Re: E-DRUG: Copy of: SEDA 19 essay, draft 7

[Anne Holbrook <holbrook@fhs.csu.mcmaster.ca>]

Andrew, 
Thanks for sending on a copy of your article. I have showed it to several 
other members of the Canadian Adverse Reaction Advisory Committee.
It was well written and made several good points - such as the potential 
benefits of submitting signals re: ADRs to wider scientific review (how 
about urgent RCTs as well?)

>From a practicing clinician and pharmepi's point-of-view, my plea would 
be for much more rigorous critical appraisal of the data at hand re: 
ADR's prior to any policy decision. To use your examples, for instance:

1. Cotrimoxazole is an extremely useful and cost-effective antibiotic. It 
is one of the first line choices for many conditions listed in our 
provincial antibiotic guidelines. TMP alone is unproven and much more 
expensive here in Canada. As the Jick paper stated, although co trimox 
clearly has increased risk of serious skin reactions, in the overall 
scheme of things compared to other antibiotics, it is very safe. For 
instance, the elderly are much more likely to die of C difficile diarrhea 
than cotrimos-related SJS. A recent NEJM article put the cotrimos-related 
serious skin reactions at 1 in 6 million incidence - at least some number 
extremely rare. GPs may manage rash, but they tend not (at least here) to 
manage the type of rash that could be fatal. We have so far advised that 
an alert reminding docs about the skin reactions and what to look for be 
issued but the drug not be restricted.

2.  the 3rd generation contraceptives. I am still trying to get the 
original studies, but as I understand teh situation all 3 studies were 
case control in design, or atleast nonexperimental (? one cohort). If so, 
it would be impossible no matter what matching or adjustments were done 
in analysis to avoid the original prescribing bias (why did these women 
get the 3rd generation drugs instead of the other OC's? Was it because 
they were believed safer and therefore prescribed to those women who 
smoked or had other unmeasured risk factors for venous t-embolism. 
Diagnostic suspicion bias is a major problem - everyone believes that OCs 
are assoc. with venous thromboembolism. As well confirmation of clot is a 
big problem in most centres. Short of venogram and duplex ultrasound, 
most methods are relatively inaccurate. Jim Douketis and a few of us have 
a paper submitted that reviews how lowsy the studies on OCs and HRT and 
thromboembolism really are. It will be very interesting to add these 3 
new ones to the brew!

In general, I think that a systematic assessment of ADR data is needed. 
Now that I sit on CADRAC, I realize how terrible a lot of the case report 
data is - often impossible to make much of, even if further info is 
requested. We should not take a totally risk averse stance otherwise no 
therapies would ever be used (although maybe nobody would smoke or drink 
either - a thought!). Policy in this area really needs to be tempered by 
the strength of the signal of ADR, the potential harm presented by the 
ADR, the benefit of the medication compared to others available, and the 
contribution of confounders. Medicolegal practices unfortunately have not 
caught on to this larger picture either.

my 2 cents' worth,

Regards, anne

***************************************************************************
Anne Holbrook,MD,PharmD,MSc,FRCPC
Centre for Evaluation of Medicines, St Joseph's Hospital, and
Department of Medicine, McMaster University, Rm 4X1
Phone: (905) 522-1155 ext 5269 (SJH) or 521-2100 ext 3371 (MUMC)
FAX: 521-6136 (SJH) or 521-4971 (MUMC)
E-mail: holbrook@fhs.csu.mcmaster.ca 

On 23 Jan 1996, Andrew Herxheimer wrote:

> From:	Andrew Herxheimer, 101364,2017
> 
> Dear e-drug colleagues,  I am sending you this draft essay for the Side Effects
> of Drugs Annual  not only because I think it will interest many of you,  but
> because I would also appreciate any suggestions for improving it.   AH
> ________________________________________________________________
> Side Effects of Drugs Essay '96 [for SEDA 19] draft7/7.1.96
> 
> SIDE EFFECTS: 
> FREEDOM OF INFORMATION AND THE COMMUNICATION OF DOUBT 
> 
> Andrew Herxheimer
> Chairman, International Society of Drug Bulletins
> 
> The nature and frequency of unwanted effects is a fundamental
> consideration in the choice of drugs and in their use.  All
> health professionals and patients who make therapeutic choices
> and then use medicines therefore need easy access to reliable
> and up-to-date information about unwanted effects.  Even where
> information about medicines is widely available from
> independent sources, as in the United Kingdom, they mostly do
> not have it.  The need has been evident since the thalidomide
> disaster, over 35 years ago, but it has not been met.  There
> are three major obstacles.
> 
> One is that the systematic collection, investigation, analysis
> and interpretation of data on adverse drug reactions and other
> "side effects" has developed slowly, and is still lagging far
> behind the development of clinical trials to assess
> potentially therapeutic effects.  The second obstacle to
> access to adequate information on adverse drug reactions is
> the unwillingness of pharmaceutical companies and drug
> regulatory authorities to disclose information that they are
> uncertain about and that might threaten a product.  Even when
> an authority has issued a warning about a product or has
> ordered its withdrawal from the market, it often does not
> disclose the data which led to the action.  Third, some
> pharmaceutical companies use litigation against editors and
> publishers to try to suppress the publication of information
> that casts doubt on the safety or effectiveness of their
> products.
> 
> The processes by which unwanted effects of drugs come to
> light, and are then investigated and analysed have been
> extensively described and discussed (eg, Inman 1986; Uppsala
> Symposium 1988).1,2  The first difficulty is to distinguish
> between adverse events that are related to drug therapy and
> those that are unrelated to it.  The nature of the event has
> to be defined, and the frequency of its occurrence estimated
> in relation to particular drugs, diseases and types of
> patient.  Only then can the attributable risk and overall
> risk-to-benefit balance be considered for a particular drug
> and indication.  One should note that this is a public health
> approach and is not valid for an individual patient: in a
> single patient the overall pharmacoepidemiological information
> is just a part of the data to be considered in making the
> diagnosis.
> 
> If an event is potentially important and might be related to
> drug treatment, then further investigation is necessary.  The
> aim of investigation is to confirm or refute the suspicion of
> a causal connection between the drug and the adverse event,
> but it commonly takes a long time to reach a clear public
> health conclusion.  The legal rule "innocent until proved
> guilty" cannot necessarily be accepted as valid in this
> situation: a reasonable suspicion of a serious event often
> requires all concerned to act to protect the users of the
> medicine before causality is proven. 
> 
> The primary responsibility for gathering, collating and
> analysing information on suspected adverse drug reactions lies
> with the individual manufacturers and their licence holders,
> and with national authorities. (The European Medicines
> Evaluation Agency [EMEA] is now responsible for integrating
> the pharmacovigilance activities of the member states of the
> European Union.)  The authorities and the manufacturers must
> also act promptly to inform and protect users and prescribers. 
> The difficulty they face is that of deciding at what stage a
> suspicion justifies a public statement and action.  To state
> that a suspicion exists is already an action; the many
> different ways of passing information to doctors and patients
> vary greatly in their effects on the recipients.  The kind of
> action and the mode and tenor of the communication should be
> determined by the seriousness of the event, the strength of
> attribution, its possible frequency and the impact on the
> relative risk benefit balance of the drug.
> 
> The WHO Adverse Drug Reaction database
> Since 1978 the WHO Collaborating Centre for International Drug
> Monitoring in Uppsala has run a worldwide early warning
> system, and now 46 countries are contributing their ADR
> reports to the common database.  For drugs that are being used
> in more than one country, the Collaborating Centre can
> generate and analyse signals earlier than a national centre,
> and its publication Signal for national pharmacovigilance
> agencies regularly alerts them to new associations.  The
> National Centres in the contributing countries also request
> specific information and have online access to the WHO
> database.  The Uppsala Centre receives around 150 requests a
> year from external inquirers - mostly from industry about
> their own drugs.  Only about half the National Centres agree
> to automatic release of data they have contributed; the others
> must be asked individually, but most accede to a reasonable
> request (Prof I.R. Edwards, personal communication, 19.12.95). 
> All data from the WHO database are accompanied by a caveat
> document that explains their limitations and insists that
> these must be stated in any publication of information
> obtained from the database.
> 
> A recent survey by the WHO Collaborating Centre has examined
> the impact of the WHO adverse reaction signals on the
> decisions of participating countries (Fucik and Edwards
> 1996).3 The majority of the National Centres that responded
> found the signals useful, but their impact varied greatly
> between the countries.  Of 278 "serious" ADR/drug associations
> published in Signal in 1990-1994, 171 were followed up by one
> or more Centres, 62 were published in national bulletins, and
> 13 led to labelling changes.  One of the Centres that did
> little remarked that the information in Signal was "too
> tentative to take action upon".  Bibliographic searches of the
> literature for letters or articles supporting the signals
> tested their relevance: 30 were found to be supported by
> publications and none were refuted.  There are therefore
> important differences in the handling of early signals.  How
> far these differences reflect variations in medical practice,
> cultural norms or differences in the indications for or the
> extent of drug use is not clear.
> 
> Regulatory deliberation, action and communication
> The greater speed and efficiency in generating and analysing
> signals does not seem to have led to faster communication by
> regulators or manufacturers.  What continues to happen is that
> nothing is said or done until the available data have been
> considered and discussed at length by the authority and the
> manufacturer.  If a decision is then made, it may be announced
> publicly, or it may merely be implemented without any obvious
> public statement, eg by merely amending a product's data
> sheet.  
> 
> For example, in 1989, over six years after dependence on
> benzodiazepines had emerged as a serious and widespread
> problem, the regulators in Britain and the manufacturers woke
> up to the fact that lorazepam had for more than a decade been
> prescribed in North America and elsewhere at about half the
> dosage recommended in the UK.  Thus the data sheet had for
> many years recommended that British patients with "mild
> anxiety" be given the same dose of lorazepam as an American
> would get for "severe anxiety".  The UK recommended dosage for
> moderate and severe anxiety was suddenly changed from "up to
> 8mg daily" to "1-4mg daily", with dosages by half as much
> again for elderly people.  This crucial change in the data
> sheet was not publicised, and most prescribers probably
> remained unaware of it.  John Abraham, a sociologist, has 
> recently documented how in the 1980s the regulatory
> authorities in the UK and the USA suspected various non-
> steroidal anti-inflammatory drugs of harmful effects, but did
> nothing for months or years.4   
> 
> Two other examples of long delayed action are the cases of
> nomifensine and glafenine.  That nomifensine did serious harm
> was first clearly apparent in 1984 when cases of hepatitis
> were reported,5  and many other publications soon followed.6 
> After more than a year of confusing statements from the
> company and regulatory authorities7 the manufacturer withdrew
> the drug worldwide when suspension of the licence in the UK
> was imminent.8  The story of glafenine is even worse.  An
> early report in 1984, of two cases of severe hepatitis in the
> Netherlands, also cited six others from the French
> literature.9  The manufacturer then approached the reporting
> physicians requesting withdrawal of the article, but they
> refused.  An analysis of 38 cases of liver injury by the drug
> with a review of the literature was rejected by the editor of
> Hepatology under pressure from the manufacturer, but then
> appeared in Liver.10  Only in 1990 was the French data sheet
> revised to restrict the use of the drug.11  The first country
> to order withdrawal of the drug was Belgium, in 1990, against
> the recommendation of the European Committee for Proprietary
> Medicinal Products.12  In France and the Netherlands it
> remained on sale until Roussel withdrew the drug worldwide in
> 1992, but generic versions may still exist in some developing
> countries.
> 
> In some countries, such as Australia, Sweden and the USA, the
> public has direct access to anonymised ADR data.  The argument
> for free access is that the data are provided by the public
> for the benefit of everybody and belong to the public, with
> the proviso that individual privacy must be protected.  Other
> major national authorities, like those in the UK, France and
> the Netherlands, remain secretive about ADR reports and allow
> little or no public access to their information and
> deliberations.  The UK Committee on Safety of Medicines (CSM),
> for example, provides only summary print-outs of reported
> reactions by drug (on request), and publishes Current Problems
> in Pharmacovigilance, a bimonthly bulletin for doctors
> summarising some of its concerns and conclusions.  Two recent
> examples from the UK illustrate the difficulties created by a
> lack of openness. 
> 
> The demotion of co-trimoxazole
> The first example concerns the decision to restrict the use of
> co-trimoxazole, the combination of trimethoprim with
> sulphamethoxazole (Current Problems in Pharmacovigilance
> 1995).13  The CSM said that it "should now only be considered
> for use in acute exacerbations of chronic bronchitis and
> infections of the urinary tract when there is bacteriological
> evidence of sensitivity to co-trimoxazole, and good reason to
> prefer this combination ... to a single antibiotic."  This
> conclusion could have been reached many years ago, and the CSM
> has not explained why its decision was so long delayed.  For
> example in 1986 the British National Formulary noted reports
> of deaths in elderly patients treated with the drug "and
> almost certainly associated with the sulphonamide component",
> and advised: "Co-trimoxazole should not be prescribed in the
> elderly unless there is no acceptable alternative." (BNF
> 1986).14  The CSM, however, cites a recent post-marketing
> study which found serious hepatic, renal, blood and skin
> disorders to be similarly rare with co-trimoxazole and
> trimethoprim (Jick and Derby 1995).15   But the study
> identified only patients referred to a specialist or admitted
> to hospital with such a disorder within 45 days of receiving
> the prescription.  Patients in whom a serious reaction was
> recognised and managed by the GP alone, as might happen
> especially with severe rashes, were not sought.  The
> implication of the CSM's statement, that the presence of
> sulphamethoxazole hardly matters, is hard to accept since
> sulphamethoxazole alone can undoubtedly cause major adverse
> effects (Dollery 1991).16
> 
> Gestodene and desogestrel: the risk of venous thromboembolism
> The second example is the CSM's warning in October 1995 that 
> oral contraceptives (OCs) containing gestodene or desogestrel
> (third-generation progestagens) carry a higher risk of venous
> thromboembolism and that women using these should consider
> changing to another brand (Carnall 1995).17
> 
> The warning was based on three unpublished studies which had
> not at the time been formally peer reviewed.  All three found
> about double the risk of venous thromboembolism (VET) with
> gestodene and desogestrel products in comparison with OCs
> containing other progestagens.  The first, a large WHO
> collaborative case-control study of cardiovascular disease and
> OCs was undertaken in 17 countries.  It was completed in July
> 1995, and analysed 829 cases of VET and 2,641 controls from
> the nine countries in which third-generation OCs were used.18 
> The second, from the Boston Collaborative Drug Surveillance
> Program, analysed the occurrence of VET in a UK general
> practice cohort of 238,130 women who had received a
> prescription for an OC containing levonorgestrel, gestodene or
> desogestrel.19 The third was a transnational case-control
> study conducted in five European countries, funded by
> Schering, the leading manufacturer of gestodene.20  It was due
> to continue until 1996, but this is now uncertain.    
> 
> The WHO study was made because the association between OC use
> and VET had not been examined since the 1970s when OCs
> contained higher doses of oestrogen and progestagen than now,
> and because none of the earlier studies had been done in
> developing countries.  The finding of a higher risk of VET
> with gestodene and desogestrel came as a surprise, and this,
> together with publicity in the media, prompted the CSM to
> commission the UK cohort study.  The transnational study was
> set up at the request of the German regulatory authority to
> follow up German spontaneous reporting data which in 1990 had
> strongly suggested higher risks of thromboembolism with
> gestodene-containing OCs.  This led to much controversy in the
> press and television.  Schering had argued that highly
> publicised deaths had stimulated selective reporting, and that
> the claimed higher risk was an artefact.
> 
> The sudden announcement by the CSM caught prescribers and
> users of OCs completely unprepared.  The "pill panic" in the
> media drove many thousands of women to consult their
> doctors,21 who had not been briefed.  A heated debate ensued
> over whether the CSM's decision was justified, whether its
> announcement should have been delayed until the data were
> published, and over the precipitate manner in which it was
> issued.  Before long there was general agreement that the
> decision was necessary and correct, but that its communication
> had been handled badly, and the health minister said that the
> government would review the incident to learn from it. 
> However, official actions in other countries ranged from the
> imposition of stronger restrictions, in Germany, to a decision
> in the Netherlands and Canada to wait and consider the
> published studies, and in the USA not to advise switching to
> other products (Carnal et al. 1995).21  The European Union's
> Committee on Proprietary Medicinal Products has also decided
> to wait and see.  
> 
> Although doubts about the relative safety of gestodene emerged
> in 1990, regulators did not publicly acknowledge them.  Nor
> did they help independent scientists to examine all the
> relevant data (including prescribing figures, essential to
> provide denominators for the calculation of risk) - perhaps
> because they were insufficient and might have been wrongly
> interpreted.  In Germany the authority privately requested
> Schering to undertake a case-control study, in Britain the
> Medicines Control Agency considered the available evidence
> inadequate and dismissed the doubts.  Whether the CSM was
> consulted at that time or only later remains a minor official
> secret.  The attitude changed early in July 1995 when the CSM
> saw the early results of the WHO Collaborative Study, and a
> highly critical television programme about gestodene-
> containing OCs was broadcast in the UK (Granada, World in
> Action 10 July).  The CSM asked workers from the transnational
> case-control study to "expedite their results", and
> commissioned the cohort study using data routinely collected
> from British general practitioners.  The analysis of VET in
> the transnational study was completed on 8 October, and
> promptly discussed with members of the MCA and CSM.  The CSM
> quickly made its decision20 and announced it on 18 October. 
> The relevant data from the WHO study and the GP cohort study
> were not published until 16 December.  
> 
> In retrospect, what could the CSM and other authorities have
> done to alert the world to the potential problem during the
> five years of growing doubts about the relative safety of
> gestodene?  To know about the doubts would have helped women
> and their doctors.  They could then have discussed among
> themselves whether to act on them or not, to weigh any real
> advantages of their gestodene OC against the doubt and 
> potential risk.  A joint statement by regulators and
> manufacturers about their plans for further investigations
> would have made it clear that there was a problem to be
> resolved, instead of appearing to ignore or deny it.  This
> would have allowed prescribers and independent experts to
> rethink their prescribing policies, for example deciding in
> what circumstances gestodene OCs should no longer be the first
> choice.  That would not have pleased the manufacturers, since
> it would almost certainly have decreased the sales of their
> leading product in an uncontrollable way.  The position of
> desogestrel creates additional complications, since its
> relative safety was not questioned until the data from the WHO
> study came out in July 1995.  If doubts had been officially
> expressed about the gestodene-containing products Femodene and
> Minulet, these OCs would meanwhile have lost some market share
> to Marvelon and Mercilon, the desogestrel products.  But
> greater openness from the start could have minimised the alarm
> and confusion.  
> 
> It may be noted that the UK and Germany where there was most
> public concern were the first countries to take strong action. 
> After Britain's action the Norwegian regulatory authority
> advised women against taking Marvelon, the only third-
> generation OC marketed in Norway.  Other national authorities
> have so far done nothing.  After December 1995, the month of
> publication of the relevant studies, they could become legally
> liable, perhaps jointly with the manufacturers, for harm
> suffered by women from thromboembolism associated with OCs
> containing desogestrel or gestodene.
> 
> Litigation to prevent publication of warnings
> A recent judgement in Germany has had the extraordinary result
> that professional and other media can no longer publish
> unfavourable official assessments of a medicine before the
> Federal Institute for Drugs and Medical Products (BfARM) has
> made a final decision on it.  The case concerned Cordichin, a
> combination of quinidine and verapamil for the treatment of
> arrhythmias marketed by Minden Pharma, a subsidiary of BASF. 
> This product, which would on the face of it seem more
> dangerous than useful, was associated with several deaths,
> leading the BfARM to proceed to withdraw the licence.  The
> first step in this process is to send the BfARM assessment of
> the product to the manufacturer and other parties formally
> involved in the assessment procedure.  This assessment is not
> confidential, and was published (as usual) by the
> Pharmazeutische Zeitschrift, the official journal of the
> German Association of Pharmacists.  However, Minden Pharma
> obtained an injunction preventing the Deutsche Arzteblatt, the
> official journal of the German medical profession, from
> publishing it, on the ground that it would be libellous.  The
> journal appealed, but the injunction was confirmed by the
> superior court.23  The legal decision conflicts with the
> statement made by the lower court that decided to end the
> thalidomide case in 1970: "The scientific uncertainty as to
> whether a drug causes an adverse effect or not must not be
> resolved at the patient's expense, because the patient's right
> to health ranks above the manufacturer's right to market his
> product without hindrance."24  The recent decision will
> certainly influence other German courts, though it is not
> binding.  The only remedy may be to change the law.   
> 
> Conclusion
> Twenty years ago Inman described the main reasons why doctors
> failed to report adverse drug reactions: he called them "the
> seven deadly sins" of voluntary reporting.25  Perhaps
> regulatory authorities also commit some similar sins in not
> disclosing their data or their deliberations.  In their case 
> complacency may be an uncritical belief in the correctness of
> the original licensing decision.  Fear of involvement in
> litigation is probably a common motive: decisions that are not
> explained are harder to attack.  Ignorance of the need for
> disclosure and clear communication, including the 
> communication of uncertainty, cannot be ruled out.  Diffidence
> is to be expected about communicating data that might cause
> controversy and might affect company profits and share prices. 
> And finally lethargy is always a problem when busy
> understaffed agencies face tasks given low priority.  But one
> sin that is probably very rare among doctors seems common
> among regulators and is more deadly: the reluctance to hurt
> industry, to "rock the boat".4
> 
> Knowledge of adverse reactions comes from the experiences of
> individual patients and their doctors.  Their reports are
> collected and analysed on their behalf for the benefit of all
> future patients, that is society as a whole.  This is a
> powerful argument for open access to the anonymised data.  To
> interpret data on adverse drug reactions correctly it is
> necessary to relate them to the population exposed to the
> drugs concerned.  Drug prescription or utilisation data should
> therefore be equally accessible.  Industry needs current
> prescription data to guide its sales efforts and pays for
> expensive market research to obtain them, but somewhat older
> data - perhaps 12 months old - are sufficient for analysing
> ADR reports and should be available to all bona fide users,
> including teachers of clinical pharmacology and therapeutics,
> research workers, and representatives of patients or
> consumers.  It is encouraging that this is now the norm in
> several leading countries, including Australia, New Zealand,
> Sweden and the USA, and may soon be adopted by all the other
> participating countries in the WHO collaborative ADR
> monitoring system.
> 
> The challenge of drug safety is clearly that therapeutic risk
> should be minimised and benefits maximised for every patient. 
> The complete availability of relevant data is essential for
> therapeutic decisions.  Not every doctor or patient will wish
> to see all the data all of the time, but it should be
> considered a basic right to have access both to the data and
> the basis for regulatory decisions on request.
> 
> 
> 
> 
> Must also put in one or more paras about Freedom of info and
> the UK Medicines Info Bill of 1994, and official & industry
> resistance to disclosure of prescribing data. Where?
> 
> 
> 
> 
> 
> 
> 
> 
> 
> REFERENCES
> 1.   Inman WHW, ed. Monitoring for drug safety. 2nd ed.
>      Lancaster: MTP, 1986.
> 2.   WHO Anniversary Symposium. Adverse drug reactions: a
>      global perspective on signal generation & analysis.
>      Uppsala: WHO Collaborating Centre for International Drug
>      Monitoring, 1988.
> 3.   Fucik H, Edwards IR. Impact and credibility of the WHO
>      adverse reaction signals. Drug Information Journal 1996;
>      in press.
> 4.   Abraham J. Science, politics and the pharmaceutical
>      industry: controversy and bias in drug regulation. 
>      London: UCL Press, 1995.
> 5.   Vaz FG, Singh R, Nuruzzaman M. Hepatitis induced by
>      nomifensine.  BMJ 1984; 289: 1268-9.
> 6.   Editorial. Trouble with nomifensine. Drug and
>      Therapeutics Bulletin 1985; 23: 98-100.
> 7.   Schonhofer PS. Germany: the nomifensine affair. Lancet
>      1991; 338: 1448. 
> 8.   Committee on safety of Medicines. CSM Update: withdrawal
>      of nomifensine. BMJ 1986; 293: 41.
> 9.   Verhamme M, de Wolf-Peters C, van Steenbergen W. Hepatic
>      injury due to glafenine. Neth J Med 1984; 27: 35-39.
> 10.  Stricker BHC [et al]. Glafenine-associated hepatic
>      injury: analysis of 38 cases and review of the
>      literature. Liver 1986; 6: 63-72.
> 11.  Editorial. Glafenine: precautions d'emploi. Revue
>      Prescrire 1990; 10: 62.
> 12.  Herxheimer A. Belgium: withdrawal of glafenine. Lancet
>      1991: 337: 102.
> 13.  Editorial. Revised indications for co-trimoxazole
>      (Septrin, Bactrim, various generic preparations). Current
>      Problems in Pharmacovigilance 1995; 21: 6.
> 14.  British National Formulary No 11, 1986: 208.
> 15.  Jick H, Derby L. Is cotrimoxazole safe? Lancet 1995; 345:
>      1118-19. [letter]
> 16.  Dollery C, ed. Therapeutic Drugs. Edinburgh: Churchill
>      Livingstone, 1991, vol 2: S135-6.
> 17.  Carnall D. Controversy rages over new contraceptive data.
>      BMJ 1995; 311: 1117-18.
> 18.  WHO Collaborative Study of Cardiovascular Disease and
>      Steroid Hormone Contraception. Effect of different
>      progestagens in low oestrogen oral contraceptives on
>      venous thromboembolic disease. Lancet 1995; 346: 1582-88.
> 19.  Jick H, Jick SS, Gurevich V, Myers MW, Vasilakis C. Risk
>      of idiopathic cardiovascular death and non-fatal venous
>      thromboembolism in women using oral contraceptives with
>      differing progestagen components. Lancet 1995; 346: 
>      1589-93.
> 20.  Spitzer WO. Data from transnational study of oral
>      contraceptives have been misused.  BMJ 1995; 311: 1162.
>      [letter]
> 21a. Armstrong JL, Reid M, Bigrigg A. Scare over oral
>      contraceptives: effect on women attending a family
>      planning clinic. BMJ 1995; 311: 1637.[letter] 
>   b. Seamark CJ. ...: effect on women in general practice in
>      Devon ... BMJ 1995; 311: 1637. [letter]
>   c. Davies AW, York JR, Jones SR. ...: ...and south Wales.
>      BMJ 1995; 311: 1637-8. [letter]
> 22.  Carnall D, Karcher H, Lie LG, Spurgeon D, Josefson D,
>      Zinn C. Third generation oral contraceptives - the
>      controversy. BMJ 1995; 311: 1589-90.
> 23.  Oberverwaltungsgericht Munster. Judgment of 20 Nov 1995.
>      Az 13B 619/95. 
> 24.  Landgericht Aachen. Contergan, Einstellungsbeschluss, 18
>      Dec 1970. Az 4 KMs 1/68: p49-50.
> 25.  Inman WHW, Weber JCP. [Post-marketing surveillance in the
>      general population:] The United Kingdom. In: Monitoring
>      for drug safety, 2nd ed. Inman WHW, ed. Lancaster: MTP,
>      1986, 37-38.
> 
>