AD HOC DATA SOURCES AVAILABLE FOR PHARMACOEPIDEMIOLOGY

STUDIES

INTRODUCTION:

Pharmacoepidemiology- By definition it is "the study of the use and effects of drugs in

populations". It is a field that applies the methods of Epidemiology to the content area of
Clinical Pharmacology.

The data sources available for Pharmacoepidemiological studies as Ad hoc sources are those
that are collected during post marketing surveillance studies. Ad hoc is something which is
for or concerned with one specific purpose i.e. providing data regarding drug safety in long
term use or which concerns about the drug aspects in a large population which is seen in post
marketing studies compared to pre marketing studies. Thus, limitations and objectives of
these studies respectively include:

Limitations of Premarketing Clinical Trials

 Size of the Patient Population Studied
 Narrow Population - often not providing for special groups:
 Elderly, children, women, ethnicity, use in pregnancy, co-morbidities
 Narrow Indications Studied
 Short Duration
 Not reflective of chronic use
Objectives of post marketing surveillance studies:
---Rare adverse effect
--- AEs in patients different from the study subjects
--- Long-term effects
--- Effectiveness of the drug for the original indications
--- Other beneficial effects of the drugs
DIFFERENT Ad hoc DATA SOURCES:
 Spontaneous reporting
 Global drug Surveillance : The WHO programme for International Drug Monitoring
 Case-Control Surveillance
 Prescription-Event Monitoring
SPONTANEOUS REPORTING:

All unsolicited reports from health care professionals or consumers, received by the FDA via
either the voluntary or mandatory route, are called spontaneous reports. A spontaneous report
is a clinical observation that originates outside of a formal study.
The United States Food and Drug Administration (FDA) is the Federal public health agency
that has regulatory responsibility for ensuring the safety of all marketed medical products,
including pharmaceuticals (i.e., drugs and biologics). In order to ensure that safe and
effective pharmaceuticals are available, the FDA relies on both the recognition, and voluntary
reporting, of serious adverse events (AEs) by health care providers and their patients and the
mandatory reporting of AEs by manufacturers as required by law and regulation.
Harm to patients from pharmaceutical use may occur due to four types of risk. Most injuries
and deaths associated with the use of medical products result from their known side effects,
some unavoidable but others able to be prevented or minimized by careful product choice and
use. It is estimated that more than half the side effects of pharmaceuticals are avoidable.
Other sources of preventable adverse events are medication errors, which may occur when
the product is administered incorrectly or when the wrong drug or dose is administered.
Injury from product quality problems is of interest to the FDA, which has regulatory
responsibility for oversight of product quality control and quality assurance during the
manufacturing and distribution process.
The final category of potential risk, those risks most amenable to identification by an
effective voluntary reporting system, involves the remaining uncertainties about a product.
These uncertainties include unexpected and rare AEs, long-term effects, unstudied uses
and/or unstudied populations, unanticipated medication errors due to name confusion or
packaging format, and product quality defects during the manufacturing process.

The main objective of the FDA postmarketing reporting requirement is to provide early
prompt detection of signals about potentially serious, previously unknown safety problems
with marketed drugs, especially with newly marketed drugs.

Terminology to be known are-

1. Adverse Event (or Adverse Experience)
Any untoward medical occurrence in a patient or clinical investigation subject
administered a pharmaceutical product and which does not necessarily have to have a causal
relationship with this treatment.
2. Adverse Drug Reaction (ADR)
A response to a drug which is noxious and unintended and which occurs at doses
normally used in man for prophylaxis, diagnosis, or therapy of disease or for modification of
physiological function.
3. Unexpected Adverse Drug Reaction
An adverse reaction, the nature or severity of which is not consistent with the
applicable product information (e.g., Investigator's Brochure for an unapproved
investigational medicinal product).
4. Serious Adverse Event or Adverse Drug Reaction
A serious adverse event (experience) or reaction is any untoward medical occurrence
that at any dose:
• Results in death,
• is life-threatening,
• requires inpatient hospitalisation or prolongation of existing hospitalisation,
• Results in persistent or significant disability/incapacity, or
• is a congenital anomaly/birth defect.
Advantages of spontaneous reporting
 Very wide spectrum: includes many different adverse effects (although mainly type A
and B), interactions and other problems
(E.g. pharmaceutical defects) and countrywide (in principle, coverage of all drugs and
all patients)
 Effective
 Rapid
 Continuous
 Comparatively cheap
Limitations of spontaneous reporting
 Causal relationship in case reports usually uncertain
 Underreporting and reporting bias
 No quantitative measurement (comparison of drugs is often difficult)
 Insensitive to type C adverse effects
Determinants of the level of performance of a spontaneous reporting system is based on-
 Reporting rate (e.g. number of case reports/million inhabitants/year)
 Reporting distribution, i.e. the proportion of physicians reporting, reporter
characteristics (e.g. general practitioners, specialists, pharmacists), geographic
distribution, reporting rates in specific populations
 Reporting quality (documentation and follow-up, e.g. The Uppsala Monitoring
Centre’s documentation grading)
 Reporting efficiency (the proportion of relevant case reports, e.g. concerning
unknown or serious adverse effects)

The data collection of spontaneous reporting through voluntary or mandatory services is done
by MEDWATCH programme.
The MedWatch program is for health professionals and the public to voluntarily
report serious reactions and problems with medical products, such as drugs and medical
devices. It also ensures that new safety information is rapidly communicated to the medical
community thereby improving patient care. All data contained on the MedWatch form will be
entered into the AERS (Adverse Event Reporting System) database.
The main goals of it are:
 The first is to increase awareness of drug, device, and other medical product induced
disease and the importance of reporting. Health professionals are taught that no drug
or other medical product is without risk and are encouraged to consider medical
products as possible causes when assessing a clinical problem in a patient. This goal
is accomplished through educational outreach, which includes professional
presentations, publications, and a continuing education program.
 The second goal of MedWatch is to clarify what should be reported. Health
professionals and their patients are encouraged to limit reporting to serious AEs,
enabling the FDA and the manufacturer to focus on the most potentially significant
events. Causality is not a prerequisite for reporting; suspicion that a medical product
may be related to a serious event is sufficient reason to notify the FDA and/or the
manufacturer.
 The third goal is to make it convenient and simple to submit a report of a serious AE,
medication error, or product quality problem directly to the FDA.

A single-page form is used for reporting suspected problems with all human-use medical
products (except vaccines) regulated by the Agency— drugs, biologics, medical devices,
special nutritionals (e.g., dietary supplements, medical foods, infant formulas), and
cosmetics. (Appendix 1, 2)
 https://www.accessdata.fda.gov/scripts/medwatch/medwatch-online.htm
 http://vaers.hhs.gov/index
Dear Health Professional Letters and safety alerts are the two means of providing updates on
adverse events reported to MEDWATCH through FDA to the health care professionals.
The safety assessment of the reported adverse events is done by THE ADVERSE EVENT
REPORTING SYSTEM (AERS). AERS is a client–server, Oracle-based relational database
system that contains all AE reports on pharmaceuticals submitted to the Agency either
directly or via the manufacturer. The mission of AERS is to reduce adverse events related to
FDA-regulated products by improving postmarketing surveillance and helping to prevent
adverse outcomes related to medical errors.
AE reports are sent to the FDA central document room, where they are tracked and forwarded
to the Office of Drug Safety (ODS) in the FDA’s CDER. When received by the ODS, these
incoming 3500 and 3500A reports (MEDWATCH forms) are assigned a permanent report
number (individual safety report), imaged, and stored in a Retrieval- Ware Imaging System;
subsequently they are entered verbatim into the AERS database. Data entry has a number of
sequential steps involving comparative entry, quality comparison of critical entry fields, and
coding and quality control into standardized international medical terminology using
MedDRA. Direct and 15-day expedited reports receive priority handling and are entered into
AERS within 14 days. Automated quality control is performed to review reports for
timeliness, completeness, and accuracy of coding. Statistical samples are also used to spot
check manufacturer performance in providing accurate and timely reports, which can be used
for compliance functions.

Every serious labelled or unlabeled adverse event report or reports describing important
medical events such as liver failure, cardiac arrhythmias, renal failure, and rhabdomyolysis
are electronically transferred into the computer inbox of the safety evaluators, who monitor
these events daily. The safety evaluators try to identify a potential “signal,” which is defined
as a previously unrecognized or unidentified serious adverse event. As a rule, a signal is a
more or less strong suspicion. It is a ‘snapshot’, a picture of the situation on a given moment,
somewhere in the course of discovery.
Various steps involved in signal generation are:
1. Signal Detection
Signal detection in Pharmacovigilance comprises the processes of: selection of a drug-
adverse event association of possible interest; the preliminary assessment of the available
evidence; and a follow-up of how the signal develops. In the automated systems currently in
use, the computer selects drug-adverse event pairs that stand out against the background of
the database, according to prefixed statistical criteria, for example using the proportional
reporting ratio (PRR) or, as is the case at the UMC (Uppsala monitoring centre), the
information component (IC) calculated by a Bayesian confidence propagation neural network
(BCPNN). Once the computer has at a given moment identified the associations that meet the
quantitative criterion, individual assessors have to select those associations that deserve
further preliminary attention. Criteria that often play a role in this selection process are listed
in table I.

Since a signal is by definition something new, first of all comes the question of whether or
not the association has previously been recognised, i.e. labelled in the product information or
described in the literature. This assessment is done by the National Centres before the
information is stored in the international database, this criterion will be automatically
included in the routine output documents of the UMC (i.e. the Combination and Association
Databases). In addition to the shift to statistical significance (a lower confidence interval
value above zero in the UMC BCPNN system), the statistical position of the association in
the database and its course over time provides valuable arguments for selecting or
abandoning the association.

The next important consideration is how expected or unexpected the connection would be
from the pharmacological point of view. Many new drugs are a close or distant relative of
one or another – chemical, pharmacological or therapeutic – family (see table II). Such
groups or subgroups often have a common profile as regards pharmacological actions (type A
adverse effects), interactions, hypersensitivity reactions (type B adverse effects) or other
problems (for example addiction). Understandably, but somewhat paradoxically, the
situations that an adverse reaction is pharmacologically unexpected or expected (but
‘unlabelled’ in the Summary of Product Characteristics) can both be a reason for selecting the
association. A third major consideration concerns the clinical characteristics of the adverse
event: is it something characteristic or non-specific, objective or subjective, rare or common
(i.e. high background frequency), is it a typical drug reaction (for example agranulocytosis)
or a recognised infectious or endogenous disorder? Also the potential relevance of the
association plays a role in the early decision to select an association for further attention.
Reasons why a signal may be likely to be relevant are reviewed in table III.
Pharmacovigilance has, first of all, an early warning function. New adverse drug reactions
should be detected as soon as possible, in particular when they are a potential threat to public
health or require for other reasons prompt regulatory action. Pharmaceutical companies in
particular are concerned with unforeseen adverse reactions that may change the
effectiveness/risk balance of a drug, threatening its market position in comparison with
competitive drugs. In addition, findings that is indicative of an unrecognised pharmacological
action or other scientific finding, even though not a regulatory concern may be pertinent to
our knowledge of the drug and to the diagnosis and management of the patients. Also, since
the case observations reported to a Pharmacovigilance centre disclose the conditions under
which adverse reactions occur in practice, they may have great educational value, even if the
association may not be new or unknown. Obviously, also the credibility, i.e. the amount of
evidence contained in the signal, contributes to the question whether a signal is likely to
require – and to justify – rapid action.
2. Signal Strengthening and Follow-Up
As has already been said, a signal is a snapshot on a given moment in the lengthy process of
the discovery of an adverse drug reaction (figure 1). Once a signal has been recognised and
assessed, it needs to be followed how it evolves over time in the database, e.g. as regards
absolute numbers of cases, the statistical parameters, exposure to the drug (utilisation) and
the persistence of the characteristicity and consistency of the reporting pattern.
Often further analysis of the database can provide a preliminary assessment of the strength of
a signal (table IV). A comparison of the reporting patterns in different countries is a logic
first step. A signal consisting of case reports from only one single country can be less
convincing than one based on reports from for example 20 different countries. A ‘best case–
worst case’ scenario can give a more balanced view of the strength of a signal. A variety of
comparisons, following from the hypothesis of the signal, using selections of the database
(e.g. different drugs, drug groups or drug combinations) can produce a better picture of the
statistical composition of the connection and of its consistency in different situations. A
nested case control study can be an attractive way of doing an ‘internal check’ of a signal
using the same database. A next step in signal follow-up is the use of data from other sources
for a search for the presence of a similar connection (table V). Sometimes also experimental
observations can be found that support the signal or provide a possible explanation, e.g. in
individual case reports or in the registration file of the company
3. Signals not Found by Disproportionality
A signal may consist of only a few case reports and not be statistically prominent and
nevertheless herald a true adverse reaction. This may in particular occur when the event is a
typically drug induced disorder, for instance agranulocytosis or erythema multiforme, and is
relatively frequently reported and ‘over-represented’ in the database. An example of it is
anaphylactic reactions to omeprazole and other proton pump inhibitors. Such signals will be
detected through the so-called Associations Database. However, they can be found in the
Combinations Database, which includes all new combinations, statistically prominent or not.
Although the results of the BCPNN signalling system of the UMC and of systems in place at
other centres show reasonable sensitivity and specificity, they should not be regarded as a
panacea.

4. The Balance of Evidence in a Signal (Credibility)

It is a highly professional and characteristic feature of Pharmacovigilance that during signal
follow- up on the right moment the balance must be made of the – usually incomplete –
evidence, in favour and against the connection, to decide the strength of the suspicion. This
evaluation is a combined assessment of the individual reports ‘case by case’, of the
aggregated data (clinically and statistically), and of information from other sources. Building
on the original thoughts of Bradford-Hill, a number of quantitative and qualitative criteria
have been identified that can give structure to the process (table VI). For obvious reasons
there is some repetition of the criteria in table VI and those already listed in table I and table
III. ‘At the right moment’ refers to a delicate weighing of the credibility of the signal, its
importance for individual users and in the public health perspective, of the measures needed,
and of the consequences if the signal proves false. In addition to the quantitative strength of
the association, the existence of a characteristic and consistent pattern in the reported events
can be a strong argument in favour of a connection.

As always in pharmacology, the presence of a dose-response relationship – i.e. a more

frequent or more severe effect when high doses are used – can represent substantial evidence.
Often the availability of a plausible hypothesis as regard the possible pharmacological or
pathological mechanism can add to the credibility, although with adverse reactions, where
often unidentified factors play a role, this argument can be a poor guide. Deliberate
experiments in humans are rarely justified. Nevertheless, in case reports often data elements
can be found that have an experimental aspect; for example the course of events after
stopping the drug (‘dechallenge’) or after an (often inadvertent) exposure, or information
regarding blood levels, metabolites, or perhaps a drug-dependent immunological
phenomenon. Pharmacokinetic or laboratory findings can be suggestive of a possible
mechanism (e.g. a drug interaction). An again different but often helpful piece of information
can be if there is a previous experience with a drug that may serve as an analogy. Finally
there is the often very influential issue of the nature of the event (objectivity, specificity), the
quality of the documentation of the observation, and the results of standardised causality
assessment of individual case reports.

5. Signal Testing
Time and again there has been confusion, for regulators, companies or the media, following
from the uncertain nature of signals generated by spontaneous reporting and the inherent
limitations of the system to produce secure and quantitative information. Spontaneous
reporting has been designed as a system for hypothesis generation in the first place. As a rule,
further study, using the most appropriate (and usually different) method, is needed to put the
hypothesis to the test. Often this is a formal Pharmacoepidemiological study, but also
pharmacological or pathological studies may give support. Nowadays, on a worldwide basis
there are far more signals found than can reasonably be tested. Further
Pharmacoepidemiological studies take time and money, and in the case of serious problems
measures need to be taken without delay. More comprehensive Pharmacoepidemiological
study programs may improve signal testing and add to the scientific basis of post-approval
regulatory decision making.

BAYESIAN NEURAL NETWORK:

The method is called a Bayesian confidence propagation neural network (BCPNN) and
implements Bayesian statistics within neural network architecture.
 It was developed for quantitative signal detection on the WHO database
It provides
 Objective initial assessment of all drug-ADR combinations
 Transparent selection of drug-ADR combinations for review
 A quantitative aid to signal assessment
 It is used to describe a wide range of different computational architectures (prediction,
classification of data sets, and data mining)
 It is made up of many simple processors (‘units’), where each unit has a small amount
of local memory.
 Communication channels (‘connections’), which carry numerical data, link the units
to each other, so that each unit only operates on local data and on inputs received via
connections from other units.
 In this application, a neural network is used to search for dependencies in the data set
Eg. All occurrences of variable x (for example a drug), all occurrences of variable y
(for example an ADR) and all occurrences of x and y together.

The strength of dependency between a drug and adverse reaction is defined by a

logarithmic measure of Disproportionality called the information component (IC)
When a drug-ADR combination is reported more often than expected relative to general
reporting of the drug, and general reporting of the ADR, it results in positive values of the IC.
IC is a distribution that changes on addition of new data and it is not a point estimate. A
Bayesian statistical approach is used to highlight unexpected dependencies in the data set;
implementation in neural network architecture. It allows the routine calculation of the
strength (or weakness) of all drug-ADR dependencies in the data set.

These calculations result in an IC and confidence intervals for each IC (more accurately
‘probability intervals’), for each relationship in the network. As the IC is also the weight in
the neural network, the computational efficiency of the method is further optimised.
Thus, BCPNN is a tool developed to enhance rather than replace traditional signal detection
procedures used on the WHO database by:
(i) Initial highlighting of associations for clinical review;
(ii) Providing a tool for further analysis of potential signals; and
(iii) Seamless extension of the method for detection of complex dependencies in the data set.

Global drug Surveillance - the WHO programme for International Drug Monitoring

WHO headquarters in Geneva is responsible for the WHO drug monitoring programme.
Operational aspects are managed by a Swedish foundation named the WHO Collaborating
Centre for International Drug Monitoring in Uppsala (now known as the Uppsala Monitoring
Centre, UMC) according to an agreement signed between WHO and Sweden in 1978. The
only regular budget contribution to the Centre is provided by the Swedish government. In
each country participating in the WHO programme there is a national centre, appointed by
the government, responsible for collecting spontaneously reported suspicions of ADRs,
originating from health professionals. National centres transform their case reports into a
specific WHO format and submit them to the UMC on a regular basis.

As of September 2009, 96 countries had joined the WHO Drug Monitoring Programme,
and in addition, 30 'associate members' were awaiting compatibility between the national and
international reporting formats.
Functions of it are:
 Identification and analysis of new adverse reaction signals from the case report
information submitted to the National Centres, and sent from them to the WHO ICSR
database. A data-mining approach (BCPNN) is used at the UMC to support the
clinical analysis made by a panel of signal reviewers
 Provision of the WHO database as a reference source for signal strengthening and
ad hoc investigations.
 Information exchange between WHO and National Centres, mainly through
'Vigimed', an e-mail information exchange system
 Publication of periodical newsletters, (WHO Pharmaceuticals Newsletter and
Uppsala Reports), guidelines and books in the Pharmacovigilance and risk
management area
 Supply of tools for management of clinical information including adverse drug
reaction case reports. The main products are the WHO Drug Dictionary and the WHO
Adverse Reaction Terminology
 Provision of training and consultancy support to National Centres and countries
establishing Pharmacovigilance systems
 Computer software for case report management designed to suit the needs of
National Centres (VigiFlow)
 Annual meetings for representatives of National Centres at which scientific and
organizational matters are discussed
 Methodological research for the development of Pharmacovigilance as a science.

The UMC has developed a set of computer programmes by which the incoming information
from participating centres is screened every 3 months. Various kinds of listings are produced
as a result of this screening and these listings are distributed to national centres for review.
The UMC has also set up an international panel of approximately 30 expert consultants who
are assisting the centre in identifying new and clinically important adverse reaction signals
within their specific area of expertise. Each consultant receives a subset of suspected drug
reactions that are identified by the computer system as being new or potentially serious. The
consultant selects and looks deeper into the potential drug problems that he/she judges as
being of greatest clinical relevance and writes short statements based on the available
information. These statements are distributed regularly to national centres as a document
called ‘Signal’ to be evaluated and acted upon according to the discretion of each centre.

Problem areas are as follows:

 Delay in reporting.
 Incompleteness of the database
 The vast number of potential signals
 Often case details supporting the drug-reaction association are scarce which makes
causality
 assessment difficult.
 Limited resources for medical assessment of potential Signals.

The dissemination of information about the ADE’s to the health professionals is done by
 Uppsala Reports
 The UMC Internet home page. This can be found at http://www.who.pharmasoft.se
 An e-mail discussion group called ‘vigimed’.
REFERENCES:

1. Bate A, Lindquist M, Edwards IR, et al. A Bayesian neural network method for
adverse drug reaction signal generation. Eur J Clin Pharmacol 1998; 54: 315-21
2. Meyboom RHB, Hekster YA, Egberts ACG, et al. Pharmacovigilance in perspective.
Drug Saf 1997; 21: 429-47
3. Olsson S. The role of the WHO programme on international drug monitoring in
coordinating worldwide drug safety efforts. Drug Saf 1998; 19 (1): 1-10
4. Strom BL. Pharmacoepidemiology. 4th ed. 2006 USA John Wiley & Sons Ltd. P.131-
203
5. Bate A, Lindquist M, Orre R, et al. Data mining analyses of Pharmacovigilance
signals in relation to relevant comparison drugs. Eur J Clin Pharmacol.
6. Websites:
 https://www.accessdata.fda.gov/scripts/medwatch/medwatch-online.htm
 http://vaers.hhs.gov/index
 www.fda.gov