2 - Study Design Flashcards
(OBJ) Describe the major types of research designs and when they are used.
- Descriptive study: no explicit comparisons; a mean, median, or probability is the goal
- -Case reports/series
- -Rates in population - Analytic study: explicit comparisons; used to measure an association
- -Observational
- -Intervention
- -use relative risks to assess the impact of some exposure on the risk of an outcome or correlations and regression to assess the relationship between two measurements
(OBJ) Describe the differences between observational and intervention studies.
- -Observational: watching people, collecting information (cohort studies, etc)
- -Intervention: randomized trial; doing something actively and watching results
(OBJ) Describe the differences between cross-sectional and longitudinal studies and discuss when each might be used.
Cross-sectional study:
- -Single population
- -No follow up
- -Measurement of PREVALENT traits and diseases (not incidence)
- -Performed at ONE point in time
- -Common for studies of biology, diagnosis
- -Direction of causality often unclear
Longitudinal studies:
- -Follow over time
- -Measure incidence (and prevalence)
- -Require much more time and effort
(OBJ) Contrast the advantages and disadvantages of case-control studies.
Identify people with/without disease and ask about previous exposures (have disease = case, no disease = control)
Advantages:
- -Relatively quick, easy, and inexpensive
- -Good for very rare diseases
- -Allows you to compute ODDS RATIO
Disadvantages:
- -Reversal of causality
- -Non-response
- -No single measure of association (**cannot calculate risk!!)
- -Selection bias: can be tough to select appropriate controls
- -Response bias: measurement issues (have outcome before measuring exposures, so may bias recall), timing issues (cholesterol and BP after an MI would not be the same as before)
(OBJ) Describe methods used to select appropriate controls in research studies.
Should represent exposure in source population for cases
Difficult to define in a hospital-based study (who WOULD have gone to DHMC vs Boston or Fletcher Allen if they got ALS?)
Easy to define in a population-based study
(OBJ) Describe the features of cohort studies (retrospective and prospective); contrast their advantages and disadvantages.
Observe disease risk in exposed and unexposed
Prospective: following people in real time
Retrospective: all in the past (look at medical records)
Advantages:
- -Can compute absolute risk/rate, risk ratio/rate ratio, and risk differences
- -Natural logic: exposures, then outcomes
- -Possible to study rare exposures
- -Relatively free of some biases of case-control studies
Disadvantages:
- -Work/time, especially for prospective (lots of subjects, lots of follow up)
- -Rare outcomes need even more follow up and larger groups
- -Drop-outs bad (want complete follow-up to observe risks)
- -Outcome ascertainment - want unbiased follow up
- -Outdated exposures with long follow up - exposures change with time (people stop smoking)
(OBJ) Contrast the advantages and disadvantages of randomized controlled trials.
Advantages:
- -No confounding
- -No exposure measurement error
- -No response biases
- -Organized follow up for both treatment groups
Disadvantages:
- -Patients must be eligible and willing for ALL study treatments & may not represent general population
- -Treatments are uniform and rigid, not like real life
- -Co-interventions
- -Drop outs
- -Difficult and expensive
(OBJ) Describe the roles of randomization and blinding in clinical trials, including concepts such as subject selection, exposure allocation, stratification, self-selection, systematic assignment. (this is incomplete)
Subject selection:
Exposure allocation: FIXED, isolates treatment effect and limits confounding
Stratification:
Self-selection:
Systematic assignment:
(OBJ) Introduce the odds ratio as a measure of association.
Cross product: AD/BC
Provides a good estimate of relative risk, but is NOT a risk ratio
(OBJ) Discuss intention-to-treat and as-treated analyses, and their impact on measures of effect.
Intention-to-treat analysis: includes everyone assigned to treatment group, regardless of drop outs or cross overs
–Must be your primary analysis, but will be more conservative
As-treated analysis: keep people in groups of what treatment they actually received (the people who completed the trial on the protocol)
(OBJ) Discuss in detail the types of bias affecting the various study designs, and methods used to minimize bias. (descriptive, case-control, cohort, clinical trial, all)
DESCRIPTIVE:
- -No comparisons; describe only what is possible, not what is better or worse.
- -Concern is the precision (or lack of it) and bias in measurements.
CASE-CONTROL:
- -Selection bias (do cases and controls correspond?).
- -Information bias (did the presence/absence of the disease affect the data obtained?).
COHORT Follow-up: (were exposed and unexposed groups followed and endpoints determined with equal intensity, and without bias?).
CLINICAL TRIAL:
- -What kind of controls (randomized is best)?
- -Were all groups followed-up in the same (unbiased) way?
- -Was the randomization concealed from investigators and subjects? (Blinding helps assure this.)
ANY STUDY:
- -Adequate sample size to detect a meaningful effect?
- -Is confounding present?
- -Did people with the exposure of interest have the same chance to be detected to have the outcome as people without?
ANY VALID STUDY:
–Generalizability
(OBJ) Explain the terms internal and external validity.
Internal validity: did study reach correct conclusions about the study population?
External validity: can results be applied beyond the study example (e.g. to the general population)?
Define bias.
A feature of a study tending to distort its findings away from the true result for the population being studied.
Define blinding.
The process of keeping clinical trial subjects and/or study personnel unaware of the treatment assignments
Also used to describe the process of keeping study participants or personnel unaware of the hypotheses under study.
Define case-control study.
A research design in which subjects with a disease or outcome (cases) are compared to subjects without the disease/outcome (controls). (Improperly called a retrospective study.)
(OBJ) Define case report. List its advantages and limitations.
An informal study in which the experience of a group of patients is described, without a comparison group.
AKA case series
–Describes what is possible, NOT necessarily what is likely
–Includes an IMPLIED inference, if any (“this is interesting”)
Advantages: trends in the data can be used to develop hypotheses that can be tested in larger, controlled studies
Limitations: lack of a comparison group makes it impossible to make statistical comparisons of risk
Define clinic-based.
A study in which all subjects are derived from patients seen at a medical facility.
A hospital-based study is an example
Typically used with regard to a case-control study
Define clinical trial.
A formal intervention study, in which the investigators determine the treatments or exposures of the subjects (usually by randomization)
Define cohort study.
A research design in which a defined group of subjects is followed over time for the occurrence of events
AKA follow-up study
Define cross-sectional study.
A research design in which subjects are evaluated at only one timepoint
Define double-blind.
A situation in which both subjects and investigators are unaware of the subjects’ treatments or exposures
Define ecological study. What problem exists with ecological studies?
An investigation in which the unit of analysis is a group; good for generating hypotheses
–Ex: correlation of trends over time in disease rates with trends in cigarette sales, or association of disease rates in different countries with cigarette sales
–PROBLEM: ecologic fallacy - inferences from group level may not apply to individual level (individuals who eat meat may not be the ones getting breast cancer)
Define efficacy analysis.
A statistical analysis of data from a randomized clinical trial in which the outcome is evaluated according to the treatment actually received
Loses some of the advantages of randomization
AKA as treated analysis
Define external validity.
The ability of a study to apply to populations other than the one actually studied
AKA generalizability
Define historical controls.
A control group comprised of subjects whose past experience is contrasted with those of currently observed subjects
Historical controls are generally considered inferior to concurrent controls.
Define information bias.
A distortion caused by information provided by one group of subjects being systematically different than that provided by another group
Usually associated with case-control studies, in which case patients may provide systematically different information than controls
AKA response bias
Define intention-to-treat analysis.
A statistical analysis of data from a randomized clinical trial in which outcome is evaluated according to the randomized treatment group, regardless of what treatment was actually received
Makes full use of the advantages of randomization
Define internal validity.
The ability of a study to reach a correct conclusion about the population actually studied
Define intervention study.
A study in which the investigators determine the exposures or treatments for subjects
Define observational study.
A study in which the investigators do not determine the exposures or treatments subjects experience
Define odds ratio.
A type of relative risk; the only type of relative risk possible in case-control studies
Calculated with a cross-product ratio from a 2 × 2 table
Define placebo.
A sham drug used in blinded studies to keep subjects and/or investigators unaware of subjects’ treatment assignments
Define prospective study.
An investigation in which the events studied occur during the time the research is conducted
Define randomization.
A way of assigning treatments so that each subject has the same probability of receiving a particular treatment. Randomization makes treatment “independent” of the characteristics of the subjects, and thus effectively “prevents” confounding
Define retrospective study.
An investigation in which all of the events studied have already occurred at the time the research begins; existing records provide the data
Define retrospective cohort study.
A study in which the exposure and outcome have already occurred: exposed and unexposed subjects are “followed” through time afterwards e.g. through review of medical records
Define selection bias.
A bias that arises because of the selection of subjects in a study
Define single blind.
A situation in a clinical trial in which either the patient or the investigator group is unaware of treatment assignment, but not both
Best used to describe the situation in which the investigators evaluating endpoints are unaware of treatment assignment
An odds ratio is the only valid estimator of relative risk in…
a. A case control study
b. A cohort study
c. Neither a nor b
d. Both a and b
e. RCT
A
The absolute risks of the outcome (in the exposed & unexposed) can be calculated in…
a. A case control study
b. A cohort study
c. Neither a nor b
d. Both a and b
B (also in RCT)
Confounding is a potential issue in…
a. A case control study
b. A cohort study
c. Neither a nor b
d. Both a and b
e. RCT
D
To study factors which affect the incidence of diseases, we could use…
a. A case control study
b. A cohort study
c. Neither a nor b
d. Both a and b
D
Subjects who have the disease under study may provide distorted risk factor information in…
a. A case control study
b. A cohort study
c. Neither a nor b
d. Both a and b
A
Data collected entirely in the past could be used for…
a. A case control study
b. A cohort study
c. Neither a nor b
d. Both a and b
e. RCT
D
Bias is virtually never a potential issue in…
a. A case control study
b. A cohort study
c. Neither a nor b
d. Both a and b
D
The causes of rare diseases are best studied in…
a. A case control study
b. A cohort study
c. Neither a nor b
d. Both a and b
A