Lecture 5 and 7 (Study Types) Flashcards
Cohort study definition:
- prospective, observational.
- start with no diseases, look at exposure and disease development over time.
- more expensive than case-control.
Statistic utilized for cohort study:
RELATIVE RISK
Cross-sectional study definition:
- single time point, observational.
- take a slice of the population
- how many people have a disease or a risk factor right now
Statistic for cross-sectional study:
POINT PREVALENCE
point prevalence = (total #sick)/(total population)
Case-control study definition:
- retrospective, observational.
- start with disease and go back retrospectively to find the risk factor
- cheaper than cohort studies
Statistic for case-control study:
ODDS RATIO
Nested case-control study definition:
- select cases and controls from a cohort study
- go back and find the risk factor
The three types of error in studies:
- random (chance)
- systematic (bias)
- confounders
Random error:
- chance
- will cancel out as sample gets larger
- more likely will lead to type 2 error
Systematic error (bias):
- cannot be corrected for regardless of sample size
- more likely will lead to a type 1 error
Confounder definition:
- must be associated with exposure
- must be associated with outcome
- cannot be in the causal pathway between exposure and outcome
How can you control for confounding?
- At the design stage:
- randomization
- restriction
- matching
- At the analysis stage:
- stratification
- multivariable adjustment (regression)
Selection bias:
- nonrandom assignment to groups
- loss to follow up
- must be controlled at design stage.
Healthy worker effect (bias):
- workers are generally healthier than the general population.
Self-selection bias:
- research volunteers different from the general population in terms of their exposure and disease status.
Withdrawal bias:
- differential loss to follow-up; people drop out
- drop outs may be more sick
Recall bias:
- knowledge of presence of disease changes subject’s response
Sampling bias:
- subjects are not representative of population, non-generalizable
Confounding bias:
- one factor distorts/confuses the effect of another closely related one.
Information/observer bias:
- researcher’s decision affected by prior knowledge of subject’s exposure.
- Correct for by blinding.
- must be controlled at design stage
What forms of bias are case-control studies sensitive to?
- recall bias
- selection bias
- observer bias
Case-control study benefits:
- good for rare diseases
- can test for a wide variety of exposures
- important in understanding new diseases
- commonly used in outbreak investigations
Cohort study benefits:
- can choose all the variables you want
- can control for bias and confounders
How to choose sample for case-control study:
- Case definition should not be too broad - misclassifies
- Case definition should not be too narrow - reduces sample size
- Cases and controls should be members of the same base population and have had equal opportunity to have been exposed (same environment).
Proces of the development and testing of a new clinical test (index test):
need to compare index test to reference test:
- select patients
- test all patients with index test
- test and diagnose all patients with reference test
- compare index test results with reference test results
Potential bias in an index test study:
- utilization of wrong reference test
- spectrum bias
- verification bias
- observer bias
Spectrum bias:
- applies to index test studies
- patients in study not the same as those found in standard practice
- i.e. using only patients more likely to test positive - will make test look better
Verification bias:
- applies to index test studies
- not all patients in study receive both the index and reference test
Analytical observational studies:
- Observe events as they happen in the ‘real world’ without playing an active role in what happens.
- “Incidence” studies
- Two-types:
- cohort
- case-control
What is a cohort?
- a group of people with something in common who are then followed over time to see what happens to them.
- should not have the outcome at enrollment, but should be at risk for it
Effect modifiers are:
- Effect modifiers are variables that change the effect of the exposure of interest on risk of disease.
- “interaction”
Randomization:
- an attempt to evenly distribute potential (unknown) confounders in study groups.
Restriction:
- excludes participants with known confounding variables
- may decrease sample size and power
Stratification:
- stratify data by a potential confounder, and then compare both groups
Matching:
- For each subject in the exposed group, one or more patients with the same characteristic (with or without the confounder, eg. smoker or nonsmoker) is chosen for the non-exposed group.
- has to be coupled with a matched analysis
Randomized control trial:
- prospective, interventional, gold standard
- treatment versus placebo — watch outcome
- statistics: NNT/NNH; multiple comparisons
- four phases
Meta-analysis study:
- retrospective, chart review, platinum standard
- pools data from multiple studies, usually RCTs
- analyzes cumulative data to determine external validity
Twin concordance study:
genes versus environment (early life)
- monozygotic = identical twin, same genetic material, same environment
- dyzygotic = non-identical twins, different genetic material, same environment
Adoption study:
genes versus environment (later life)
- come from genetically sick family, raised in completely different environment, but still get sick. Hints to powerful effect of genes.
RCT Phase 1:
SAFETY
- small # healthy volunteers
- toxicity
- pharmacokinetics
- in cancer, phase I studies will include critically ill/terminally ill patients
RCT Phase 2:
EFFICACY
- small # patients with disease
- dosing
- adverse reactions
RCT Phase 3:
TREATMENT VERSUS CONTROL
- large # patients with disease
- compares standard treatment to novel treatment
RCT Phase 4:
post-marketing SURVEILLANCE
- after drug has been approved
- watches for rare, long term effects
