Pharmacoepidemiology

Basic Principles and Methodology

Professor Saad Shakir
MB ChB LRC&S FRCP FFPM FISPE MRCGP
Director - Drug Safety Research Unit
Southampton
Pharmacoepidemiology is the study of the use
and effects of medicines in large numbers of
people.
or
The application of epidemiological methods to
study the effects of medicines in large numbers of
people.
Pharmacovigilance
Monitoring drug safety in clinical practice
Identification of drug safety hazards
Evaluation of the issues
Conducting risk/benefit assessment
Taking actions to optimise drug safety (risk
management), when needed
Monitoring that actions taken were effective
Pharmacovigilance is supported by
pharmacoepidemiology
Scope of pharmacoepidemiology
Includes all observational (i.e. real
world) research studies in man relevant to
use of drugs
In observational studies, the investigator does
not assign interventions
Also occasionally includes randomised
clinical trials (usually simple clinical trials)

PHARMACOEPIDEMIOLOGY
Is mostly based on observational data
Links exposure and outcome in terms of adverse events,
therapeutic gain, length of life or quality of life, or for any
appropriate groups or subgroups in any selected
medicines
Also includes drug utilisation studies
Has developed mainly for exploration and quantification
of drug risks
Is now being developed for comparative effectiveness
Bridges medicine, statistics, demography, clinical
pharmacology and epidemiology



Pharmacoepidemiology - Why?
Limitations of clinical trials in assessing
drug safety
Too few patients
Too short
Too healthy (exclusions)
Too few other drugs
Not too many young or old
Too few women
Too specific

Number of patient exposures
required to detect ADEs
ADE
incidence
1 Event 2 Events 3 Events
1 in 100 300 480 650
1 in 200 600 960 1300
1 in 1000 3000 4800 6500
1 in 2000 6000 9600 13000
1 in 10000 30000 48000 65000

*Statistical power: the probability of detecting an ADR if it
really occurs in the population under study (95%)
Record Linkage
The use of automated databases to link
exposure to medicines with health
outcomes
Examples
Insurance databases
Health management databases e.g. Medicaid
Health information databases e.g. GPRD,
Mediplus
The advantages of
pharmacoepidemiological studies
Represent real life
Relatively inexpensive
Increasing number of databases
Increasing academic and regulatory interest
BUT BEWARE PITFALLS
Some problems with
pharmacoepidemiological studies
Prone to bias and confounding
Missing data
Misclassification
In many cases relies on data which have been
collected for another purpose (however, follow-
up can improve the quality of the data)
The trade off between double blind
clinical trials and observational studies
Address bias and confounding
but sometimes there are
valid concerns
regarding external validity
Represent real world usage
Usually large sample sizes
but issues regarding
bias and confounding
RCTs Observational studies
Definitions
Populations and Samples
Population - a large group of people in a
defined setting
e.g. because of a characteristic such as
presence of a disease, relatively unselected
Sample - a subset of a population selected
from it.
Sample
Population
Measures of Associations
Relative risk
Attributable risk
Odds ratio
Relative Risk
The ratio of the incidence rate of an
outcome in the exposed group to the
incidence rate in the unexposed group.
More than 1
1
Less than one
Use with confidence interval
Attributable Risk
The arithmetic difference between the
incidence rates in the exposed and
unexposed groups.
Use confidence intervals
Very important in considering the public
health impact of an association.
Oral Contraceptive Pills and
Venous Thrombo-embolism
Third Generation COC
Second Generation COC
RR
Attributable Risk
25/100,000 per year
15/100,000 per year
1.66
10/100,000 per year
Pregnancy
Healthy non pregnant
women
60/100,000 per year
5/100,000 per year



Odds Ratio
Odds - If an event has a probability Pr(E),
the odds of the event is defined as
Pr(E)/{1-Pr(E)}
Odds ratio - If two events, E
1
and E
2
, have
the respective probabilities Pr(E
1
) and
Pr(E
2
), the odds ratio comparing E
1
with
E
2
is [Pr(E
1
)/{1-Pr(E
1
)}]/Pr(E
2
)/1-Pr(E
2
)}],
namely the ratio of the odds of E
1
to the odds of
E
2

Odds ratio
For rare events the odds ratio is an
acceptable estimate of relative risk
The relative risk and the odds
ratio are measures of the
strength of an association.
Neither measure necessarily
indicate causation
Number Needed to Harm
NNH
1/Attributable risk
e.g. risk of DVT in users 3 rd generation
OCP is 10/100,000 women years
NNH 1/10/100,000
NNH 10,000
Very useful number for communications
Methods
Types of epidemiological studies
case reports
case series
cross sectional surveys
Case-control studies
Cohort studies
Clinical trials
Cohort Study
Cohort is a group of subjects identified at a
certain point in time and followed to
determine the incidence of a disease
In pharmacoepidemiology a cohort includes
patients exposed to a medicinal product
Compare the frequency with which the
disease in question (ADRs) develops with the
frequency in an unexposed group

Case-control study
Start with patients who already have the
disease in question (case patients) e.g.
ADR
Compare frequency of past exposure to
the risk factor in question e.g. medicinal
product with the frequency of exposure in
a group without the disease (controls)





Data collection from Cohort
and Case-Control Studies
a
b
c d
present
(cases)
absent
(controls)
present
(exposed)
Study
Direction
S
t
u
d
y

D
i
r
e
c
t
i
o
n

absent
(not exposed)
Drug
Exposure
Odds ratio (OR) = (a/c)/(b/d)
Relative risk (RR) = (a/a+b)/(c/c+d)
C
o
h
o
r
t

s
t
u
d
i
e
s

Case control studies
Disease
Cohort and Case-Control
Studies MEASURES OF ASSOCIATION
Cohort study

Relative risk (RR) = (a/a+b)/(c/c+d)
The risk of the outcome in exposed subjects compared to
unexposed subjects.
Case-control study

Odds ratio* (OR) = (a/c)/(b/d)
The odds of exposure in cases compared to the odds of
exposure in controls.
It is an estimate of the relative risk.


Case Control Studies
DVT and oral contraceptives
Myocardial infarction and oral contraceptives
Clear cell vaginal adenocarcinoma and
diethylstilbesterol
Oestrogens and endometrial cancer

Cohort Studies
Advantages
Can calculate rates of outcomes
Can study many outcomes
Can elucidate temporal relationships
between exposure and outcome
Bias can be reduced in the acertainment of
exposure
Cohort Studies
Disadvantages
Insufficient for rare outcomes
Liable to selection bias
In pharmacoepidemiology confounding is
common
Even with record linkage requires
availability of medical records
Case Control Studies
Advantages
Relatively quick and inexpensive
Good design for rare outcomes
Suitable for conditions with long latency

Case Control Studies
Disadvantages
Particularly prone to bias compared with
other methods in particular selection and
recall bias
Can not directly compute incidence in the
comparative groups
Sometimes it is difficult to establish the
temporal relationship between disease
and exposure
Clinical Trials
In clinical trials each subject has an equal
chance of being included in either group


A
B
RANDOMISED CLINICAL
TRIALS
Randomisation
Intervention
Not the first choice in pharmacoepidemiology
ADEs generally infrequent
unworkable methodologically
prohibitively expensive
? always valid to extrapolate to real life groups and to
individual patients

Types of associations between
factors under study
No Association
Apparent Association
When there is an apparent
association
False
Chance (unsystematic variation, random error)
Bias (systematic variation)
Confounding
Causal (true)
Chance
Bias
Confounding
Chance
Statistical methods measure the role of
chance
BIAS
A process at any stage of inference which
produces results that depart
systematically from true values.
Examples:
Selection bias
Measurement bias
Information bias
SELECTION BIAS
When comparisons are made between groups of
patients that differ with respect of determinants of the
outcome other than those under study.

Example:
patients with hypertension attending a renal clinic

Measurement Bias
Methods of measurement are different for
some patients
Example - patients who receive ACE
inhibitors for hypertension may get more
frequent biochemical measurements than
those who receive beta blockers.
INFORMATION BIAS
When comparisons are made between groups
where there are systematic differences in the
information in respect of exposure or outcome
Examples:
Mothers of healthy babies forget about use of
drugs during pregnancy
Patients after MI give better dietary history than
controls
Bias- Points to remember
Represent error in study design
Statistical significance is not a protection
Proper study design is the best protection.
ONFOUNDCING
A confounder is a variable other than the risk
factor which is associated independently with
exposure and outcome.

It may create an apparent association or mask a
real association e.g. smoking and cervical cancer.

CONFOUNDING
Exposure
coffee
drinking
Outcome
heart disease
Confounder
smoking
TYPES OF CONFOUNDING
Confounding by indication
the indication for the drug is the confounding
variable i.e. to prescribe a given drug to
patients who have a poorer prognosis than
those who not receive the drug

Confounding by association
smoking is an association between lung
cancer and ulcer healing drugs
Approaches for controlling confounding
Random allocation
Subject allocation
exclusion
matching
Data analysis
stratification
mathematical modelling e.g logistic
regression
Epidemiological study designs
Case reports
Case series
Analysis of secular trends
Case-control study
Cohort study
Randomised clinical trial
Pharmacoepidemiology in
Drug Development
Define disease burden
prevalence and incidence
cost and disability
Examples studies in migraine, asthma and
Alzheimers disease
Pharmacoepidemiology in
Drug Development
Descriptive epidemiology
the natural history
complications of the disease and long term
sequel (description and frequency)

Background frequency of
adverse events of interest
Cardiac arrest and ventricular arrhythmia in patients taking antipsychotic
drugs:cohort study using administrative data
Hennesy et al. BMJ.2002;325:1070
Cohort of patients with schizophrenia treated
with antipsychotic drugs, a control group with
glaucoma and a control group with psoriasis
Outcome measure diagnosis with cardiac
arrest or ventricular arrhythmia
Patients with schizophrenia had higher rates of
cardiovascular death than controls ranging from 1.7 to 3.2
Epidemiological Data Sources
National statistics
Healthcare databases (GPRD, MEMO,
United Health Care)
Clinical databases (diabetes, HIV)
Registries
Monitoring systems (PEM)
Teratology centres

Record linkage is linking exposure with outcomes
Drug Utilisation studies
Study the marketing, distribution,
prescription and use of drugs in a society,
with special emphasis on the resulting
medical, social and economic
consequences. (WHO)
Drug utilisation studies
Measuring drug use in populations
Understanding the reasons for variations
in drug usage within and between
populations
Exploring the determinants of drug use
Assessing the appropriateness of prescribing
Some uses of drug utilisation studies in
pharmacovigilance and risk management
Any differences between pre- and postmarketing
populations and whether these differences
impact on safety
e.g. differences in age profile or disease severity
Off label use
Misuse
Is the drug being prescribed and taken
appropriately, e.g. drug interactions

Prescription Event Monitoring (PEM)


Technique of PEM

Aim to monitor all new products which are
expected to be widely used in general practice
Study starts as soon as possible after drug
marketed in UK
Identify all patients prescribed the drug by general
practitioners (GPs) in England.
DSRU notifies PPA of new drug to be studied
Patient takes prescription to pharmacist
Pharmacist dispenses drug and forwards
prescription to PPA for reimbursement purposes
PPA sends prescription data to DSRU in
confidence
Patient and GP identified
DSRU send GP questionnaire
GP returns questionnaire to the DSRU
Data from questionnaire entered on DSRU database
Follow-up
Selected events Pregnancies Deaths
Write to GP Outcomes Certificate
consultant records
assess causality
An outline of the PEM Process
Analysis of Event Data
PEM provides:-
number of events reported - numerator

number of patients exposed to the drug - denominator

duration of exposure for each patient - days of treatment
Compare event rates in the 1
st
month of treatment with 2
nd -
6
th
months
of treatment
Compare event rates during and after drug exposure
Identify Reasons for Stopping
Identify events considered as ADRs by doctor
Compare with incidence without drug exposure (literature, other
databases)
Compare event rates between drugs
- of same class
- with similar indications


Hypothesis/Signal Generation
PEM Analysis
Compare event rates in the 1st month of treatment with
2nd - 6th months of treatment
Compare event rates during and after drug exposure
Identify Reasons for Stopping
Identify events considered as ADRs by doctor
Compare with incidence without drug exposure (literature,
other databases)
Compare event rates between drugs
- of same class
- with similar indications



Hypothesis/Signal Generation

Follow-up
events of medical importance
premarketing
postmarketing in other countries
signals raised during the study
all deaths of uncertain cause
all exposure during pregnancy
children
Recently completed/ongoing
PEM studies
Tadalafil - ED
Vardenafil - ED
Ciclesonide - asthma
Rosuvastatin - hyperlipaemia
Modafanil - sleep disorder
Yasmin - COP
Strontium - osteoporosis
Intrinsa - testosterone
patch
Sitagliptin - type II DM
Varenicline - smoking
cessation



Tacrolimus - eczema
Pimecrolimus - eczema
Atomoxetine - ADS
Pregabalin neuropathic pain,
epilepsy
Duloxetine - depression,
urinary
incontinence
Quetiapine - antipsychotic
Rimonabant - obesity
Ibandronate - osteoporosis
Ivabronate - angina
Solifenacin - bladder
dysfunction

Full list www.dsru.org
M-PEM
For certain drugs, the information collected using standard prescription-
event monitoring (PEM) methodology is not sufficient to fulfil the aims
of the study
M-PEM methodology is an extension of the standard PEM methodology

M-PEM studies can monitor both safety, detailed drug utilisation and
health outcomes in patients who were prescribed a particular drug.
M-PEM methodology can be used to compare the events reported before
drug use to the events reported after drug use in patients who were
prescribed a particular medication

Information on drug utilisation and the characteristics of patients using
particular medications can also be collected using M-PEM

The DSRU has used M-PEM methodology to monitor the introduction of
new drugs onto the market, requesting the patients previous medical
history, concomitant medication use and concordance to the prescribed
medication under study.
M-PEM
The Effects of Risk Management
Carvedilol in the treatment of heart failure
Interim report in 847 patients
Acharya N, Wilton LV, Shakir S. Int J Clin Pharmacol Ther.2005.43;1:1-6.

Treatment initiated by hospital specialists in 735
(87%)
Supervision under shared care 595 (70%)
>90% started carvedilol in the recommended dose
Grades of cardiac failure at start of treatment
Grade II 281 37%
Grade III 297 43%
On treatment with carvedilol
improvement in NYHA was reported for 364 (43%)
20 <2.5% deteriorated
OK!
Exclusive: Britney's Little Helpers?
Potential for misuse
Pharmacoeconomics and outcomes
research
Disciplines related to PE with some
common methodologies
Pharmacoeconomics
Health outcomes research

Post-authorisation safety studies
(PASS)
Any study relating to an authorised medicinal
product conducted with the aim of identifying,
charactersing or quantifying a safety
hazard,confirming the safety profile of a medicinal
product or of measurement of risk management
measures
PASS
General principles
Characterise the safety profile
Provide assurance about absence of a
safety concern
Investigate potential or identified risk, e.g. characterise
the incidence rate, estimate rate ratio or rate difference
in comparison to non-exposed population and
investigate risk factors and effect modifiers
Evaluate the risks of a medicinal product used in
authorised indications by patient groups not studied in
the pre-authorisation phase (e.g. pregnant women,
elderly patients)
To assess patterns of utilisation and use of the medicinal
product that may have an impact on its safety (e.g. co-
medication, medication errors)
To evaluate the effectiveness of risk minimisation
activities
Sabine Strauss
PASS
At first authorisation
Post-authorisation
PASS is a condition of the authorisation and is legally binding
In the event that the safety concern applies to more than one
medicinal product, the EMA/National Competent Authority shall
encourage the MAHs to conduct a joint post-authorisation
swafety study.
Post authorisation Efficacy Studies
PAES
At authorisation: where concerns related to some aspect
of the efficacy of the product are identified or can be
revealed only after the product has been marketed.
Post authorisation: when the understanding of the disease
or the clinical methodology indicate that the previous
efficacy evaluations have to be modified significantly.
EU Commission may adopt implementing measures
regarding situations when PAES may be required
EMA shall adopt scientific guideline (not expexed before
end of 2012)
S Strauss
Scenarios for PAES might be:

Effectiveness studies or studies with increased external
validity
Outcome studies following initial evaluation on
biomarkers/surrogate
endpoint
Studies in sub-populations
Long-term/sustained efficacy
Products with narrow benefit/risk

EMA plans a public consultation for Spring 2012


A Spooner
All scientific work is incomplete - whether it is
observational or experimental. All scientific
work is liable to be upset or modified by
advancing knowledge. That does not confer on
us the freedom to ignore knowledge we already
have, or to postpone the action that it appears
to demand at a given time.
Sir Austin Bradford Hill
saad.shakir@dsru.org
www.dsru.org