Journal ofNeurology, Neurosurgery, and Psychiatry 1992;55(Suppl):26-31

Hither neurology: research

Charles P Warlow

AbstractNeurological disability may be preven-

ted, or it may be alleviated if preventionis impossible or ineffective. Research intoprevention and alleviation can be"laboratory" or "clinical", the latterbeing no less scientific than the former.All proposed treatments must beproperly evaluated to ensure that effec-tive interventions are widely adoptedand ineffective ones abandoned. Unless anintervention has a major effect on out-come (which most do not), the mostefficient assessment is by random alloca-tion of patients to the new interventionversus the old. Although there were, andstill are, forces opposed to the proper

evaluation of treatment, there are strongeconomic clinical arguments in itsfavour, which will lead to appropriatetargeting of scarce health resources.

"One should be delightfully surprised when any

treatment at all is effective, and always assume

that treatment is ineffective unless there isevidence to the contrary". A L Cochrane, Effec-tiveness and efficiency (1971).

Department ofClinicalNeurosciences,Western GeneralHospital, Edinburgh,UKC P Warlow

Some equate science, and therefore research,with laboratory based activities. There is an

argument that rats and molecules providescientific solutions to human problems and theresults can be applied by clinicians, who are

themselves seldom "scientific". One hears talkof clinical research as though it is somethingother than science. This view may come from:basic scientists most of whom are notmedically qualified and know little, and maycare even less, about the difficulties, rules andsolutions of clinical science; from clinicianswho feel intellectually inferior to basic scient-ists and find solace, and perhaps profit, in the"art" of medicine; and from clinicians whohave become bewitched by basic scientists,join them in their laboratories and yet struggleto keep up because of their clinical commit-ments.

I must confess to being intimidated by basicscience and basic scientists but only because Ido not have their training, aptitude or back-ground; I have the same problem with winetasting. Of course, this does not mean that Iam against basic science, or wine tasting, far

from it. What I am against, however, is theuncritical application to patients of ideas,hypotheses and solutions generated by basic

scientists (and others) without proper evalua-tion; what works in the rat does not neces-sarily work in humans. The application oftreatments and diagnostic techniques which,according to the basic scientisits, or others,ought to work but do not, has caused harmwhile the introduction of treatments and diag-nostic techniques which really work (accord-ing to the clinicians), but no one knows why,has been beneficial. An example of the formeris the use of high concentrations of oxygen inpremature babies which seemed a good ideabut turned out to cause retrolental fibroplasia,'and an example of the latter is the use ofelectroconvulsive therapy to alleviate depres-sion.' Many of the recent spectacular succes-ses in medicine have come from basic andclinical scientists working together, neitherignoring each other nor attempting to do eachother's work; patients provide the problems,which clinicians identify and place in perspec-tive, basic scientists find possible solutionswhich are then tested on patients by the clin-icians. Beta blockers, thrombolytics, H2-antagonists, CT scanning and foetal ultra-sound were all introduced this way.With a few notable exceptions, clinical

neurologists (and clinicians generally) do nothave the training, facilities or time to build upand orchestrate a large laboratory-basedresearch programme. Usually this is carriedout much better by full-time basic scientistsworking with neurologists, maybe within thesame university department or pharmaceuticalcompany. Most clinicians tend to be muchmore comfortable with clinical researchbecause they are familiar, and competent todeal with patients. Clinicians are capable ofmeeting the intellectual demands of any sortof science, clinical or basic, if the criteria forUnited Kingdom medical school entry areanything to judge by. It seems a mistake for ayoung neurologist to work for a couple ofyears in a laboratory and then obtain a consul-tant position looking after countless patientswithout any prospect of working in alaboratory again; but often this is preciselywhat happens. Surely it is preferable to giveequal encouragement and prestige to youngneurologists to do clinical research, at least inthe sense of evaluating diagnostic techniquesand treatments, because that is what they willbe able to do as consultants, given a snmallamount of time, a microcomputer, and reason-able contact with other researchers.The main contribution of most clinical

neurologists is our ability, and facilities, to doclinical research with our patients. For

26

Hither neurology: research

neurological disability, our major currentobjectives must be to evaluate rigorously theprevention, cure, alleviation and palliation ofthe common disabling disorders such asstroke, epilepsy, multiple sclerosis andParkinson's disease. Whether the suggestions,often from the basic scientist, are concernedwith vaccines (such as, polio), drugs (such as,vigabatrin), surgery (for example, brain tissuetransplantation), psychology (for example,cognitive rehabilitation), physiotherapy (suchas the Bobath technique), or servicerearrangements (for example, stroke units), inthe end they all require proper evaluation.Competent evaluation is the greatest con-tribution clinicians have to offer both science,and their future patients.

WHY BOTHER WITH EVALUATION?Evaluation is fashionable, particularly amongmanagers whose sights may be set no higherthan "before and after" comparisons of costwhich is so much easier to measure than qualityof outcome. There are, however, a number ofvalid scientific reasons for encouraging evalua-tion, which is clinical research into the care ofpatients.1 Limited resources should be appropriatelydirected. What is deemed appropriate dependson whether one wishes to concentrate on broadbrush public health measures (such as, vaccina-tion or changing the nation's diet), on morefocussed efforts with individuals (for example,treating patients with severe hypertension), oron a combination of both. However, before anysuch philosophical and political decisions arepossible it is necessary to known what works,how much it costs, what are the adverse effects,and what is the overall likely outcome if thetreatment is introduced into routine clinicalpractice. We must conserve resources by rejec-ting the useless, while implementing usefultreatments.2 Clinical practice should be directedappropriately so that clinicians neither wastetheir time nor expose patients to unnecessaryrisk by providing treatments, ofany sort, whichare ineffective, even though they shouldtheoretically work. For example, carefulevaluation has quite rightly inhibited theuncritical introduction or continuation ofextracranial-to-intracranial by-pass to preventstroke,3 bed rest to improve the outcome oftwin pregnancy,4 the use of a new "improved"

Table I The chance of a prison sentence for criminaldamage: 17-20 years old

Powys 5-1%Surrey 5-9%Suffolk 8-1%Norfolk 8-3%Gloucestershire 9-8%Nottinghamshire 21 1%South Yorkshire 21 9%Northumbria 25-3%Kent 26 1%Hertfordshire 31 6%

Home Office statistics for Crown courts in 1988 (report fromThe Guardian 23 January 1990).

suture material for episiotomy,5 hyperbaricoxygen to improve the prognosis of multiplesclerosis,6 haemodilution for the treatment ofacute ischaemic stroke,7 and antifibrinolyticdrugs to reduce the risk of rebleeding afteraneurysmal subarachnoid haemorrhage.83 Large and unacceptable variations in clin-ical practice must be reduced. It cannot be rightto treat the same patient in one way in one placeand in quite another way in another place. Yetthis is what often happens in clinical practice,9as it does in legal practice (table 1). Suchvariation is not usually a reflection of ignoranceon the part of the clinician (or judge), but lackof evaluation combined with inadequate data.Without proper evaluation, the clinician has tomake a decision whether to give, or not to give,a particular treatment to a particular patientand may jump one way or the other dependingon whim, the result in the previous patient,philosophy, pressure from managers, pressurefrom peers, pressure from salesmen, or possi-bly profit. After proper evaluation which caninfluence clinical practice there is, rightly,much less room for manoeuvre and for non-scientific influences on which patients shouldreceive what treatments.4 Even at worst, evaluation protects us fromthe riducule of future generations. I would notwant any treatment that I introduced to beincluded in books such as "Need your doctorbe so useless?"'.10 nor used as teaching materialabout how not to do something, such as theepidemic of unnecessary tonsillectomy in themiddle of the twentieth Century."

WHY ARE NEW TREATMENTS NOT WIDELYADOPTED?There are several reasons why a new treatmentis not taken up in routine clinical practice: thetheoretical basis may be regarded as sounsound that no evaluation is carried out (forexample, copper deficiency as a cause foratheroma); the advantage may be perceived(rightly or wrongly) as rather modest, and notenough to outweigh the risks (such as, carotidendarterectomy in the prevention of stroke);financial and political pressure, or just fashion,which may explain the controvery surroundingthe introduction of front seat belts and thecontinuing controversy about the introductionof rear seat belts, government policy on smok-ing, and heart transplantation; the innate con-servatism of most doctors (for example, thedifficulty ofintroducing very early operation foraneurysmal subarachnoid haemorrhage); andlastly prejudice against the investigator who isintroducing the treatment.

All these reasons for doing nothing arereinforced when a new treatment is suggested ifthe "experts" cannot agree with each other,and "experts" usually do disagree whenevaluation is inadequate. After one, or prefer-ably more, excellent randomised clinical trialsit is very difficult for anyone to resist, for anyreason, the introduction of an effective treat-ment. For example, after decades of debatingon the inadequate data, thrombolysis was even-

27

Warlow

tually introduced as the routine treatment foracute myocardial infarction as a result of twovery large clinical trials which overwhelmedany theoretical, political, financial or otherarguments.'2 Of course, to influence clinicalpractice, any evaluation has to be scientificallyconvincing and free from conflict of interest.For drugs, there is no question that the mostcost-effective and convincing method to com-pare treated and untreated patients is therandomised trial. Of necessity, economic andpolitical policy must make do with unreliable"before and after" comparisons'3 but inmedicine we have the great advantage of ran-domisation. When randomisation is not used,there is so often continuing and sterileargument such as whether consultation with aneurologist improves the outcome for patientswith non-organic headache. Although the gen-eral practitioner consultation rate for headachemay fall after seeing a neurologist,'4 the totalnumber of consultations does not.'5 However,we do not know what this means as thesecomparisons were "before and after" theneurological consultation, whilst headachesmay remit spontaneously. Furthermore with-out any comparison between neurologists and,for example, general physicians we do not knowwhether neurologists are the best type ofspecialist to empty general practitioners' sur-geries if that is the outcome of interest. Ofcourse, in practice, curing the headache is moreimportant. Extraordinarily, these two studiesare almost the only ones where the effectivenessof neurologists has been examined so it ishardly surprising that we still do not knowwhether neurologists are really effective, and ifso whether they are more effective than generalphysicians. Naturally, neurologists themselvesbelieve this to be the case, at least for themanagement of rare neurological disordersrather than common diseases such as strokewhich so many neurologists seem to want toleave to other people.

Unfortunately, although random allocationis now the accepted method for the evaluationof drugs, there is still resistance to its use in theevaluation of surgery and even more for theevaluation of most of the treatments used inneurological rehabilitation. This is particularlyunfortunate since surgery and rehabilitationcan be very costly, and even dangerous, and it isan indefensible double standard to insist onrandomisation for the evaluation of drugs butnot of other treatments. However innovative,and theoretically attractive, treatments may notbe taken up into clinical practice if they are notproperly evaluated in randomised clinical trialsby clinical scientists whose business it is to testout whether theories work in practice. Alter-natively, a theoretically attractive treatmentmay be used for years before a proper evalua-tion shows it is useless. As Thomas Huxleywrote, "the great tragedy of science the slay-ing of a beautiful hypothesis by an ugly fact".'6In this context, the ugly fact may be a com-petently performed but negative clinical trialon an adequate number of patients. Or, if thereare several clinical trials addressing the sameissue, a competently performed but negativeoverview (or meta) analysis.

RESISTANCE TO EVALUATIONDespite its vogue, proper evaluation is con-stantly under threat from several directions.This is not the place to debate in detail theethics of randomisation, but there is no doubtthat ethical arguments can inhibit randomisedtrials if the new "improved" treatment must begiven because it is bound to work (that is, theorysays it will). This may have been true for thevery few treatments which have major andobvious effects (for example, penicillin formen-ingococcal meningitis) but most treatmentshave modest, and yet potentially useful, benefitsas well as modest, and yet potentially harmful,adverse effects and the balance of benefit andharm cannot be predicted from theory alone.Moreover, non-randomised comparisonsprovide no guarantee that like (patients) arebeing compared with like.The great ethical advantage of randomised

trials is not just for future patients but also forpatients in the trial; not only do patients inrandomised trials tend to fare better thanexpected (perhaps as a consequence of selectionof better risk cases for the trial, better ancillarytreatment during the trial, and a mistakengloomy view of the natural history before thetrial) but also those randomised to the oldtreatment will be spared the adverse effects of anew treatment which simply does not work, oris so unpleasant it cannot be used in routineclinical practice. Without randomisation, atoxic and/or ineffective treatment which oughtto work may be given to thousands of patientsbefore it is abandoned (for example, tonsillec-tomy for sore throats) or an effective treatmentcan be abandoned in error so denyingthousands of patients an improved prognosis(which nearly happened with thrombolysis foracute myocardial infarction). Clumsy attemptsat detailed "informed consent" may be sodistressing to patients that they will not enterrandomised trials and become so confused andanxious that they cannot even decide whichtreatment they want from among the alter-natives offered.

Basic scientists may sometimes be so over-whelmed by the elegance of their solutions tohuman problems that they insist that thecrucial step of proper evaluation can be omit-ted. The task of the clinical scientist is to resistsuch pressure and to remind the basic scientistwhat Thomas Huxley had to say, and indeedwhat Richard Asher, that consummate generalphysician, wrote: "One of the most importantthings about treatment is that it should beeffective-notmerely that it oughtto be effective.A remedy which is known to work, thoughnobody knows why, is preferable to a remedywhich has the support of theory without theconfirmation of practice"." Also, some basicscientists tend to deride clinical research yetthey may dominate the research funding bodiesso attracting disproportionate funding to basicrather than clinical science. Of greater concernis that clinical trials can take so long that thebasic scientists have changed their proposedtreatments several times before the first one hasbeen properly tested. Another and irritatingproblem, for the clinical researcher, is the basicscientist who simply uses the wrong model to

28

Hither neurology: research

predict the consequences of, and solutions tohuman diseases.

It can be difficult to obtain funding forrandomised clinical trials of treatments likecoronary care, physiotherapy, surgery, aspirin,or service rearrangements when suchendeavours are considered "unscientific" bythe research councils, or "not profitable" byindustry, or "someone else's business" byhealth authorities. Such problems are com-pounded where there is no international sourceof funding for international multicentre trials,other than the multinational pharmaceuticalcompanies. Therefore, if multicentre inter-national trials are required to obtain largenumbers of patients quickly there is little hopeof funding unless the treatment is a patenteddrug. Institutions such as the European Com-munity ought to provide the resources but, sofar, they have not.The NHS reforms may be very dangerous

for clinical research, especially for the evalua-tion of treatment, unless the purchasers of care(the health authorities) and/or providers(hospitals) see it as part of their remit toevaluate treatment rather than merely to pur-chase or provide it. The market solution wouldbe to make it clear, particularly to theproviders, that investment in evaluation wouldbe to their advantage since the introduction of acost-effective treatment would attract contractsfrom the purchasers of health care in thecompetition with other providers who may beusing ineffective, out-dated or costly treat-ments. The problem is that unless such atreatment is kept a commercial secret it wouldbe shared by all the competing providersbecause clinical scientists are in the habit ofsharing their ideas and results. Indeed, it wouldbe unethical if a new treatment discovered by aprovider hospital was kept secret so that hos-pital could obtain a contract, and yet this iswhat the NHS reforms might in practiceencourage. An acceptable alternative would beto take evaluation out of the purchaser/provider relationship by insisting that it isfunded either centrally by Government, or, bytop slicing a proportion of the budget of apurchasing authority and insisting it is spent onresearch. Furthermore, the NHS reforms maybe damaging if every last item of the currentlyhidden costs of clinical research is charged for(for example, providing a copy of a patient'sdischarge summary to a reseacher), if academicclinical departments are seen as unprofitableand unaffordable luxuries, if potential clinicalacademics are tempted by inflated salaries to dojust clinical work for institutional or personalprofit, and if there are so few medical staff thatthere is simply no time for research.

Paradoxically, the mushrooming of multi-centre clinical trials, and now overviews, mayreduce the perceived contribution of anindividual clinician so much that beinginvolved as a minute cog in a large evaluationmachine becomes profoundly unattractive.One only hopes that this problem can bealleviated by making multicentre trials lessarduous, more fun to take part in, and by givingthem an educational component. Personally, I

learnt a great deal about epilepsy by myinvolvement in a multicentre trial of anticon-vulsant drug withdrawal and I hope that otherslearn about cerebrovascular disease by theirinvolvement in the trials which I organise.Clinicians find it hard enough to find the time tocontribute to multicentre trials already andmight abandon them completely if clinicalscience is "rubbished" by basic scientists, iftheclinicians are overwhelmed by routine clinicalwork, and if they are not given credit for beingthe resource for the proper evaluation of treat-ment for their patients. Without clinical scient-ists there would be a jungle of untested andtoxic treatments, some effective and othersineffective, and no-one would know whichtreatment to use although, as clinicians, theywould have to do something.

Perhaps the main distinction between thebasic scientist and clinician is that the formercan admit to uncertainty and do an experiment,while the latter can admit to uncertainty yet hasto treat the patient. However, the scientificclinician can take advantage of uncertaintywhen it is as ethical, or even more ethical,randomly to allocate the patient in a clinicaltrial which is large and effective enough toresolve the treatment dilemma.

THE MOVE TOWARDS EVALUATIONGiven the need to evaluate new treatments sothat clinical practice is improved, and despitethe difficulties, it is worth emphasising thatthere are several strands which are now strong-ly encouraging evaluation. Not only is evalua-tion fashionable among managers (although asnoted above they need to be educated how to doit properly), but clinical trial methodology hasbecome better, simpler and cheaper. Indeed,the fact that randomisation is an option makesevaluation very cost-effective in a climate inwhich resources for everything are limited.The challenge is to persuade the purchasersand providers of health care that evaluationmust be considered in their budget so theirresources can be directed appropriately. Also,there is increasing patient pressure for properevaluation and this is worth encouraging. It isvery cheering to hear patient representatives onethical committees criticising non-randomisedtreatment comparisons as being unscientific andtherefore unethical, and medical charitiesencouraging their members to enter random-ised clinical trials.There is no doubt therefore that there are

plenty of forces to be harnessed when the goinggets tough and the demoralised and under-funded clinical scientist starts dreaming ofearly retirement, the golf course or privatepractice. One must not be daunted by the costof clinical trials because they do seem to becost-effective in strictly economic terms, evenwhen carried out in North America,'8 nor bythe consequences of introducing new treat-ments which do not necessarily have to be moreexpensive than the old (for example, earlymobilisation after myocardial infarction,phenothiazines for schizophrenia, streptokin-ase for acute myocardial infarction, aspirin toprevent stroke, no bed rest for twin pregnancy,

29

Warlow

Table 2 Association of Medical Research Charities;expenditure on biomedical research 1989-90

£ million

Cancer charities* 85 0British Heart Foundation 14-6Arthritis and Rheumatism Council 11 0Cystic Fibrosis Research Trust 2 6Multiple Sclerosis Society 2-4British Diabetic Association 2-1Muscular Dystrophy Group 1 9Chest, Heart and Stroke Association 1 0Parkinson's Disease Society 0 9

*Cancer Research Campaign, Imperial Cancer Research Fund,Ludwig Institute for Cancer Research, Tenovus Cancer Fund.

no bed rest to prevent post lumbar punctureheadache, and angioplasty rather than surgeryto relieve certain arterial obstructions).

Indeed, in the United Kingdom the NationalHealth Service (NHS) gives us a particularadvantage in clinical research because healthcare is free at the point of delivery, andtherefore patients are not financially disadvan-taged by being in clinical trials, and it is equallyavailable to all patients. In the NHS, medicalrecords are held by general practitioners, withwhom almost all patients are registered, andthese records follow the patients when theymove. This provides a unique file includinghospital letters and discharge summaries aswell as the general practitioner notes. Thesecan be used for detecting outcome events suchas stroke, or antecedent events such as vaccina-tions or fractures. Moreover, follow up can bearranged with general practitioners withoutpatients necessarily having to go to a hospitalclinic. In the UK, doctors are still trustedbecause they do not, usually, make a directprofit out of their patients who are likely tocollaborate with well explained sensible andnecessary clinical research.

Clinical research is still regarded as a reason-able use ofNHS resources to be added to thoseprovided by third party sponsors such asuniversities, research councils, industry andcharities. However, such sponsors cannot fundeverything, such as additional follow upappointments or investigations, which may bemore than is strictly required for routinepatient care. The new NHS reforms mightsweep away these advantages, or lead toattempts to cost every last item so sinkingclinical research in complex demarcation dis-putes.

WHO WILL PAY?Many countries cannot, or will not, finance anymedical research at all so in the UK we arefortunate to have a long and secure enough

Table 3 Interventions being evaluated or requiringevaluation in the Edinburgh department of ClinicalNeurosciences

Nurse practitioner for multiple sclerosis patientsFamily care officer for motor neuron disease patientsComprehensive stroke serviceProgrammed investigation unitDistrict based neurology clinicsAspirin/heparin/both/neither for acute ischaemic strokeCarotid endarterectomyEssential amino acids for motor neuron diseaseCognitive rehabilitation

tradition of medical research to ensure that,come what may, it will continue. However, thepattern of funding is changing'9 with relativelymore coming from industry and medicalcharities and relatively less from Governmentalsources through the research councils anduniversities. The disadvantages of leaving mat-ters to the industrial market place are thatresearch will be done only for profit rather thanbeing led by scientific and public healthpriorities; underfunded clinical departmentswill be tempted into patented drug researchwhich they do not really want to do just to keepsolvent; clinical researchers will be temptedaway from the universities and NHS to indus-try; non-patented interventions will beignored; and sometimes the results, or theirinterpretation, will be presented primarily tosuit the marketing rather than the medicaldepartment of the sponsoring company. Ofcourse one cannot, and should not, expect acompany responsible to its shareholders ratherthan to the public health to do anything otherthan strive for profit; it would be unrealistic toexpect Roche to support research intobehavioural treatments of insomnia when theymanufacture hypnotics.At the same time, the medical charities do

not have the resources to fund truly majorresearch, except perhaps for the cancercharities, and in any event their success isdriven so often by forces which tend to favourcancer and children rather than public healthpriorities which might, perhaps, favourneurological disorders, or the elderly (table 2).So it remains with Governmental agencies toidentify good clinical research and to allocateresources based on a mixture of public healthdemands and what is scientifically likely tosucceed. In other words, there must be acooperative plan rather than an unmanagedmarket place with short term commercialobjectives. This is a major national challenge.

In the meantime, clinical scientists have theirown challenges since there is no shortage oftreatments which need evaluating (table 3). It isour task to encourage research into evaluationof current practice, of the large number ofnewideas coming from basic scientists and others,and ofwhat our managers would like us to do inthe future. It is an exciting time for clinicalresearch so, "hither clinicians" who areinterested. Do not be overly intimidated by thebasic scientists; overwhelmed with routineclinical work, tempted by the pharmaceuticalindustry, excessive personal profit or an easylife.

1 Silverman WA. Retrolental fibroplasia a modern parable.London: Academic Press, 1980.

2 Crow TJ, Johnstone C. Controlled trials of electroconvul-sive therapy. In: Malitz S, Sackem HA, eds. Electroconvul-sive therapy, clinical and basic research issues, Vol 462. NewYork: Ann NY Acad Sci, The New York Academy ofSciences: 1986:12-29.

3 Editorial. Extra-cranial to intra-cranial bypass and theprevention of stroke. Lancet 1985;2:1401-2.

4 Crowther CA. Hospitalisation for bed rest in twin preg-nancy. In: Chalmers I, ed. Oxford database of perinataltrials. Version 1.2, Disc issue 5, February 1991, Record3376.

5 Grant A, Sleep J, Ashurst H, Spencer D. Dyspareuniaassociated with the use of glycerol-impregnated catgut torepair perineal trauma. Report of a 3 year follow up study.

30

Hither neurology: research

Br J Obstet Gynaecol 1989;96:741-3.6 Barnes MP, Bates D, Cartlidge NEF, French JM, Shaw DA.

Hyperbaric oxygen and multiple sclerosis: final results ofaplacebo controlled, double-blind trial. J Neurol NeurosurgPsychiatry 1987;50:1402-6.

7 Italian Acute Stroke Study Group. Haemodilution in acutestroke: results of the Italian haemodilution trial. Lancet1988;1:318-21.

8 Lindsay KW. Antifibinolytic agents in subarachnoidhaemorrhage. JNeurol 1987;234:1-8.

9 UK-TIA Study Group. Variation in the use of angiographyand carotid endarterectomy by neurologists in the UK-TIA Aspirin Trial. BMJ 1983;286:514-17.

10 Malleson A. Needyour doctor be so useless? London: Allen andUnwin, 1973.

11 American Child Health Association, New York. SchoolHealth influence on Tonsillectomy. In: Physical defects;the pathway to correction. New York: US Department of

Health 1934:80-96.12 Collins R, Julian D. British Heart Foundation surveys

(1987-89) ofUnited Kingdom treatment policies for acutemyocardial infarction (submitted).

13 Campbell DT. Reforms as experiments. American Psy-chologist 1969;24:409-29.

14 Fitzpatrick R, Hopkins A. Referrals to neurologists forheadaches not due to structural disease. J Neurol,Neurosurg Psychiatry 1981;44:1061-7.

15 Grove JL, Butler P, Millac PAH. The effect of a visit to aneurological clinic upon patients with tension headache.The Practitioner 1980;224:195-6.

16 Huxley TH. Collected essays, viii. Biogenesis and Abiogen-esis.

17 Asher R. Apriority. Lancet 1961;2:1403-4.18 Detsky AS. Are clinical trials a cost-effective investment? J

Am Med Assoc 1989;262:1795-800.19 Smith R. All change for Research. BMJ 1987;295:1177-82.

31