Principles of pharmacovigilance
Marc Girard
Independent pharmaceutical consultant
Address for correspondence:
Dr Marc Girard, MSc, MD
76 route de Paris
78760 Jouars-Pontchartain
France
Tel: +33-(0)1-3489-4229 Fax: +33-(0)1-3489-7608
Email: agosgirard@free.fr
Abstract
After a short historical survey of adverse drug
reactions, this paper summarises the current
organisation of pharmacovigilance in Europe, with
its principles of reporting, assessing, understanding
and preventing the adverse effects of medications.
Issues related to contrast media are included to
exemplify the principles. Respective responsibilities of
the various stakeholders are described: the European
Agency, EEC member states, manufacturers and, last
but not least, health professionals.
The paper then focuses more specifically on
spontaneous reporting, with the limits and biases
that result from inevitable under-reporting. In the
event of a new disease likely to be ascribed to a
drug or a class of drugs, a conceptual framework
for investigation is proposed, articulated on the
criteria of:
In spite of its awkward methodological limitations,
the author concludes that the principle of
spontaneous reporting may be appropriate for
investigating a new pathological entity such as
nephrogenic systemic fibrosis (NSF).
A brief historical survey
Physicians and, perhaps more particularly,
governments have long been more concerned with the
safety of drugs than with their efficacy. Hippocrates'
principle of 'first not to harm' is widely remembered
and it would be easy to find in the historical textbooks
on pharmacy a number of regulatory measures taken
even in ancient times in the hope of protecting people
against iatrogenic hazards due to chemicals. Indeed,
whatever doubts that could reasonably hang over the
genuine benefits of medicines in antiquity, their toxic
nature was clear for many: in ancient Greek, the word
pharmakon referred not only to medicines and drugs
but also to poisons.
Over and above the intrinsic toxicity of substances
given to patients, the risks inherent in their
preparation have received attention. To some extent,
it has been easier for governments to attend to
the quality of medicines (and to take appropriate
regulations about their manufacture, conservation
and supply) than to investigate directly the safety
of their (supposedly) 'active' principles.
In fact, the story that initiated the modern
organisation of drug monitoring occurred precisely
at the encounter between a drug that was genuinely
active - a sheer 'miracle' at the time - and a pitiful
mistake in the manufacturing process. After German
researchers discovered the antibacterial properties
of the dye Prontosil in 1932 and, soon afterwards
their French colleagues identified sulphanilamide as
the active portion of the product, a new era began
in the treatment of infectious diseases. At the end of
1937, a US company financed a massive advertising
campaign to launch the 'Elixir of Sulfanilamide' -
whose main defect was its formulation in a solvent
containing 72% diethylene glycol [1]. The death
toll was over 100 and many of the victims were
children.
This health disaster led to important changes in the
US law and, as early as 1938, the Federal Food, Drug
and Cosmetic Act was passed. This required a certain
amount of toxicological testing for drugs. According
to the same law, the Food and Drug Administration
(created in 1927 under the initial name of Food,
Drug, and Insecticide Administration and shortened
to FDA 3 years later) was given the power of seizing
or withdrawing from the market products that
proved to be toxic. However, it was not until 1962
before the FDA also insisted that drugs should be
effective before their introduction to the market.
Among other safety incidents that impacted
significantly on the organisation of pharmacovigilance,
the most important are probably the
thalidomide disaster (1959-61) [2,3] and the practolol
story (1974-76) [4,5]. The former is too well known to
require thorough repetition but it may be useful to recall that, besides the brutal recognition of a fact
to which little attention had been paid until then -
namely that chemicals might have a toxic potential
in the development of the human foetus - in the
UK it also resulted in the passing of the Medicines
Act (1968). This embraced all aspects of the control
of medicines regarding safety, quality and efficacy.
Directive 65/65/CE - the first European Directive on
pharmaceutical products - was also a reaction to this
incident.
In contrast with the above-mentioned 'Elixir of
Sulfanilamide' story, which really concerned a 'miracle'
drug, the thalidomide tragedy strikes us by its horrific
contrast between the fairly trivial expected benefit
(relief of nausea/vomiting in early pregnancy) and the
atrocity of its safety consequences (phocomelia -
i.e. defect of long bones in the limbs - in babies
exposed in utero). This introduces us to the pivotal
issue of the benefit:risk ratio.
The story of practolol is less well known as the
drug, a recent chemical entity from ICI, was not
then registered in a number of countries outside
the UK. Essentially, this beta-blocker induced an
'oculo-muco-cutaneous syndrome' under the form
of a fibrous transformation of mucosal tissues,
resulting in particular in corneal perforations with
subsequent blindness and, even worse, in potentially
fatal plastic peritonitis. Yet the significant feature
of this situation for the history of drug monitoring
was that practolol - the first cardioselective beta-blocker
- was developed and advertised as a marked
improvement in drug safety, ironically exemplifying
that drug monitoring is, par excellence, the field of
unexpected events.
Together, these introductory stories summarise a
large part of the historical developments in drug
monitoring. For example, the need for a strong
guarantee in the process of manufacturing is
unfortunately not an archaic concern in a time
where counterfeiting in medicines has become a
major commercial and regulatory issue. Likewise, the
question of a disproportionate benefit:risk ratio for a new chemical entity regularly fuels controversies
in modern pharmacy [6,7]. Again, the recent Vioxx story
illustrates that the risk of hidden toxicity is still
present for any drug first promoted as an advance
in patient safety [8].
The last prototype example in the history of drug
safety is probably that of benoxaprofen [9-12], a new
anti-inflammatory drug whose sales exploded as
soon as it was introduced on the market, in the early
1980s. Within a few months, dozens of patients died
of liver failure as a result of insufficient warning
about a marked reduction in hepatic clearance of
the product in the elderly. Although the cause of
this 'unexpected' toxicity was perfectly known and,
most probably, was quite easy to manage with
appropriate recommendations for prescribing, the
manufacturer chose a blunt withdrawal of the drug.
This example highlights the possible threat hanging
over every potential blockbuster as a consequence
of a scale change in exposure: a toxicity unnoticed (if
not undetectable) in the small populations exposed
to a new chemical entity during its development
may quickly become a major health problem if its
promotion is too successful.
Responsibilities of the various stakeholders
in pharmacovigilance
Although one may wonder whether the
distinction has any relevance, modern regulations
and procedures have consistently distinguished
between the assessment of drug safety before the
marketing authorisation and the post-marketing
monitoring of drugs. The latter is usually referred to
as 'pharmacovigilance' (a Greek-rooted neologism
of dubious value - it means exactly the same as the
term 'drug monitoring').
The respective responsibilities of the various
stakeholders in pharmacovigilance have been defined
by a number of texts, in particular Regulation (EC)
No 726/2004 and Directive 2001/83/EC as amended
by Directive 2004/27/EC.
In order to ensure the adoption of appropriate
regulatory decisions regarding medicinal products
within the European Community, the member states
must take all appropriate measures to encourage
health professionals to report suspected adverse
reactions to the competent authorities. They operate
a pharmacovigilance system to collect and assess
information likely to impact the benefit:risk ratio of
drugs, such as adverse reactions, misuse or abuse. In
collaboration with member states and the European
Commission, the central agency (EMEA) runs a data-processing
network to facilitate the exchange of
information and to allow all competent authorities
to share the information at the same time.
The marketing authorisation holder has the duty
of continuously ensuring post-marketing surveillance
of its products and must have an appropriately
qualified person at his disposal. The holder must
record, report and keep a database of all suspected
adverse reactions brought to his attention either
directly (by the reporting of a health professional) or
indirectly (e.g. by the medical literature). The basis
for reporting may be:
The marketing authorisation holder may also
undertake post-authorisation studies, e.g. in order to
assess more accurately the safety of its product in
varied situations (such as specific sub-populations).
Finally, the holder must ensure that any request
from the regulatory authorities is answered fully and
promptly. A number of guidelines are available for
expedited reporting, PSURs and company-sponsored
post-authorisation studies.
Due to their financial cost, their expected
duration and their logistic burden, safety post-authorisation
studies are more the exception than
the rule. Therefore, it is clear that the main source
of pharmacovigilance data remains spontaneous
reporting. This gives a pivotal responsibility to health professionals and, more specifically, to physicians and
nurses who are often in the best position to witness
suspected drug reactions. In the imaging world,
radiologists, cardiologists, nuclear medicine physicians
and other health professionals using imaging agents
have the professional duty of reporting significant
reactions of which they may be aware, even if such
reactions are already recognised as expected hazards.
This is the main way to catch a potential alert signal.
In some countries (such as France), the law may
even include a mandatory requirement for health
professionals to report serious or unexpected
reactions. However, this may be regarded as a
provision of dubious relevance, as it is difficult to
conceive any form of sanction in the case of non-reporting.
The report should be made to the 'Centre
Régional de Pharmacovigilance' in France and to
similar official collecting bodies in other European
countries. It is useful to inform the manufacturer at
the same time.
Under-reporting
Under-reporting is the main limitation of
pharmacovigilance based upon spontaneous reporting
as illustrated by the following story: during the 1960s,
it was estimated in the UK that nearly 3000 young
asthmatics had died from excessive use of pressurised
aerosols containing bronchodilating agents, whereas
only six doctors in that country had reported a
potential association between overuse of aerosols and
sudden death [13]. Even today it is estimated that, at
best, the rate of reporting does not exceed one report
out of 10 actual drug reactions and may be far below
this figure (<1/100) in a number of instances.
Reasons for under-reporting are many and include
overwork, inadvertence or negligence, lack of
recognition of the iatrogenic nature of a disease
(particularly if there is a long time interval between
drug intake and the reaction), reluctance to report
suspicions without a definite proof of cause, and
fear of litigation. Inasmuch as the absolute number
of reports corresponds only to a small minority of
the actual cases, the frequencies of adverse events
calculated on the basis of spontaneous reporting
are often open to massive biases: for example, an
individual physician interested in (or obsessed by)
a particular reaction can easily make a significant
difference to the data generated. Possibly induced by
vested interests or manipulated by pharmaceutical
promotion, such biases may account for selective
reporting of reactions related to one product even if it corresponds to a class effect, thereby contributing
to a distorted perception of a higher occurrence for a
specific product in the class.
The causal assessment of a drug-related reaction
relies, above all, on an exclusion diagnosis (the
disease under consideration was not caused by a
non-drug 'natural' aetiology such as an infection,
a vascular or an auto-immune disease, etc.). It is
of paramount importance to understand that the
performance of spontaneous reporting in recognising
actual iatrogenic complications decreases in proportion
to the 'background noise' - that is, the frequency of
similar clinical pictures unrelated to the drug under
consideration. Markedly reduced as it is by under-reporting,
the observed frequency of a drug-induced
disease may be quite easily obscured in the expected
frequency of the same disease when it is not caused
by the medicinal product. This is generally the case
with complications such as hepatitis, bleeding, stroke,
phlebitis, renal failure and depression as their expected
number independently of any drug exposure is already
high, particularly in some subpopulations such as
geriatric patients. As a typical example, autism, the
reported hazard of MMR vaccines is not unusual in the
general, non-vaccinated paediatric population [14].
So, inasmuch as under-reporting generally conceals
the actual fluctuations of frequency in an iatrogenic
disease, the only situation where spontaneous
reporting may quickly have a clear significance
in terms of drug-induced causation is when the
baseline incidence of the reported disease is
extremely low. This was the case with the above-mentioned
phocomelias after thalidomide or with
vaginal adenocarcinomas in young females after
diethylstilbestrol.
Investigation of an unexpected drug safety
problem
In its principle of continuous surveillance over the
whole life of a medicinal product, pharmacovigilance
obviously extends beyond the scope of unexpected
reactions. For example, there is probably much room
for improvement in the assessment and practical
management of well recognised drug hazards
such as those related to iodinated contrast media.
Although these adverse reactions are comprehensively
discussed in pharmacological textbooks, reporting
by practitioners (radiologists, in particular) is still
topical and may contribute, for example, to a better
recognition of potential risk factors or to optimisation
of their medical management.
This notwithstanding, early recognition of as
yet unrecognised drug reaction is obviously the
major justification for continuous post-marketing
surveillance.
Thus, when a new disease is reported as potentially
ascribable to a drug or a class of drugs, what should
be the intellectual framework to interpret any
available data? Essentially, analysis of the reporting
input will rest on the criteria of:
Credibility
New as it seems, the new disease must be
precisely characterised. It must also comply with a
minimum of medical or pharmacological consistency:
a cancer, for example, cannot occur within a few days
after drug exposure. Further, the reporters (as well as
any opinion leaders commenting on the situation)
must also have at least a minimum of credibility and
must not be suspected of vested interests in ascribing
specific toxicities to the competing drugs of their
sponsors.
Causality
Regarding causation, Hill's criteria [15] are classical
and may be found in any textbook of epidemiology.
However, an important issue in pharmacovigilance
is often that, although such criteria may not be
fulfilled, clinical or regulatory decisions must be
taken in practice without undue delay.
In pharmacy, an important issue is whether a
new hazard is ascribable to a specific drug or to a
pharmacological/therapeutic class. Likewise, when a
new pathological entity is described, another question
is that of the confounding factors: are there any
changes in environment that could account for or
contribute to this new disease? For example, one may
wonder whether a condition such as nephrogenic
systemic fibrosis (NSF) may be influenced by any
kind of recent change in the exposure of dialysed
patients. This condition, first diagnosed in 1997 [16],
occurs in patients with advanced renal disease, is
mainly characterised by skin thickening (and far more
rarely, may involve lungs, liver, heart or muscles) and
may be associated with previous use of gadolinium-containing
magnetic resonance imaging (MRI)
contrast agents [17].
Likewise, the question of biological plausibility is a
tricky one: the history of therapeutics is full of products
which proved to be devoid of the benefits that should
have existed on the basis of theoretical arguments of
pharmacological plausibility, whereas other drugs were
withdrawn from the market because of their indubitable
but pharmacologically unexplained toxicity. To sum up,
the message is quite clear: it is not because it appears
biologically plausible that a drug hazard exists, and,
conversely, it is not because it is unexplainable on a
pathophysiological basis that it cannot exist.
Benefit:risk ratio
The re-assessment of the benefit:risk ratio of a
drug includes an evaluation of the severity and
the irreversibility/lethality of its hazards, as well as
the definition of the populations at risk and of any
therapeutic alternatives if they exist. Actually, a severe
and unexpected toxicity may prove to be acceptable
if there is no alternative in a serious indication.
Course of action
Finally, the thorough assessment of a new iatrogenic
hazard presupposes that each of the concerned
stakeholders complied with its duties. Namely, the
health professional reported their suspicions, the
manufacturer(s) transmitted without delay and with
all due details the cases that were brought to their
attention. Meanwhile the regulatory authorities took
all appropriate measures and elaborated relevant
recommendations to reinforce patient safety and
ensure that the benefit:risk ratio remained medically
and ethically acceptable.
Conclusion
The preceding conceptual framework is certainly
appropriate to the assessment of a new entity such
as the cases of nephrogenic systemic fibrosis (NSF)
recently associated with gadolinium-based contrast
agents for MRI [16,17]. Of paramount relevance, in
particular, is the need to characterise as precisely
as possible the clinical, para-clinical and laboratory
features related to this new entity, the potential risk
factors (Renal failure? Other pathological co-factors?
Medicinal treatments?), as well as the identification
of products likely to be involved.
In spite of the complexity of the issue, it is important
to emphasize that the rarity as well as the novelty
of the disease fulfil the criteria that maximise the
value of spontaneous reporting in pharmacovigilance,
since the 'background noise' for such a clinical entity
is near zero. This means that every spontaneous
report, so long as it is precisely documented, is
likely to represent a significant contribution to the
investigation, the recognition and the practical
management of this intriguing new disease.
Conflict of interest: Dr Girard works as an independent
consultant for the pharmaceutical industry, including
GE Healthcare.
- From its inception through to the present day, the history of therapeutics has been marked by unexpected reactions to medicines, some of them accounting for health disasters
- Regulatory authorities, manufacturers and health professionals are the main stakeholders in the organisation of post-marketing surveillance of drugs (pharmacovigilance)
- Although much pharmacovigilance relies on spontaneous reporting from health professionals, the system may be skewed by under-reporting
- When a new disease emerges as potentially ascribable to a drug or a class of drugs, its investigation should meet the criteria of credibility, causality, benefit:risk assessment, and course of action
- For hazards such as nephrogenic systemic fibrosis (NSF), whose spontaneous frequency is negligible, spontaneous reporting may be an efficient tool of investigation
References
- Gelling E, Cannon P. Pathological effects of Elixir of Sulfanilamide (diethylene glycol) poisoning. JAMA 1938;111:919-96.
- McBride W. Thalidomide and congenital abnormalities. Lancet 1961;ii:1358.
- von Lenz W. Die Thalidomide embryopathy. Dtsch Med Wochenschr 1961;86:2555.
- Girard M. Testing the methods of assessment for adverse drug reactions. Adverse Drug React Acute Poisoning Rev 1984;4:237-42.
- Nicholls JT. Adverse effects of practolol. Ann Clin Res 1976;8:229-31.
- Moynihan R, Cassels A. Selling sickness. How drug companies are turning us all into patients. Crows Nest, Australia: Allen & Unwin, 2005.
- Moynihan R, Smith R. Too much medicine? BMJ 2002;324:859-60.
- Graham DJ. COX-2 inhibitors, other NSAIDs, and cardiovascular risk: the seduction of common sense. JAMA 2006;296:1653-6.
- Lessons from the benoxaprofen affair. Lancet 1982;2:529-30.
- Lessons from benoxaprofen. Lancet 1982;2:667.
- Brooks PM, Birkett D. Release of new drugs. The lessons of benoxaprofen. Med J Aust 1983;1:251-2.
- Naude SP, Anderson R. Benoxaprofen - could the side-effects have been prevented? S Afr Med J 1982;62:427.
- Inman WHW. Detection and investigation of adverse drug reactions. In: Davies DM, ed. Textbook of adverse drug reactions, 3rd edition. Oxford, UK: Oxford University Press, 1989: pp. 49-62.
- Taylor B. Vaccines and the changing epidemiology of autism. Child Care Health Dev 2006;32:511-9.
- Hill AB. The environment and disease: association or causation? Proceedings of the Royal Society of Medicine 1965;58:295-300.
- Kuo PH, Kanal E, Abu-Alfa AK, et al. Gadolinium-based MR contrast agents and nephrogenic systemic fibrosis. Radiology 2007;242:647-9.
- Grobner T, Prischl FC. Gadolinium and nephrogenic systemic fibrosis. Kidney Int 2007;72:260-4. Epub 2007 May 16.
12-2007 BUY1159388/JB2916/MB002521/CMC INT'L ENG 15th edition
Evidence-based medicine and clinical trials - from a clinical trials unit perspective
