Observational Studies Flashcards
When might an observational study be indicated?
Hint - a scenario where it may not be possible/ethical to randomise patients to a particular exposure
Many examples. They include exposure to particular harms e.g. smoking and its relationship to lung cancer. Patients cannot be randomised to smoke.
What is the difference between a descriptive and analytical observational study?
Descriptive studies make no comparison between an observed exposure and a comparison group in relation to its effect on the measured outcome. In simple terms. How, what and where but not why.
Analytical studies make a comparison with the observed exposure and a comparison group to assess their effects on the outcome measured. How, what, where and why
List three advantages of cross-sectional studies
Cheap
Quick
Simple
What are cross-sectional studies?
Studies which collect information from individuals in order to estimate things such as prevalence (not incidence) at a designated point in time
What is the defining criteria of a confounding factor?
It must be associated with the exposure and outcome of interest.
It must not be on the causal pathway
It is of greater relevance in observational studies than in randomised trials
Outline the utility of cross-sectional studies
Provide important descriptive information on the distribution and burden of exposures/outcomes in a population.
Used as a first step in assessing possible exposure-outcome relationship.
Generally used for hypothesis generation
What is the remit of a case-control study?
Investigation of a group of people with an outcome of interest with retrospective assessment of the cases and controls on their past exposure to one or more factors of interest.
Where might the source data come from in a case-control study?
Interviews, medical records, analysis of previously stored biological samples.
What is the rationale behind case-control studies?
If the cases report a (statistically significantly) greater level of exposure than the controls then this exposure might be a risk factor for the disease
How are controls selected for case-control studies?
Selected from a group of individuals who would have been eligible for the case group if they had developed the disease during the study period.
The selection of controls should also be unrelated to the exposure(s) under investigation. i.e. representative of the greater population.
What are cohort studies?
Definition of a healthy people by their exposure status. These participants are followed up over time to see which one develops a disease or condition.
More closely resembles an RCT however there is no randomisation.
What effect does loss to follow-up have on observational studies?
Introduces bias (more significantly than RCTs as there is no mitigating randomisation in the first place)
What is an ecological study?
Examines the association between exposures and outcomes by using grouped data. The average (or other summary statistic) exposure is plotted against the rate of outcome for the population. This is repeated for several populations.
What are the limitations of an ecological study?
It is not possible to draw conclusions about exposure on the individual level. Only on a population scale.
What are the benefits of using registry databases for observational studies?
- Increase external validity (more applicable to wider population)
- Not being restricted by limited number of patients and being able to study rare endpoints rather than surrogate markers
- Offering insight into rates of adverse outcomes associated w intervention in usual practice
What is the main limitation of using registry databases?
They cannot control for selection bias. Interventions are based on clinical judgment not a process of randomisation.
What critical questions must be asked if an observational trial indicates an association between an exposure and an outcome?
- Could the association be due to bias?
- Could the observed association be due to confounding?
- Random chance?
- Is there a causal relationship or not?
What is selection bias as it pertains to observational studies?
Bias which arises when the study population is not a random selection from the target population for which conclusions are made
What is information bias as it pertains to observational studies?
This occurs when measurements or classifications of exposures/disease are inaccurate i.e they do not reflect the true values.
E.g. blood samples sitting for a long time before analysis affecting the measured variable
How is the role of random chance assessed in observational trials?
Significance testing and confidence intervals
What are the Bradford-Hall criteria for causality?
Temporal Plausible Consistency Strength Dose-response relationship Specificity Reversibility Coherence Analogy
What are the strengths of cross-sectional studies?
Cheap
Quick
Simple
What are the limitations of cross-sectional studies?
Cannot assess causality
What are the benefits of case-control studies?
Simple, cheap and quick
Can assess multiple exposures
Good for rare outcomes
Unaffected by loss to follow-up
What are the limitations of case-control studies?
Subject to many biases including selection and information
Cannot establish causality
Not suitable for rare exposures
Cannot be used to measure prevalence or incidence of disease (only prevalence of the exposure among those with/without the outcome)
What are the benefits of cohort studies?
Simple (quick and cheap)
Can measure multiple outcomes
Suitable for rare exposures
Can be used to measure incidence of disease in exposed and unexposed groups
What are the limitations of cohort studies?
Can be expensive and time consuming for prospective studies
Exposure status and diagnostic criteria may change
Ascertainment of outcome might be influenced by the exposure
Loss to follow up introduces serious bias
Less suitable for rare outcomes
What are the benefits of an ecological study?
Quick , efficient, cheap, easy to interpret
What are the limitations of ecological studies?
Cannot assess causality
Conclusions are made at the population level (rather than an individual one)
Cannot control confounders
