M.Sc.

PROGRAMME IN EPIDEMIOLOGY
PRINCE OF SONGKLA UNIVERSITY

COURSE : EPIDEMIOLOGICAL METHODS II

MODULE : CLINICAL TRIAL

MODULE CONTENTS
 Definition of a clinical trial
 Phases of clinical trials
 Rationale for conducting a phase III clinical trial
o Discussion 1: What are advantages and disadvantages of clinical trials?
o Discussion 2: Examples of clinical trials
 Type of clinical trials
 Hypotheses of clinicals
 Choosing Participants, Intervention, Control and Outcome in a clinical trial
o Inclusion and exclusion criteria for participants
o Choosing interventions
o Type of controls
o Outcome measurements
o Randomization
o Blinding
 Common problems in clinical trials
o Compliance
o Contamination
o Co-intervention
o Discussion 3: What are some problems in conducting a clinical trial?
 Designing and Conducting a clinical trial
o Clinical record form
o The SPIRIT statement
o Data analysis
o The CONSORT Statement
 Ethical considerations in clinical trials
 Clinical trial registration
EXERCISES.
 2

OBJECTIVES
After completing of this module, you should be able to:
1. Describe phases of clinical trials
2. Describe the rationale for conducting a clinical trial
3. Describe the steps in designing a clinical trial
4. Understand how to design and conduct a clinical trial
5. Describe the possible problems in conducting a clinical trial
6. Understand ethical considerations in conducting a clinical trial
7. Know how to register a clinical trial

PERIOD REQUIRED

1 period

TEACHING AND LEARNING ACTIVITIES

1. Read attached notes and references

2. Practice an exercise
3. Group discussion

SOURCE REFERNCES

1. Cummings SR, Grady D, Hulley SB. Designing a Randomized Blinded Trial. In:
Hulley SB, Cummings SR, eds. Designing clinical research, 4th Ed. Philadelphia:
Lippincott Williams & Wilkins, 2013: 137-150.
2. Grady D, Cummings SR, Hulley SB. Alternative Clinical Trial Designs and
Implementation Issues. In: Hulley SB, Cummings SR, eds. Designing clinical
research, 4th Ed. Philadelphia: Lippincott Williams & Wilkins, 2013: 151-166.
3. Efird J. Blocked randomization with randomly selected block sizes. Int J Environ
Res Public Health. 2011;8(1):15-20. doi:10.3390/ijerph8010015
4. The CONSORT Statement:. http://www.consort-statement.org/
5. The SPITIT Statement: https://www.spirit-statement.org/
6. Assanangkornchai S, Nima P, McNeil EB, Edwards JG: Comparative trial of the
WHO ASSIST-linked brief intervention and simple advice for substance abuse in
primary care. Asian J Psychiatr, 2015, 18:75-80.
7. Tantirangsee N, Assanangkornchai S, Marsden J: Effects of a brief intervention for
substance use on tobacco smoking and family relationship functioning in
schizophrenia and related psychoses: a randomised controlled trial. J Subst Abuse
Treat 2015, 51:30-37.
 3

INTRODUCTION

The clinical trial is an experimental study in human beings which investigates the
clinical effects of an intervention. It is the best way to obtain information on the efficacy of

a treatment or prophylaxis – the knowledge about which is essential for clinical practice.
While a community experiment involves the community as a whole, a clinical trial
focuses mainly on treatment and its effect on individual subjects. In a clinical trial, the

response to exposure of each individual is considered to be independent from that of others.

The principle of clinical trial can be applied to test the effectiveness of both curative
and preventive measures. Efficacy implies clinical impacts under ideal conditions whereas

effectiveness is the impact under real conditions.

Clinical treatment includes medication, surgery, radio- therapy etc. Clinical

prevention includes improvement of environment, host immunity, chemo-prophylaxis etc.

Although education programme for individual subject can be included in clinical trial, it is
more difficult to standardize the manoeuvre and contamination is difficult to avoid. Thus

one can study the effectiveness of an education programme but not its efficacy.

PHASES OF CLINICAL TRIAL

The standard clinical trial can be divided into three phases.

Phase I involves a small number of subjects without any control. The general

objective is to test for immediate safety and possible benefit. Subjects can be either healthy

volunteers or terminal cases who might benefit and get relatively little harm from the trial.

The number of tested subjects is usually below 100.

Phase II involves more subjects and no control. The general objective is to expand
the test of safety to more subjects and to adjust appropriate dose levels before a controlled
trial (phase III) begins. The number of subjects in this phase ranges between 100 to 200.
main objective

1. test the effica

2. the safety 4

Phase III is the standard randomized controlled trial. There must be a concurrent

control group and the allocation of subject is based on chance ( so called randomized

allocation). The general objective is to test safety and efficacy of a standardized treatment

before it is approved for general use. The outcome of interest may include longer-term effects

such as mortality if the disease of interest is fatal e.g. heart attack, cancer.

After the treatment has been approved for general use, the next phase of study,
aiming to look at longer-term effects, is called “Post-marketing surveillance”. It can be a non-

randomized cohort study (since randomization may be more difficult after the result of phase

III study has been established). Some specific long-term rare effects such as teratogenicity

and carcinogenicity should be studied using a case-control design.

Other trials

Non-standardized trial includes open (no control or historical control) trial and non-

randomized concurrent control trial. In an open trial, without concurrent control, the causal
relationship between the treatment and the outcome is weak since the outcome of a disease
is also determined by various factors such as those in the host, environment and co-

intervention, which change over time. In a non-randomized trial (actually, a cohort study)

although treatment can be standardized, allocation may be subject to serious biases.

 5

RANDOMIZED CLINICAL TRIAL OR RANDOMIZED CONTROLLED TRIAL

(RCT)

The strength of RCT lies in the random allocation of subjects into groups of
treatment, which will minimize confounding and allocation bias. If randomized allocation
does work, distribution of different known and unknown prognostic factors in the two
groups will be balanced. Any difference in outcome of the two groups would be explained

by chance or difference in treatment only.

Rationale for RCT

RCT is more difficult to conduct than other kinds of studies. It sounds relatively
simple to compare the outcome of the two or more groups based on the difference in
artificial exposure or treatment. But, on a more detailed level, there are many procedures to

be followed in order to minimize biases and to ensure that the study will be conclusive. Since

RCT needs a lot of effort, the investigator must have good justifications for undertaking it.

The following are some examples:

- a medication has been proved to be useful from Phase I and Phase II trial
- a treatment is potentially beneficial for a serious disease but may itself be
potentially harmful (in this case, a placebo control is needed)

- outcome of a disease is not predictable (a placebo control is also needed)

- there has not been any solid conclusion as to the effect of the treatment, for
example, no RCT has been done before or the previous trials have serious pitfalls.
 6

TYPE OF CONTROL IN RCT

There are several possible control groups. The appropriateness of each kind of

control group depends on existing knowledge.

A placebo control group is conventionally the first choice. It is used to adjust for the
placebo effect in a pharmacological trial when the subjects tend to be psychologically
influenced by any kind of treatment. It is also very appropriate for testing a new treatment

of disease which has no known acceptable standard in the past. A new drug which is

promisingly effective may also have toxicity and do more harm than good. Bad outcomes of
many serious diseases may come from the disease itself or from the adverse effects of
treatment. The placebo controls are used as base-line group subject to the manoeuvre but

without any pharmacological effects.

When there is an accepted standard treatment (which should have been proved to be

better than placebo), it becomes unethical to include placebo control in the next trial. The

standard treatment is the best control to test against the newer promising treatment.

When the disease is chronic and candidate treatments do not have carry-over effect,

a cross-over design may be most appropriate. Examples of treatments with carry-over effects

are surgery, pharmacological products which may render permanent physiological,

pathological and/or immunological changes. In a cross-over design, there are two parts of

trials. In the first part, the treatment group and the control group are treated as in any RCT.

After the first part, a “wash-out period” follows to allow the subjects to recover to their initial

status. Then, the second part of the trial follows. In this part, the treatment group and the

control group are alternated. The major comparison in this kind of study is therefore within

the same subject. In other words this is a self-control study. The results are more precise

because comparisons are made within the same individual and the study needs fewer
subjects.
 7

RECRUITMENT OF SUBJECTS

Inclusion criteria

Generalizability of the result is perhaps the first thing to Consider in a clinical study.

One would usually want his/her study to be applicable to other patients. This is however

limited by dissimilarity between his/ her patients and those of others’ . An expert in a

particular clinical field usually works in a referral center. The status of his/her patients is

usually more serious than that of others. They have been filtered through the health care
system hence the result of the treatment in these difficult cases may not be directly
applicable to a setting in peripheral health care.

Homogeneity is another important issue. To test a hypothesis, subjects recruited

should be relatively homogeneous at least in race, age group and severity. Homogeneity can

help avoiding unexpectedly abnormal results which lead to difficulties in interpretation.

Convenience for follow-up is also another key issue in selecting subjects. With a high

proportion of loss to follow- up, information on practice outcome cannot be obtained.

Conclusion is then impossible.

Generalizability or external validity of a clinical trial comes after internal validity. It

is necessary to have a relatively homogeneous sample and to ensure that follow-up is not

too difficult. Otherwise, the internal validity will be poor and external validity will be

impossible.

Exclusion criteria

Exclusion criteria are set up to exclude any patients who have any contraindications
for the treatment as well as those who might give an extraordinary result due to some
abnormalities. Examples of subjects to be excluded from the study are: subjects with a history
 8

of hyper-sensitivity to any of the treatments, subjects who have concurrent abnormality of

anatomy or physiology which might influence the result of the treatment.

In writing a proposal, an inclusion criteria should be applied to cover the majority of
a homogeneous group of patients whereas exclusion criteria should be applied to exclude a
few extraordinary subjects.

BIASES IN RCT

A poorly designed clinical trial may be subject to a number of kinds of biases.

Allocation bias

Allocation of a subject is potentially biased if the subject selects the choice of

treatment by him/herself (self-selection bias). If allocation is “randomized” but not strictly so,

allocation bias can still play a role. For example, if treatment is alternated by day of the

week, the therapists or the patients may know this and try to come to the hospital only in the
patients may know this and try to come to the hospital only in the day of preferred choice.
Even if this secret is well kept, the patients who come on Monday may have some different
background from those who come on the other days. A subject may change his/her mind

after he/ she knows his allocation. To overcome allocation bias, it is recommended that

randomized allocation should be as strict as possible. It should take place after the patient

has given full consent, immediately before the treatment is given. More details of

randomized allocation are dealt with in the Biostatistical session.

Bias due to co-intervention and monitoring

During or after the treatment, there is usually some kind of concurrent intervention
such as nursing care, supportive therapy. There are also different kinds of monitoring.

Awareness of the therapist and the patient of their treatment group tends to introduce biases.
 9

Difference in co-intervention and monitoring between the two groups may occur. One group

may be visited more often than the other. As a result, there may be a difference in

psychological support or unwanted effects may be detected more often than usual.

Bias due to evaluation

If the outcome of the treatment is subjective such as pain, mood, performance,

quality of life, there can be biases from the evaluator.

Bias in analysis and publication

The most common bias in analysis is in choosing a subgroup which shows the
expected results. “ Negative” results (that is, results showing no difference between groups)

tend to be published less.

Blinding

The above biases ( except biases in analysis and publication) can be effectively

minimized by applying a blinding strategy in the trial.

A trial is single blind if the therapist knows about the allocation of the treatment but
the patient does not.

A trial is double blind when neither the therapist nor the patient know the allocation.

A trial is triple blind when the therapist, the patient and the evaluators all do not
know the allocation.

Blinding eliminates many biases. Double blinding eliminates allocation bias and co-

intervention bias. Triple blinding prevents biases in monitoring and evaluation.

 10

Blinding is very logical but often not possible. When two treatments are obviously

different, e. g. , radiation vs surgery, blinding is impossible. Even when the therapist is

supposedly blinded, side effects of one treatment may show up and become known to the
therapist. Placebo control and sham operation (mostly considered as unethical nowadays)

have been introduced to make double blinding possible.

COMPLIANCE

Compliance is often a problem in clinical trials. In principle, an experiment should

be designed so that the compliance will be at maximum. In practice, one hundred per cent

compliance is usually not achievable. Lack of compliance interferes with interpretation of

the results. A treatment may not be effective either because it has side effects highly
unacceptable to patients or the motivation of the patients may be too low to complete the
study. Consequently, the study cannot demonstrate efficacy of the treatment as planned.

During the study, compliance of the subjects should be monitored. Examples of methods to

measure and promote compliance are: diary keeping, pill counting, urine test for metabolites

etc. Compliance should be included in the report.

In the statistical analysis, it is generally advised that each subject should be
categorized according to “intention to treat” regardless of his compliance and whether he has

changed groups or not. The general research question for RCT is therefore “what would

happen if a patient is allocated in group A?” not “ what would happen if a patient complies

to group A treatment”?
 11

STATISTICAL ISSUES

There are several statistical issues in RCT. They are dealt with in depth in the

Biostatistics sessions. However, they will also be briefly mentioned here.

Balance in randomization

When the sample size is large, randomization should yield balanced groups, i.e., the

number of subjects in the two groups should not be too different. Distribution of basic

characteristics in the two groups should be similar or the causal relationship between
treatment and outcome will be difficult to conclude.
In practice, it may be appropriate to set strata of key prognostic factors, such as sex,
age group, staging and hospital, in advance. To ensure balance within each stratum, block

randomization may be applied. However, in practice, the size of block should not be known

by the therapist or by the patient so as to avoid allocation bias.

Type I and Type II errors

If the level of difference is high, it is quickly and easily detected. The study will be

repeatable and the results will be quickly accepted (such as efficacy of vitamin C in treatment

of scurvy and penicillin in treatment of pneumococcal infection). Unfortunately, most new

therapies do not have great or unequivocal superiority. A new therapy without any improved

effectiveness may show a certain degree of advantage just due to chance. On the other hand,
a useful new therapy may not be proved to be statistically superior to the standard one
because the number of the subjects in the trial is too small.

When there is no difference between the two therapies but the study shows
difference (due to chance), there is a Type I error.
When there is a clinically significant difference but the study cannot statistically
significant difference, (mainly due to inadequate sample size), there is a Type II error.
 12

The power of a study is the probability to detect statistically significant difference if

there is a clinically significant difference. Power depends upon, among many factors, the

sample size. A large sample size usually has higher power than a small sample size. A good

study design can minimize “noise” and increase power.

Since one always deals with a sample of patients not the whole population, error due
to chance always occurs. A conclusion declaring a difference always has some probability
of Type I error whereas a conclusion declaring no difference always has some probability
of Type II error.

Type I error leads to accepting an ineffective therapy as a better choice. Type II error

leads to rejecting superior therapy. The designer of a clinical trial must specify probability

of Type I and Type II errors (alpha and beta) and decide the level of clinically significant

difference below which any difference will be ignored. Alpha is usually set at a lower level

than beta, indicating the conservative approach of clinical science. Error in accepting new
therapy which is not better, is considered more harmful than error in ignoring new therapy
which is actually more effective. Based on levels of these errors, the probability of successful
outcome in each group and the ratio of sample size of the two groups, an appropriate sample
size can be computed (see related module in Biostatistics).
 13

FOLLOW-UP AND CENSORSHIP

The most difficult part of a clinical trial is in follow-up. A good clinical trial is usually

based on a good special clinic which is trusted by the patients. Home visit should be

supplementary rather than a main mechanism for follow-up.

In a trial for curable disease such as infection, the outcome of interest is success rate,
a dichotomous variable. On the other hand, in a trial on chronic non-curable disease such as

cancer, survival time may be more relevant. The latter type of study must often be terminated

at some time when the follow-up period of the subjects may not be equal (subjects recruited

early must have been follow-up longer than those recruited later). Moreover, in some trials,

such as those on IUD usage, there may be more than one final outcome: pregnancy,

spontaneous expulsion, removal on request, and death. These are all mutually exclusive

events. When pregnancy is considered as failure, others such as death, spontaneous

expulsion and removal on request are considered as censored. On the other hand,

spontaneous expulsion may be considered as failure and so on.

Statistical methods for survival analysis can help to get the most out of the available
information prior to censoring of follow-up. Such analysis is used mainly for analysis of

survival time for chronic diseases in which loss to follow-up may be, to a certain degree,

considered as equivalent to other kinds of censoring. However, for a trial in which definite

outcome such as successful treatment is mandatory, loss to follow- up cannot be

compensated for by this kind of analysis. Lastly, as survival decrease with time, a reasonably

long period of follow-up is also necessary.

 14

TERMINATION OF THE TRIAL AND INTERIM ANALYSIS

In some RCT’s on acute disease, the outcome is seen very soon after allocation and

treatment of the subjects. If a treatment shows obvious superiority over than its competitor

it is unethical to continue the study since a number of future subjects will allocated to the
inferior treatment. In this instance, termination comes early not because of statistical reasons

but from ethical concern. In order to know the difference early, interim analyses may be

done. Frequent interim analyses, however, jeopardize the final result. The initial hypothesis

will be repeatedly tested by statistical procedures and the probability of Type I error
increased. Such a conflict between ethical concern and requirement of statistical precision

can be pre-solved by well planned interim analysis or sequential trial. (See details in related

Biostatistics module).

DATA COLLECTION FORM AND COMPUTERIZED DATABASE

MANAGEMENT

A computerized data collection form is very useful but complicated in RCT. A cross-

sectional survey or a case control study usually needs one or just a few records for one case.

A cohort study and RCT need a more complicated form. The following categories and their

subcategories are examples:

A. Background of a clinical trial

1. Eligibility checklist
2. Demographical data
3. Medical history
B. Physical and other examinations
C. Laboratory results
D. Adverse events
E. Treatment record
F. Follow-up form
 15

1. Evaluation of outcome
2. Compliance
3. Patient diaries
G. Patient termination form
H. Miscellaneous form
1. Global impression of the therapists
2. Additional and unscheduled patient visit
3. Conclusion of treatment
4. Form for other comments
Example of these forms can be obtained from the reference.

ETHICAL CONSIDERATIONS

Of all studies on health problems, the clinical trial may cause the most ethical
concern.
In general, it is perhaps wrong to accept any treatment as standard without a prior
well-designed RCT. This is especially true when the treatment has some potentially harmful

effects. But the benefit and human right of the patient must be more important than the

scientific knowledge which might be gained from the study. The patient must have a right
to refuse participation without any effect on the quality of care provided by the therapist
who is a concurrent researcher.

However, firstly, in designing a RCT, the researcher usually has a hypothesis that
one treatment is better than the other. It is often considered unethical to have some patient

allocated to the presumed inferior treatment.

Secondly, who should make the decision on whether a particular patient should
participate a study? A patient usually experts the best treatment from the therapist. He is

practically in his hands. Although it is globally accepted that the patient should be well

informed and make the decision by him/herself, some groups of patients cannot be informed

due to ignorance, illness, extreme age etc. There may be some argument for and against

appropriateness of a relative making such a decision. A lot of clinical trials thus often try to
 16

avoid these groups. As a result, some knowledge on appropriateness of a treatment for these

groups may be lacking. For example, most contraceptive trials were conducted exclusively

in women aged between 18 and 35. Little has been learnt about their effects on a girl below

18 years of age or on a women older than 35. Prescribing practice thus based on

generalization of data from the other age groups.

In conclusion, a researcher who is going to conduct a RCT should be prepared to
provide answer for a serious ethical review. A successful RCT not only has a beauty in

design, data collection and analysis, but also an ethical justification.

 17

Exercise 1

Disease X is a common disease among elderly in your country. It has episodic

course, of which an episode usually lasts for one month, and remits spontaneously. During
an episode, patients are suffering with severe headache everyday, making them unable to
work or perform daily functions. A natural history study found that patients with this disease
have 4-6 episodes per year.

Drug A is a newly discovered drug and has just been investigated in Phases I and
II clinical trials to be safe and help reducing headache severity, shortening the duration of
each episode and decreasing number of episodes. Drug A is in an oral tablet form, to be
taken 1 tablet once daily for a period of 28 days. To be effective, it has to be taken within 3
days after the onset of an episode. No other drug has been proved to effectively treat
patients with Disease X, only some behavior modification methods such as routine physical
exercise and relaxation technique are claimed to be effective in reducing the severity of
headache and delaying the onset of new episode. However, cigarette smoking and alcohol
drinking can worsen the symptoms.

Based on the SPIRIT Statement checklist (https://www.spirit-statement.org/), design a

clinical trial for Drug A in treatment of Disease X in your country, following the items below.

1. Title
2. Objectives
3. Trial design
4. Study setting
5. Eligibility criteria
6. Interventions
7. Outcome
8. Recruitment
9. Allocation
10. Blinding
11. Duration of study
18