In April 2003,[1] the UK Committee on the Safety
of Medicines set up an expert working group to consider the
safety of selective serotonin reuptake inhibitors (SSRIs) and
in December 2003 declared that the expert group '...has now
studied all available evidence and found that the risks of
treating depressive illness in under 18s with certain SSRIs
outweigh the benefits of treatment' (Department of Health
2003). Also quoted was the chair of the UK's Medicines and
Healthcare Regulatory Agency (MHRA), who said that 'The
majority of SSRIs...- the most commonly prescribed type of
anti-depressants - are not suitable to be used by under 18s'.
The regulatory language of the MHRA advice was intended as a
'strong signal', while falling short of an outright ban;
however it has been treated as an effective ban by many
practitioners and trusts around the country, and widely
interpreted as such in the media. Related action by the US
Food and Drugs Administration (FDA) based on the same evidence
took a less assertive line, issuing 'supplementary warnings'
only. These announcements have initiated a period of heated,
often partisan debate and media attention, intensified by
strongly held positions in relation to drug companies, and
some confusion in clinicians about their right course of
action. The MHRA has published abstracts of the evidence base
on which their decision was based and the FDA has minuted
further documentation of ongoing review work.[2]
Further academic reviews of published and unpublished
studies[3,4] have put intensive focus on the
evidence base, raising key issues about the internal and
external validity of the trials undertaken, the nature of data
from unpublished trials and the reporting of adverse
events.
At the time of writing (April 2004), the issues are far
from settled, although the more work that is undertaken, the
more questionable seems the evidence base on which inferences
about safety and effectiveness of SSRIs in young people have
been made. Whatever the final outcome, the furore has raised
welcome and long overdue questions about the conduct of drug
trials and the evidence base for child and adolescent
psychiatry psychopharmacology as a whole. Highlighting and
addressing these issues could be a real value to future
practice in the field.
The medicine of childhood (including child psychiatry and
paediatrics) has long suffered an anomalous position in
relation to prescribing. The general absence of licensing for
most medications in under 18s has led child and adolescent
specialists to extrapolate from adult data in treating their
patients. In child psychiatry there has been
characteristically a period of 3 or 4 years while medications
introduced into adult practice trickle down into child and
adolescent practice - often in an idiosyncratic and patchy
fashion and dependent on local and national peer consensus.
Given the considerable increase in general Child and
Adolescent Mental Health Services (CAMHS) prescribing over the
last decade and an increasingly litigious environment, it is
not surprising that this working arrangement (unsatisfactory
as it always was) has come under scrutiny.
The fundamental problem is the lack of good quality and
purposefully designed psychopharmacological research in
childhood and adolescence. Detailed evidence to the FDA
specialist advisory committee[5] confirms that the
FDA actually issued invitations to drug companies to undertake
the studies in childhood and adolescence, which form the basis
of the current debate, as a condition of conferring continuing
licensing exclusivity with this age group. This may have been
their motivation, but major questions have since been raised
about the internal and external validity of the resulting
studies. Regarding adverse effects, it was only when patient
level rather than summary data from these trials were
disclosed that the FDA was alerted to increased levels of
'suicidality' reported by patients during the trials. This had
not been apparent in the summary data because the studies had
classified the phenomena as 'emotional lability', a term which
did not signal increased risk.[5] Quite apart from
this apparent reporting bias, recent reviews have highlighted
the poor quality of the basic research design that elicited
these unwanted effects in the first place.[2]
Because of these flaws, Laughren[5] points out that
it might equally be possible to make a reciprocal error of
overestimating the signal of risk from these data. No
completed suicide has been reported in studies on a total of
1717 patients[6]; the actual symptoms described
cover a broad range of more or less specific phenomena from
agitation, arousal, depressive thinking, thoughts of
self-harm, to active suicidal ideation and self-harming
behaviour. These are important symptoms and include an
'activation syndrome' with SSRIs, which has been a common
observation in the clinical literature for many years. But it
is still an open question as to whether all these symptoms
necessarily signal increased 'suicidality'. The
FDA[2] has commissioned an extensive reevaluation
of the data in relation to adverse effects, to see to what
extent the phenomena initially reported as emotional lability
need reclassifying, whilst acknowledging that the original
data are often so poor that even this exercise is likely to be
inconclusive, a conclusion also reached by Whittington et
al..[4]
However, it was not just the presence of apparent adverse
effects that led to the MHRA guidance, but the presence of
these in the context of a relative absence of evidence of
effectiveness for all the drugs in the SSRI class bar
fluoxetine. Both published and unpublished studies have been
subject to searching reviews in this regard. One
review[3] is highly sceptical of the internal
validity and presents evidence to suggest that the authors
(whom they note were funded by pharmaceutical companies)
exaggerated the positive effects in the studies by selective
quoting of the data, post hoc revision of primary outcomes and
failure to report serious adverse effects. Some of these
serious criticisms are justified by detailed reexamination of
primary data, but in this partisan area one has to note that
the senior authors themselves declare conflicts of interest as
members of an international organization 'aiming to reduce
harm from misleading drug promotion'. A helpful
metaanalysis[4] compared data from published and
unpublished studies and showed evidence of publication bias in
favour of positive results: data indicating effectiveness of a
number of SSRIs in the published studies disappeared when
unpublished data were added. The authors concluded that the
evidence base is not methodologically sufficient to make
conclusions regarding adverse reactions and suicidality, but
that in the absence of evidence for efficacy the precautionary
principle should discourage their use.
While this evidence compromising claims of effectiveness
seems compelling, Whittington et al. do not address
remaining concerns about the possible lack of external
validity in these studies. The sampling frame was poorly
characterized and may have excluded many children and
adolescents who would have been prescribed these drugs in the
UK (e.g. those with more complex problems). Data were gathered
over a huge developmental age range (5-18 years) with no
purposive stratification in relation to age or severity, both
likely predictors of outcome. The upper age criteria were
based on administrative factors in contracting rather than
testing any biologically plausible hypothesis about
age-related responses. Only five studies in childhood across
the whole of the SSRI group meet the methodological inclusion
criteria for Whittington et al.'s review. Given the
modest effect sizes in adult SSRI studies,
Laughren[5] argues that there would be a 70%
probability of such a small number of studies showing negative
results even if SSRIs in childhood actually had the same
effectiveness as in adulthood.
Against such an uncertain evidence base, the use of the
precautionary principle, while perhaps reasonable for interim
action pending further data, may provide a poor basis for
definitive judgement, and can throw up paradoxical and
unintended consequences.[7] Without properly
designed studies, we run the risk of throwing the baby out
with the bathwater, of overlooking clinical situations and
patient subgroups in which some of these medications may well
be valuable (as much clinical experience would suggest) and
depriving clinicians of treatments for some of their most ill
patients. While the clinical decision-making regarding these
drugs in CAMHS previously relied on extrapolation from
multiple adult studies, this new recommendation is made on the
basis of a small number of methodologically unsatisfactory
studies in the under 18s.
What are the Lessons We Can Learn From This?
The first lesson is the pressing need for properly designed
and powered pragmatic studies of medications in children and
adolescents, which could test the concerns raised by the MHRA
and FDA data. Any such studies would need to be purposively
designed around age and severity stratification (clinical
experience would suggest that it is likely that over 15s with
severe disorders would be more likely to benefit from
antidepressants), and include sophisticated measurement
techniques that would allow contextualization of the unwanted
effects and other mental state phenomena reported. Such trials
would arguably best involve independent academic institutions
rather than being conducted solely by the pharmaceutical
industry. Unfortunately, there is little tradition of the
independent funding of such studies in the UK, which surely is
no longer a viable position to be in.
It would be very unfortunate if the debate around this
issue polarized and resulted in pharmaceutical companies
withdrawing from this area of activity. A variety of
medications are increasingly used in this age group, mostly
off label. Surely we need a partnership approach which
encourages responsible pharmacological research in children
and adolescents. We should no longer have to rely on dubious
extrapolations from adult practice, but generate our own
research base. Not to do so risks depriving children of the
potential benefits of scientific medicine. We may be able to
look to new European legislation requiring drug companies to
undertake this kind of research.
Given the concern about non-reporting of negative results
in these studies (a concern of course which is not confined to
drug trials alone), there would be an argument for allocation
of central funds to conduct replications of drug company
trials. Furthermore, trial methodology should include phase II
dosage-finding as well as primary studies. None was undertaken
in the trials reviewed by the MHRA.
Also highlighted is the anomaly whereby research undertaken
by drug companies can be withheld from public view on the
basis of commercial confidentiality. Potentially this means
that both serious adverse effects and potential benefits of
drugs are not disclosed to the wider community and that bodies
such as the National Institute of Clinical Excellence, who are
attempting to evaluate the evidence base, are working on
partial or corrupted information.[4] There can
surely be no ethical justification for this. Suggested
solutions have included the mandatory registration of all
trials undertaken or a condition made of ethical approval that
the results should enter the public scientific domain. The
scientific community is moving towards greater systemization
of trial methodologies and the inclusion of registration for
drug company trials could allow methodology and results to be
open to public view.
There is a complex issue of evaluating what is significant
harm. In effectiveness trials we are now used to thinking in
terms of 'numbers needed to treat' or clinically relevant
effect sizes rather than simple measurement of change as the
most relevant endpoint analysis. Perhaps similar steps need
considering in relation to the assessment of risk, with some
metric for what constitutes 'clinically significant harm'.
Otherwise, an increasingly litigious culture may make us set
the threshold criteria for risk in a rather different way to
those for effectiveness.[7] In the SSRI case, what
exactly constitutes the risk is open to debate - to what
extent are the phenomena intrinsic to the disorder, a
manifestation of increased arousal as a well recognized
unwanted effect of the drugs, or true suicidality? It would
equally be quite wrong, however, to minimize the concerning
findings in these studies. I suggest the correct response is
for more carefully designed and fit for purpose studies to be
carried out in the target clinical populations to give
practitioners a good evidence base for their practice. It is
time for a step change in the way we go about funding this
central area of our work and for child and adolescent
psychiatry to cease to be a poor relation in terms of
psychopharmacology treatment trials.
