Taxonomy of

Clinical Research
Design
THE LANCET • Vol 359 • January 5, 2002;
Grimes and Schulz
Look at this study
https://bmcpediatr.biomedcentral.com/a
rticles/10.1186/s12887-019-1594-4
Were the design claim appropriate with
the method? Why?
 Pivotal Observatio
nal

Healthy Patient Patient

Drug Clinical
Trial
Read More:
https://www.fda.gov/patients/drug-development-
process/step-3-clinical-
research#Clinical_Research_Phase_Studies Dose, Info for Treatment
Administrati Phase III Benefit
on design

https://www.novartis.co.uk/our-work/clinical-trials
Experimental Study
Parallel groups design
European Patients’ Academy
on Therapeutic Innovation

https://www.eupati.eu/clinical-development-and-trials/clinical-trial-designs/

5
Cross-over design (2 x 2)
European Patients’ Academy
on Therapeutic Innovation

6
Matched-pair design
European Patients’ Academy
on Therapeutic Innovation

7
Factorial Design

▪ Test → combination of treatment ( 2 or more treatments

simultaneously)
▪ Example 2x2 factorial design → Treatment A and Treatment B
▪ Group: Treatment A alone, Treatment B alone, Combination of
treatment A and B, Neither Treatment A or B (control)
Types of Comparison in Clinical Trial

▪ Superiority comparison trials demonstrate that the investigational

medicine is better than the control.

Sometimes drug not better → safer, cheaper, tolerable so :

▪ Equivalence comparison trials demonstrate that the endpoint measure is
similar (no worse, no better) to the control.

▪ Non-inferiority comparison trials demonstrate that the investigational

medicine is not worse than the control.

▪ Dose-response relationship trials demonstrate various dose parameters

including starting dose and maximum dos.
Bias and Variability

https://catalogofbias.org/biases/
Randomization

Random Non-Random
▪ distribution of study subjects at ▪ Pre-assigned to group
random (i.e. by chance)
▪ Selection/allocation bias!
▪ All subjects → have the same
chance to be allocated to ▪ May able to explain association
treatment group → but the association may be
strongly influenced by a third
▪ Eliminate → allocation/selection variable
bias
▪ No systematic difference
between the groups
Stratification
Stratification

▪ occurs at about the same time as randomization, and it occurs

before subjects receive the experimental or control treatments.
▪ study population are classified according to stage of disease,
gender, age, location of the clinic, prior therapy (if any),
cytogenetics, biomarker expression, etc.
▪ For each of the stratified subgroups, roughly equal numbers of
subjects are assigned to the experimental treatment and to the
control treatment
▪ Advantage: minimize confounding, allow sub-group analysis →
strong conclusion
 Randomization, Stratification

A combination of ascorbic acid and α-tocopherol to test the effectiveness and safety in the Advantage: minimize confounding, allow sub-group
fragile X syndrome: Study protocol for a phase II, randomized, placebo-controlled trial analysis → strong conclusion
Confounding

Factor 1: Outcome : Lung

Gender Cancer

Factor 2: Smoking

Picture credit:
https://commons.wikimedia.org/wiki/user:Ultrabem
Eligibility Criteria
 Eligibility Criteria

• Association → high statistical power → homogenous patient, remove noise (internal validity)
• However → results should be generalizable (external validity)
• Thus → balance: inclusion and exclusion criteria
Blinding
 Blinding

Definition: one or more parties in a trial are kept unaware of which treatment arms participants have been
assigned to

https://www.eupati.eu/clinical-
development-and-trials/concept-
blinding-clinical-trials/

Question:

Why perform blinding?

Is it always possible?
Why Blinding

https://catalogofbias.org/biases/
Endpoints
 Endpoints

• Definition: primary outcome that is being measured by a

clinical trial
• Why patients take their medication:
• survival
• Benefit detectable by patients
• decreased probability of developing a condition or diseases
complication
• Primary endpoints should be directly measuring one of these
three
Types of Endpoints

1. DIRECT ENPOINTS /CLINICAL ENDPOINT

▪ Clinically meaningful measures → endpoint characterize clinical
outcome of interest
▪ Clinical endpoints → most relevant → Study subject is aware or
afraid of, such as death, a heart attack, loss of vision
▪ Example → Life-threatening disease → endpoint: Overall survival
→ The length of time from either the date of diagnosis or the start
of treatment for a disease, such as cancer, that patients diagnosed
with the disease are still alive
▪ Drug approval based on this endpoint (most of the time)
Types of Endpoints

▪ Direct Endpoints can be:

▪ Objective: survival, disease exacerbation, clinical event
(e.g. stroke)
▪ Subjective: “Health Related Quality of Life”
Types of Endpoints

2. Surrogate Endpoints
▪ Usually used → natural progression of the disease: long time, or
window of drug therapy : extremely long
▪ Laboratory measure, physical sign → substitute for a clinically
meaningful endpoint
▪ Changes induced by a therapy on the surrogate endpoint reflect
changes in the clinically meaningful endpoint.
▪ Example: LDL-cholesterol to determine impact of treatment on
atheroschlerosis
 Composite Endpoints

▪ Considerations:
▪ Each component should itself be clinically meaningful.
▪ Ideally, each component would be approximately equally
meaningful.
▪ “Success” should not be concluded if driven by a less
meaningful component, if there is evidence of a therapeutic
disadvantage on a more meaningful component.
▪ The composite should not include individual components for
which a treatment effect is not expected.
▪ May complicate communication of the established benefit of a
drug.

https://www.fda.gov/media/84987/download
Choosing Endpoint

▪ TIMING is also important → when to measure endpoints

▪ Sometimes → choice of timing can cause bias
▪ Example : comparing rapid acting analgesic to longer
duration analgesic
Analysis
Intent to Treat Analysis vs. Per Protocol
Analysis
Intent to Treat Analysis (ITT)

▪ population includes all enrolled subjects, and encompasses those

(at the time of enrollment) who were believed to meet all the
inclusion/exclusion criteria, and those who, some time after
enrollment, were found not to have actually met these criteria on
the enrollment date
▪ all subjects who complied with the terms of the Clinical Study
Protocol through the duration of the clinical trial, as well as
subjects who deviated from the Clinical Study Protocol (failure to
sign the consent form, failure to take the study drug or control
according to the required schedule, failure to make all the
scheduled assessment visits, and dropping out of the trial)
Per Protocol Analysis

▪ including, in the analysis, only subjects who were enrolled in the

clinical trial, and who actually complied with all of the substantive
instructions set forth in the Clinical Study Protocol
▪ More intuitive, easy to understand
▪ BIAS!! E.g., Subject dropout → side effects
▪ Reasons for drop out should be investigated

▪ Perform both for sensitivity analysis

 ASSIGNMENT

1. Look at this study

▪ https://www.thelancet.com/journals/lanres/article/PIIS2213-2600(13)70070-
5/fulltext
▪ Summarize the design:
Type of Comparison, Randomization, Eligibility Criteria, Blinding, endpoints, design
(treatment allocation)