The Way We Make Progress Against Disease
• ‘Any research study that prospectively assigns human participants or groups of
humans to one or more health-related interventions to evaluate the effects on
… (The World Health Organization)
• “A carefully and ethically designed experiment with the aim of answering some
precisely framed question”.
… (Hill A. B., 1951)
Phase 0 Trials
• In recent years, a new trial phase term has emerged- the so called phase 0 (zero)
or micro-dosing trials.
• This is a interim step between preclinical research and phase I studies.
• Small number of human volunteers take small dose of experimental test article so
there is little risk of toxicity.
The design specification should be able to reflect the:
• type of treatment and number of treatments
• method of randomization
• type of blinding
• type of study question
• study medication
During Protocol Development
Bakhai et al, 
Source of patients
Phase I and II pharmaceutical studies usually involve small numbers of patients and
are therefore frequently performed in specialist centers (e.g., asthma laboratories in
hospitals with dedicated research units). – ‘Center of excellence’.
Investigators and centers
Investigators must have an understanding of what is required for the trial so that
they can clearly communicate this information to potential participants.
Enrollment of more centers, not economical, given the resources needed to train
Bakhai et al, 
Define patients with the disease or condition under investigation.
Eligibility criteria are usually based on:
• patient characteristics
• diagnostic test results
• disease duration
• disease severity
Inclusion criteria are the factors (or reasons) that allow a person to participate in a
clinical study. While exclusion criteria is vice-versa. Informed Consent.
Study to Find Out if Intensive Diabetes Clinic and Continuous Glucose Monitors Help
Teenagers With Diabetes
Inclusion Criteria Ages Eligible for Study:
Genders Eligible for Study:
10 Years to 18 Years
Type 1 diabetes mellitus of
at least 12 months duration
Most recent HbA1c >= 8.5%
Patients and families must
be willing to come to diabetes
clinic once a month for 4
Inability to understand
and/or speak the English
It is important to select patients who are ill enough to improve with the intervention,
but some patients might be too ill to participate in the study.
Selecting patients who have not been treated previously is a common requirement.
A balance needs to be achieved between scientific and practical issues. For example,
individuals aged >65 years.
Special consideration should be given to children or neonates as they can respond to
medications differently to adults.
Randomization is the unpredictable allocation of a patient to a particular treatment
strategy in a clinical trial.
Common types of randomization methods are:
• simple randomization
• block randomization
• stratified randomization
Tossing an unbiased coin, e.g., heads for treatment A and tails for treatment B. When
the next subject is to be assigned, previous allocations are not considered.
Bakhai et al, 
Consider an example trial with 12 patients. While there is an equal chance of being
allocated treatment A or treatment B, the number of subjects randomly assigned to
each treatment ends up being 5 and 7, respectively.
In cases where there are few patients, there is a need for other methods of
A block randomization method can be used to periodically enforce a balance in the
number of patients assigned to each treatment.
A block randomization can be implemented in three steps:
Step 1: Choose the block size and the number of blocks needed to cover the number of
patients in the study.
Step 2: List all possible permutations of treatments in a block.
Step 3: Generate a randomization code for the order in which to select each block.
Step 1: Given a sample size of 24 and using a block size of 4, we need six blocks.
Step 2: There are six possible permutations that allow two As and two Bs in each box:
AABB, ABAB, ABBA, BAAB, BABA and BBAA.
Step 3: The randomization code for blocks can be generated by producing a random-
number list for permutations 1–6.
Stratified randomization takes the balance correction suggested by blocking one step
Step 1: Choose the prognostic factors that could impact on the primary endpoint.
Experience of earlier trials and literature show that atopy, forced
expiratory volume within 1 second (FEV1), and age are the most
important determinants of time to first respiratory exacerbation.
Step 2: Determine the number of strata for each factor.
When several prognostic factors are chosen, a stratum for
randomization is formed by selecting one subgroup for each factor (continuous
variables such as age are split into meaningful categorical ranges). The total number
of strata is therefore the product of the number of subgroups in each factor.
Table 3 describes the strata for stratified randomization in the CF-WISE study. In this
example, the total number of strata is 2 (atopy) ×3 (FEV1) ×2 (age) = 12.
Step 3: Generate randomization codes.
This is done by generating a randomization list for each stratum and then combining
all the lists.
It has not been widely used because of the practical difficulties associated with
implementing it. Gaining importance recently in larger trials due to technology.
Randomization can minimize bias, but it can still occur, however, if study personnel and
patients know the identity of the treatment, due to preconceptions and subjective
judgment in reporting, evaluation, data processing, and statistical analysis.
With respect to blinding, there are four general types of blinded studies in clinical trials
• single blinded
• double blinded
• triple blinded
Clinicaltrials.gov, Bakhai et al, 
If the new intervention is a surgical treatment and is being compared with tablets
then the difference between the two is difficult to hide.
Advantage : simple, fairly inexpensive, and a true reflection of clinical practice.
• Knowing which treatment is being given are numerous.
• Patients may underreport adverse effects of the new treatment.
• Possibility that local investigators might supply different
amounts of concomitant treatments (e.g., only giving analgesics to the surgical
The patient should be unaware of which treatment they
are taking, while the investigators are aware of whether the treatment is new,
standard, or placebo.
The design is relatively simple and allows investigators to exercise their clinical
judgment when treating participants.
• Patients might under- or over report treatment effects and side-effects, based on
some influence or response from the investigators.
• Investigators may give advice or prescribe additional therapy to the control group
if they feel that these patients are disadvantaged in comparison to the active group -
In double-blinded studies, neither the patient nor the investigator knows the identity
of the assigned intervention.
This reduces the ability of the investigators to monitor the safety of treatments.
Specific problem of Double Blind :
Drug unblinded if medication not identical in appearance. Matched medication
required, especially in crossover study.
In triple-blinded studies, as well as the investigators and participants, all members
of the sponsor’s project team (e.g., the project clinician, statistician, and data
manager) are blinded.
Lessens the chance that the trial may stop early to favor either treatment, and
makes evaluation of results more objective.
Lessens investigator’s ability to monitor safety and efficacy.
Appropriate for studies in which the risk of adverse events due to the new or
standard treatment is low, and should not be
used for treatments where safety is a critical issue.
Ronald Fischer (1890-1962)
After a randomized controlled trial is conducted, statisticians help
determine whether any observed difference between outcomes
in the experimental and control arms are real, or simply chance
The process to which Fisher’s methods can be applied is called
In hypothesis testing a null hypothesis (H0) is articulated which is
typically a statement of no difference between experimental and
control patient populations.
The null hypothesis is about patient populations—all current and future patients
with specific conditions who receive specific interventions. Statistical inference is
used to decide whether or not the null hypothesis is true, based on a sample of
patients in a clinical trial.
Francis Bacon (1561-1626)
Evidence-based medicine depends on the
systematic accumulation of information about
how different treatments affect patients.
Using observations from a patient
sample to draw conclusions about its patient
The scientific method is schematized in Figure
Parallel group designs
• In a parallel study design, each subject is randomized to one and only one
treatment. Most large clinical studies adopt this approach.
• During the trial, participants in one group receive drug A "in parallel" to
participants in the other group receiving drug B.
• Subjects are randomized to sequences of treatments (A then B or B then A)
• Uses the patient as his/her own control
• Often a “wash-out” period (time between treatment periods) is used to avoid a
“carry over” effect (the effect of treatment in the first period affecting outcomes in
the second period)
• Can have a cross-over design with more than 2 periods
Bakhai et al, , ICH E9
• Treatment comparison is only subject to within-subject variability not
• Reduced sample sizes
• Strict assumption about carry-over effects
• Inappropriate for certain acute diseases (where a condition may be cured
during the first period)
• Drop outs before second period
• Period effect
Appropriate for conditions that are expected to return to baseline levels at the beginning
of the second period
• Treatment of chronic pain
• Comparison of hearing aids for hearing loss
• Mouth wash treatment for gingivitis
In early drug development, especially in Phase I pharmacokinetic, bioequivalence,
dose-proportionality, and dose escalation studies (for investigating the maximum–
tolerated dose), and in Phase II pharmacodynamic studies.
In later phases of drug development, as well as in other clinical studies, a crossover
design is suitable for trials that involve relatively stable conditions such as asthma,
rheumatism, migraine, mild-to-moderate hypertension, and angina.
Individuals are randomly assigned to two separate interventions (e.g.,
interventions A and B) and these interventions are each compared with their
corresponding control(s) .
• N/ 4 individuals are allocated to no treatment (control group).
• N/ 4 individuals are allocated to intervention A only.
• N/ 4 individuals are allocated to intervention B only.
• N/ 4 individuals are allocated to the combination of A + B simultaneously.
An important concept for these designs is interaction (sometimes called effect
Interaction: The effect of treatment A differs depending upon the presence or
absence of intervention B and vice-versa.
Bakhai et al, 
• If no interaction, can perform two experiments with less patients than
performing two separate experiments
• Can examine interactions if this is of interest.
• Cost: It is possible to evaluate multiple treatments within the same trial using
fewer patients than individual comparisons.
• Potential for adverse effects.
The possibilities on considering interactions between the treatments are as follows:
• To compare three active treatments with control and to show that any of the
treatment combinations is effective compared with the control.
• To compare two active treatments with control and to show that either
intervention is effective on its own compared with the control.
• To make six pair-wise comparisons between all four groups.
A trial to compare the altered versus the original compound or drug to demonstrate
that there has been no loss of effectiveness or increase in side-effects.
Bakhai et al, 
Design issue for equivalence trials
Equivalence trials can be of parallel or crossover design. Subject validity has a large
statistical influence on the equivalence result.
Both types of variability are present in each trial, crossover design more efficient in
terms of sample size.
If a parallel design were used, more volunteers would be needed to reach
equivalence with the same power.
Equivalence studies have important uses in clinical trials:
• They are used for comparing similar treatment compounds.
• They are used for comparing the efficacy of the same treatment compound in
differing formulations or in different cohorts of patients.
The key things to remember are that:
• They can have either an equivalence or a non inferiority endpoint.
• The outcomes can be clinical or pharmacokinetic.
Equivalence trials (Continued)
Non Inferiority Trials
A non inferiority trial aims to demonstrate that the effect of a
new treatment is as good as, or better than, that of an active
This is assessed by demonstrating that the new treatment is not worse than the
comparator by more than a specified margin (the non inferiority margin [δ]).
The following study will illustrate issues relating to the design and analysis of non
inferiority clinical trials.
Bakhai et al, 
When are the Non inferiority trials used?
Non inferiority trials are employed in situations where efficacious treatments already
The new treatment might be tested to establish that it matches the efficacy of the
standard, and at the same time has secondary advantages (e.g., in terms of safety,
convenience to the patient, or cost-effectiveness).
Alternatively, it might have potential as a second-line therapy to the standard (in
cases where the standard fails or is not tolerated).
Cluster Randomized Trials
Cluster randomized trials use a group of individuals, a hospital, or a community as
the unit of randomization.
Bakhai et al, 
Design of cluster randomized trials
Patients being counseled might report fewer side-effects of other therapies or be
more compliant, or be receiving support from other patients who are also in
counseling – contamination of outcomes.
This exchange of information will bias the effect of the intervention, and so it is
easier to offer all of the patients of one hospital counseling, while those in a similar,
nearby hospital receive no counseling.
It allows a large number of patients to be recruited in a shorter time; the results are
more generalizable and contemporary to a broader population at large; and such
studies are critical in trials involving patients with rare presentations or diseases.
A multicenter trial is a trial that is performed simultaneously at many centers
following the same protocol. The activity at these centers is synchronized from a
single command center – the coordinating center.
Bakhai et al, 
Why are multicenter trials conducted?
Multicenter rather than single center trials are carried out for several reasons:
• When studying rare diseases, there will be a larger pool of patients to recruit from
when using a multicenter trial. Therefore, the patient recruitment target will be
reached more quickly than in a single-center study.
• For diseases with low event rates, treatments are likely to have a small absolute
benefit and so large numbers (thousands) of patients might be needed in order to
see a significant benefit.
• Treatment benefits are not dependent on one specific center and, therefore,
should be reproducible at other centers.
• Any bias that might be related to the practice methods of a single unit – where
methods may be tailored to address local issues – will be reduced.
• Using many investigators to simultaneously evaluate a treatment gives more
sources of feedback, allows more doctors and healthcare professionals to gain
experience and confidence with the experimental intervention, and highlights any
• In certain cases, a difference in therapeutic approaches is being tested and
multiple centers that have different facilities or access to treatments are needed.
Innovations in Trial Design
Thomas Bayes (1702-1762)
The Bayesian Concept
Bayesians build on prior knowledge, rather than viewing each trial in isolation.
The larger the trial and the greater the difference between the experimental and
control arms, the greater the influence of the trial data on the new belief about the
On being asked to talk on the principles of research, my first thought was to arise
after the chairman’s introduction to say, “Be careful”, and to sit down.
Jerome Cornfield (1959)
1. Bakhai, A., Wang, D., (2006), Clinical Trials: A Practical Guide to Design, Analysis
and Reporting, Remidica, London, UK, pp: 1-153.
2. ICH E9 guidelines.
3. Kalliomaki, J., Miller, F., Kagedal, M., Karlsten, R., (2012), Early phase drug
development for treatment of chronic pain—Options for clinical trial and program
design. Contemporary Clinical Trials, 33 (689-699), [Accessed on: 15 Nov 2013]
4. Lloyd, J., Raven, A., (1994), Handbook of Clinical Research, In: Lloyd, F., Clinical
Trial Design, 2nd Edition, Longman Group, UK Limited, pp: 92-113.
4. Meinert, C.L., Tonascia, S., (1986), Clinical Trials, Design, Conduct, and Analysis,
Oxford, UK, Monograph in Epidemiology and Biostatics, Vol: 8, pp: 63-68, 113-118.
5. Perlmutter, J., Understanding Clinical Trial Design: A Tutorial for Research
Advocate, Research Advocacy Network’s Advocate Institute.