BRITISH MEDICAL JOURNAL 21 FEBRUARY 1976 419

PAPERS AND ORIGINALS

Oral anticoagulants-the totem and the taboo*

A BRECKENRIDGE

British Medical Journzal, 1976, 1, 419-423

For many years one of the main aims of pharmacology hasbeen to promote rational drug treatment. Most new drugs thatare introduced today originate in a chemist's test tube and entertherapeutics via an all-embracing animal screening programmeand studies in normal volunteers. A few drugs have a slightlyless stereotyped origin, which brings me, of course, to the oralanticoagulants.The discovery of these drugs by Karl Paul Link' has been

cited as a triumph of logic. His group identified, isolated,purified, and synthesised bishydroxycoumarin, the first oralanticoagulant. After a brief study of animal toxicity, whichwould have appalled drug regulatory authorities today, bis-hydroxycoumarin "was grabbed from Link's hands by theclinicians"2 and given to man in 1941.3The first long-term studies using bishydroxycoumarin in

myocardial infarction were reported in 1946.4 5 The controversyabout the efficacy of oral anticoagulants then began and thelogic of their discovery rather vanishes. But over the years, inthe quieter backwaters of pharmacology, oral anticoagulantshave become a model system for the study of drug action,6 andI shall address myself to both these facets-the pharmacologicaland the therapeutic. Our work with the oral anticoagulantwarfarin started in 1967 and I shall describe here some of thelessons in clinical pharmacology that I have learnt from our

studies.

Is there a clinical problem to investigate?

Many research programmes in clinical pharmacology havetheir beginnings in apparently isolated clinical incidents involv-ing drug treatment, and this was certainly the case with our oral

* Based on the Goulstonian lecture delivered at the Royal College ofPhysicians, London, on Thursday 6 November 1975.

anticoagulant work. In the summer of 1967 our attention was

drawn to two patients taking warfarin who were readmitted toHammersmith Hospital 10 days after discharge because of life-threatening haemorrhage. Their histories were remarkablysimilar. Both had been stabilised on warfarin when inpatientssome 10 days earlier, and on discharge their anticoagulant con-

trol was immaculate. But during this first admission each hadbeen given a nightly hypnotic, dichloralphenazone (Welldorm)in one case and amylobarbitone in the other. On discharge theystopped the hypnotic in their quieter home environment, butwarfarin was continued in its previous dose.Today every candidate for the MRCP will recognise this as a

prescription for disaster, but in 1967 information on the fre-quency and time course of this type of drug-drug interaction inman was small. We therefore planned a series of studies to definewhich hypnotics and sedatives could be safely given to patientson warfarin without altering anticoagulant control and whereinteractions did occur to study their mechanism, magnitude, andtime course. The patients we studied were already stabilised on

warfarin and were about to end their course of treatment.Since dichloralphenazone had been one of the drugs incrimi-

nated we gave this to a group of patients7 and showed that its

1i0i

EE 30]40z

1201c 1.5

~r1.0

o065E0 ~~~~~~~~~~~~M

6 8 1Ib2434 48 64in8e 6 104Time (days)

FIG 1-Administration of dichloralphenazone (1300 mg nightly) for 33 days.Effect on Thrombotest result and plasma warfarin concentration.

University of Liverpool, Liverpool L69 3BXA BRECKENRIDGE, MD, FRCP, professor of clinical pharmacology

on 1 October 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

r Med J: first published as 10.1136/bm

j.1.6007.419 on 21 February 1976. D

ownloaded from

420

administration was associated with a fall in plasma warfarinconcentrations and a decrease in anticoagulant effect (fig 1). Onstopping the dichloralphenazone it took some 20 days for plasmaconcentrations and effect to return to previous levels. Dichloral-phenazone-as its name implies-contains chloral hydrate andphenazone (better known as antipyrine) in the ratio of 2:1.Phenazone is added to chloral to improve its stability and, moreimportantly from the patient's point of view, to mask the bittertaste and gastric irritation of chloral hydrate.When chloral hydrate alone was given to patients on warfarin,

there was no change in the anticoagulant response, but therewas a fall in plasma warfarin concentrations. This apparentparadox was explained by Sellers and Koch-Weser.8 Chloralhydrate is metabolised to trichloracetic acid, 9 which accumulatesin plasma. Since it is more strongly protein-bound than warfarin,the anticoagulant is displaced and there are two consequences.Firstly, the increased free (unbound) warfarin temporarilyenhances the anticoagulant response, but, secondly, at the sametime the free drug is metabolised more readily by the liver. Thusone sees only a temporary increase in effect but a sustained fallin plasma warfarin. Nevertheless, the clinical importance and themagnitude of this interaction between chloral hydrate and war-farmn remain in question. "1, lWhen phenazone was given both plasma warfarin concen-

tration and its therapeutic response fell, as we had seen withdichloralphenazone, and these changes followed the same timecourse. This effect of dichloralphenazone and phenazone wasdue to a stimulation of the rate of warfarin metabolism in theliver, by the process of liver microsomal enzyme induction, andwe produced evidence for this in both man and rats.- Contraryto previous reports we found no evidence in either man or ratto suggest that chloral hydrate stimulated rates of drug meta-bolism." None of the three benzodiazepines that we studied-chlordiazepoxide, diazepam, and nitrazepam-had any effecton either anticoagulant response or warfarin kinetics."3 Thesecompounds, of course, are the most widely used tranquillisersand hypnotics today and may safely be given to patients takingwarfarin.

Pharmacological principles in experimental design

If the stimulus to studies in clinical pharmacology often comesfrom clinical medicine then these studies must be designed alongaccepted pharmacological lines. Two basic principles in phar-macology were important in planning studies with oral anti-coagulants. Firstly, different subjects may respond in differentways to the same dose of a drug-a topic discussed by Dolleryin his Bradshaw Lecture of 1974." Secondly, if one increasesthe dose of a drug one must expect increasing effects.When quinalbarbitone 100 mg nightly for 33 days was given

to six patients the fall in plasma warfarin ranged from 6", to650', which represented variations in the extent of inductionin the different subjects.15 This has been noted by others,'6 butthe reason for the variation is not clear. Environmentalists claimthat our drug metabolising enzymes are constantly stimulatedby the polluted air we breathe and the food additives we con-sume'7 and thus it is tempting to postulate that those subjectswho respond least to barbiturate administration are those whohave already been exposed most to inducing agents, but thereis little evidence to support this. Neither have we found acorrelation between the rate of elimination of the inducing agentand the degree of induction produced.When we gave an increased dose (200 mg for a similar period)

to two subjects who showed a small fall (50' and 15-3°0) insteady state plasma warfarin concentrations on 100 mg quinal-barbitone the percentage fall in plasma warfarin concentrationswas greater (40.50' and 5100 respectively). When one of thesetwo subjects took 300 mg quinalbarbitone for a further 30 days,there was no further fall in plasma warfarin (49°,,) (fig 2). Inother words, enzyme induction may have dose-dependant charac-

BRITrISH MEDICAL JOURNAL 21 FEBRUARY 1976

teristics, like many other pharmacological responses, and onecan construct a dose-response curve in individual subjects.18There are differences in this dose-response relationship with

different agents.'5 Groups of rats were treated with increasingdoses of four inducing agents-phenobarbitone, quinalbarbitone,

80

*- 60 -b00

3a 40'Ea

0

0 20-0-

.

o J

100 200 300Dose of quinolborbitone (mg)

FIG 2-Percentage fall in steady state plasma warfarin concentration in sixpatients given 100 mg quinalbarbitone for 33 days, in two patients given200 mg, and in one patient given 300 mg. (Reproduced by kind permissionof Raven Press, New York.)

amylobarbitone, and phenazone. Preparations of liver micro-somal enzymes were made from these rats and the rate ofmetabolism of the substrate ethylmorphine examined. Witheach of the four drugs, the greater the dose the greater the rateof ethylmorphine metabolism (fig 3). Furthermore, pheno-barbitone was a better inducing agent than amylobarbitone,quinalbarbitone, or phenazone. This related partly to the elimina-tion half life of the inducing agent and partly to the liver :plasmaratio. We did not find that relative lipid solubility was importantin our experiments, although this has been challenged.'9

o380

°3400

i3CO04-

260uo 220

E

c 180'

2 140>0c 100

3rbitone

Qui nolbarbitone

A Lx.'I...X..X.

A AI1--

-0--l*-1Amylobarbitone Antipyrine

3 10,umol Inducinq agent/kq

Io dooFIG 3-Change in in-vitro microsomal enzyme activity in rats pretreated withvarying doses of four enzyme-inducing agents. (Reproduced by kind per-mission of the editor of Clinical Pharmacology and Therapeutics.)

The lesson from these studies is that to classify a drug as aninducing agent-in man or any other species-on the basis of asingle dose level is unwise. What may be an inducing dose of abarbiturate in one patient may have no effect in another. Enzymeinduction is not an all-or-nothing phenomenon.

Importance of chemical expertise

Clinical pharmacology is a hybrid discipline. If ideas originatein clinical medicine, and studies are carried out in accordance

I~~ *_

I --

0

on 1 October 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

r Med J: first published as 10.1136/bm

j.1.6007.419 on 21 February 1976. D

ownloaded from

BRITISH MEDICAL JOURNAL 21 FEBRUARY 1976

with pharmacological principles, then unless one works closelywith a chemist interpretation of one's data may be fallacious.This is important from two points of view. Warfarin is meta-bolised by enzymes of the liver and to a lesser extent by othertissues such as the kidney.20 22 Metabolites are inactive or havean activity considerably less than the parent drug but accumu-late in plasma,e 23 and thus in studies to correlate plasma andtissue drug concentrations with effect specific analytical tech-niques are mandatory. Early studies of the kinetics of warfarinused assays for warfarin that failed to distinguish drug frommetabolite,2' and in our initial studies we too used non-specificmethods. But we were soon corrected by our chemist colleaguesand we adopted the method of Lewis et al,'2 in which drug isseparated from metabolites before assay, and these metabolitescan themselves be measured.

Specificity is one aspect of methodology, sensitivity is another.In Liverpool we are currently engaged on studies in which smalldoses of warfarin are given to subjects and we must thus measurewarfarin in plasma in nanogram quantities, and thus even moresensitive chemical assays must be devised.

Chemical expertise is not necessary only from the analyticalstandpoint. Warfarin has an asymmetric centre, and this impliesthat there are two isomers, or enantiomers.26 In the rat Swarfarin is about five times more potent than R warfarin.2 7When we examined the rate of elimination of these enantiomersin the rat we found that S warfarin was eliminated less rapidlythan R warfarin. The difference in the rate of elimination wasonly twofold, whereas the difference in potency was some five-fold to sixfold. Thus the intrinsic anticoagulant effect of S war-farin was greater. 8 9 This was studied by measuring the inhibi-tion of the rate of clotting factor synthesis, as originally describedby Nagashima et al1' in rats given R or S warfarin. S warfarinhad double the effect of R warfarin.

Which species to study?

A perennial problem in clinical pharmacology is to decidehow much reliance should be placed on data derived fromanimal studies and, further, which animal species should beused. We made extensive use of measurements of rates ofin-vitro drug metabolism in rat liver preparations in studies ofenzyme induction with hypnotics, and the results were used tocomplement clinical observations, but even in this field thereare significant differences between man and the rat.'3 Thereare other circumstances when this type of extrapolation may beunwise, as is illustrated with studies of the enantiomers ofwarfarin.

Commercial warfarin is a racemic mixture. In many respects theenantiomers are different drugs. In man they differ in potency (Swarfarin being about three to four times more potent than R warfarin),rate of elimination (S warfarin being eliminated more rapidly than R,unlike in the rat), and pathway of metabolism.' 22 :11 .2 The principalmetabolic product of S warfarin is 7-hydroxywarfarin, while that ofR warfarin is a warfarin alcohol, formed by reduction of a ketonegroup on the warfarin side chain (fig 4). Since the configuration ofwarfarin clearly influences its pharmacological effect, it seemed rele-vant to explore further the relation between the configuration andmetabolism of warfarin. We thus re-examined a well known druginteraction-that of phenylbutazone and warfarin. Phenylbutazoneadministration in man will augment the anticoagulant effect of war-farin."1-1 The explanation usually proposed for this potentiation is thatphenylbutazone will compete with warfarin for binding sites on circu-lating albumin and displace it, leading to an enhanced effect. Inin-vitro studies when phenylbutazone is added to plasma containingwarfarin as much as 30% of warfarin is in the free form.'4 But, ifdisplacement from albumin were the sole mechanism of the druginteraction, logically there should be more rapid metabolism andelimination of warfarin in vivo. There was, however, no change in theclearance of racemic warfarin in the presence of phenylbutazone.APlasma concentrations of 7-hydroxywarfarin were, however, con-sistently lower in phenylbutazone studies than controls when racemic

10000

5000

1000

E 500

100

so

* = Warforin alcohol ix=S Warfari n o = Warforin alcohol 2Y = R Warfari n A = 7- hydroxyworfari n

100000farin 35mg R War

1000

500

100

50

1010 '0Days

02 3

2 3

FIG 4-Patient was given 35 mg S warfarin and 35 mg R warfarin on separateoccasions. S warfarin was oxidised to 7-hydroxywarfarin and reduced pri-marily to warfarin alcohol 2. R warfarin was reduced to warfarin alcohol 1.(Reproduced by kind permission of the editor ofJournal of Clinical Investiga-tion.)

100

0

04--C

FacX

E

a

0-

CL

10 1R WarfarinS Warfarin +

phenyl butazone

S Warfarin

R Warfarin +phenyl butozone

0 2 4

Days6

FIG 5-Single doses of R and S warfarin were administered before and afterphenylbutazone 100 mg three times daily for 10 days. Concentrations ofisomers are represented as percentage of extrapolated initial plasma con-centration. (Reproduced by kind permission of the editor of Journal ofClinical Investigation.)

warfarin was given, and plasma concentrations of warfarin alcoholswere higher. The explanation for these differences was found instudies where the isomers were given separately. The fractional clear-ance of S warfarin (the more potent) was slowed down, while that ofR warfarin (the less potent) was increased by phenylbutazone (fig 5).32

Differences in rates of elimination have been shown between theoismers of methadone,3 propranolol,'6 amphetamine,37 and hexo-barbitone.8 It does not seem unreasonable to assume that druginteractions with these agents, like those with warfarin, might bestereochemically dependent, and where the isomers show differencesin pharmacological potency such changes may have clinical signifi-cance.When one examines drug interactions with the isomers of warfarin

in other species, such as the rat and the dog, the results are quanti-tatively and qualitatively different from those described above, andclearly definitive studies must be done in man.

More than just a study of pharmacokinetics

Dollery39 has paraphrased the suggestion that there is more toclinical pharmacology than just a study of pharmokinetics byposing the question "Pharmacokinetics-servant or master ?"

421

aa:-Z--(71;L

on 1 October 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

r Med J: first published as 10.1136/bm

j.1.6007.419 on 21 February 1976. D

ownloaded from

422

He implied that there has been an increasing tendency by manyclinical pharmacologists to regard the study of pharmacokinetics-that is, the way the body handles drugs-as an end in itself,rather than to consider it as complementary to a study of whatdrugs do to the body. I am uncertain if he had our studies withoral anticoagulants in mind when he coined this phrase, butif so let me hasten to redress the balance and consider the natureof the response to oral anticoagulant treatment.

If one measures the plasma warfarin concentration in patientswho have all had the same optimal anticoagulant response afivefold to sixfold range is found.40 These differences in plasmaconcentration may be equated with differences in tissue (receptor)sensitivity, provided there are no appreciable differences in theliver:plasma warfarin ratio.

Over the last few years several groups4'-44 have examined the natureof response to warfarin administration and the warfarin "receptor."Warfarin interferes with the hepatic synthesis of clotting factors II,VII, IX, and X. Their synthesis is closely linked to a shuttle involvingvitamin K, (which is active) and an inactive metabolite vitamin K, 2:3epoxide (k, oxide). The cycle is mediated by two enzymes-a reduct-ase and an epoxidase-both of which are membrane-bound and foundin the microsomal fraction of liver homogenate.

According to this scheme, generation of clotting factors occursduring the conversion of K, to the epoxide while the reductase isconcerned with the regeneration of vitamin K,. This evidence hasbeen deduced from studies in rats, but is supported by studies in man.Shearer et al4 have shown that the metabolic pattern of vitamin K,is altered in man by therapeutic doses of warfarin in that there is anaccumulation in plasma of K, oxide (fig 6). The clearance of vitaminK, from the plasma is not affected.

10000-

1000-0E

-aE

100.

10

22

.6

A--A

(a)

2 3 4 5 6 7 8 22 23 24Time (hours)

FIG 6-Plasma concentration of radioactivity, expressed as disintegrations perminute per ml of plasma, after intravenous injection of 1 mg (11, 21,- HO)-vitamin K- in one subject. 0 =Control studies. A =Studies after warfarinadministration. (Reproduced by kind permission of the author and the editorof British Journal of Haematology.)

By administering various doses of warfarin to different subjects andmeasuring the accumulation of vitamin K, oxide in plasma, it ispossible to study the relationship between doses of warfarin and thevitamin K,-K, oxide response.A dose-response curve can be constructed which has a sigmoid

shape, indicating that above a critical dose of warfarin (equivalent toabout 70 mg/day) there is no corresponding increase in K, oxideaccumulation in plasma but that over the therapeutic range therelationship is linear.43While clearly the action of warfarin is mediated through vitamin

K, many important questions remain unanswered. Can one postulatethat there is a receptor for warfarin ? Is this receptor a series of enzymesor a single enzyme in the vitamin K,-K, oxide cycle? Is geneticresistance to warfarin, well documented in the rat and in man,6 amanifestation of change in this enzyme system? The answers tothese and many other questions might be forthcoming more easily ifvitamin K could be measured. Surprisingly, vitamin K cannot be

BRITISH MEDICAL JOURNAL 21 FEBRUARY 1976

measured and we have to rely on studies with labelled vitamin K,.This aspect of oral anticoagulant treatment is of great importance andshows the potential we have for inonitoring a biochemical aspect ofthe response to oral anticoagulants.

Importance of clinical trial design

So far I have addressed myself to problems of the pharma-cology of oral anticoagulants, but now I must turn to a morecontentious aspect-namely, their use in clinical medicine. Ithas been said that perhaps the greatest use that anticoagulantsmay yet have in clinical medicine is their ability to cause con-troversy.46 There are those who claim that oral anticoagulantsare effective rat poisons and there their usefulness ends. At theother end of the spectrum are the workers who find that anti-coagulation produces a fourfold decrease in mortality in patientswith acute myocardial infarction.'7 48Few would now gainsay their place in the treatment of venous

thrombosis and pulmonary embolism, established by the classi-cal studies of Sevitt and Gallacher49 and Barritt and Jordan.50The place of these drugs in treatment of patients with myocardialinfarction and cerebrovascular disease, however, remains con-troversial. When one examines critically the design of theclinical trials on which entrenched attitudes have been adoptedabout the efficacy or otherwise of oral anticoagulants in theseconditions, many of these trials are deficient by modernstandards. Although statistics were first used in the scientificevaluation of treatment as long ago as 1828,-' in a study toshow the inappropriateness of blood letting as a therapeuticmeasure, the principles of conducting therapeutic trials werenot laid down until the work of Bradford Hill in the 1930s.52Moreover, studies designed and executed in the 1940s and 1950sare no longer acceptable in the 1970s.

Gifford and Feinstein 3 and Gross et a154 have recentlyanalysed all trials of anticoagulants in acute myocardial infarc-tion reported in the English language between 1948 and 1972.Forty-three studies were reviewed. Seven criteria were chosenwhich would today be generally acceptable as necessary in thedesign of a therapeutic trial. Not one of these 43 studies fulfilledall criteria, and, interestingly, the lower the standard of thetrial the more common was the finding that oral anticoagulanttreatment was better than no treatment. The number of studiesfulfilling each of the criteria is shown in the table.

Numnber of trials fulfilling each of seven criteria considered necessary to thedesign of a trial

Diagnostic criteria for:Myocardial infarction . . ..Pulmonary embolus

Prospective clinical trialConcurrent controls . . ..Controls treated in same hospital as test group ..Random allocation to test or control groupsDouble blind methodology ..Allocation of bad-risk patients to test and control groups .

|No

1311

243426113

19

302

567961267

44

With hindsight, of course, it is an easy task to criticise thesestudies, which assessed many thousands of patients. If one doescriticise the next question must be: is a new properly designedtrial now justified ?

In my opinion the answer is no. If the introduction of oralanticoagulants was logical, perhaps their administration topatients with arterial thrombosis is not entirely rational. Thefundamental role of the platelet in the genesis of arterial throm-bosis and the failure of oral anticoagulants to alter plateletfunction suggest that on theoretical grounds alone, these drugsmight be ineffective.55 Nor is a possible diminution in thrombo-embolic events a valid reason for resuscitating large-scale trialsof oral anticoagulant treatment in myocardial infarction. The

on 1 October 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

r Med J: first published as 10.1136/bm

j.1.6007.419 on 21 February 1976. D

ownloaded from

BRITISH MEDICAL JOURNAL 21 FEBRUARY 1976 423

main causes of death in acute myocardial infarction are pumpfailure, cardiac arrhythmias, cardiac rupture, and thrombo-embolism. Only the last of these would be influenced by anti-coagulants, and they account for some 200' of the deaths. Toprevent deaths due to thromboembolism would reduce theoverall death rate only from 20o1 to 16%o, and to detect thisthe size of the trial and its complexity would indeed have tobe great.56

If there is controversy about the use of anticoagulants in acutemyocardial infarction the same arguments rage in cerebrovascu-lar disease. By using the same criteria as those proposed formyocardial infarction, 16 studies of the use of oral anticoagulantsin studies of cerebrovascular disease have been analysedrecently.5 None fulfilled all the criteria. But, as in trials inmyocardial infarction, the better the design58 59 the morecommon was the conclusion that anticoagulant treatment wasof no benefit.

Totem and taboo

The totem, Freud's guardian spirit,60 I have equated with thepharmacology of oral anticoagulants. Because of the precisionof measurements of both effect and concentration, these drugshave become a pharmacologist's touchstone. The taboo, Freud'sill-understood and dangerous being,60 is obviously their placein therapeutics. To improve treatment with these drugs, notonly must their pharmacokinetics be understood, the role ofvitamin K in determining their response must be appreciated,trials must be appropriately designed, and, of course, clinicaljudgment must be exercised.

I wish to express my appreciation to those colleagues with whomthese studies were carried out and for the many helpful discussions wehave had, particularly, Dr Michael Orme, Dr Donald Davies, ProfessorColin Dollery, Dr Martin Shearer, Dr Richard Lewis, and ProfessorMalcolm Rowland. This work was supported by the Medical ResearchCouncil, the Wellcome Trust, and Roche Pharmaceuticals Limited.

References1 Link, K P, Harveian Lecture, 1943, 39, 162.2 Link, K P, Circulation, 1959, 19, 97.3Butt, H R, Allen, E V, and Bollman, J L, Proceedings of the Staff Meeting

of the Mayo Clinic, 1941, 16, 388.4 Nichol, E E, and Page, S W, Journal of the Florida Medical Association,

1946, 32, 365.5 Wright, I S, American Heart Journal, 1946, 32, 20.6 O'Reilly, R A, and Aggeler, P M, Pharmacological Reviews, 1970, 22, 35.7Breckenridge, A, et al, Clinical Science, 1971, 40, 351.8 Sellars, E M, and Koch-Weser, J, Clinical Research, 1970, 18, 344.9 Marshall, E K, and Owens, A H, Bulletin of Johns Hopkins Hospital,

1954, 95, 1.1 Lasagna, L, Proceedings of the International Symposium on Clinical

Pharmacology, p 236. Brussels, 1970.11 Boston Collaborative Drug Surveillance Programme, New England

Journal of Medicine, 1971, 286, 53.12 Cucinell, S A, et al, Journal of the American Medical Association, 1966,

197, 366.13 Orme, M L E, Breckenridge, A, and Brooks, R V, British Medical_Journal,

1972, 3, 611.

14 Dollery, C T, in Advanced Medicine, Topics in Therapeutics, ed A Brecken-ridge, vol 1, p 95. London, Pitman Medical, 1975.

15 Breckenridge, A, et al, Clinical Pharmacology and Therapeutics, 1969, 10,22.

16 Vessell, E S, and Page, J G, Journal of Clinical Investigation, 1969, 48,2202.

17 McLean, A E M, Medicine, 1972, 4, 331.18 Breckenridge, A, et al, in Drug Interactions, ed P L Morselli et al, p 223.

New York, Raven Press, 1974.19 Pelkonen, 0, and Harki, N T, Chemico-Biological Interactions, 1973, 7, 93.20 O'Reilly, R A, et al, Thrombosis et Diathesis Haemorrhagica, 1962, 8, 82.21 Brodie, B B, Pharmacologist, 1964, 6, 12.22 Hewick, D S, and McEwan, J, Journal of Pharmacy and Pharmacology,

1973, 25, 458.23 Lewis, R J, et al, Journal of Laboratory and Clinical Medicine, 1973, 81,

925.24 Corn, M, and Berberich, R, Clinical Chemistry, 1967, 13, 126.25 Lewis, R J, Ilnicki, L P, and Carlstrom, M, Biochemical Medicine, 1970,

4, 376.26 West, B D, et al, J7ournal of the American Chemical Society, 1961, 83,

2676.27 Eble, J N, West, B D, and Link, K P, Biochemical Pharmacology, 1966,

15, 1003.28 Breckenridge, A, and Orme, M L E, Life Sciences, 1972, 11, part 2, 337.29 Hewick, D,Jfournal of Pharmacy and Pharmacology, 1972, 24, 661.30 Nagashima, R, O'Reilly, R A, and Levy, G, Clinical Pharmacology and

Therapeutics, 1969, 10, 22.31 O'Reilly, R A, Clinical Pharmacology and Therapeutics, 1974, 16, 348.32 Lewis, R J, et al,,Journal of Clinical Investigation, 1974, 53, 1607.33 O'Reilly, R A, and Aggeler, P M, Proceedings of the Society of Experimental

Biology and Medicine, 1968, 128, 1080.34 Solomon, H M, and Schrogie, J J, Biochemical Pharmacology, 1967, 16,

1219.35 Elison, C, et al,,Journal of Medicinal Chemistry, 1963, 6, 237.36 George, C F, et al, European Journal of Clinical Pharmacology, 1972, 4,

74.37 Debackere, M, and Massart-Leen, A M, Archives Internationales de

Pharmacodynamie et de Therapie, 1965, 155, 459.38 Furner, R L, et al,J'ournal of Pharmacology and Experimental Therapeutics,

1969, 169, 153.39 Dollery, C T, European J'ournal of Clinical Pharmacology, 1973, 6, 1.40 Breckenridge, A, and Orme, M L E, in Biochemical Effects of Drugs in

Relation to their Plasma Concentrations, ed D S Davies, and B N CPrichard, p 145. London, Macmillan, 1973.

41 Olson, R E, Science, 1964, 145, 926.42 Bell, R, and Matschiner, J T, Nature, 1972, 237, 32.43 Shearer, M, McBurney, A, and Barkhan, P, British Journal of Haema-

tology, 1973, 24, 471.44 Suttie, J W, Science, 1973, 179, 192.45 Shearer, M, Barkhan, P, and Breckenridge, A, European Journal of Clinical

Pharmacology, in press.46 Gilchrist, A R, and Tulloch, J A, Scottish Medical Journal, 1956, 1, 1.47 Modan, B, et al, New England Journal of Medicine, 1975, 292, 1359.48 Tonascia, J, Gordis, L, and Schmerler, H, New England Journal of Medi-

cine, 1975, 292, 1362.49 Sevitt, S, and Gallacher, N G, Lancet, 1959, 2, 981.50 Barritt, D W, and Jordan, S C, Lancet, 1960, 1, 1309.51 Louis, P C A, Archives Generales de Medicine, 1828, 18, 321.52 Bradford Hill, A, Statistical Methods in Clinical and Preventative Medicine.

Edinburgh and London, Livingstone, 1962.53 Gifford, R H, and Feinstein, A R, New England3Journal of Medicine, 1969,

280, 351.54 Gross, H, et al, American Journal of Medicine, 1972, 52, 421.55 Douglas, A S, Anticoagulant Therapy, Oxford, Blackwell, 1962.56 Ebert, R V, Circulation, 1972, 45, 903.57 Cervantes, F D, and Schneiderman, L J, Archives of Internal Medicine,

1975, 135, 875.58 Marshall, J, and Shaw, D A, Lancet, 1960, 1, 995.59 Hill, A B, Marshall, J, and Shaw, D A, British Medical Journal, 1962, 2,

1003.60 Freud, S, Totem and Taboo, trans James Strachey. London, Routledge

and Kegan Paul, 1960.

on 1 October 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

r Med J: first published as 10.1136/bm

j.1.6007.419 on 21 February 1976. D

ownloaded from