Glossary for Study
Design Taxonomy
‘Analytic’ study
Unhelpful
descriptor, since many studies involve analysis and could be described as
analytic.
Basic Research
The primary outcomes are not measured in living human
subjects. Includes animal studies; cell, biochemical, and genetic investigations;
and studies on the properties of drugs and materials.
Before-After study
The same outcomes were measured before and after the
intervention/exposure in one group (uncontrolled). Before-After studies are
quasi-experimental. It is misleading and incorrect to use the word “trial” to
describe a “quasi-experimental” study (such as ‘uncontrolled trial’ or
‘single-arm-trial’). Even if the order of assignment of interventions was
random, this was not a randomized trial. Even if the subject(s) receive blinded
placebo or intervention, this was not a randomized trial. Sometimes,
Before-After study is called a Pre-Post study.
Interrupted Time Series: A Before-After study where 3 or more
measures were taken before and 3 or more measures were taken after (interrupted
by) the intervention/exposure. Risk of bias includes regression to the mean.
N-of-1 study: A Before-After study in only 1 subject.
Case Control study
Subjects were selected based upon the presence (case) or
absence (control) of known outcome(s) of interest (disease). Exposure status is
then collected based on the subjects’ past. Subject group characteristics
should be similar and controls must be representative of those individuals who
would have been selected as cases, had they developed the disease. Case Control
studies are always retrospective.
Nested Case Control study: Subjects were
selected from the population of a prospective cohort study.
Case
Report
Presents the outcome(s) of one to
few individual case(s). May or may not describe an intervention/exposure.
Case
Series
Descriptive report of an aggregation of
individual cases. The cases were defined by similar diagnoses/disease
state or having undergone the same intervention/procedure. Case series are non-comparative
(uncontrolled) and therefore cannot test, report, or conclude about treatment
safety or effectiveness. Calculation of an absolute risk or a rate for the
outcome is not possible in a case series.
Ideally, case series report on consecutive cases from a
well-described study population with a well-defined intervention using
validated outcome measures. Case series are a commonly reported study design,
but the label "case series" is rarely used, and sometimes
incorrectly.
Prospective Case Series are unusual, and require that the
investigator designed the study, obtained ethics board approval, and patient
consent to participate in a clinical research study, prior to commencing any of
the interventions. Prospective case series often have a small sample size and
are described as exploratory or pilot. Prospective case series resemble before-and-after
studies or time series (uncontrolled). If the investigator compared the same
primary outcomes before-and-after the intervention, then the study would
qualify as a Before-After study.
Retrospective Case Series are the most common type. The
investigator may have used clinical records, registries, or databases.
Case series are valuable for
studying symptoms and signs, creating case definitions, hypothesis generation,
or detection of rare adverse events.
Clinical Research
The primary outcomes were measured in living human subjects.
Clinical research includes epidemiologic research.
Cluster Randomized Controlled Trial (see Randomized Controlled Trial)
Cohort
A group of individuals (or other
organizational units) who have one or more common features/traits at the
time that they assembled (examples: birth year, place of employment, medical
condition, place or time period of medical treatment, military unit). This term
should not be used in the report of a study that was not a Cohort Study.
Cohort Study
Prospective Cohort Study
A research study that pre-specified
2 or more groups of individuals (who were alike in many ways but differed by a
certain characteristic), after designing the study, and then followed them
for a period of time, and compared an outcome of interest among the groups. The
target group(s) and the control group(s) were created at the same time
(concurrent).
The groups may be well-matched,
poorly matched, or moderately well-matched. Corrections may be made, using
regression analysis statistical techniques, to “control for” or adjust for
differences in the groups that could tend to bias the differences (or lack of
difference) in outcomes.
Retrospective Cohort Study
A research study that created 2 or
more groups of individuals (who were alike in many ways but differ by a certain
characteristic), using information that was available before designing
the study (determined in the past), and then compared an outcome of interest among
the groups. The data/information may be from clinical records, registries, or
databases. The target group(s) and the control group(s) may have been created
using contemporaneous data/information (concurrent). The controls may be
historic or non-concurrent.
The groups may be well-matched,
poorly matched, or moderately well-matched. Corrections may be made, using
regression analysis statistical techniques, to “control for” or adjust for
differences in the groups that could tend to bias the differences (or lack of
difference) in outcomes.
A retrospective cohort study might
utilize data/information that had all been documented prospectively (by entry
into a medical record or database). It would be misleading and incorrect to use
the descriptive term ‘prospective’ in the Methods if the study was designed
after ‘the event’.
Consent
There are three
different consents relevant to study design. Patient consent to treatment is
not sufficient for reporting human research. Patient specific consent for the use of
patient images is essential for a case report or other study that includes such
images in the report. Consent to be a participant in a clinical research study
is essential for any experimental study. Consent to be a participant in a
clinical research study is usually necessary for a prospective cohort study.
Clinical research
studies require an ethics approval or waiver, with exceptions for a case report
or a small case series, where specific consent has been obtained from the
patient(s).
Controlled study
There were 2 or more
groups, with different intervention/exposure, and outcomes were measured in all
groups and compared between groups.
Controlled Before-After Study
The same outcomes were measured before and after the
intervention/exposure in 2 or more groups. A type of
non-randomized trial.
Controlled Interrupted Time
Series: A Controlled
Before-After study where 3 or more measures were taken before and 3 or
more measures were taken after (interrupted by) the intervention/exposure.
Cross-sectional study
Both the exposure and the outcome status in the subjects were
assessed concurrently. Causation cannot be determined. Commonly examine the relationship between
disease and other variables of interest at one particular point in time. A
common tool is the survey.
‘Database’ study
Unhelpful
term that should not be used in Methods or to describe the study design. Likely to be
retrospective. Could be comparative or non-comparative. The database is
simply the repository of the data/information (rather than a paper clinical
record, or an electronic medical record, or a registry).
Descriptive
studies
Descriptive studies are:
cross-sectional studies, case series, case reports, trend studies, and focus
group studies. They are non-comparative and may not be used to determine
association, causation, or treatment effectiveness. They are important for
hypothesis generation for future research, and are often the first emerging
evidence.
Diagnosis/screening study designs
Diagnostic studies may be comparative or non-comparative; comparative
may refer to between tests or between groups.
The most relevant patient-centered outcome is optimal for
assessing the impact of diagnostic tests on patient outcomes; often a
previously validated surrogate was used. Alternatively, sometimes disease
modeling studies may be used to examine patient outcomes.
Diagnostic accuracy case-control
study
An observational design for
diagnostic/screening studies; the subjects were selected by known diagnosis, (or
clinical suspicion, or clinical outcome, or disease state), followed by
inclusion/exclusion criteria to select the study group suspected to have the
target condition. The control group was selected using similar
inclusion/exclusion criteria, with a critical difference to identify a group
presumed not to have the target condition. Both an index test and a reference
standard test were performed (either in series or simultaneously).
There is a level of uncertainty
regarding the diagnosis (outcome) in the target group at entry; this may be very
low (likely true/established diagnosis, most approximates case control study),
or this may be moderate (an estimate based on symptoms and clinical suspicion,
most resembles a cross-sectional study).
Diagnostic accuracy cohort study
Two or more groups were selected
from a representative population, based upon inclusion/exclusion criteria that
included test results (either dichotomous or above and below a cut-point of a
continuous test result). Each subject then underwent the experimental test and
the reference standard. Testing may have been done in series or in parallel.
There may or may not have been blinding of the reference result. Results are
restricted to diagnostic accuracy (sensitivity, specificity, likelihood ratios,
diagnostic odds ratios). Efforts to minimize confounding should be documented. Diagnostic
accuracy cohort studies are not directly tied to patient outcomes. Only when
diagnostic accuracy is coupled with effective therapy (or noxious therapy) can
outcomes be influenced.
Diagnostic accuracy
cross-sectional study
One group was selected from a
representative population, based upon inclusion/exclusion criteria of a defined
clinical presentation. The subjects may have been consecutive, or
non-consecutive. Each subject then underwent the experimental test and the
reference standard. Testing may have been done in series or in parallel. There
may or may not have been blinding of the reference result. Results are
restricted to diagnostic accuracy (sensitivity, specificity, likelihood
ratios).
Diagnostic randomized controlled
trial (RCT)
Randomized comparisons of two
diagnostic interventions (one experimental and one reference standard control) were
made with identical therapeutic interventions, based on the results of the
competing diagnostic interventions (for example, disease: yes or no), and with
the study outcomes being clinically important outcomes consequent to diagnostic
accuracy.
The study may have been blinded or
un-blinded; in a blinded study, both tests are done for both groups but only
the allocated test result is known for decision making until the study ends.
Diagnostic reliability study
A comparison is made between different
machines, laboratories, observers, raters for variability in results. A type of cross-sectional study, without a reference standard.
Human observers and raters may or may not have been blinded to each other’s determinations.
Results are reported as kappa values.
Diagnostic yield study
Non-comparative observational
report of one test performed in one group of pre-selected patients. Reports of
the number of cases identified among the study population. A
type of case series study. Usually considered to be an exploratory or
pilot study. If there is only one test performed in one subject, the study is a
case report.
Non-randomized diagnostic trial
Non-randomized comparisons of two
diagnostic interventions (one experimental and one reference standard control) were
made with identical therapeutic interventions, based on the results of the
competing diagnostic interventions (for example, disease: yes or no), and with
the study outcomes being clinically important outcomes consequent to diagnostic
accuracy.
The study may have been blinded or
un-blinded; in a blinded study, both tests are done for both groups but only
the allocated test result is known for decision making until the study ends.
Experimental study design
Investigators had
direct control over the assignment of the intervention (or exposure), and the timing and
selection of outcome measures. All experimental studies are prospective. Examples
include randomized controlled trial, nonrandomized trial, quasi-experimental
study.
Interrupted time
series (without a comparison group) - see Before-After study
Nested Case Control
study – see Case Control study
Non-comparative
observational studies
All subject(s) received the same intervention(s)/exposure,
and did not have the same primary outcome(s) measured before as after. These
are always retrospective because the design and conduct of the study commenced
after the intervention or exposure occurred. Includes: retrospective case
series, case report(s), time series, and focus group.
Non-concurrent cohort study
A retrospective study in which 2 or more groups of
individuals were identified on the basis of common features, but at
different time points. Individuals in each group were classified according to
exposure status (exposed or unexposed) at the time the groups existed or were
created. They were followed to determine if the development of the outcome of
interest was different in the exposed or unexposed groups.
Non-randomized controlled
trial
An experimental study in which individuals
or groups of individuals were assigned to the intervention or control by a
method that was not random (date of birth, date of admission, judgment of the
investigator, personal preference, etc.). Individuals or groups were
followed prospectively to assess differences in the outcome(s) of interest.
The objective of the study could be to determine
superiority, equivalence, non-inferiority, or to disprove the null hypothesis.
Selection bias is frequently an issue in nonrandomized
studies. Because the assignment was not randomized, assessment of matching
between groups is important. The groups may be well-matched, poorly matched, or
moderately well-matched. Corrections may be made, using regression analysis
statistical techniques, to “control for” or adjust for differences in the
groups that could tend to bias the differences (or lack of difference) in
outcomes.
Observational study design
A
study in which the investigator did not control/assign the exposure/
intervention status of the study participants. Some are analytic (prospective cohort) and
some are descriptive (case series). Examples are: Case Control, Cohort,
Cross-sectional, Case Series, and Case Report.
Prospective study
The subjects were enrolled in the study after the study had
been designed and approved, and the subjects were followed for some period of
time.
Prospective Case
Series – see Case Series study
Prospective Cohort
study - see Cohort study
Pseudo-primary
research
The same investigator used data from one or more previous
studies for the current study.
Quasi-experimental
study - see Before-After study
Randomized Controlled
Trial (RCT)
An experimental comparative study in which participants are
randomly assigned to intervention or control and followed over time to measure
differences in outcomes. Concealment of the allocation of the randomized
assignment is important. Known and unknown prognostic factors (confounders) should
be evenly distributed between the groups.
There are many sub-classifications of RCTs:
-The unit
of allocation and analysis may be the individual, or the group (cluster = hospitals, schools, neighborhoods, or entire
communities).
-Variations
in methodology include factorial, cross-over, parallel, stepped wedge, and
Solomon four-group.
-RCTs may be un-blinded (open
label) or single/double/triple/quadruple-blinded. Except for outcomes that do
not involve any subjective judgment, blinding of outcome assessors, patients, healthcare
providers, and investigators is important to minimize bias.
-The objective of
the study could be to determine superiority, equivalence, non-inferiority, or
to disprove the null hypothesis.
-The design could be explanatory (traditional)
or pragmatic (naturalistic).
A disadvantage of explanatory RCTs
is that study populations are often highly selected, the setting well-defined
and controlled, and results therefore sometimes represent the maximum effect
and lack generalizability (external validity). The pragmatic RCT is designed to
test interventions in the full spectrum of everyday clinical settings in order
to maximize applicability and generalizability, and usually has a large sample
size, simple design, and diverse settings.
-The RCT could be registered or not
registered. The International Committee of Medical Journal Editors (ICMJE)
acknowledged the vested interests, academic biases, and market forces
(including insurance coverage policies and reimbursement issues) that influence
the selective reporting of RCTs. The ICMJE member journals agreed to require
the registration of any RCT submitted for review
Retrospective Case
Series – see Case Series
Retrospective study
The study started after the data/information had been collected.
Retrospective Cohort
study – see Cohort study
Trend
study
A study in which data are collected at a
series of points in time on the same population to observe trends in the
outcome(s) of interest.
There is no intervention/exposure, and no comparative group.
Trial
A term restricted to research
studies with control group(s): randomized or nonrandomized controlled trials.
