- DM
- adiposity
What makes the identified causes of CHD ‘casual’?
- strong evidence
- RR associated with exposure are high
- risks are widespread in population
What happens when there are multiple causes of CHD co-existing?
Risks MULTIPLY together
What are the other contributing factors that INCREASE risk of CHD?
- DM (>2x elevated risk, F>M)
- obesity
What are the other contributing factors that REDUCE risk of
- physical activity
- EtOH (light to moderate)
- high HDL-chol.
What are the other potential causes of CHD?
- poor foetal growth and nutrition
- factor related to inflammation
- micronutrient status (low vit D or C or low folate)
BUT the evidence is inconsistent
What is a disease cause?
a factor which, of itself, increases
the risk of a disease occurring
What is a risk factor?
any characteristic which
IDENTIFIES a group at increased (decreased)
risk of disease now or in the future
What are the non-modifiable causes of CHD?
Age
Sex
Ethnicity
FHx
What are the modifiable causes of CHD?
- BP
- blood cholesterol
- smoking
- obesity
- T2DM
What are the main potential biases in cohort studies?
- SELECTION BIAS: marked and selective loss to follow up
- INFO BIAS
misclassification of outcome status (disease or not)
usually influences by knowing their expo status
What can be done to ensure that bias is limited in cohort studies?
- make follow-up as complete as possible
- standardised reference for assessment of outcome
- blinding to exposure status
What is the validity of case-control vs cohort study?
- cohort is a stronger study design: exposure (potential cause) is studies prior to disease onset)
- bias can occur in either study type but is much more likely in case-control
What can be added to cohort studies to strengthen the design?
- v. large cohorts
- pooling of data (statistical power and precision)
- nested case-control within cohort study
- retrospective cohort studies
What is advantage of using ver large cohorts in cohort studies?
good for studying less strong associations
good for studying interplay of multiple factors
What is the UK biobank prospective cohort?
500, 000 UK men and women
aged 40-69
extensive baseline questions, assessment and stored samples
What is a nested case-control study design?
= prospective case-control
4 groups:
exposed vs non-exposed
disease vs no disease
What is a historical cohort study?
exposures have already occurred and been assessed
cohort established whilst follow up is occurring or has occurred
What are the main advantages of cohort studies?
- exposure measured before disease onset (reduced bias)
- temporal sequence of events
- multiple outcomes/disease can be studied from one exposure
- incidence can be measures (expos vs. non-expo)
What are the main pitfalls of cohort studies?
- slow, expensive and complex
- need large number for long time
- data collection may change participant behaviour
- expo knowledge may impact outcome ascertainment
- selection bias: losses in follow up