Lecture 2. Types of Study Design Flashcards
What was the research into scurvy in sailors a good example of?
One of the earliest and best recorded experimental studies
How are epidemiological studies split?
Between experimental and observational study designs
What is a type of an experimental study design?
Randomised control trial (RCT)
Give four types of analytical observational study designs?
Cohort (longitudanal)
Case-control
Cross-sectional
Ecological (population-based)
What is an ecological/population-based study and why isn’t it used as often?
A study design where the unit of inference is a group
Plotting proportions based on aggregation
Subject to the most bias
What would be an example of an experimental study?
Clinical trail for a new antibiotic or impact of putting fluoride in the water
What are the requirements for a study to be experimental?
Observational units are randomly assigned to one of two levels of an exposure
All other exposures are identical between the two groups (controlled)
Subjects are as similar as possible or stratified
They maximise a researcher’s ability to observe one effect independent of other variables
What would be an example of an observational study?
Natural exposures: compare disease in smokers vs. nonsmokers?
Historical controls: Compare obesity rates over time?
What makes observational studies different from experimental studies?
Data are collected without randomisation or artificial manipulation of predictor variables
More difficult to identify causes because it is hard to tease out individual variable effects
There are fewer ethical constraints on the research
How are observational studies split?
Between being descriptive or analytical
What are examples of descriptive observational studies (estimation)?
Proportion of women with breast cancer (prevalence)
The rate of cases of measles in one year (incidence)
Case studies describing a new disease or treatment
What do descriptive observational studies describe but not identify?
Describe what is occurring and typically quantify a variable
Do not identify causes
Used to estimate disease outbreaks, but don’s say why they occur
What are analytical or testing observational studies?
Studies where statistical comparisons are made to identify associations
Can allow causal inference
What are examples of analytical studies?
Comparing the symptoms and progression of Parkinson’s in patients taking a new vs old drug
Comparing pesticide concentrations (chlordecone, an endocrine disruptor) in individuals with and without prostate cancer
What is required for all analytical studies?
Identification of variables that are associated with a disease by comparing groups
Interpretation: Is the association ‘real’ or due to confounding?, Is the association causal? Risk factors are real associations that are not necessarily casuals
Consideration of the risk of bias and confounding factors throughout the investigation
What are the key features of a randomised control trial?
Explicit randomisation of treatment or exposure
Defined target population
Following patients forward in time to see if they improve or not
What is an example of an experimental study?
How do intestinal nematodes (and their treatment) influence susceptibility to TB, mortality from TB and its spread within African buffalo?
When are experimental studies carried out?
After we are fairly sure what is causing the differences that we have found in comparative, analytical studies
What does experimental design allow us to conclude?
Differences in the dependent variable are due to differences in the treatment
What happens in experiments studies that do not allow for random allocation of treatments?
Remains subject to bias
What are cohort studies?
Follow groups of individuals through time until the exposure and outcome happens in a defined proportion of the sample
How are cohort studies different from randomised control trials?
Cohort studies have a non-random assignment of exposures due to ethical reasons
What is an example of a cohort study?
Framingham Study (1948)
Study population: Age 30-62 in Framingham
Hyp: Incidence of Coronary Heart Disease (CHD) increases with age and is higher in men
Hyp: High cholesterol, weight, tobacco use is associated with increased risk of CHD
What are the properties of a cohort study?
Good at identifying causes because the change occurs after the observations (time sequence)
Data on exposures are gathered as they happen (no recall bias)
Expensive and time consuming - “loss to follow-up” (people might say they don’t want to be in the study anymore)
Difficult to analyse (repeated observations - info is not independent, it is reliant on using the same people every time)
What are case control studies?
Take individual with disease (cases) and compare them with individuals from the same population without the disease (controls)
Calculate proportion of cases that were exposed & proportion of controls that were exposed
What is an example of a case control study?
Breast cancer in women
Select cases - women with breast cancer
Select controls - women from the same referral unit with another non-linked disease, the same GP practice list, the same town
Compare the diet, lifestyle, life experiences of the two groups
Match on age, race, social group
What are the properties of case control studies?
Looks backwards (vs. cohort studies look forwards)
Begins with identifying cases (vs. cohort study identifies exposure groups)
More difficult to identify causes compared to cohort study but better at identifying causes than cross-sectional studies because you can still assess the temporal sequence of risk vs. disease
Are cheap and quick to do
They rely heavily on recall or past records, which may be inaccurate or bias
Rely on selection of cases
What are cross sectional studies?
Predictor and response variables are measured at one point in time
What are examples of cross sectional studies?
Questionnaire designed to evaluate the factors associated with alcoholism that also asks about lifestyle and feelings
Randomly sampled African buffalo populations designed to evaluate if co-infection influenced nematode burdens
What is the big problem with cross-sectional studies?
Which came first? They demonstrate association but cannot attribute cause
Did alcoholism lead to lifestyle changes of vice versa?
Did the coccidia influence susceptibility to nematodes or vice versa?
Did a confounding factor influence them both?
What are the properties of cross-sectional studies
Poor at identifying causes (vs. case-control and cohort studies that consider timing)
Good at generating hypotheses (good to do when there is little known about a subject)
Quick and relatively cheap
Consequently, the literature and media abound with ad hoc cross-sectional surveys
