560 Journal of the Royal Society of Medicine Volume 85 September 1992

Food allergy -fact or fiction: a review

R Finn MD FRCP Royal Liverpool Hospital, Prescot Street, Liverpool L7 2XP

Keywords: food allergy; food sensitivity; food intolerance

SummaryFood sensitivity is a common condition presentingwith various clinical syndromes including migraine,urticaria, gluten enteropathy, Crohn's disease andirritable bowel syndrome. It is a heterogeneouscondition affecting different organ systems and is alsoaetiologically diverse with subgroups due to allergy,pharmacological reactions, enzyme deficiencies andpsychological causes. Clinical acceptance of foodsensitivity has been delayed by the use of dubiousdiagnostic techniques by a minority of practitionersand the lack oflaboratory diagnostic tests, but severaldouble blind studies have now fully validated theexistence of food sensitivity syndromes. More wide-spread recognition of food sensitivity would be costeffective for the National Health Service.

IntroductionThe question 'Food Allergy - Fact or Fiction' wasoriginally raised by Finn and Cohen in 19781. Thisarticle received extensive publicity and led towidespread public interest. More than 12 years haveelapsed and food allergycontinues to generate muchpublic interest and medical scepticism. The question,therefore, arises as to why more than a decade later,food allergy still remains a controversial subjectdespite a major report from The Royal College ofPhysicians entitled 'Food Intolerance and FoodAversion' 2. This article explores the reasons for thissituation and assesses the current state ofknowledgein this field.

Diagnostic methodsUndoubtedly some ofthe techniques used to 'diagnose'food allergy in the last decade have done much to bringthe subject and its practitioners into disrepute3'4.Methods used have included symptom provocation byfood extracts, hair analysis, dowsing, cytotoxic testingand various electrical methods including the Vegamachine. None of these methods has been validatdand most are scarcely crediible. Patients who hadpreviously seen a clinical ecologist and subsequentlyconsult an allergist iaight desaibe having been testedwith one of these methods, and this has the effect ofalienating the physician to the whole subject of foodallergy. These techniques are mainly used by laypractitioners, but unfortunately a small number ofmedical practitioners lhave used similar methods.An unfortunate side effect consequent on the use

of these false diagnostic methods is the phenomenonof 'guilt by association' which to some degree effectsall of those working in food allergy and to some extentgeneral allergy, and tends to discourage young menfrom entering this field.The only accepted method of diagnosis is an

elimination diet followed by test reintroduction of

foods. Practitioners offood allergy must restrict theirdiagnostic methods to those that are accepted andhave a reasonable theoretical basis, if their subjectis to escape the controversy that has surrounded itin the last decade. Elimination diets are troublesomeand tedious and a continued search for more acceptablediagnostic methods must continue. Intradermaltesting deserves further study and validation. It is avariation ofthe classic prick test which is only ofvaluein IgE food allergies. Food antigens are injectedintradermally and growth of the wheal over 10 minis defined as a positive reaction. It is an objective testwhich does not lend itself to placebo responses. It isused by many physicians particularly in the UnitedStates, and merits further validation, but at thepresent time, despite several trials5, it has to beregarded as non-proven.

Scientific studiesIn the past decade scientific studies have been carriedout which should reduce the credibility gap for foodallergy as a genuine clinical phenomenon6.Acute allergic reactions to such foods as eggs, nuts

and shell fish are well recognized by appropriatediagnostic methods such as prick tests and RAST andhave an accepted mechanism involving IgE antibody.These reactions are easily diagnosed by the patienthimself and are uncommon. The problem is whetherthere are other food-related immunological conditionsmediated by other immunoglobulins (IgG and IgA),immune complexes or cellular mechanisms.Food allergy is common in paediatric practice with

an incidence of up to 7%7. This may be due to apoorly developed mucosal barrier in infants whichnormally protects against food antigens8. Bovineserum albumen absorption from the gut is increasedby achlorhydria9 and gastric acid is low in infants.Food antigens are complexed to secretory IgA in thegut and this limits absorption, and it is, therefore,significant that IgA levels in serum and gut arereduced in infancy. The extensive gut associatedimmune system is, however, continuously exposed tofood antigens, but most produce oral tolerance inducedby suppressor T cells derived from Peyers Patches.Nevertheless IgG food antibodies are regularly foundin adults. Our group has worked with ovalbumen,casein and lactalbumen and has established that therelevant IgG antibodies are commonly found innormal adults. The frequency of these antibodies isslightly increased in classical allergic conditionsincluding eczema, but there is no definite evidencethat these antibodies cause disease":'6. Neverthe-less, despite oral tolerance there is now no doubtthat immunological reaction to foods as shownby the production of IgG antibodies is a commonphenomenon.

0141-0768/92090560.05/$02.00/0© 1992The RoyalSociety ofMedicine

Journal;of the Royal Society of Medicine Volume 85 September 1992 561

Coeliac disease is an immunological reaction togluten and there is a lymphoid infiltration of theintestinal mucosa with villous atrophy. CirculatingIgA antibodies to gliadin can often be demonstrated.Possibly due to damage to the barrier function ofthe intestinal wall the frequency of other IgG foodantibodies is increased. Milk allergy in infantscauses diarrhoea with patchy areas of villous atrophyand some lymphoid- iufiltrationV7. A proctocolitiswith bloody diarrhoea, eosinophilic infiltration andcrypt abscesses is described. Milk allergy can causerhinorrhoea and sneezing and' Heiner's syndromecomprises recurrent- pneumonia, pulmonary haemo-siderosis, anaemia, failure to- thrive' and intestinalblood loss' . Milk allergy can- also- cause urticariaand atopic dermatitis'9. Breast fed infants with milkallergy improve when the mother avoids milk21.These conditions are usually- transient and -mostinfants will usually tolerate milk by one year.

It is generally acceptedthat food allergydefined asan immunologically mediated clinical syndrome thatdevelops after the ingestion of a dietary product22 iSless common in adults. Gluten enteropathy occurs inadults and the recent finding that IgA antibodies tobaker's and brewer's yeast (Sacchai-omyces cereviswe)are found in Crohn's disease, but not in ulcerativecolitis or normal controls23'24 raises the possibilitythat Crohn's disease may also be an allergic condition.Food allergy is not a common cause of adult rhinitis,asthma or eczema, but occasional cases respond to anelimination diet.

It has to be concluded that there is a gap betweenthe experimental evidence and the claim that foodallergy is common in adults. Well controlled clinicaltrials have, however, shown that elimination diets canbe helpful in irritable bowel Wyndrome,2, Crohn'sdisease26, migraine27, hyperkinesis28 and rheumatoidarthritis29'30 but this does not constitute evidencethat these conditions are due to food'allergy, ratherthan some other food-related reaction.

TerminologyAcademic acceptance of the reality offood allergy hasbeen delayed because of the blanket use by somepractitioners of the term 'Food allergy' to cover allfood-related reactions. The subject was clarified by theRoyal College Physicians report on 'Food Intoleranceand Food Aversion'2 which divided food reactionsinto defined aetiological groups. Food intolerance hasbeen suggested as a general term to cover all typesof reaction, but food sensitivity is probably preferableas food intolerance does not emphasize the critical factthat only a small number of sensitive subjects willdevelop food reactions.

Classification of food sensitivityThis classification of food sensitivity is based on theRoyal College of Physicians Report2:

Food allergyAcute IgE mediated reactions are uncommon, butwould include allergy to shellfish, eggs and nuts. Theyare usually acute urticarial or gastrointestinalreactions, but can, particularly with peanuts, causefatal anaphylaxis. Non-IgE food allergy includesgluten enteropathy and possibly Crohn's disease23.Eczema, urticaria, asthma and rhinitis are occasionallydue to food allergy in adults. There is little evidencethat food allergy is responsible for other diseases, but

this area requires further study. The gut- associatedimmune system comprises the largest single collectionof lymphocytes in the- body and the potential mustexist for other diseases caused by as yet unrecognizedfood induced immunological reactions. IgA- nephro-pathy3' and giit associated arthropathies are possiblecandidates.

Pharmacological reactionsCertain foods, particularly those containing caffeineprodiuce pharmacological Eactions iftaken im excessivequantities. Thus coffee, tea and coca-cola can produceheadaches, palpitations and anxiety in-subjects predis-posed to these symptomis'. Headaches are particularlyproduced by caffeine withdrawal32 which may be afactor in weekend migraine, *hen the amount ofcoffee taken regularly at work is suddenly reduted.Other reactive amines such as tyramine in cheese,phenylethylamine in chocolate and octopamine incitrs-fruits can exacerbate migraine.Aspirin and salicylates in foods such as fruits can

induce urticaria, rhinitis, nasal polyps and asthmain susceptible subjects33-34. Aspirin causes a pseudo-allergic reaction by inhibiting the cyclo oxygenaseenzyme leading to inhibition of prostaglandin forma-tion. Certain food- additives such as tartrazine,benzoate and ascorbic acid act in a similar fashionand can produce pseudo-allergic symptoms. T-heseadditives act synergistically and thus the effects ofdiet oiu symptoms are very variable, and testingof single substances may be misleadinge. Certaincolouring agents and salicylates can cause hyper-activity in children, but the mechanism is obscure2.The syidrome of cyclical oedema with mental

symptoms offatigue and irritability is often mistakenfor allergic angioedema. It almost invariably occursin women tn long-ter-m diuretics who vary their foodintake considerably with- periods of partial fastingfollowed by binging on foods high ii sodium andcarbohydrate. Sodium retention following a largecarbohydrate intake is well documented. Stopping-diuretics and avoiding carbohydrate hinges is usuallyeffective, although the oedema is often worse for ashort periods.

Partial enzyme deficienciesLactase deficiency which is to some extent dosedependant can cause intestinal disorders. Poorbulphoxidation is common in subjects with foodreactions37 and G-6-PD deficiency is associated withhaemolytieanaemia caused by fava beans. It is likelythat partial enzyme deficiencies will be shown to bean important cause of food sensitivity and furtherwork in this area is indicated.

Psychological reactionsPsychological reactions include anorexia nervosa andbulimia.

It is now evident that food allergy is responsible foronly a proportion of patients with food sensitivity.Pharmacological reactions are probably more common,and further work may show that partial enzymedeficiencies constitute the most important cause offood sensitivity.

DiagnosisThe only proven method of diagnosing food sensitivityis by elimination diet and challenge. An elimination

562 Journal of the Royal Society of Medicine Volume 85 September 1992

diet consists of a small number of safe foods whichclinical experience has shown rarely produce reactions.Typical diets would exclude additives and salicylates;or the main food allergens such as milk, eggs andwheat. Alternatively a synthetic elemental diet can beused. The diet is given for about 10 days. Withdrawalsymptoms may occur in the first 3 days, but thepatient should be free of symptoms within 10 days.If symptoms persist, food sensitivity as a cause ofthepatients symptoms can be excluded and the patientshould be informed of this fact and returned to a

normal diet. If symptoms are relieved the omittedfoods are reintroduced to the diet sequentially todetermine which produce the initial symptoms. Therate of reintroduction is usually one food every 24hours. If no reaction is produced the food is returnedto the diet. A strong reaction may take several daysto settle and further testing should be delayed. Afteravoiding a food for about 10 days, sensitivity isincreased and a genuine reaction is usually very

obvious.Elimination dieting is tedious and requires

enthusiasm by both patient and physician. Thus itshould only be considered for severe, incapacitatingand intractable symptoms which have failed torespond to other therapy. With minor symptoms thetreatment is worse than the disease and patientcompliance understandably is poor. With severe

intractable symptoms the patient is usually only tooanxious to attempt anything to obtain relief, and inthese circumstances there is rarely anything less thantotal co-operation. Conversely the physician should beenthusiastic and knowledgeable and hence physicianswho use the technique only occasionally and withsome scepticism are less likely to be successful. Thisraises a logistic problem as to who should manage foodsensitivity. As the symptoms of food sensitivity can

affect many organ systems the appropriate specialistis a possibility, but they usually lack the clinicalexperience or interest to master a difficult clinicaltechnique. Allergists are often only willing to treatspecific IgE food allergy, which, as has been indicatedonly accounts for a small proportion of cases of foodsensitivity; furthermore there are few allergists inthe NHS, which has forced many patients into an

uncontrolled private sector.Elimination dieting is tedious and the search

continues for objective laboratory tests for foodsensitivity. It should be realized t-hat no single testis possible as there are many different causes of foodsensitivity. Acute IgE reactions can be confirmed byskin testing and RAST, but no reliable tests are

available for non-IgE allergy. IgG and IgA antibodiesto various foods can be detectedll"16, but they do notcorrelate closely with symptoms. Intradermal testinghas been used by some physicians38 but has not yetbeen fully validated. It is likely that investigationswill become available to identify partial enzymedeficiencies leading to difficulty in handling certainfoods, but these are now only available on an

experimental basis.The lack of objective laboratory tests has undoubt-

edly delayed the acceptance of food sensitivity as a

genuine clinical phenomenon, and this has beencompounded by the use of less than credible techniquesby some alternative practitioners. Thus the diagnosisof food sensitivity remains dependant on clinicalrather than laboratory methods, but it should beemphasized that this situation also applies in

other specialties. There are no objective laboratorytests for various psychiatric disorders such asschizophrenia, but no one would doubt that they existas genuine clinical entities.

TreatmentThe treatment of food sensitivity is exclusion. AcuteIgE sensitivities are permanent but do not present aproblem as they usually only involve a single food.Allergy to dairy products or cereals involves avoidinga group of foods, but suitable alternatives can beadvised. Many other forms of sensitivity are dose-related and simply reducing the quantity of a foodeaten repetitively and to excess is often effective. Thisapplies to most pharmacological reactions. Other foodsensitivities are transient and will resolve after ashort period on an elimination diet. Patients whoclaim to be allergic to multiple foods should be testedunder double blind conditions as the problem is oftenpsychological. Attempts to desensitize to foods havebeen made, particularly with low dosage regimens,but these methods require further validation39.

Food sensitive conditionsHow important is food sensitivity in clinical medicine?Is the diagnosis of food sensitivity only helpfulin a few rare cases or does it have wider relevance?Double blind studies25-30 in migraine, irritable bowelsyndrome, Crohn's disease and rheumatoid arthritishave shown that food sensitivity does exist as agenuine entity.In order to determine the spectrum ofconditions due

to Food Sensitivity a successful result has beendefined as at least 70% improvement in symptomsusing a visual analogue scale. To exclude a placeboeffect which is usually transient the improvement hadto be maintained for at least 6 months. Over thelast decade a record has been kept of over 300highly successful cases fulfilling these criteria andfrom this data it is now possible to state whichconditions are likely to respond to dietary management(Table 1). It should be pointed out that many ofthesecases had been ill for many years and had seenseveral physicians, before seeking treatment for foodsensitivity. Migraine occurs in 10% of the populationand is severe in 1% causing much morbidity andloss of time from work comparable with industrialaction. Severe irritable bowel syndrome probablyoccurs at about the same frequency, and is alsoresponsible for much morbidity and cost due torepeated clinical investigations. Ifone simply considers

Table 1. Food sensitive conditions

Group A Group B

Headache including migraine AsthmaIrritable bowel syndrome EczemaUrticaria and angio-oedema Rheumatoid arthritisCrohn's disease RhinitisGluten enteropathy Hyperactivity

Anxiety and depressionAphthous ulcers

In group A, food sensitivity is a common cause of theconditions listed, and at least 50% will improve withelimination diets. In group B, food sensitivity is a lesscommon cause ofthe conditions listed, and the success ratewith elimination diets is lower

-Journal of the Royal Society of Medicine Volume 85 September 1992 563

these two conditions, at least 2% or 100 000 ofthe UKpopulation have a major fbod sensitivity which mightbe amenable to fairly simple dietary modification. Itmay, therefore, be concluded that food sensitivity isclinically important, and should be considered inintractable cases of the conditions shown in Table 1,when other methods of treatment have failed. Itshould also be emphasized that successful treatmentof these cases would be highly cost effective to theNHS by preventing further morbidity and expensiveinvestigations.

Short and long term reactionsBy definition food sensitivity reactions occur acutelyand can be reversed within a week by avoidance ofthe particular food or additive. These reactions, whichare the subject of this review should be clearlydistinguished from those that take longer to producesymptoms. Thus diet may produce heart disease andhypertension in genetically susceptible individualswith an incubation period of several decades, compar-able to the tobacco induction ofbronchial carcinoma.Subjects with motor neurone disease may have poorsulphur oxidizing and conjugating enzymes40, andthis could make them sensitive to a food componentor toxin over a prolonged period. There are more thanhalf a million natural compounds in plant foods, andthere is a suspicion that some may have long-termside effects in man41.Food components or toxins could therefore be of

considerable importance in late onset degenerativediseases, but by definition this is distinct from foodsensitivity which is a more acute and reversible process.

ConclusionsTen years on, the question posed in the original article'Food Allergy - Fact or Fiction' can be answered inthe affirmative, with the proviso that food sensitivitywould be a better general description of the clinicalphenomenon. Many patients with otherwise refractorysymptoms can be helped by relatively limited dietarymanipulation.The scepticism associated with this subject has led

to its rejection by many in the academic community,and this in turn has restricted research. The useof frankly dubious techniques by a small number ofalternative practitioners, should not be allowed tonegate the work of serious clinicians and investigatorsin this field. A more liberal view by the academiccommunity could lead to the opening up of animportant area of clinical science.Finally, who should manage patients with food

sensitivity? At the present time there are minimalfacilities within the NHS and many patients are forcedinto an uncontrolled private sector. Perhaps clinicalallergists should widen their horizon from IgEreactions and also accept responsibility for themanagement of food sensitivity in its broad sense.

References1 Finn R, Cohen HN. Food allergy: fact or fiction. Lancet

1978;i:426-82 Food Intolerance and Food Aversion. Ajoint report ofthe

Royal College of Physicians and the British NutritionFoundation. JR Coll Physicians 1984;18:83-122

3 Jewett DL, Fern G, Greenberg MH. A double blind studyof symptom provocation to determine food sensitivity. NEngi J Med 1990;323:429-33

4 Ferguson A. Food sensitivity or self-deception. N EngiJ Med 1990; 323:476-8

5 Scadding G, BrostoffJ. Low dose sublingual therapy inpatients with allergic rhinitis due to house dust mite.Clin Allergy 1986;16:483-91

6 Sehreiber RA, Walker WA. Food allergy: facts andfiction. Mayo Clin Proc 1989;64:1381-91

7 Kajosaari M. Food allergy in Finnish children aged1 to 6 years. Acta Paediatr Scand 1982;71:815-19

8 Walker WA. Pathophysiology of intestinal uptake andabsorption of antigens in food allergy. Ann Allergy1987;59:7-16

9 Kraft AC, Rothberg RM, Knaver CM, Suoboda AC,Monroe LS, Farr RS. Gastric acid output and circulatinganti-bovine serum albumin in adults. Clin Exp Immunol1967;2:321-30

10 Allensmith M, McClellan BH, Butterworth M, MaloneyJR. The development ofimmunoglobulin levels in man.J Paediatr 1968;72:276-90

11 Barnes RMR, Barton PG, Doig JE, Finn R, Harvey MM,Johnson PM. Distribution of serum antibodies towheat gliadin and bovine milk in atopic and non-atopichealthy adults. J Clin Lab Immunol 1983;12:175-8

12 Finn R, Harvey MM,-Johnson PM, Verbov JL, BarnesRMR. Serum IgG antibodies to gliadin and other dietaryantigens in adults with atopic eczema. Clin ExpDermatol 1985;10:222-8

13 Barnes RMR, Harvey MJ, Blears J, Finn R, Johnson PM.IgG subclass ofhuman serum antibodies reacting withdietary problems. Int Arch Allergy Appl Immunol1986;81:141-7

14 Lewis-Jones MS, Barnes RMR, MacFarlane AW, CurleyRK, Johnson PM, Finn R. Frequency and isotopedistribution of serum antibodies reactive with dietaryproteins in adults with chronic urticaria. Clin ExpDermatol 1987;12:419-423

15 Barnes RMR, Johnson PM, Harvey MM, Blears J,Finn R. Human serum antibodies reactive with dietaryproteins - IgG subdose distribution. Int Arch AllergyAppl Immunol 1988;87:184-8

16 Barnes RMR, Johnson PM, Harvey MM, Blears J,Finn R. Human serum antibodies reactive with dietaryproteins - antigenic specifically. Int Arch Allergy ApplImmunol 1988;87:189-93

17 Walker-Smith JA. Food sensitive enteropathies. ClinGastroenterol 1986;15:55-69

18 Bahna SL, Heiner DC. Allergic to milk. New York:Grune & Stratton, 1980

19 Sampson HA, McCaskill CC. Food hypersensitivity andatopic dermatitis: evaluation of 113 patients. JPaediatr1985;107:669-75

20 Warner JO. Food allergy in fully breast fed infants. ClinAllergy 1980;10:133-6

21 Lake AM, Whitington PF, Hamilton SR. Dietary protein-induced colitis in breast fed infants. J Paediatr1982;101:906-10

22 Anderson JA, Sogn D. Adverse reactions to foods.American Academy of Allergy and ImmunologyCommittee on Adverse Reactions to Foods, NationalInstitute of Allergy and Infectious Diseases. NIH pub-lication No. 84-2442. Washington DC. US GovernmentPrinting Office, 1984:4

23 Main J, McKenzie H, Yeaman GR, et aL Antibody toSaccharomyces cerevisiae (baker's yeast) in Crohn'sdisease. BMJ 1988;297:1105-6

24 Barnes RMR, Allan S, Taylor-Robinson CH, Finn R,Johnson PM. Serum antibodies reactive with Saccharo-myces cerevisiae in inflammatoryr bowel disease. Is IgAantibody a marker for Crohn's disease. Int Arch AllergyAppl Immunol 1990;92:9-15

25 Alun Jones V, McLaughlan P, Shorthouse M, WorkmanE, Hunter JO. Food intolerance: a major factor in thepathogenesis of irritable bowel syndrome. Lancet1982;ii:1115-17

26 Hunter JO. The dietary management of Crohn's disease.In: Alun Jones V, Hunter JO, eds. Food and the gut.London: Bailliere Tindall, 1985:221-37

564 Journal of the Royal Society of Medicine Volume 85 September 1992

27 Egger J, Carter CM, Wilson J, Turner MW, Soothill JF.Is migraine food allergy? A double blind controlled trialof oligo antigenic diet treatment. Lancet 1983;ii:865-9

28 Egger J, Carter CM, Graham PJ, Gumley D, Soothill JF.Controlled trial of oligo antigenic treatment in thehyperkinetic syndrome. Lancet 1985;i:540-45

29 Darlington LG, Ramsey NW, Mansfield JR. Placebo-controlled, blind study of dietary manipulation therapyin rheumatoid arthritis. Lancet 1986;i:236-8

30 Kjeldsen-Kragh J, Haugen M, Borchgrevink CF, et al.Controlled trial of fasting and one year vegetarian dietin rheumatoid arthritis. Lancet 1991;338:899-902

31 Sato M, Kojima H, Takayama K, Koshikawa S.Glomerular deposition of food antigens in IgA nephro-pathy. Clin Exp Immunol 1988;73:295-9

32 Dusseldorp M, Katan MB. Headache caused by caffeinewithdrawal among moderate coffee drinkers switchedfrom ordinary to decaffeinated coffee: a 12 week doubleblind trial. BMJ 1990;300:1558-9

33 Chafee FH, Settipane GA. Asthma caused by FD andC approved dyes. J Allergy 1967;40:65-72

34 Supramainiam G, Warner JO. Artificial food additiveintolerance in patients with angio-oedema and urticaria.Lancet 1986;ii:907-9

35 Williams WR, Pawlowicz A, Davies BH. Aspirin-likeeffects of selected food additives and indusial sensitizingagents. Clin Exp Allergy 1989;19:533-7

36 MacGregor CA, Markandu ND, Roulston JE, Jones JC,De Wardener HE. Is 'idiopathic' oedema idiopathic.Lancet 1979;i:397-400

37 Scadding GK, Ayesh R, BrostoffJ, Mitchell SC, WaringRH, Smith RL. Poor sulphoxidation ability in patientswith food sensitivity. BMJ 1988;297:105-7

38 Miller JB. Food allergy: provocative testing and injectiontherapy. Springfield Ill: Charles C Thomas, 1972

39 Miller JB. Intradermal provocation - neutralizing foodtesting and subcutaneous food extract injection therapy.In: Brostoff J, Challocombe CJ, eds. Food allergy andintolerance. London: Bailliere Tindall, 1987:932-46

40 Steventon G, Waring RH, Williams AC, Pall HS, Pall HS,Adams D. Xenobiotic metabolism in motorneurondisease. Lancet 1988;ni:644-7

41 Spencer PS, Nunn PB, Hugon J, et al. Linkage ofGuamamyo-trophic lateral sclerosis/parkinsonism-dementiato a plant excitant neurotoxin. Science 1987;237:517-20

(Accepted 30 October 1991)