EPIDEMIOLOGY

DEFINITION: quantitative study (scientific, systematic and data driven) of the distribution (frequency, pattern) and determinants (causes, risk factors) of health-
related states and events in specified populations. It is the application of this study to the control of health problems

Clinical epidemiology – investigates (presumed) patients

Population epidemiology – investigates presumably healthy subjects

Techniques can be:

Observational:
 they do not change the natural course of events nor the established practice
 they just state the measurement procedure for the wanted assessment

Experimental:
 limited to prevention, diagnosis and treatment
 they alter the clinical practice in a deliberate way (and state the appropriate measurement procedure)
 Experimental studies are more ethically problematic - participants approval and collaboration needed

Probably the biggest difference between observational studies and designed experiments is the issue of association versus causation. Since observational
studies don't control any variables, the results can only be associations. Because variables are controlled in a designed experiment, we can have conclusions
of causation.
TYPES OF OBSERVATIONAL STUDIES:

 routine data analysis

 register
 cross-sectional studies
 longitudinal studies
o case-control studies
o cohort studies

Cross-sectional Studies: analyses data from a population, or a representative subset, at a specific point in time

COHORT STUDY (can be retrospective or prospective): The cohort study design already identifies a people exposed to a particular factor and a comparison
group that was not exposed to that factor and measures and compares the incidence of disease in the two groups.

A higher incidence of disease in the exposed group suggests an association between that factor and the disease outcome. This study design is generally a good
choice when dealing with an outbreak in a relatively small, well-defined source population, particularly if the disease being studied was fairly frequent.

CASE-CONTROL STUDY (are retrospective): Here researchers identify people with an existing health problem/disease (“cases”) and a similar group without the
problem (“controls”) and then compare them with respect to the frequency of one or more past exposures.
If the cases have a substantially higher odds of exposure to a particular factor compared to the control subjects, it suggests an association. This strategy is a
better choice when the source population is large and ill-defined, and it is particularly useful when the disease outcome was uncommon.

Researchers attempt to select homogeneous groups, so that on average, all other characteristics of the individuals will be similar, with only the characteristic in
question differing.
RANDOMISED CONTROLLED CLINICAL TRIALS

 Randomized controlled trial: (RCT) A study in which people are allocated at random (by chance alone) to receive one of several clinical interventions.

 One of these interventions is the standard of comparison or control. The control may be a standard practice, a placebo ("sugar pill"), or no intervention at
all. Someone who takes part in a randomized controlled trial (RCT) is called a participant or subject.

 RCTs seek to measure and compare the outcomes after the participants receive the interventions. Because the outcomes are measured, RCTs are
quantitative studies.

 Randomisation is the best way of ensuring that the results of trials are not biased by the way participants in each group are selected.

 In summary, RCTs are quantitative, comparative, controlled experiments in which investigators study two or more interventions in a series of individuals
who receive them in random order. The RCT is one of the simplest and most powerful tools in clinical research.

Treatment allocation. What’s so special about randomisation?

 The method used to assign interventions to trial participants is a crucial aspect of clinical trial design.
 Random assignment is the preferred method; it has been successfully used regularly in trials for more than 50 years.
 Randomisation has three major advantages.
o First, when properly implemented, it eliminates selection bias, balancing both known and unknown prognostic factors/confounders, in the
assignment of treatments. Without randomisation, treatment comparisons may be prejudiced, whether consciously or not, by selection of
participants of a particular kind to receive a particular treatment.
o Second, random assignment permits the use of probability theory to express the likelihood that any difference in outcome between
intervention groups merely reflects chance.
o Third, random allocation, in some situations, facilitates blinding the identity of treatments to the investigators, participants, and evaluators,
possibly by use of a placebo, which reduces bias after assignment of treatments.
 Of these three advantages, reducing selection bias at trial entry is usually the most important.

Successful randomisation in practice depends on two interrelated aspects—adequate generation of an unpredictable allocation sequence and concealment of
that sequence until assignment occurs. A key issue is whether the schedule is known or predictable by the people involved in allocating participants to the
comparison groups. The treatment allocation system should thus be set up so that the person enrolling participants does not know in advance which treatment
the next person will get, a process termed allocation concealment. Proper allocation concealment shields knowledge of forthcoming assignments, whereas
proper random sequences prevent correct anticipation of future assignments based on knowledge of past assignments.
RATIONALE OF CLINICAL EXPERIMENTS

 To ensure the experimental treatment is superior to the standard one – it is necessary to compare the results obtained by both

 If a difference is found – its origin has to be ascertained to be the experimental factor, thereby excluding the concomitant circumstances of the experiment

 Because of the unavoidable random variation of the studied outcome  it’s possible non-random improvement can only be found statistically  therefore
clinical experiments have to enrol many units (e.g. patients)

A CLINICAL TRIAL IS ETHICALLY FEASIBLE UNDER 3 MAIN CONDITIONS

1. The experimental treatment is meant to improve the health conditions of the experimental subjects
2. The experimental subjects have understood the nature and implications of the trial they will undergo and have signed their consent – when the trial
subjects are minors or unconscious, the informed choice to participate is made by their care takers
3. The experimenter is convinced that the treatment is potentially useful to the subjects

PHASES OF A CLINICAL TRIAL

PHASE 1

 Trial conducted on healthy volunteers (50-100)

 Assesses adverse side effects
 Assesses maximum tolerated dose
 PK properties
 No control group – but there can be different dosage groups

PHASE 2

 Trial conducted on a small number of patients (100-600)

 Demonstrate therapeutic properties (efficacy)
 Choose optimally effective dose
 Characterise PK-PD relationships
 Can be controlled (and randomised) or not
 Not meant to be final proof of effectiveness of treatment

PHASE 3

 Trial conducted on a large group of patients (500-5000)

 Should conclusively prove the worthiness of the experimental treatment (efficacy)
 Definitive assessment of how effective a drug is compared to gold standard treatments
 Controlled, randomised and double blind if possible
 Should be adequately powered and analysed following the “by intention to treat” protocol
 Should mimic usual routine practice, where the treatment will be administered if approved
 Conducted only if phase 2 trials turn out to be consistent with treatment effectiveness

PHASE 4

 Trial conducted once treatment is on the market

 Controlled and randomised
 Aims to find new indications for treatment
 Assesses continuing validity in new situations and patient population

MAIN SPECIFIC CHARACTERISTICS OF A WELL CONDUCTED PHASE 3 CLINICAL TRIAL

1. Concomitant control group

2. Randomised allocation of treatments
3. Double blindness (if feasible)
 Adequate power is also desirable

FLAWS OF NON-CONCOMITANT CONTROLS

 Non-concomitant control patients can have a better or worse prognosis, due to different general conditions or socio-demographic characteristics.

 They can even have a different disease from the experimental subjects, due to outdated diagnostic criteria
SOME REASONS TO RANDOMLY ALLOCATE PATIENTS TO THE COMPARED TREATMENTS

 If the experimenter or the patients themselves choose the treatments they can, on purpose or inadvertently, allocate to the experimental treatment
individuals with a poorer (or better) prognosis than the control.

 So called systematic association (e.g. Patients enrolled on Monday undergo the experimental treatment, those of Tuesday undergo the control treatment),
doesn’t allow blinding.

TECHNIQUES AND METHODS OF RANDOMISATION (they are not mutually exclusive)

 Randomly permuted blocks

 Within strata randomisation (e.g. within gender)
 Adaptive randomisation: the probability of allocating a subject to the experimental group changes among strata and during the course of recruitment
 Cluster randomisation: groups instead of single units are randomised
Simple randomisation

- patient is assigned a treatment without regard for previous assignments

- similar to flipping a coin
- problem = small probability of assigning the same number of subjects to each treatment group  severe imbalance in numbers assigned to each
treatment is a critical issue with small sample sizes
- problem = lead to imbalance among treatment groups with respect to prognostic variables that affect the outcome variables

Randomly permutated blocks

one approach to achieve balance across treatment groups

scheme consists of a sequence of blocks such that each block contains a pre-specified number of treatment assignments in random order
therefore, the randomisation scheme is balanced at the completion of each block
Stratified randomisation

Another type of constrained randomization is called stratified randomization. Stratified randomization refers to the situation in which strata are constructed
based on values of prognostic variables and a randomization scheme is performed separately within each stratum. For example, suppose that there are two
prognostic variables, age and gender, such that four strata are constructed:

The strata size usually vary (maybe there are relatively fewer young males and young females with the
disease of interest). The objective of stratified randomization is to ensure balance of the treatment
groups with respect to the various combinations of the prognostic variables. Simple randomization will
not ensure that these groups are balanced within these strata so permuted blocks are used within each
stratum are used to achieve balance.

If there are too many strata in relation to the target sample size, then some of the strata will be empty or
sparse. This can be taken to the extreme such that each stratum consists of only one patient each, which
in effect would yield a similar result as simple randomization. Keep the number of strata used to a
minimum for good effect.
Adaptive randomisation

Adaptive randomization refers to any scheme in which the probability of treatment assignment changes according to assigned treatments of patients
already in the trial. Although permuted blocks can be considered as such a scheme, adaptive randomization is a more general concept in which treatment
assignment probabilities are adjusted.

One advantage of permuted blocks over adaptive randomization is that the entire randomization scheme can be determined prior to the onset of the study,
whereas many adaptive randomization schemes require recalculation of treatment assignment probabilities for each new patient.

Urn models provide some approaches for adaptive randomization. Here is an exercise that will help to explain this type of scheme. Suppose that there is
one "A" ball and one "B" ball in an urn and the objective of the trial is equal allocation between treatments A and B. Suppose that an "A" ball is blindly
selected, so that the first patient is assigned treatment A. Then the original "A" ball and another "B" ball are placed in the urn so that the second patient has
a 1/3 chance of receiving treatment A and a 2/3 chance of receiving treatment B. At any point in time with nA"A" balls and nB"B" balls in the urn, the
probability of being assigned treatment A is nA/(nA+ nB). The scheme changes based on what treatments have already been assigned to patients.
This type of urn model for adaptive randomization yields tight control of balance in the early phase of a trial. As n A and nB get larger, the scheme tends to
approach simple randomization, so the advantage of such an approach occurs when the trial has a small target sample size.
SIMPLE VS CLUSTER RANDOMISATION

 In cluster randomisation, instead of single subjects, small sets (named clusters) of units are allocated randomly to different treatments.

 Example of clusters used for randomisation are:

o Hospital wards allocated to different nursing protocols
o The sets of patients who booked a surgical intervention a given date
o School classes in dental prevention trials
o Town communities in anti-malaria intervention trials

ALLOCATION CONCEALMENT

 Study researchers are unaware of the treatment allocated to any study subject until the actual administration begins
 At that time, they learn it (e.g. by opening a sealed envelope)

The allocation concealment should not be confused with blinding. Allocation concealment seeks to prevent selection bias, protects the assignment sequence
until allocation, and can always be successfully implemented.

In contrast, blinding seeks to prevent performance and ascertainment bias, protects the sequence after allocation, and cannot always be implemented.

HOW TO BLIND A TRIAL

 Make the different treatment appearance indistinguishable, for the treating doctor and the patient
 If an arm has no treatment – administer a placebo
 If the compared treatments have differing assumption protocols (e.g. pills vs injections), administer dummies
THE INTENTION-TO-TREAT (ITT) PRINCIPLE

The intention-to-treat principle defines that every patient randomized to the clinical study should enter the primary analysis. Accordingly, patients who drop
out prematurely, are non-compliant to the study treatment, or even take the wrong study treatment, are included in the primary analysis within the respective
treatment group they have been assigned to at randomization (“as randomized”).
Consequently, in an analysis according to the ITT principle, the original randomization and the number of patients in the treatment groups remain unchanged,
the analysis population is as complete as possible, and a potential bias due to exclusion of patients is avoided. Thus, the patient set used for the primary
analysis according to the ITT principle is called “full analysis set”.

There are only some specific reasons that might cause an exclusion of a patient from the full analysis set:
 no treatment was applied at all
 there are no data available after randomization

THE PER-PROTOCOL (PP) PRINCIPLE

While an analysis according to the ITT principle aims to preserve the original randomization and to avoid potential bias due to exclusion of patients, the aim of a
per-protocol (PP) analysis is to identify a treatment effect which would occur under optimal conditions; i.e. to answer the question: what is the effect if
patients are fully compliant? Therefore, some patients (from the full analysis set) need to be excluded from the population used for the PP analysis (PP
population).
Usually, this applies to patients fulfilling any of the following criteria:
 any major protocol deviations (e.g. intake of a concomitant medication affecting the primary endpoint)
 non-availability of measurements of the primary endpoint
 non-sufficient exposure to study treatment
There might be further criteria for selecting a PP population; however, the following approaches are essential:
 The assignment to the PP analysis set needs to take place prior to the analysis (if possible in a blinded manner).
 Deviations that might be affected by the actual treatment should not be used as exclusion criteria: e.g., “premature discontinuation from the study”
might not be a good choice of criterion for exclusion from the PP analysis, if this discontinuation was due to lack of efficacy (and therefore associated
with the treatment received).
EXPERIMENTAL DESIGNS

Diverse experimental designs are possible for clinical trials: when applicable, they offset some components of the unexplained variability of the observations,
thereby reducing it.

1. The design choice determines the appropriated statistical analysis, this in turn determines the sample size required to get the wanted power of the trial

2. The possible gain of a design vs. another is measured by the relative efficiency

In simple cases, the relative efficiency of two designs can be expressed as the ratio of the sample sizes required to achieve a given objective

PARALLEL GROUP DESIGN

The most common design is the parallel-group trial. Each subject is randomised to one and only one treatment and followed-up to determine the effect of each
treatment in parallel groups.

Advantages:
- Simple and easy to implement
- Applicable to acute conditions
- Analysis is less complicated and interpretation is straight forward
Disadvantages:
- Does not take into account the inter-individual variability

CROSSOVER DESIGN

In a crossover design, each participant is randomized to a sequence of two or more treatments therefore the participant is used as his or her own
control. Crossover trials produce within participant comparisons, whereas parallel designs produce between participant comparisons. For example, in the
simplest case, participants are randomised to receive either intervention A followed by intervention B or randomised to intervention B followed by intervention
A. This design is called a two-period crossover design.

Crossover trials can be used to investigate chronic conditions, such as asthma, where the objective is to investigate the participants’ short-term response to
therapy. The condition must also be stable, so that the circumstances at the beginning of each period are more likely to be the same. Clearly, not all
interventions can be studied in crossover designs. For example, comparing surgical procedures or evaluating long term outcomes such as 5-year survival where
it is impossible to crossover to another intervention.
The advantage of using crossover trials is a more precise estimate of the treatment effect and therefore the trial requires fewer participants.
- Each patient serves as his own control
- Avoids between participant variation in estimating intervention effect
- Requires a small sample size

The disadvantages are:

- the difficulty in avoiding carryover effects (the influence of the first treatment phase ‘carrying over’ to the second treatment phase)
- patient withdrawals complicating interpretation and analysis
- difficulties with assigning adverse events which occur in later treatment phases to the appropriate intervention
- not useful for acute disease

Use of cross-over design: bioequivalence studies, phase I

Depending on the involved outcomes and explanatory variables, a two parallel group experiment can be analysed by:
 Independent sample t test
 X2 test
  Multivariable models (e.g.: general linear model, logistic regression, Cox model) containing the treatment factor plus the putative prognostic variables

Depending on the outcome and the explanatory variables, a two treatment two group cross-over experiment can be analysed by:
 Paired sample t-test
 McNemar test
 A multivariable model (e.g.: mixed effect general linear model, mixed effect logistic regression) that contains the treatment factor, the putative
prognostic variables and the blocking factor (i.e. the patient identity)

CONFOUNDING means the distortion of the association between the independent and dependent variables because a third variable is independently
associated with both.

A causal relationship between two variables is often described as the way in which the independent variable affects the dependent variable. The independent
variable can take different values independently, and the dependent variable varies according to the value of the independent variable.

So, let’s say you want to find out how alcohol consumption affects mortality…

You decide to compare the mortality rates between two groups – one consisting of
heavy users of alcohol, one consisting of teetotalers. In this case alcohol
consumption would be your independent variable and mortality would be
your dependent variable.

If you find that people who consume more alcohol are more likely to die, it might seem
intuitive to conclude that alcohol use increases the risk of death. In reality, however, the
situation might be more complex. It is possible that alcohol use is not the only mortality-
affecting factor that differs between the two groups.

People who consume less alcohol might be more likely to eat a healthier diet or less likely to smoke, for example. Eating a healthy diet or smoking might in
turn affect mortality. These other influencing factors are called confounding variables. If you ignore them and assume that any differences in mortality must be
caused by a difference in alcohol consumption, you could end up with results that don’t reflect reality all that well. You might find associations where in reality
there are none, or fail to find associations where they do in fact exist.

How to minimise the effects of confounding during study design

If you are investigating the effects of an intervention, you can randomly assign people to an intervention and control group. The aim of randomization is to
evenly distribute the known and the unknown confounders between the two groups. The groups might still differ in potential confounders by chance but
randomization minimises these differences.

In other types of studies you can address confounding through restriction or matching. Restriction means only studying people who are similar in terms of a
confounding variable – for example, if you think age is a confounding variable you might only choose to study people older than 65. (This would obviously limit
the applicability of your results to other groups). Matching means pairing people in the two groups based on potential confounders.

How to minimise the effects of confounding during statistical analysis

After completing the study you can minimise the effects of confounding using statistical methods.

If there is only a small number of potential confounders you can use stratification. In stratification you produce smaller groups in which the confounding
variables don’t vary and then examine the relationship between the independent and dependent variable in each group. In the example we used before, for
example, you might want to divide the sample into groups of smokers and non-smokers and examine the relationship between alcohol use and mortality within
each.

If there is a larger number of potential confounders you can use multivariate analysis, for example logistic or linear regression.
QUESTIONS

WHY IS IT NECESSARY THAT A CLINICAL EXPERIMENT ENROL MORE THAN JUST ONE PATIENT?

 Because of the unavoidable random variation of the studied outcome  it’s possible non-random improvement can only be found statistically  therefore
clinical experiments have to enrol many units (e.g. patients)

WHAT ARE THE ELIGIBILITY CRITERIA FOR PARTICIPANTS IN CLINICAL TRIAL?

 In a clinical trial, the eligibility criteria aim to define the study population. The study population is a subset of the population with the medical condition in
which the intervention is to be tested. Eligibility criteria are related to patients’ safety and the anticipated intervention effect. The eligibility criteria will
have a significant impact on the generalizability of the study results, as well as on recruitment rates of the study; using eligibility criteria that are too
restrictive can lead to difficulty in recruiting sufficient patients.

Eligibility criteria are categorized into two groups: inclusion criteria and exclusion criteria. Appropriate justifications should be given for the use of certain
inclusion and exclusion criteria.

HOW IS IT POSSIBLE TO BLIND A CLINICAL TRIAL WHEN ONE OF THE TWO SAMPLES TO COMPARE RECEIVES NO ACTIVE TREATMENT?

 If one group receives no treatment  administer a placebo that is indistinguishable from the treatment
 If the compared treatments have differing administration protocols (e.g. injection and tablet)  administer dummies (e.g. doubled dummy trial)
WHAT IS THE PURPOSE OF THE USE OF “DUMMIES” IN CLINICAL TRIALS?

 Dummy medications are placebos that mimic the experimental drug and thereby enabling the blinding of the trial

WHAT IS A DOUBLE DUMMY CLINICAL TRIAL?

 Double dummy is a technique for retaining the blind when administering supplies in a clinical trial, when the two
treatments cannot be made identical. Supplies are prepared for Treatment A (active and indistinguishable placebo)
and for Treatment B (active and indistinguishable placebo). Subjects then take two sets of treatment; either A
(active) and B (placebo), or A (placebo) and B (active).

Double dummy is a method of blinding where both treatment groups may receive placebo. For example, one group
may receive Treatment A and the placebo of Treatment B; the other group would receive Treatment B and the
placebo of Treatment A.

 The double-dummy technique is used to compare drugs with very different appearance (for example, different dosage forms). For this purpose, subjects
take two types of drug, one of which is a placebo. For example, a tablet is to be compared to a juice, a study group receives the tablet with active drug and
a placebo juice and the comparison group receives a juice with active ingredient and a placebo tablet

WHY IS IT ADVISABLE TO RANDOMLY ALLOCATE THE SUBJECTS OF A CLINICAL TRIAL TO THE COMPARED TREATMENT, INSTEAD OF LETTING THE PHYSICIAN
CHOSE EACH SUBJECTS TREATMENT?

 Randomisation has three major advantages.

o First, when properly implemented, it eliminates selection bias, balancing both known and unknown prognostic factors, in the assignment of
treatments. Without randomisation, treatment comparisons may be prejudiced, whether consciously or not, by selection of participants of a
particular kind to receive a particular treatment.
o Second, random assignment permits the use of probability theory to express the likelihood that any difference in outcome between
intervention groups merely reflects chance.
o Third, random allocation, in some situations, facilitates blinding the identity of treatments to the investigators, participants, and evaluators,
possibly by use of a placebo, which reduces bias after assignment of treatments.
 Of these three advantages, reducing selection bias at trial entry is usually the most important.
LIST TWO TYPES OF RANDOMIZED CLINICAL TRIAL DESIGN AND SKETCH THE DIFFERENCE BETWEEN THEM.

EXPLAIN THE DIFFERENCE BETWEEN SUPERIORITY (I.E. USUAL) AND A NON-INFERIORITY CLINICAL TRIAL.

 A superiority study aims to show that a new drug is more effective than the comparative treatment (placebo or current best treatment)
 A non-inferiority trial aims to demonstrate that the test product is not worse than the comparator by more than a small pre-specified amount. This amount
is known as the non-inferiority margin

WHAT IS A “PLACEBO”?

 A placebo is an inert substance or treatment which is not designed to have a therapeutic value

WHY IS A LARGE SAMPLE SIZE BETTER SUITED THAN A SMALLER ONE (ALL OTHER CONDITIONS BEING EQUAL) TO IMPROVE THE QUALITY OF AN
EPIDEMIOLOGICAL STUDY?

 Increasing sample size increases power

 As sample size increases, the confidence in the estimate increases, uncertainty decreases and there is greater precision.
 Increasing the sample size tends to reduce the sampling error; that is, it makes the sample statistic less variable. However, increasing sample size does not
affect survey bias.
 This uncertainty is called sampling error and is usually measured by a confidence interval.
SURVIVAL

Survival analysis is a model for time until a certain “event.” The event is sometimes, but not always, death. For example, you can use survival analysis to model
many different events, including:
 Time the average person lives, from birth.
 Time after cancer treatment until death.
 Time from first heart attack to the second.
 Time from HIV diagnosis to AIDS development.
 Healing time of a bone fracture

Survival analysis can also be used to compare time-to-event for multiple groups. In medicine, two groups with different attributes (like normal/overweight,
diabetic/non-diabetic, high/low cholesterol) could be compared to see how those factors contribute to survival time in patients with heart disease, cancer or
other diagnosis.

Survival analysis is used to compare groups when time is an important factor. Other tests, like the independent samples t-test or simple linear regression, can
compare groups but those methods do not factor in time. In addition, survival times are usually positive numbers; many other methods would have to
transform data in some way in order to maintain positive numbers.

CENSORING

A distinctive feature of survival data is that observations may be censored: often the event
of interest (death, failure, recovery) has not occurred by the end of the study. Hence all is known for these subjects is that the lifetime is at least some
value.

Typically, survival data isn’t completely observed. Instead, some of the data is censored.

Censoring refers to missing data in a study such as subjects dropping out of trials or data that is otherwise lost. These are called “right censored” subjects. They
are usually counted as alive (or disease free) for the duration of the study for purposes of data analysis. Another type of censoring happens when a subject
simply doesn’t experience the event in question during the study. It doesn’t necessarily mean that the patient will never experience the event, just that the
event didn’t happen under observation. In other words, the time to event is incomplete.

 Type I: n subjects are observed for a fixed time. The number of censored units is then random.
 Type II: n subjects are observed until r events occur (the study is ended when a certain number of individuals have experienced the event of interest.)

The typical situation in medical studies is type I right censoring.

Condition of applicability of the usual survival analysis methods

The right censoring should not be caused by, or associated with the patient prognosis, within each of the groups compared

Survival analysis is a set of statistical approaches for data analysis where the outcome variable of interest is time until an event occurs.

Survival data are generally described and modelled in terms of 2 related functions:
 the survivor function representing the probability that an individual survives from the time of origin to some time beyond time t. It’s usually estimated
by the Kaplan-Meier method.

 the hazard function gives the instantaneous potential of having an event at a time, given survival up to that time. It is used primarily as a diagnostic tool
or for specifying a mathematical model for survival analysis.