Celebrity's death from cancer resulted in increased calls to CancerBACUP

Article (PDF Available)inBMJ Clinical Research 317(7164):1016 · November 1998with35 Reads
DOI: 10.1136/bmj.317.7164.1016 · Source: PubMed
The Swiss heroin trials
Trial is needed comparing
decriminalisation of heroin with existing
policy of prohibition
Farrell and Hall seem to have mis-
understood the importance of the Swiss
trials of heroin on prescription for addicts.
The call for a clinical trial of heroin versus
methadone is irrelevant as these drugs cater
for different segments of the addict popula-
tion; no one suggests stopping methadone
clinics. It is self evident that prescribing
heroin will attract addicts who need the
“buzz” and will not switch to methadone.
These include dealers and pushers and
those who succeed in obtaining funds
through crime. Methadone clinics attract
newer rather than hard core addicts. A logi-
cal policy for decriminalising heroin under
medical supervision would have four steps:
giving prescriptions of heroin to all addicts
in or out of prison (which would gradually
put criminals out of business); providing
methadone clinics for those who will switch;
weaning the addicts off the drugs; and
providing a follow up programme to
minimise relapse. The trial that is needed
would compare a city region or country
adopting this approach with a similar
community continuing the existing policy of
prohibition. This policy has already failed
for the same reason that prohibition failed
in the United States: it created an oppor-
tunity for the criminal mafias who dominate
the drug scene. The end points of a
comparative trial should not be narrowly
defined as conceived by Farrell and Hall;
they should include the numbers of new
addicts, mortality and morbidity among
addicts and former addicts, the impact on
spread of HIV infection and hepatitis B both
inside and outside prisons, and statistics for
drug related crime (allegedly reduced by
60% in the Swiss trials). The economic gain
to the community from heroin clinics will
include the street price forgone by the
clinics’ clients, which would otherwise be
stolen from members of the community.
This is a massive gain over and above the
similar gain from methadone clinics. The
time has come for the medical management
of heroin addicts to be submitted to the dis-
ciplines of clinical pharmacology and epide-
miology, including, ideally, randomised con-
trolled trials.
Apart from the impact on problems
caused by hard drugs, the new approach will
be essential for resolving issues surrounding
soft drugs. Marijuana is safer than alcohol or
tobacco, but legalisation is inhibited by the
fear that pushers of hard drugs can recruit
G R Venning Consultant in pharmaceutical medicine
Pharmaceutical Research Services, High Wycombe,
Buckinghamshire HP13 6QG
1 Farrell M, Hall W. The Swiss heroin trials: testing
alternative approaches. BMJ 1998;316:639. (28 February.)
Further studies of heroin treatment are
On the basis of a trial in Switzer-
Farrell and Hall conclude that
medically prescribed heroin is likely to have
only a limited role in the future.
It is hard to
know how any judgment can be made about
this approach after only one randomised
control trial,
one case-control study,
the large but uncontrolled study that the
authors commented on.
Farrell and Hall are right to emphasise
the difficulty of disentangling the contribu-
tion of comprehensive social and psycho-
logical inventions in the impressive results
obtained in this population.
After more
than three decades of research evaluating
the effectiveness of methadone maintenance
treatment it is safe to conclude that such
interventions improve outcomes but diffi-
cult to be much more precise than this.
Social and psychological interventions
probably improve results of virtually all
interactions between patients and the
healthcare system.
The authors say that the proposal to
conduct a heroin trial in Australia damaged
support for harm reduction. This is argu-
able. The mood in Australia has generally
been more conservative over the past two
years. The controversy over the heroin trial
in Australia probably contributed to major
improvements in a local methadone pro-
gramme and certainly led directly to a com-
mitment and funding for research on an
expanded range of pharmacological treat-
ments for heroin dependence.
The authors question whether Australia
is in the middle of a national heroin crisis. If
a sixfold increase in mortality from drug
overdose during the past 16 years
is not a
crisis it is hard to know what would be.
The Swiss heroin trial may have been
more expensive than routine prescription
methadone. But trials of new treatment are
invariably more costly than providing well
established treatments as a routine. Heroin
and methadone are both cheap to produce.
The cost of methadone represents under 5%
of the costs of methadone maintenance
treatment. There is no good reason to
believe that heroin treatment will be consid-
erably more expensive than methadone
Most important, Farrell and Hall
endorse studies that will determine the
comparative usefulness and cost effective-
ness of injectable heroin and methadone
maintenance treatment. The 71% support
for heroin maintenance shown in a referen-
dum in Switzerland in September 1997
(with majorities in all 26 cantons) suggests
that the Swiss were not confused about these
Rigorously conducted trials of medically
prescribed heroin must be conducted soon
lest we repeat the history of methadone
maintenance treatment: the failure to con-
duct such studies when they were still pos-
sible proved costly.
Alex Wodak Director
Alcohol and Drug Service, St Vincent’s Hospital,
Sydney, NSW 2010, Australia
1 Uchtenhagen A, Gutzwiller F, Dobler-Mikola A, eds.
Programme for a medical prescription of narcotics: final report of
the research representatives. Summary of the synthesis report.
Zurich: University of Zurich, 1997.
2 Farrell M, Hall W. The Swiss heroin trials: testing
alternative approaches. BMJ 1998;316:639. (28 February.)
3 Hartnoll R, Mitcheson M, Battersby A, Brown G, Ellis M,
Fleming P, et al. Evaluation of heroin maintenance in con-
trolled trial. Arch Gen Psychiatry 1980;37:877-84.
4 McCusker C, Davies M. Prescribing drug of choice to illicit
heroin users: the experience of a UK community drug
team. J Subst Abuse Treatment 1996;13:521-31.
5 Hall W, Darke S. Trends in opiate overdose deaths in Australia
1979-1995. Sydney: University of New South Wales, 1997.
(NDARC technical report.)
Authors’ reply
We used the Swiss trial of prescrib-
ing heroin to counter the unrealistic
expectations of its impact exemplified by
Venning, who asserts that heroin prescrib-
ing will eliminate the black market, drug
related crime, and the recruitment of new
users of heroin.
Venning’s letter is lacking in supporting
evidence, relying on appeals to “self evident”
facts and “lessons” from the history of alco-
hol prohibition that are uninformed by
recent scholarship (for example, the article
by Tyrell).
A confusion between heroin pre-
scribing as a therapeutic option of last resort
and the repeal of prohibition of heroin
one of the reasons why Australia did not
proceed with a trial of heroin prescribing.
Wodak contests our assertion that the
Swiss model of on-site heroin prescribing is
likely to be an expensive, minority treatment
option. It is costly to ensure that heroin is
not diverted between manufacture and
Website: www.bmj.com
Email: letters@bmj.com
1011BMJ VOLUME 317 10 OCTOBER 1998 www.bmj.com
administration, and for staff clinics to
provide extended hours to supervise self
treatment with heroin. The costs are
calculated to be between five and 10 times
those of oral methadone treatment. The
high level of supervision that was necessary
to address public anxieties about heroin
prescribing and the risks of diversion was an
interesting aspect of the trial.
The increased rate of deaths in people
taking heroin in Australia certainly presents
a major public health problem, but we doubt
that “crisis” is the best way to describe it. The
increase has occurred over two decades, and
most deaths are now among people who
started using heroin a decade or more ago.
“Heroin crises, like “drug wars, tend to
prompt ill considered and disproportionate
policy responses.
The relation between the debate over a
trial of heroin and the trial of alternative
pharmacotherapies in Australia is more
complex than Wodak suggests. A large trial
of buprenorphine predated the debate over
trials of heroin, as did plans for trials of
other agents. The failure to proceed with the
heroin trial did produce a commitment to
partially fund these other trials, which would
not have happened if the trial had
proceeded. But the only trials that have been
funded to date (with A$1.4m (£518 500))
have been three small trials of naltrexone
maintenance (after accelerated withdrawal
under general anaesthesia), a treatment of
uncertain efficacy whose prominence in
Australia has owed much to the debate
about a heroin trial.
Wayne Hall Executive director
National Drug and Alcohol Research Centre,
Sydney, Australia
Michael Farrell Senior lecturer
National Addiction Centre, Institute of Psychiatry,
London SE5 8AF
1 Tyrell I. The US prohibition experiment: myth, history and
implications. Addiction 1997;92:1405-9.
2 Farrell M, Strang J. Confusion between the drug
legalisation and drug prescribing debate.Aust Drug Alcohol
Rev 1990;9:364-8.
Recovered memories of
childhood abuse
We must tell patients that they were not
to blame
Pope welcomes the recent report
from the Royal College of Psychiatrists,
which counsels caution when treating
patients with “recovered memories” of
sexual abuse in childhood.
This report may
encourage disbelief of patients presenting
with any memory of childhood abuse.
Child sexual abuse is common in the
United Kingdom.
It is associated with long
term physical and psychological sequelae.
Its victims are reluctant to disclose their his-
tories to doctors: only one in 20 will try to do
They wait an average of 17 years after
the abuse has occurred, and expect to be
disbelieved and blamed.
Since 1896, when Freud radically altered
his opinions and pronounced that his
patients’ stories of incestuous abuse were the
stuff of fantasy, society has found it difficult
to confront the reality of child abuse. There
is a national organisation that represents
people who allege that they have been
falsely accused of perpetrating abuse, but
there is no one body to which adult
survivors can turn for help and advice.
Should any of our patients be desperate
enough to entrust us with their stories of
childhood sexual abuse then the least we can
do is to believe them and tell them that they
were not to blame.
J K Ilsley General practitioner
Nunwell Surgery, Bromyard, Herefordshire
1 Pope HG Jr. Recovered memories of childhood sexual
abuse. BMJ 1998;316:488-9. (14 February.)
2 Baker A, Duncan SP. Child sexual abuse: a study of preva-
lence in Great Britain. Child Abuse & Neglect 1985;9:
3 Arnold PR, Rogers D, Cook DAG. Medical problems of
adults who were sexually abused in childhood. BMJ
4 Lechner ME, Vogel ME, Garcia-Shelton LM, Leichter JL,
Steibel KR. Self-reported medical problems of adult
female survivors of childhood sexual abuse. J Fam Pract
5 Kendall-Tackett MA, Simon AF. Perpetrators and their
acts: data from 365 adults molested as children.Child Abuse
& Neglect 1987;11:237-45.
People with memories of abuse must be
given reassurance
I agree with the recommendations
by the Royal College of Psychiatrists
reported in Pope’s editorial.
One problem
with memories of sexual abuse in childhood
is the absence of empirical evidence. In cer-
tain cases the “recovered” memories have
been shown to be based on hysteria, fantasy,
malicious invention, or lies. These cases,
however, are surely the exception that
proves the rule.
There is an epidemic of child sexual
abuse in the world today. In the United
States, several million cases of child abuse
and neglect are “indicated” each year, and
this figure continues to climb. These cases
are backed up by the authority of the
government entities entrusted by us all to
investigate child abuse.
Such a widespread phenomenon could
not have simply sprung into existence in the
past 10 or 20 years, alongside the survey
instruments that now prove its existence.
Childhood sexual abuse is far older
perhaps as old as the hills themselves. What
is new is society’s willingness to look at the
problem; the masses of adults waking up to
the terrible reality that violence was perpe-
trated against them at the most vulnerable
time in their lives; the millions of adults
throughout the world speaking out on their
own behalf and for the children of today
who are being abused in equal or greater
numbers, yet continue to be silenced or
ignored. There may be disputes over
methodology during the initial phase of
studying the problem, but that must not dis-
tract us from marshalling all necessary
resources for the prevention, treatment, and
punishment of all cases of abuse.
Traumatic memory occurs on many dif-
ferent levels. The verbal component is what
we are certain of in a rational sense
this, said that, saw them, etc. Trauma is hard
to explain. however, because it overwhelms
the senses. Survivors of a traumatic experi-
ence may fragment their memories into
various components
the touch, the smell,
the sight, and the sound. Once a memory is
fragmented, it must be even harder to put it
in words.
Many of us who have survived pro-
foundly traumatic experiences spend the
rest of our lives piecing all those fragments
only to be attacked by misguided
psychiatrists and researchers as people with
“false memories. It is a dark day for the psy-
chiatric profession when a consensus seems
to have been constructed against the telling
of stories by survivors of violence. If we are
ever to turn back the tide against the abuse
of children we must begin by listening to
these stories
and giving the story tellers the
benefit of the doubt.
Carl-David Birman Director
Safety, New York, NY 11218-4315, USA
1 Pope HG Jr. Recovered memories of childhood sexual
abuse. BMJ 1998;316:488-9. (14 February.)
Author did not declare his affiliation
Pope’s editorial on recovered
memories of sexual abuse in childhood
not inform readers of his bias in the area of
dissociative amnesia. He is a board member
of the False Memory Syndrome Foundation,
an organisation that advocates for people
accused of sexually abusing children.
I have experienced dissociative amnesia
and recovered memories of sexual abuse in
childhood, and I found enough corrobora-
tion to meet the necessary legal standard of
evidence to prove my charges in a landmark
civil suit. My father (who at trial admitted to
fondling another child) and his second wife
have reported paying membership dues to
the False Memory Syndrome Foundation.
Details of other cases of recovered
memories of childhood abuse which have
found strong corroboration can be found on
the following website:
Jennifer Hoult Professional musician
New York, NY 10025, USA
1 Pope HG Jr. Recovered memories of childhood sexual
abuse. BMJ 1998;316:488-9. (14 February.)
Some memories may be false, but some
Pope’s editorial on the Royal Col-
lege of Psychiatrists’ recommendation on
how to treat people with recovered memories
of childhood sexual abuse
conformed to the
views of the false memory societies and some
of their scientific advisers, but it was at
variance with the current scientific consensus
among memory experts.
The apparent forgetting and recovery of
memories of trauma is not a recent observa-
tion confined to patients having treatment
for actual or suspected abuse in childhood.
1012 BMJ VOLUME 317 10 OCTOBER 1998 www.bmj.com
The forgetting of traumatic incidents during
this century’s two world wars has been
repeatedly described by some of Britain’s
most eminent psychologists and psychiatrists.
Consistent with this, recent research indicates
that the forgetting and subsequent recovery
of memories is associated with a wide range
of traumatic incidents, including the witness-
ing or experiencing of injury, violence, and
death. Memory recovery often seems to occur
spontaneously, before the involvement with
therapists. Corroborative evidence of the
essential accuracy of the recovered memories,
sometimes of high quality, has been found in
substantial numbers of cases.
Pope offers no
explanation for these data.
Moreover, these repeated observations
are supported by biological studies of
trauma and memory. Whereas heightened
states of arousal often improve memory
consolidation, extraordinarily high concen-
trations of catecholamines or other neuro-
peptides at the time of the trauma, perhaps
in combination with a failure to release suf-
ficient cortisol, may produce amnesia.
Several studies have shown reduced hippoc-
ampal volume in traumatised subjects, a
potentially reversible anatomical change
that may have important implications for
memory functioning. Cognitive psychology,
too, recognises that ordinary memory relies
for its efficiency as much on the ability to
inhibit unwanted material as on the ability to
gain rapid access to relevant material.
Experimental studies show the inhibition of
memory retrieval and the existence of a sub-
group of people with poor memories for
negative experiences.
The implication that recovered memo-
ries are invariably the product of “recovered
memory therapy” is circular and lacks
empirical support.
Likewise, Pope’s discus-
sion of existing methods and findings relies
on its own set of questionable assumptions.
Rather than simply attack the idea that trau-
mas may be forgotten, we need to develop
theories that can account for all the available
data. At present it looks as though one such
theory will have to admit the essential accu-
racy of some recovered memories while
acknowledging that others may be false.
Chris R Brewin Professor of psychology
Department of Psychology, Royal Holloway,
University of London, Egham, Surrey TW20 0EX
1 Pope, HG Jr. Recovered memories of childhood sexual
abuse. BMJ 1998;316:488-9. (14 February.)
2 Brewin CR. The scientific status of recovered memories. Br
J Psychiatry 1996;169:131-4.
3 Brewin CR, Andrews B. Recovered memories of trauma:
phenomenology and cognitive mechanisms. Clin Psychol
Rev (in press).
4 Yehuda R, Harvey P. Relevance of neuroendocrine altera-
tions in PTSD to memory-related impairments of trauma
survivors. In: Read JD, Lindsay DS, eds. Recollections of
trauma: scientific evidence and clinical practice. New York:
Plenum Press, 1997:221-42.
5 Pope KS. Memory, abuse and science: Questioning claims
about the false memory syndrome epidemic. Am Psychol
There is no evidence that therapists
implant memories
Pope states in his editorial that he
has found no evidence of repression in clas-
sical or scientific literature.
The same
argument holds true for the opposite stance.
There is no evidence that therapists have
implanted false memories of childhood
trauma in classical (or modern) literature
nor is there clear evidence of this in scientific
The popular belief that therapists can
implant memories “does not spare investiga-
tors the burden of providing a rigorous,
methodologically convincing demonstra-
tion of its existence, (to use Pope’s words).
Lynn Crook Editor, Treating Abuse Today
Richland, WA 99352, USA
1 Pope HG Jr. Recovered memories of childhood sexual
abuse. BMJ 1998;316:488-9. (14 February.)
Author’s reply
I strongly agree with Ilsley and Bir-
man that sexual abuse in childhood is both
common and deplorable. It does not
logically follow, however, that such abuse can
create “repressed memories. Nor is my
scepticism about the validity of repressed
memories an attempt to refute the extent of
sexual abuse itself. I am not a board member
of the False Memory Syndrome Foundation,
but rather a member of the foundation’s sci-
entific advisory board, which I volunteered
to join after witnessing the tragedies perpe-
trated by “recovered memory therapy.” This
position creates no conflict of interest, since
I receive no payment of any sort from it, nor
does the foundation have any influence over
what I do or write. The scientific validity of
repressed memory is not improved simply
because the concept is venerable and
endorsed by personal testimonials; because
it seems plausible that survivors would frag-
ment their memories; or because there
exists a brain mechanism which might
explain it if, hypothetically, it did occur. I
have discussed elsewhere the fallacies inher-
ent to these arguments.
The existence of
repressed memory will be established only
by sound prospective studies meeting basic
minimum methodological standards.
Here I diverge from Brewin, who argues
that such evidence already exists and that
the “consensus among memory experts”
goes against me.
I would like to point out
the following:
1) in an unselected review of 63 studies,
comprising more than 10 000 victims of all
manner of traumas, none reported
repressed memories
2) in a review of 45 other studies of the
sequelae of childhood sexual abuse, examin-
ing 3369 victims, amnesia is nowhere
3) in a survey of 301 board certified
American psychiatrists that I carried out
only 23% answered that there was “strong
evidence” for the validity of “dissociative
amnesia, with 19% answering “little or no
evidence” (unpublished data);
4) American courts in the last several
years have overwhelmingly ruled
“repressed” and “recovered” memory scien-
tifically inadmissible
5) numerous scholarly reviews have
questioned the validity of these concepts.
In the light of these observations, it
seems doubtful that I, or the Royal College
of Psychiatrists, or the false memory
societies, voice an idiosyncratic or minority
Harrison G Pope Jr Associate professor of psychiatry
Harvard Medical School, Boston, MA 02115, USA
1 Pope HG Jr. Psychology astray: fallacies in studies of “repressed
memory” and childhood trauma. Boca Raton, FL: SIRS, 1997.
2 Pope HG Jr, Hudson JI, Bodkin JA, Oliva P. Questionable
validity of ‘dissociative amnesia’ in trauma victims.
Evidence from prospective studies. Br J Psychiatry
3 Brewin CR. The scientific status of recovered memories. Br
J Psychiatry 1998;172:216-7.
4 Kendall-Tackett KA, Williams LM, Finkelhor D. Impact of
sexual abuse on children: a review and synthesis of recent
empirical studies. Psychol Bull 1993;113:164-80.
5 Johnston M. Spectral evidence: the Ramona case: incest, memory
and truth on trial in Napa Valley. Boston: Houghton Mifflin,
Mouth care and skin care in
palliative medicine
We advocated a pragmatic
approach to mouth and skin care in the
ABC of Palliative Care.
Not all the
suggestions in the subsequent correspond-
ence followed the same principle.
Lucas and Roberts rightly point to the
evidence for chlorhexidine in reducing oral
bacteria. They fail to point out that
chlorhexidine alters taste and can damage
oral mucosa and that some randomised
trials have shown it to be no better than
water in preventing mucositis caused by
treatments such as chemotherapy.
with the oral discomfort that occurs on
application, chlorhexidine is not the pana-
cea in palliative care that they would have us
Pemberton and Thornhill reiterate the
continuing uncertainty over the origin of
apthous ulceration,
but they contradict
themselves in complaining about treating
apthous ulcers as infective and then
admitting the importance of secondary
infection and the appropriateness of using
antimicrobial agents.
Davies, along with several of the
correspondents, disagrees with our manage-
ment of a dry mouth.
Hyposalivation is a
problem in patients, and two correspond-
ents recommend salivary stimulants. Chew-
ing gum, suggested by two correspondents,
has not proved helpful in our patients with a
persistently dry mouth. Davies et al have
shown that pilocarpine stimulates salivation
but at the cost of significantly more adverse
When patients were given a choice,
they thought pilocarpine was no better than
a substitute for saliva based on mucin.
In a
previous study patients found porcine
mucin to be no better than water.
seem to find sprays containing water or cool,
pleasant drinks most helpful.
Dixon and Hockley chastise us for
failing to be sufficiently multidisciplinary by
not suggesting surgical debridement in
managing pressure sores and malignant
We agree that this approach has a
role in some patients, but for nearly all
patients with very advanced disease it is
1013BMJ VOLUME 317 10 OCTOBER 1998 www.bmj.com
inappropriate because they are too ill or
they are adamant that they wish no further
intervention at the end of their lives.
Treatments must be chosen with
patients. In reality, however, patients who are
asked for their preferences do not always
choose what professionals believe to be the
correct treatment. In addition, when evi-
dence is contradictory, it can be difficult to
accept that the simplest approaches may be
the most appropriate. Perhaps it was the
simplicity of our approach that caused these
correspondents difficulty.
Claud Regnard Consultant in palliative medicine
Sarah Allport Clinical nurse manager
Lydia Stephenson Senior physiotherapist
St Oswald’s Hospice, Newcastle upon Tyne
1 Regnard C, Allport S, Stephenson L. ABC of palliative
care: Mouth care, skin care, and lymphoedema. BMJ
2 Lucas VS, Roberts GJ, Dixon JM, Hockley J, Pemberton M,
Thornhill MH, Davies A. Mouth care and skin care in pal-
liative medicine [correspondence]. BMJ 1998;316:1246-7.
(18 April.)
3 Dodd MJ, Larson PJ, Dibble SL, Miaskowski C, Greenspan
D, MacPhail L, et al. Randomized clinical trial of chlorhex-
idine versus placebo for prevention of oral mucositis in
patients receiving chemotherapy. Oncology Nursing Forum
4 Davies AN,Daniels C, Pugh R, Sharma K. A comparison of
artificial saliva and pilocarpine in the management of
xerostomia in patients with advanced cancer. Palliat Med
5 Sweeny MP, Bagg J, Baxter WP, Aitchison TC. Clinical trial
of a mucin-containing oral spray for treatment of xerosto-
mia in hospice patients. Palliat Med 1997;11:225-32.
Number needed to harm
should be measured for
The concept of number needed to
treat was an attempt to introduce both
simplicity and objectivity into the evaluation
of treatment. It is helpful both on a large
scale when treatments are compared and
care is commissioned and on the personal
level when doctor and patient choose the
management for the individual. The
number needed to treat gives a comparison
of treatments and outcomes that both
doctors and patients should understand.
One problem of this unimodal numeri-
cal scale for measuring the likely outcome of
a treatment is that it oversimplifies the issues
and may overvalue the treatment by
ignoring its risks. Treatments may harm
patients in various ways, the importance of
which will depend on the disorder being
treated as well as the nature of the harm. In
a minor illness a potentially fatal treatment
would not be acceptable even if the risk were
fairly small. If a condition is uniformly fatal if
untreated the risk of death or disability from
the treatment is likely to be acceptable and a
period of pain or discomfort may be a small
price to pay. Not only are all adverse effects
not equal but their importance depends as
much on their context as their nature.
Another problem with adverse effects is
that we can seldom be as accurate in
guessing risk as we are in measuring benefit.
The therapeutic effect of a drug is usually
unimodal and obvious. Controlled trials
should separate the therapeutic effect from
the placebo effect and allow an objective
measurement of the real therapeutic benefit.
One drug can, however, have many potential
adverse effects, only some of which can be
anticipated from its pharmacology or have
been recognised when it is licensed. The
incidence of common problems can be
found from the results of clinical trials, but
the recognition of more subtle effects often
takes years. Even death and disability may
not be recognised as therapeutic misadven-
ture if the prevalence is small or the onset
delayed. The risk will always be under-
We must not ignore the risk of treatment
even if it is hard to measure. It may not be
possible to devise a unimodal number
needed to harm measurement, but a
compensating negative measure is essential
if we are not to delude ourselves and our
patients about the value of treatment.
A possible solution might be to separate
adverse effects into several grades, on the
basis of severity, reversibility, and usual dura-
tion. These might be: number needed to kill,
number needed to disable, number needed
to make you ill, and number needed to
annoy. The concept of attaching a price list
to the therapeutic menu should not stop
with the cost of the pills.
Arnold Zermansky General practitioner
Dib Lane Practice, Leeds LS8 3AY
Systematic review of trials
comparing antibiotic with
placebo for acute cough in
Data do not justify study’s conclusions
Chapman has highlighted the
confusion caused by the different interpreta-
tions of the report by the International
Agency for Research on Cancer on the
effects of passive smoking.
By ignoring the
size and direction of the effect and focusing
on the lower limit of the confidence interval
the agency came to the erroneous conclu-
sion that passive smoking does not cause
lung cancer. Unfortunately, Fahey et al have
fallen into the same trap in reporting the
results of a systematic review of the use of
antibiotics in acute cough.
They state categorically in their discus-
sion: “This systematic review shows that
antibiotic treatment has no effect on the
resolution of acute cough. This conclusion
is not justified by the data in their review.
Two of the outcomes measured
the resolu-
tion of productive cough and clinical
show a pooled effect that
favours antibiotics but does not reach
significance at the 95% level when a random
effects model is used. The authors seem to
have confused the significance of these find-
ings with the size of the effect. There is
around a 1 in 40 chance of this pooled result
arising because of random variation rather
than because of a real difference between
antibiotic and placebo; this is hardly
grounds to claim that the review shows that
antibiotics have no effect.
The authors do not show an even
handed approach when they deal with the
data concerning the efficacy of antibiotic
and side effects of treatment. In the case of
efficacy they state that “antibiotic treatment
was no better than placebo,” and in the case
of side effects they state that the data showed
“a non-significant increase in the risk of side
effects from antibiotics.” They then proceed
to exclude the only trial that showed more
side effects in the placebo group than the
antibiotic group (on the grounds that this
exclusion reduces heterogeneity) and find
that the pooled result is then significant.
Excluding a study because of the
direction of its result is not an acceptable
method of generating significance.
Heterogeneity in this outcome was not
excessive (÷
= 7.80, df = 5, P > 0.01), and the
same technique could have been used to
generate a significant benefit for the
outcome of resolution of cough by exclud-
ing Williamson’s trial. Might the review
process have been influenced by the conclu-
C J Cates Editor, Cochrane Airways Review Group
Manor View Practice, Bushey Health Centre,
Bushey, Hertfordshire WD2 2NN
1 Chapman S. The hot air on passive smoking. BMJ
1998;316:945. (21 March.)
2 Fahey T, Stocks N, Thomas T. Quantitative systematic
review of randomised controlled trials comparing
antibiotic with placebo for acute cough in adults. BMJ
1998;316:906-10. (21 March.)
Too few subjects were studied for useful
conclusions to be drawn
Objective assessment of the results
of systematic reviews is essential, as clini-
cians may place greater emphasis on the
conclusions reached in such reviews than
they would on those of any of the individual
trials. Fahey et al state that their quantitative
systematic review shows that “antibiotic
treatment has no effect on the resolution of
acute cough” and that “treatment with
antibiotic may incur side effects in a few
We do not believe that this
conclusion can be reached on the basis of
the results of the review.
The study reported results with regard
to three main outcome measures. The
relative risks obtained for the effect of
antibiotic treatment compared with placebo
on the resolution of cough was 0.85 (95%
confidence interval 0.73 to 1.00), for the
effect on clinical improvement at re-
examination it was 0.62 (0.36 to 1.09), and
for the effect on side effects it was 1.51 (0.86
to 2.64). Despite the differences found in the
review, none of the results reaches signifi-
cance as all the 95% confidence intervals
include 1.00. It is important to note how
wide the confidence intervals are for each
result. This reflects the small numbers of
patients available for comparison for each
outcome measure.
Of particular interest is the finding that,
despite its large width, the 95% confidence
interval favours antibiotics for an effect on
1014 BMJ VOLUME 317 10 OCTOBER 1998 www.bmj.com
both resolution of cough and clinical
improvement at re-examination; this
suggests a trend favouring the use of anti-
biotics over placebo.
For resolution of
cough there is a 95% chance that with more
patients a benefit of antibiotics would be
found for resolution of acute cough
somewhere between a relative risk of 0.73
and one of 1.00. There may also be a benefit
in terms of clinical improvement at
re-examination of up to 0.36, or a detriment
of up to 1.09 (the 95% confidence interval).
For side effects the 95% confidence interval
favours placebo, but a study with larger
numbers might find that antibiotics have
fewer side effects than placebo, with a
relative risk of up to 0.86.
An objective analysis of these results
suggests that the number of subjects studied
was too small for any useful conclusions to
be drawn; a potentially important benefit
favouring the use of antibiotics in acute
cough cannot, however, be excluded.
Whether any benefit found would be
clinically relevant (after the magnitude of
benefit, cost, and side effects has been taken
into account) is also yet to be answered. Fur-
ther trials are warranted.
Thomas P Shakespeare Specialist registrar in
radiation oncology
Liverpool Hospital, Liverpool, NSW 2170, Australia
Rachael C Bourke Medical student
University of Adelaide Medical School, Adelaide,
South Australia, 5001, Australia
1 Fahey T, Stocks N, Thomas T. Quantitative systematic
review of randomised controlled trials comparing
antibiotic with placebo for acute cough in adults. BMJ
1998;316:906-10. (21 March.)
2 Braitman LE. Confidence intervals assess both clinical sig-
nificance and statistical significance. Ann Intern Med
Quality scores showed poor agreement
Fahey et al report their meta-
analysis of the literature on the use of anti-
biotics for acute cough in adults
but would
have done better to stick to a good literature
review, albeit that this has been done before.
They left their inclusion criteria broad,
presumably to increase the numbers of
papers to review. I remain to be convinced,
however, that it is sensible to combine trials
that may have included 12 year olds with a
one day history of cough and no findings on
examination with trials in 90 year olds with a
three week history as well as fever, malaise,
purulent sputum, and findings on ausculta-
tion. They excluded patients with chronic
obstructive pulmonary disease but did not
consider those with asthma.
In order to combine six small trials (of
45-207 patients, total 700) they chose
outcome measures that necessitated the
exclusion of a trial in 829 patients. This
makes no sense. Small trials are more likely
to be of poor quality,
and so it is important
that this is properly assessed. Their kappa
scores, however, show poor agreement on
scoring quality, particularly on the most
important category, selection bias. Scrutiny
of the data given on the website makes me
doubt their assessment even more. For
example, a trial of 45 patients with only a
20% recruitment rate and a 29% rate of loss
to follow up scores 9 out of 12 for quality.
The authors relaxed their original crite-
ria in order to include a trial in patients as
young as 8, which they give a quality score of
only 7. This study of 72 patients goes on to
contribute a weight of 41.3% in one of the
outcome meta-analyses, and no mention is
made of the potential for bias.
The authors overstate their conclusions
by saying that resolution of cough was not
affected by antibiotic treatment but that side
effects were more common in the antibiotic
group. In fact, neither result reaches
The use of systematic reviews and meta-
analysis has brought a valuable new dimen-
sion to clinical research
and encouraged
the introduction of evidence based medi-
It is still in its youth, and such
published research must be of high quality
so that all doctors become convinced of its
Jessica Harris General practitioner
Bacon Road Medical Practice, Norwich NR2 3QX
1 Fahey T, Stocks N, Thomas T. Quantitative systematic
review of randomised controlled trials comparing
antibiotic with placebo for acute cough in adults. BMJ
1998;316:906-10. (21 March.)
2 MacKay DN. Treatment of acute bronchitis in adults with-
out underlying lung disease. J Gen Intern Med 1996;11:557-
3 Mulrow CD, Oxman AD. Cochrane collaboration hand-
book. In: Cochrane Library [database on disk and CD
ROM]. Issue 2. Oxford: Update Software, 1996.
4 Chalmers I, Altman D, eds. Systematic reviews.London: BMJ
Publishing, 1995.
5 Sackett D, Richardson WS, Rosenburg W, Haynes RB.
Evidence-based medicine. London: Churchill Livingstone,
Authors’ reply
Cates bases his criticism on
manipulation of the pooled effect estimates
and our attributing the non-significant trend
towards antibiotic as evidence of efficacy.
Because of the substantial likelihood of bias
we were deliberately cautious in attributing
benefit to antibiotic.
The danger in system-
atic reviews, particularly those based exclu-
sively on small trials, is not of statistical
precision but of systematic bias and the pro-
duction of false positive results.
nately, there is no statistical solution to this
Funnel plot asymmetry from our
systematic review (figure) shows that esti-
mates of efficacy were far greater in the two
smaller trials that contributed to the
Thus we believe that we are
correct in our cautious interpretation of the
pooled results.
In suggesting that there is a trend
favouring the use of antibiotics over placebo,
Shakespeare and Bourke fail to account for
all the evidence that we presented. They are
right that the number of subjects (700) in
which the outcome of productive cough was
based is small. They ignore, however,the 829
patients in the trial by Howie and Clark,
which reported no difference between
antibiotic and placebo but which we did not
include in the pooled estimate because
analysis was by episode of illness. Further-
more, as we reported, the trials with more
positive results were smaller trials with
substantial losses to follow up.
We would have liked to include Howie
and Clark’s trial in the pooled analysis; we
reported its results in detail because we
wanted to emphasise its importance in the
context of the evidence presented in our
review. We agree with Harris’s anxieties
about the quality of individual studies, and
this accounts for our cautious conclusions
concerning pooled estimates. We presented
losses to follow up because this allows read-
ers to judge for themselves the quality of
trials that contributed to the review. We
agree with Harris that quality criteria and
scoring systems may not distinguish high
and low quality trials.
The results of our review should be seen
in the context of the high prescribing of
antibiotics for acute chest infection in other-
wise healthy people throughout the devel-
oped world. We reported the number
needed to treat (n = 11) and number needed
to harm (n = 15) for clinical improvement
even if we accept the trend towards
antibiotic as evidence of efficacy
to show
that the clinical benefit of antibiotic is prob-
ably marginal. Further trials in higher risk
groups are more likely to show important
benefits for patients.
In collaboration with American col-
leagues we will update this review for the
Cochrane Library. We will make explicit our
concerns about potential biases in the
updated version.
Tom Fahey Senior lecturer in general practice
Nigel Stocks Clinical lecturer in general practice
Division of Primary Care, University of Bristol,
Bristol BS8 2PR
1 Egger M, Davey Smith G, Schneider M, Minder C. Bias in
meta-analysis detected by a simple, graphical test. BMJ
2 Egger M, Davey Smith G. Misleading meta-analysis. BMJ
3 Egger M, Schneider M, Davey Smith G. Meta-analysis:
Spurious precision? Meta-analysis of observational studies.
BMJ 1998;316:140-4.
4 Dunlay J, Reinhardt R, Donn L. A placebo-controlled,
double blind trial of erythromycin in adults with acute
bronchitis. J Fam Pract 1987;25:137-41.
5 King D, Williams CW, Bishop L, Shechter A. Effectiveness
of erythromycin in the treatment of acute bronchitis. JFam
Pract 1996;42:601-5.
0.3 0.5 1.0 2.0 3.0
Odds ratio
Funnel plot of six trials that contributed to outcome
of productive cough at follow up. Pooled odds ratio
(fixed effects model) is 0.78 (95% confidence
interval 0.56 to 1.08)
1015BMJ VOLUME 317 10 OCTOBER 1998 www.bmj.com
Celebrity’s death from cancer
resulted in increased calls to
It is known that a patient with can-
cer can be influenced in her choice of treat-
ment by the choice made by a famous
person with the same disease. When Nancy
Reagan, the wife of the then American
president, chose a mastectomy in 1987,
there was a subsequent 25% decrease in the
number of women choosing breast conserv-
ing surgery.
A celebrity’s death can have a
similar powerful effect. The death of Diana,
Princess of Wales, resulted in unprecedented
public and media response.
The death of
Linda McCartney in April has also shown
that patients may be substantially affected by
such a tragedy.
CancerBACUP’s telephone information
service answered 40 715 calls (daily average
160) in April 1997 to March 1998, 11 675 of
which were related to breast cancer. When
all lines are busy, calls are directed to a call
logging machine; on average 500 calls a day
are logged. BT provides data on “ineffective”
calls to the freephone
that is, those in
which the caller obtains the engaged signal
when all the nurses and answering machines
are busy. About 5500 calls are ineffective.
In the week after Linda McCartney’s
death (announced on Sunday 19 April), calls
and inquiries about breast cancer increased
substantially. Additional resourced lines
increased the daily average number of calls
to 204 (44 more calls). Seventy two calls daily
(28 more calls) were about breast cancer,
representing a 64% increase in the number
of answered calls about breast cancer. In the
same week, despite the extra open lines, the
daily average number of calls logged to the
answering machines doubled to 1008, the
highest daily number being 1262 (figure).
Altogether 1090 (14%) of the weekly calls
obtained the engaged signal. As far as we
can ascertain there were no factors other
than Linda McCartney’s death that affected
use. The next highest number of logged calls
outside that week was 1106 (on 24
February), the day after the announcement
that cancer had overtaken heart disease as
the major cause of death. Other publicity,
such as the American tamoxifen prevention
trial (early April) and the role of anti-
angiogenesis agents (early May), had a
minor effect compared with that about
Linda McCartney’s death.
The death of a celebrity from a disease, so
publicly explored, presumably forced women
to face their own mortality. The result was a
tremendous need for information, emotional
support, and reassurance. Such a response
indicates, perhaps, that however well a
woman copes with a diagnosis of breast can-
cer, anxiety is never far from the surface.
Markella Boudioni Research officer
Jean Mossman Chief executive
CancerBACUP, London EC2A 3DR
Alison L Jones Consultant oncologist
Department of Clinical Oncology, Royal Free
Hospital, London NW3 2QG
Geraldine Leydon Research fellow
Klim McPherson Professor of Public Health
Cancer and Public Health Unit, Epidemiology and
Population Health, London School of Hygiene and
Tropical Medicine, London WC1E 7HT
CancerBACUP exists to meet information needs of
patients with cancer and their relatives and friends.
1 Wise J. Celebrities influence the public. BMJ 1998;316:801.
(14 March.)
2 Nattinger AB, Hoffmann RG, Howell-Pelz A, Goodwin JS.
Effect of Nancy Reagan’s mastectomy on choice of surgery
for breast cancer by US women. JAMA 1998;279:762-6.
3 Morgan-Jones R, Smith K, Oakley P. The “Diana effect.
BMJ 1998;316:1750-1. (6 June.)
4 Shevlin M, Brundsen V, Walker S, Davies M, Ramkalawan
T. Death of Diana, Princess of Wales. BMJ 1997;315:1467.
Sumatriptan is not the only
analgesic used inappropriately
As Gaist et al report, the problem of
overuse of sumatriptan is of considerable
This is particularly so in the context
of chronic daily headache.
Catarci et al have
suggested that sumatriptan may replace
ergotamine in overuse syndromes.
Staff at
the headache clinic at the National Hospital
in London have now seen patients who use
sumatriptan, naratriptan, or zolmitriptan
daily. Only overuse of sumatriptan has been
observed in patients who have not previously
used another acute antimigraine compound,
but it seems reasonable to speculate that the
problem may be seen with any triptan.
The core of this problem is that daily or
near daily use of any triptan is, with few
exceptions, inappropriate. Moreover, daily or
near daily use of any acute antimigraine com-
pound available in the United Kingdom may
lead to appreciable management problems
in particular, headache associated with anal-
The commonest cause of this among
patients seen in the headache clinic is overuse
of compound analgesics, particularly those
that include codeine phosphate. If daily or
very frequent headache is as common in the
United Kingdom as it is in the United
in 4-5% of the population
of various painkillers is probably a consider-
able public health issue and has not received
adequate attention. Our experience suggests
that the issues raised by Gaist et al are just
part of an epidemic.
Peter J Goadsby Reader in clinical neurology
Institute of Neurology, London WC1N 3BG
1 Gaist D, Tsiropoulos L Sindrup SH, Hallas J, Rasmussen
BK, Kragstrup J, et al. Inappropriate use of sumatriptan
population based register and interview study. BMJ
1998;316:1352-3. (2 May.)
2 Silberstein SD, Lipton RB, Sliwinski M. Classification of
daily and near-daily headaches: a field study of revised IHS
criteria. Neurology 1996;47:871-5.
3 Catarci T, Fiacco F, Argentino C, Sette G, Cerbo R.
Ergotamine-induced headache can be sustained by
sumatriptan daily intake. Cephalalgia 1994;14:374-5.
4 Mathew NT. Drug-induced headache. Neurol Clin
5 Scher A,Stewart WF, Liberman J,Lipton RB.Prevalence of
frequent headache in a population sample. Headache
Suicide in patients with stroke
Depression may be caused by symptoms
affecting lower urinary tract
In their paper on suicide in
patients with stroke Stenager et al com-
mented on the prevalence of depression
after stroke.
If the underlying causes of this
depression can be elucidated and treated
then the risk of suicide may be lowered.
One such cause may be symptoms affect-
ing the lower urinary tract (urinary inconti-
nence, frequency, nocturia, and urgency), as
self reported depression is related to such
A total of 3592 questionnaires
that were developed for a large Medical
Research Council study on incontinence in
Leicestershire have been analysed. The
prevalence of self reported depression
increased from 15% in patients with stroke
without urinary symptoms to 32% in those
who had had a stroke and urinary symptoms.
Symptoms affecting the lower urinary
tract should not be discounted when assess-
ing survivors of stroke in the community
because they can be treated and hence could
play an active part in reducing depression
and therefore suicide.
K R Brittain Research associate
Division of Medicine for the Elderly, University of
Leicester, Leicester LE1 7RH
C M Castleden Professor
Leicester General Hospital, Leicester LE5 4PW
Weekly periods
Daily No of answerphone calls
27 Apr
20 Apr
13 Apr
6 Apr
30 Mar
23 Mar
4 May
18 May
11 May
Daily number of calls received by answerphone (that is, not answered by nurses) on each working day
Monday 23 March to Friday 22 May 1998. Vertical lines represent Mondays; two Mondays were public
1016 BMJ VOLUME 317 10 OCTOBER 1998 www.bmj.com
1 Stenager EN, Madson C, Stenager E, Bolsden J. Suicide in
patients with stroke: epidemiological study. BMJ
1998;316:1206. (18 April.)
2 Brittain KR. Urinary symptoms and depression in stroke
survivors. Age Ageing 1998;27(suppl):72-3.
Future studies should consider the role of
cerebrovascular disease
The epidemiological study by
Stenager et al on the risk of suicide after
stroke is the first study of its kind and high-
lights the prime importance of detecting
mood disorder in patients with stroke.
However, the conclusion that society should
take more interest in the psychosocial
aspects of living with the impairment
imposed by stroke underplays the role of
organic brain damage imposed by cer-
ebrovascular disease. As the prevalence of
depression is higher in stroke than in disor-
ders with a similar degree of disability,
closer scrutiny of the brain injury may be
required. Indeed, one study of suicidal idea-
tion after stroke found anterior brain
lesions to be associated with suicidal
thinking immediately after a stroke.
possible explanation for the likely over-
representation of suicidal thinking and
completed suicide in patients with stroke is
the high prevalence of lesions disrupting
frontal and subcortical brain circuitry,
either as lacunar infarcts or strategic
cortical infarcts affecting the frontal lobe(s).
Such lesions are associated with both
and impulsivity.
Although the role of disability and social
disadvantage after stroke cannot be under-
stated, the contribution of damage to
specific brain areas should continue to
attract research. In this way, the study of
cerebrovascular disease can complement
epidemiological research into socio-
demographic risk factors.
Rahul Rao Senior lecturer in old age psychiatry
Guy’s Hospital, London SE1 9RT
1 Stenager EN, Madson C, Stenager E, Bolsden J. Suicide in
patients with stroke: epidemiological study. BMJ
1998;316:1206. (18 April.)
2 Folstein MF, Maiberger R, McHugh PR. Mood disorder as
a specific complication of stroke. J Neurol Neurosurg
Psychiatry 1977;40:1018-20.
3 Kishi Y, Robinson RG, Kosier JT. Suicidal plans in patients
with stroke: comparison between acute-onset and delayed-
onset suicidal plans. Int Psychoger 1996;8:623-34.
4 Robinson RG, Morris PL, Fedoroff JP. Depression and
cerebrovascular disease. J Clin Psychiatry 1990;51:26-31.
5 Kaye K, Grigsby J, Robbins LJ, Curzon B. Prediction of
independent functioning and behaviour problems in geri-
atric patients. J Am Geriatr Soc 1990;38:1304-10.
Early diagnosis of cystic
fibrosis can improve children’s
Wald and Morris
seem to have
examined carefully our article on the
nutritional benefits of neonatal screening in
cystic fibrosis
but they misunderstood some
points and failed to grasp the potential
advantages of preventing malnutrition in
patients with cystic fibrosis.
They comment that our “study design is
an ingenious one, but the analysis of the
results is problematic. Ironically, the most
ingenious feature is the very element they
criticise with their comment that “a difficulty
that is not discussed . . . is that the data in
children under 4 years are subject to
selection bias.” Selection bias was one of the
challenges we overcame in our study design.
Our randomisation protocol was designed
to include a group that had been screened at
birth and a standard diagnosis (control)
Once the controls were identified by the
unblinding process at 4 years of age, their
anthropometric indices of nutritional status
since birth were obtained by reviewing
records; this information was incorporated
into a “what if analysis
a statistical analysis
to determine how the control group would
have compared with the screened group if
all patients had been identified at birth.
Although this was described in the original
version of our paper it was deleted during
editing but the paper still showed that
significant differences persisted.
Wald and Morris ask for “a separate
analysis restricted to follow up after the first
four years.” We completed the unblinding of
controls last April and have accumulated
sufficient data to show that significant differ-
ences persist. We have shown that if this
analysis starts at 4 years of age the
proportion of patients with cystic fibrosis
who have heights above the 10th centile
remains significantly greater in the screened
group (P = 0.042).
Finally, the comment that “this trial pro-
vides no evidence of any benefit of
screening” is inappropriate; not only are the
nutritional advantages obvious from our
study but also the psychological advantages
of genetic counselling and the potential pul-
monary benefits are evident.
I strongly disagree with the concern
implied by Wald and Morris that “early
knowledge of a serious disorder will cause
more harm than good if there is no effective
remedy. Our assessment shows that
malnutrition can be prevented by neonatal
screening. Most doctors caring for patients
with cystic fibrosis believe that interventions
used to treat the respiratory disease are
Philip M Farrell Dean
University of Wisconsin Medical School, Madison,
WI 53706, USA
On behalf of the Wisconsin Cystic Fibrosis Neonatal
Screening Group
1 Wald NJ, Morris JK. Neonatal screening for cystic fibrosis.
BMJ 1998;316:404-5. (7 February 1998.)
2 Farrell PM, Kosorok MR, Laxova A, Shen G, Koscik RE,
Bruns T, et al. Nutritional benefits of neonatal screening
for cystic fibrosis. N Engl J Med 1997;337:963-9.
3 Baroni MA, Anderson YE, Mischler EH. Cystic fibrosis
newborn screening: impact on early screening results on
parenting stress. Pediatr Nurs 1997;23:143-51.
4 Farrell PM, Shen G, Splaingard M, Colby CE, Laxova A,
Kosorok MR, et al. Acquisition of Pseudomonas aerugi-
nosa in children with cystic fibrosis. Pediatrics 1997;100:e2.
Available from http://www.pediatrics.org/.
5 Mischler EH, Wilfond BS, Fost N, Laxova A, Reiser C,
Sauer CM, et al. Cystic fibrosis newborn screening: impact
on reproductive decision making and implications for
genetic counseling. Pediatrics 1998;102:44-52.
Smoking and risk of
myocardial infarction
Statistical and biological interactions
should not be confused
Prescott et al report that smoking
increases the risk of myocardial infarction
significantly more in women (relative risk
2.24) than in men (relative risk 1.43).
actions between components of smoke and
hormonal factors were suspected.
Readers may conclude from this study
that men and women do not differ at all. On
the basis of data on the prevalence of smok-
ing (table 2 in Prescott et al’s paper) and
from reported relative risks, we can calculate
that in women the risk of developing
myocardial infarction during follow up is
5.88% (380/6461) in smokers and 2.63%
(132/5011) in non-smokers; in men, the risk
is 10.62% (902/8490) in smokers and 7.38%
(349/4701) in non-smokers. These are best
estimates based on published data; the
figures would change slightly if former
smokers were removed from the group of
non-smokers. The difference that is attribut-
able to smoking was therefore 3.25% in
women and 3.24% in men. Over 12 years,
smoking caused an additional myocardial
infarction in one person out of 31
distributed between men and women.
This shows that statistical interaction
should not be confused with biological
interaction. Statistical interaction concerns
the modelling of combined effects of two or
more risk factors for a disease in popula-
tions, and biological interaction refers to
biochemical reactions in an individual.
Whether statistical interaction exists or not
depends on the specification of the model
that is applied to data
“interaction” means
that a model that simply adds the effects of
two risk factors (in this case sex and
smoking) does not accurately describe their
joint effect (the risk of myocardial infarction
in men who smoke).
Prescott et al used a multiplicative model
and found a significant interaction; I used an
additive model and found none. It is not
uncommon to find a positive interaction on
the additive scale and a negative interaction
on the multiplicative scale. Models of
absolute and relative risk have their respec-
tive merits and disadvantages, neither is
wrong or right, and neither has anything to
say about the biology of the phenomenon
under study. Prescott et al may be right in
their hypothesis that components of smoke
interact with hormones in causing myocar-
dial infarction, but their data do not show
that this is happening.
Thomas V Perneger Medical epidemiologist
Institute of Social and Preventive Medicine,
University of Geneva, CH-1211 Geneva 4,
1 Prescott E, Hippe M, Schnohr P, Hein HO, Vestbo J. Smok-
ing and risk of myocardial infarction in women and men:
longitudinal population study. BMJ 1998;316:1043-7.
(4 April.)
1017BMJ VOLUME 317 10 OCTOBER 1998 www.bmj.com
Studying relative risk is not enough
Prescott et al show that women
who smoke have a high relative risk of myo-
cardial infarction, which is highest in those
aged under 55 and decreases with age.
These age specific relative risks are higher
than those for men at the same age. Prescott
et al hypothesise that tobacco may be more
harmful to women because of antioestro-
genic effects. The greater risks in women
and young people could, however, have
alternative explanations.
As in all other populations studied, the
absolute risk is much less in women than in
men and in young than in old people, as
Prescott et al show in their figure 1.
additional absolute risk caused by smoking
is small among young women and relatively
high only because of the low baseline level.
Should Prescott et al rather suggest that
smoking is less harmful in women?
Relative risk measures the effect of a risk
factor for a disease compared with the total
effects of all other risk factors for the same
disease in a population. A greater relative
risk may simply be a result of studying a
population that has few other risk factors for
the disease.
It is a misinterpretation of
heterogeneity of relative risk to consider a
greater relative risk to be more harmful
when its absolute clinical and public health
impact is small and more appropriately
measured by attributable risk.
The relative
risk due to smoking, for example, is much
higher for lung cancer than heart disease
lung cancer has few other causes, but smok-
ing kills more people through heart disease
because this disease is much more common,
so the additional absolute risk (the attribut-
able risk) of heart disease is much greater.
Epidemiological studies show that the
relative risk of raised blood pressure and
raised blood cholesterol concentrations is
much greater in young people than in mid-
dle aged or elderly ones. Does this mean that
these factors are more harmful in young
people? Clinical and preventive measures
have generally been targeted at older people
as the benefits from intervention in young
people are too small to be worth while.
The greater relative risk that Prescott et
al observe in women is due only to the fact
that women before the menopause have
much fewer or lower levels of risk factors
than men of the same age. To postulate a
new effect is not warranted on the basis of
these data.
Jin-Ling Tang Associate professor in community
James A Dickinson Professor of family medicine
Department of Community and Family Medicine,
Chinese University of Hong Kong, Hong Kong
1 Prescott E, Hippe M, Schnohr P, Hein HO, Vestbo J. Smok-
ing and risk of myocardial infarction in women and men:
longitudinal population study. BMJ 1998;316:1043-6.
(4 April.)
2 Rothman K, Poole C. A strengthening programme for
weak associations. Int J Epidemiol 1968;17(suppl):955-9.
3 Greenland S, Rothman KJ. Concepts of interaction. In:
Rothman KJ, Greenland S, eds. Modern epidemiology.2nd
ed. Philadelphia: Lippincott-Raven;1998:329-42.
Smoking is a feminist issue
Prescott et al’s finding that women
who smoke are at 50% greater risk of
myocardial infarction than men who smoke
is important and deserves wide publicity.
Cardiovascular disease is, however, not the
only risk associated with smoking. There is
already strong evidence that smoking is
20-70% more likely to result in lung cancer
and the risk of oral
cancer is more than doubled in women.
Although lung cancer has recently over-
taken breast cancer as the commonest cause
of death from cancer in British women, the
epidemic of smoking related deaths in
women is only just beginning.
Prescott et al comment that the excess
risk of death from cardiovascular disease
falls rapidly when people stop smoking. The
same is not true for lung cancer, where the
risk falls slowly. Smoking causes clonal
mutations in the respiratory epithelium that
persist for years.
Consistent with this,
smoking in youth is more dangerous than
smoking in old age.
Women know that they are more
susceptible to alcohol than men. They need
to learn that they are also at greater risk
from smoking than men. On a visit to the
United States last month I was horrified to
see one brand of cigarettes advertised on
billboards as “a woman thing. The cynical
marketing of cigarettes to women in
developing countries will ensure that the
lessons so painfully learnt here are repeated
thoughout the world.
Penella J Woll Senior lecturer
Department of Clinical Oncology, Nottingham City
Hospital, Nottingham NG5 1PB
1 Prescott E, Hippe M, Schnohr P, Hein HO, Vestbo J. Smok-
ing and risk of myocardial infarction in women and men:
longitudinal population study. BMJ 1998;316:1043-6.
(4 April.)
2 Zang EA, Wynder EL. Differences in lung cancer risk
between women and men: examination of the evidence.
J Natl Cancer Inst 1996;88:183-92.
3 Muscat JE, Richie JP, Thompson S, Wynder EL. Gender
differences in smoking and risk for oral cancer. Cancer Res
4 Mao L, Lee JS, Kurie JM, Fan YH, Lippman SM, Lee JJ,
et al. Clonal genetic alterations in the lungs of current and
former smokers. J Natl Cancer Inst 1997;89:857-62.
5 Wistuba I, Lam S, Behrens C, Virmani AK, Fong KM,
LeRiche J, et al. Molecular damage in the bronchial
epithelium of current and former smokers. J Natl Cancer
Inst 1997;89:1366-73.
Authors’ reply
We do not agree with Perneger et
al that statistical interaction and biological
interaction are separate issues. Statistical
interaction depends on specification of the
model that is applied to data. Epidemiologi-
cal analysis, however, usually attempts to go
beyond statistical modelling,
and, provided
the model fits the data, statistical interaction
may well have a biological interpretation.
When we found a statistical interaction
between smoking and subjects’ sex we
wondered whether this was a result of our fit-
ting a multiplicative model on data that
should be described in an additive model. We
discussed this in our paper. We find it biologi-
cally plausible that cardiovascular risk factors
should act in a multiplicative fashion. If, how-
ever, effects of risk factors are additive, relative
risks from a multiplicative model will vary
between men and women simply because
baseline risks vary. If this was the case we
would, firstly, expect to find higher relative
risks in women for all of the cardiovascular
risk factors studied (cholesterol concentra-
tion, blood pressure, body mass index, physi-
cal activity, etc). Relative risks for all of these
risk factors were, however, similar in men and
women. Secondly, we would expect the sex
the interaction term
to diminish
with increasing age since baseline risks
increase. This was also not found. From this
we cautiously suggested possible differences
in the mechanisms of action of tobacco in
causing cardiovascular disease in men and
Our calculation was based on an additive
model that was briefly described in the paper.
We calculated the difference in risk
of additional myocardial infarctions in smok-
ers caused by smoking
on the basis of age
specific relative risks for smoking and
baseline incidence. We concluded that the
difference in risk was higher in men up to age
65 and in women over age 65. The model
suggested by Perneger et al cannot pass as an
additive model. It ignores age, time under
risk, and the fact that the female smokers in
our study population were on average much
younger than the female non-smokers.
Finally, it is important to recognise that
from a public health point of view the
relative risk associated with smoking is
higher in women, since calculations of the
impact of smoking on women’s health
should not be based on relative risks
calculated from studies of men.
Eva Prescott Senior research fellow
Merete Hippe Research fellow
Institute of Preventive Medicine,
Kommunehospitalet, DK-1499 Copenhagen K,
1 Rothman KJ. Interactions between causes. In:Rothman KJ.
Modern epidemiology. Boston, MA: Little, Brown; 1986:
Advice to authors
We prefer to receive all responses electronically,
sent either directly to our website or to the
editorial office as email or on a disk. Processing
your letter will be delayed unless it arrives in an
electronic form.
We are now posting all direct submissions to
our website within 72 hours of receipt and our
intention is to post all other electronic
submissions there as well. All responses will be
eligible for publication in the paper journal.
Responses should be under 400 words and
relate to articles published in the preceding
month. They should include <5 references, in the
Vancouver style, including one to the BMJ article
to which they relate. We welcome illustrations.
Please supply each author’s current
appointment and full address, and a phone or
fax number or email address for the
corresponding author. We ask authors to declare
any competing interest.
Letters will be edited and may be shortened.
1018 BMJ VOLUME 317 10 OCTOBER 1998 www.bmj.com
  • [Show abstract] [Hide abstract] ABSTRACT: There is a considerable knowledge base about the information needs of patients with cancer (and their relatives and friends). Those needs will vary according to the disease, the stage of disease, the patient and his or her age, social class and culture. Lack of information may lead to increased anxiety and distress, may impact negatively on the patient's satisfaction and may influence a patient's treatment choices. Other articles in this special edition deal with psychosocial interventions and complementary therapies for cancer patients and explore their efficacy. The reality is that these are unlikely to be made available to all cancer patients for reasons of cost and practicability. Information, however, is a relatively cheap intervention that could--and should--be part of standard care. This article explores some of the research about the provision of information for cancer patients.
    Article · Nov 1999
  • [Show abstract] [Hide abstract] ABSTRACT: To explore why cancer patients do not want or seek information about their condition beyond that volunteered by their physicians at times during their illness. Qualitative study based on in-depth interviews. Outpatient oncology clinics at a London cancer centre. 17 patients with cancer diagnosed in previous 6 months. Analysis of patients' narratives to identify key themes and categories. While all patients wanted basic information on diagnosis and treatment, not all wanted further information at all stages of their illness. Three overarching attitudes to their management of cancer limited patients' desire for and subsequent efforts to obtain further information: faith, hope, and charity. Faith in their doctor's medical expertise precluded the need for patients to seek further information themselves. Hope was essential for patients to carry on with life as normal and could be maintained through silence and avoiding information, especially too detailed or "unsafe" information. Charity to fellow patients, especially those seen as more needy than themselves, was expressed in the recognition that scarce resources-including information and explanations-had to be shared and meant that limited information was accepted as inevitable. Cancer patients' attitudes to cancer and their strategies for coping with their illness can constrain their wish for information and their efforts to obtain it. In developing recommendations, the government's cancer information strategy should attend to variations in patients' desires for information and the reasons for them.
    Full-text · Article · May 2000
  • [Show abstract] [Hide abstract] ABSTRACT: OBJECTIVE: To explore why cancer patients do not want or seek information about their condition beyond that volunteered by their physicians at times during their illness. DESIGN: Qualitative study based on in-depth interviews. SETTING: Outpatient oncology clinics at a London cancer center. PARTICIPANTS: 17 patients with cancer diagnosed in previous 6 months. Main outcome measures Analysis of patients' narratives to identify key themes and categories. RESULTS: While all patients wanted basic information on diagnosis and treatment, not all wanted further information at all stages of their illness. Three overarching attitudes to their management of cancer limited patients' desire for and subsequent efforts to obtain further information : faith, hope, and charity. Faith in their doctor's medical expertise precluded the need for patients to seek further information themselves. Hope was essential for patients to carry on with life as normal and could be maintained through silence and avoiding information, especially too detailed or "unsafe" information. Charity to fellow patients, especially those seen as more needy than themselves, was expressed in the recognition that scarce resources-including information and explanations-had to be shared and meant that limited information was accepted as inevitable. CONCLUSIONS: Cancer patients' attitudes to cancer and their strategies for coping with their illness can constrain their wish for information and their efforts to obtain it. In developing recommendations, the government's cancer information strategy should attend to variations in patients' desires for information and the reasons for them.
    Full-text · Article · Aug 2000
    GM LeydonGM LeydonM BoultonM BoultonC MoynihanC Moynihan+1more author...[...]
Show more