Study Designs for Clinical and Epidemiological Research
Carla J. Alvarado, MS, CIC
University of Wisconsin-Madison (608) 263-2177 carla.alvarado@mail.admin.wisc.edu
An Overview of Study Designs
purpose of most medical research is to find causes of disease and ways to treat or prevent them Some studies are as simple as the astute observation of a new disease or problem - the case report Others are complex and expensive like the clinical trial
The
Research steps to prepare for selecting an appropriate study design
the problem State the research question(s) Review the literature Select a study type or design Develop the protocol
Define
Research Steps (continued)
test your study protocol Collect data Analyze data Write the report and disseminate findings
Pilot
Stating the Research Question
specific terms “specific aims” of the trip
what who where when
Clear,
An Overview of Study Designs
study design has its own strengths and limits Each study design type has associated problems that can distort your conclusions and lead you astray!
Each
Experimental Studies
by the fact members of the study population are assigned to either a treatment or control group Treated and untreated groups are followed prospectively to see whether the two groups subsequently differ Design most analogous to laboratory experiments
Distinguished
Experimental Studies
enough money to permit every interesting question to be answered In the case of an adverse outcome, it is unethical to assign patients to the “control” group Rare events require thousands of persons to be exposed Even experiments are subject to bias
Not
Randomized Clinical Trial
best design to look at cause and effect Subjects without the outcome are:
to different exposures receive different interventions followed over time to determine occurrence of the outcome
assigned
The
Randomized Clinical Trial - 2
Advantages
accurate
Disadvantages
cost,
information on exposure accurate information on outcome confounding less a problem gives the strongest evidence of causality
time and complexity no one wants to be control people want new treatments people (including medical staff) don’t want selection by chance ethical issues
Observational Studies
in which the treatment or exposure of interest is not assigned, but occurs by choice or happenstance Depend on people’s natural or voluntary exposures to factors not ordinarily randomly assigned Always the possibility that the exposed and unexposed differ in important ways other than “exposure”
Study
Observational Study: The Case Report
small number of persons with an unusual disease or change in disease possibly related to single cause The first statement of clinical hypothesis While case reports may lead to important findings, they cannot be a “finding” themselves because the observation may be due to chance
Described
Observational Study: The Descriptive Study
data that are routinely collected for various purposes to study the occurrence of disease and its possible causes among groups of groups Most appropriate for preliminary exploratory studies Inferences drawn from studying groups are not necessarily true for individuals
Uses
Observational Study: The CrossSectional Study
cross-sectional (prevalence) study looks for an association between an outcome and possible causes by studying a group at one point in time Can demonstrate an association between a factor and an outcome, but cannot say which came first Can provide preliminary exploration Relatively cheap and easy
A
Observational Study: The CrossSectional Study - 2
Advantages
relatively
Disadvantages
cannot
inexpensive relatively more simple can represent a target population (generalizability) good starting point for research
determine the cause (chicken v. egg) recall bias
Observational Study: The Case Control Study
with cases who have the disease of interest and a comparison group (controls) without the disease Previous exposures of both groups are investigated If previous exposure is more common among the cases, this is evidence that exposure caused the disease
Begins
Observational Study: The Case Control Study – 2
from other observational designs in that investigation begin with “ill” and works back to document possible causes of their illness This approach is a refinement of the case report – adds group without the illness Problems are by what criteria the comparison group is chosen
Differs
Observational Study: The Case Control Study – 3
Advantages
relatively
Disadvantages
selection
quick and inexpensive requires fewer subjects good for diseases with long latency periods good for rare outcomes can look for more than one exposure
bias “over matching” “under matching” recall bias
Observational Study: The Cohort Study
with one group exposed to a factor of interest and another group unexposed Groups are then observed at a later time to see if they develop differences that might be attributable to exposure Most often prospective: however, can be retrospective in use of past data
Starts
Observational Study: The Cohort Study
Advantages
selection
Disadvantages
long
bias generally not a problem recall bias not a problem can look at more than one outcome lends strength to causality because of temporal sequence
follow-up may require many subjects expensive exposure may change over time
Potential problems within a study design
The
“Confounder”
What
looks like a casual relationship may be due to another factor not taken into consideration Confounding factor is associated with bother hazard and the disease in a way the hazard and disease look related The observed association between two conditions is actually secondary to the influence of a third factor
Potential problems within a study design-2
The
“Matchmaker”
Under-matching
is the failure to select cases and controls sufficiently alike in important characteristics; can demonstrate a spurious association between a disease and a supposed cause Over-matching is the error of selecting controls that excessively resemble cases; can fail to discover an association that is present and real
Potential problems within a study design-3
The
Cohort effect
Tendency
for persons born in certain years to carry a relatively higher (or lower) risk of a given disease throughout their lives Age distribution of the disease under study may be spuriously affected by differences in past experiences among various birth cohorts in the study population