Gut, 1990,31, 1150-1155

SPECIAL REPORT

Future needs for ERCP: incidence of conditionsleading to bile duct obstruction and requirements fordiagnostic and therapeutic biliary procedures

MW L Gear, N A Dent, D G Colin-Jones, J H Lennard-Jones, J R T Colley

AbstractAlthough the development of endoscopicmethods of treatment for biliary obstructionhas proceeded rapidly in recent years, endo-scopic retrograde cholangiopancreatographic(ERCP) services are patchily distributed. Arecent survey by the British Society of Gastro-enterology has shown that almost half thedistrict general hospitals questioned did nothave a sphincterotomy service availablelocaily. To assess the level of provisionrequired, two investigations have been under-taken. Firstly, an epidemiological study of bileduct obstruction has been carried out in theSouth Western Region. Secondly, the actualsurgical and endoscopic workload in treatingobstructive jaundice has been analysed in twohealth districts. Using present incidence andtreatment rates at least 50 ERCPs per 100 000of the population per year are estimated to berequired in the future. Surgical treatment ratescan be expected to fall as the number oftherapeutic ERCPs increases. The implica-tions of this estimate in equipment and staffingterms are discussed.

Obstruction of the bile duct produces severe andobvious clinical features that may includejaundice, pain, or fever. Thus, the patient willusually be referred for hospital investigation andtreatment. The common causes of bile duct

Department of Surgery,Gloucestershire RoyalHospital, GloucesterMW L Gear

Department ofEpidemiology andCommunity Medicine,Canynge Hall, BristolN A DentJ R T Colley

Department of Medicine,Queen AlexandraHospital, PortsmouthD G Colin-Jones

St Mark's Hospital,LondonJ E Lennard-JonesCorrespondence to:Mr M W L Gear, Departmentof Surgery, GloucestershireRoyal Hospital, GloucesterGL1 3NN.Accepted for publication1 December 1989

obstruction are gall stones and tumours. Beforethe development of endoscopic methods of treat-ment for bile duct obstruction, most patientsrequired operation - exploration of the duct andextraction of stones, resection, or bypass fortumours.The development of endoscopic retrograde

cholangiopancreatography (ERCP) was followedby sphincterotomy for gall stone in 1974' andretrograde insertion of a biliary endoprosthesisfor tumours (stenting) in 1981.1 Within theUnited Kingdom these methods are being usedincreasingly in some centres and not at all inothers. A recent survey by the British Society ofGastroenterology has shown that almost half thedistrict general hospitals questioned did not havea sphincterotomy service available locally.3 Inthe South Western Regional Health Authority,for example, excluding Bristol, only three ofeight district general hospitals currently have anERCP service.At first, endoscopic sphincterotomy for duct

stone was reserved for elderly and frail patientswho had had previous biliary surgery. Morerecently, the indications have been widened toinclude older patients with gall bladder in situ,thus avoiding surgery altogether in manyinstances." Endoscopic stenting for malignantbile duct obstruction was reported after thepercutaneous transhepatic methods had beenused for some time. A recent trial from theMiddlesex and the London Hospitals has sug-

TABLE I Origins and derivation ofdata (1981 population base)

Disease Incidence Workload Data problems

Cancer: Cancer registration and death Theatre or endoscopy record. Overestimate of potential workloadDuct/ampulla certificates SWRHA (population Bypass or resection or stent (two due to confusions in coding duct asPancreas 3066 000) districts population 812 000) distinct from gall gladder, and

head of pancreas as distinct frombody/tail. Some bypass surgerydone for non-malignant conditions(about I in 10 in HAA)

Duct stones HAA exluding readmissions Exploration ofCBD or endoscopic Under-recording of subsidiarywithin 3 months (3 066 000) sphincterotomy. Theatre or diagnosis in HAA. Overestimate

endoscopy record (812 0000) of cases due to uncertain matchingof readmissions

Other duct pathology, HAA excluding readmissions Any of the key workload categories Diagnostic categories not adequateobstructive within 3 months (3 066 000) associated with other diagnoses or not adequately recorded injaundice HAA. Incidence is procedure

(workload) led and includesapparently negative cases, egECBD

Relevant codes: ICD 9 and operations:Stones ICD 574.X; duct stones ICD 574. 3,4,5; biliary tumours ICD 156; pancreatic tumours ICD 157; exploration ofcommon bile duct510; choledocholithotomy 51 1; Bypass operations 513,523,524; pancreatectomy 531; partial pancreatectomy 532; ERCP 539.Office of Population Census and Surveys surgical codes 4th revision were brought into use in Gloucester in 1988. These have resulted incloser agreement between theatre registers and HAA.HAA=Hospital Activity Analysis; CBD=common bile duct; ECBD=exploration ofcommon bile duct; ERCP=endoscopic retrogradecholangiopancreatography; SWRHA=South Western Regional Health Authority.

1150

on June 23, 2021 by guest. Protected by copyright.

http://gut.bmj.com

/G

ut: first published as 10.1136/gut.31.10.1150 on 1 October 1990. D

ownloaded from

http://gut.bmj.com/

Future needsforERCP

TABLE II 1981 Populations by age (years), numbers in thousands (%)

S54 -55 S65 :75 Total

Gloucester and Portsmouth 602 6 90-2 72-1 47 0 811-9Health Districts (74 2) (11 1) (8 9) (5 8) (100-0)

England 34393 3 5261-5 4372-3 2758 7 46785-8(73-5) (11-2) (9 3) (5 9) (100 0)

Source - Office of Population Census and Surveys, population estimates. 13

gested that the endoscopic method is safer andmore successful in relieving jaundice than thepercutaneous route.7

Surgical resection of tumours causing obstruc-tion of the bile duct is only possible in up to10-20% of patients and even simple palliativebypass procedures are associated with substantialmorbidity and a mortality ofup to 30%.8 Further-more, patients with tumours are often elderlyand considered unfit for surgery. Thus, non-operative methods of palliation in these patientsprovides an attractive alternative treatment.9The advent of endoscopic methods for investi-

gating and treating bile duct obstruction hasconsiderable implications for hospitals through-out the country. They provide a potential savingin operating theatre requirements, but a needfor extra endoscopic and radiological facilities.In addition, skilled and trained medical andnursing staff are required.

This paper estimates the potential endoscopicworkload for treatment of obstructive jaundicein two ways. Firstly, the incidence of conditionsleading to bile duct obstruction is assessed andsecondly the actual surgical and endoscopicworkload in two health districts has been deter-mined. Additionally, the effects of endoscopictechniques on surgical practice have beenanalysed in these districts. From these data, anattempt is made to predict future requirementsfor diagnostic and therapeutic ERCP facilitiesfor each 100 000 of the population.

MethodsThe methods used in this study are set out anddiscussed in the Appendix. The data sources,together with brief details of the problemsencountered, the population sizes, and agestructure are summarised in Tables I and II.

Results

REGIONAL INCIDENCE OF CONDITIONS LEADING TO

BILE DUCT OBSTRUCTIONTumours. The incidence of carcinoma of thepancreas and biliary system in the South WesternRegion has been rising slowly over the 10 yearsfrom 1976 (Fig 1). In recent years approximately480 new patients have presented each year,equivalent to 16 in a population of 100 000.Neoplasms of the bile ducts or gall bladderaccount for about one in six of the cases (TableIII). A striking feature of the age distribution isthat 76-8% (1869 of 2433) of the whole group areover the age of 65 years, and indeed 42 9% (1045of 2433) are 75 years or older.Not all patients with carcinoma of the pan-

creas will have obstruction of the bile duct. Someassessment of the minimum numbers receiving

7001

600

500

D) 400E

z 300

200

100*

Hospital admissions

Cancer registrations

Bypass and resectionoperations

1976 78 80 82 84 86Year

Figure 1: Annual hospital admissions and operationscompared with cancer incidence (pancreas and bile duct)forSouth Western Regional Health Authority between 1976 and1986.

*Estimate based on death certificates.

surgical treatment can be derived from the dataon treatment episodes and admissions (Fig 1).Hospital admission rates for cancers exceed theannual incidence because a proportion ofpatients are admitted more than once, despitethe fact that the two year survival rate is onlyabout 4% for cancer of the pancreas.'0 Thus thereare approximately 650 hospital admissions peryear for these cancers in this region. The numberof operations, however, is about half the admis-sion rate, at approximately 300 per year - that is10 per 100 000 of the population. This is partlybecause not all carcinomas of the pancreas affectthe bile duct and partly because some patientsare so old and frail that surgery is not possible.

Bile duct stones. The hospital admission ratefor all gall stones in the period 1979-86 in thisregion was approximately 100 per 100 000. Ductstones were present in about one in eight patientsgiving an approximate admission rate of 13 per100 000 (Table IV). The proportion of episodesin which duct stones were mentioned increasedfrom 9% below the age of 64 years to 19% in the65+ age group. Fifty eight per cent of all cases ofduct stones (1882 of 3253) occurred in patientsover the age of 65, and 29% were over 75 years.

For reasons discussed in the Appendix(mainly problems in ascertainment and coding)these figures are likely to be an underestimate ofthe true incidence of episodes involving ductstones. If the proportions of missed diagnosticcodes are similar to those found in the districtstudies for exploration of the common bile ductoperations (Fig 2) the true figure may be about20% higher - say 16 per 100 000.

Ninety per cent of all readmissions in patientswith duct stones which were identified inHospital Activity Analysis (HAA) within twoyears of a first admission occurred within threemonths of that admission. Taking the view that

TABLE III Numbers ofpatients with carcinoma of thepancreas and biliary system by age (yrs) registered in the fiveyear period 1979-83 in the South Western Regional HealthAuthority (population 3 066 000)

No (rates/100 000 peryear)

Tumour(ICD9) 65 ¢65 All ages

Gall bladder and bile ducts(156) 100 (0 8) 312 (11-6) 412 (2 7)

Pancreas(157) 464(3-7) 1557(57-8) 2021 (13-2)Totals 564(4-5) 1869(69-4) 2433(15-9)

Source - Cancer Registry. '°

1151

on June 23, 2021 by guest. Protected by copyright.

http://gut.bmj.com

/G

ut: first published as 10.1136/gut.31.10.1150 on 1 October 1990. D

ownloaded from

http://gut.bmj.com/

Gear, Dent, Colin-Jones, Lennard-Jones, Colley

TABLE IV Hospital admission ratesfor biliary stones, by age(yrs), between 1979 and 1986 in the South Western RegionalHealth Authority (population 3 066 000)

No (rates/I10 000 peryear)

Future needsforERCP

TABLE VI Exploration ofcommon bile duct (ECBD) and therapeutic endoscopic retrogradecholangiopancreatography (ERCP)for stone by age (yrs) and period in Gloucester andPortsmouth Health Districts combined

65 All ages

ECBD ERCP ECBD ERCP ECBD ERCP Total

No:1977-81 360 33 303 73 663 106 7691982-6 273 73 227 204 500 277 777

Total 1977-86 633 106 530 277 1163 383 1546All treatments (%) 739(48) 807 (52) 1546(100)Rates (all treatments)/100 000 per year:

1977-81 11-3 63-2 18-91982-6 10 0 72-4 19-1

Total 1977-86 10-7 67-8 19-0Therapeutic ERCP as % of all treatments:

1977-81 8-4 19-4 13-81982-6 21-1 47-3 356

Source - local theatre and endoscopy records.

technique such as therapeutic ERCP there areseveral major problems. Firstly, there aredifficulties in analysing routine data, in part dueto failure to record some diagnoses and treat-ments completely (see Appendix). Secondly,with rapidly changing technologies such as theadvent of lithotripsy, the extent to which thera-peutic ERCP is used in the years ahead must bein part speculation. A recent leading report,however, has recommended ERCP as the bestmeans of treatment for most patients with bileduct obstruction." In the future more patientsare likely to be treated overall, as ERCP is usedincreasingly for those who would be refusedsurgery. There are already signs that this isoccurring in both Gloucester and Portsmouth(Figs 3 and 4).The present situation in the management of

bile duct obstruction is unsatisfactory in thatERCP is available in some district hospitals andnot in others3 despite evidence that ERCP,particularly for older patients, is associated witha shorter hospital stay and lower complicationrate.9

Using regional data, the incidence rate forcancer of the pancreas and bile ducts is 16 per100 000. There are also a small number ofpatients who have biliary obstruction due toother secondary tumours. Hitherto, at least onethird of these patients has been unsuitable forany surgical treatment. It seems reasonable toestimate that in future more patients will besuitable for endoscopic palliation and that theproportion treatable may rise to about 12 per100 000, eight being treated endoscopically andthe remainder being treated surgically or beingunsuitable for either mode of treatment.

Also from regional data, the minimum

TABLE VII Overall treatment rates for malignant obstructionand bile duct stones by age (yrs). All casesfor the 10years1977-86 in Gloucester and Portsmouth Health Districtscombined (mid-population 812 000)

No (ratesl/00 000 peryear)-64 ¢65 All ages

Exploration ofCBD orsphincterotomy for stone 739 (10-7) 807 (67-8) 1546 (19-0)

Bypass, pancreatectomy, orstents 205 (3 0) 442(37-1) 647(8 0)

All cases 944 (13-7) 1249 (104-9) 2193 (27-0)

Source - local theatre and endoscopy records.CBD=common bile duct.

120-

110-

100-

90-

80-

0CD

-

Ez

70-

60-

50-

40-

30-

10-

0

Exploramon of 'common bile duct,panceatectomy,or bypass

1976 78 80 82 84 86 88Year

Figure 3:Annual workloadfor surgical and endoscopic biliaryprocedures in Gloucester 1977-88.ERCP=endoscopic retrograde cholangiopancreatography.

165

150-

135

120-

105-

0

00

C.)0

Ez

75.

45-

30-

15-

0

panceatectomy,or bypass

TherapeuticERCP

19 I6.880 82 841976 78680 2 84 86 88Year

Figure 4: Annual workloadfor surgical and endoscopic biliaryprocedures in Portsmouth 1977-88.ERCP=endoscopic retrograde cholangiopancreatography.

1153

on June 23, 2021 by guest. Protected by copyright.

http://gut.bmj.com

/G

ut: first published as 10.1136/gut.31.10.1150 on 1 October 1990. D

ownloaded from

http://gut.bmj.com/

Gear, Dent, Colin-Jones, Lennard-Jones, Colley

TABLE VIII Overall treatment rates by age for all cases for1988 in Gloucester and Portsmouth Health Districts combined(population 841 500)

No (ratesIlOO 00 peryear)

-64 :65 All ages

Exploration ofCBD orsphincterotomy for stone 62 (8-7) 117 (42-1) 179 (21-3)

Bypass, pancreatectomy, orstents 16 (2 2) 61 (48 0) 77 (9-2)

All cases 78 (10-9) 178 (140-2) 256 (30 4)

Source - local theatre and endoscopy records.CBD=common bile duct.

incidence rate (including readmissions) for ductstones is 13 per 100 000. Our district studies leadus to believe that this is an underestimate andshould be of the order of at least 16 per 100 000because of the data problems discussed in theAppendix. Therefore the combined incidencerates for treatable conditions would be 28 per100 000.This estimate is supported by the analysis of

treatment rates in two health districts based on10 years' data which showed a treatment rate of27-0 per 100 000.The most recent data for Gloucester residents

show that the proportion ofinterventions for bileduct obstruction which are carried out endo-scopically has increased from 50% to 80% in thelast two years. Similar trends are evident inPortsmouth. About one in five treatments arerepeat procedures - a readmission rate similar tothat described above for surgery. First endo-scopic treatments in 1988 occurred at a rate of 23per 100 000 but repeat treatments raised thisfigure to 30 per 100 000.

Although there will always be some who

120

110-

100-

Total80- ERCP

6 70

ho 60-E Diagnostic

50

require surgical treatment, for example youngerpatients with malignancy or those with empyemaof the gall bladder and duct stones, there is littleevidence yet that the transition from surgery toendoscopic management is complete. Thus, upto 33 to 35 endoscopic treatments per 100 000may well be required. Although some of thesemay be at the expense of diagnostic ERCP, thelatter has remained steady at 15 per 100 000 inGloucester in the past 10 years while endoscopictreatments have risen (Fig 5). Consequently,new and repeat treatments and diagnosticERCPs together equate to an eventual rate ofaround 50 per 100 000. To take some account ofdiffering age structures in other districts, analternative way of expressing needs might be interms of those aged over 65. Based on treatmentrates of 140 per 100 000 (Table V) future require-ments, including diagnostic ERCP, will be inexcess of 200 per 100 000 over 65.

If the population served is around 500 000(that is, two adjoining districts) then 250 ERCPswill be required annually - equivalent to twosessions each of approximately three to fourpatients most weeks. Ideally these would beprovided by two consultants, a physician, and asurgeon, thus allowing cover for holidays and forjoint management of acutely ill patients. Radio-logical support, including ultrasound and inter-ventional techniques, are also required. Theadvent of the combined percutaneous and endo-scopic stenting technique, which increases theendoscopic success rate in our experience and inthat of others'2 to close to 100%, reinforces theneed for specialist radiological skills. Teachinghospitals, with referral from several healthdistricts as well as research commitments, mayrequire more sessions than is suggested by theirnominal catchment area.ERCP procedures are technically difficult,

with a long 'learning curve.' In the early stagesthere is no substitute for a substantial period ofapprenticeship in an established unit, often ateaching centre, as well as attendance at trainingcourses.3 For a small district it is doubtfulwhether an ERCP service can be justified. Thebest solution would seem to be for adjoiningdistricts to cooperate in setting up a combinedunit so that sufficient staff, expertise, and equip-ment is acquired in both the endoscopy andradiology departments. Alternatively, districtsclose enough to a regional centre may referpatients there.

Appendix

DATA COLLECTIONIntroduction. The interpretation of morbidityrates based on routine National Health Servicestatistics is hampered by many difficulties.Both data collection and data analysis may beselective. Hospital readmission and recurrencemay complicate attempts to measure diseaseincidence.

For patients with cancer, the NationalRegistration Scheme provides a simple singlerecord of incidence. Individuals with cancer ofthe pancreas or biliary tract have a short lifeexpectancy,'" and there is thus a close correlation

YearFigure 5: Total endoscopic retrogradecholangiopancreatography (ERCP) workload - diagnosticand therapeutic -for Gloucester 1973-88.

1154

on June 23, 2021 by guest. Protected by copyright.

http://gut.bmj.com

/G

ut: first published as 10.1136/gut.31.10.1150 on 1 October 1990. D

ownloaded from

http://gut.bmj.com/

FutureneedsforERCP 1155

40-Theatre register

30- > =t; _HAA

E20E

10 Bypass surgery and pancreatectomy10

1977 79 81 83 85Year

Figure 6: Companrson oftheatre register and Hospital ActivityAnalysisfor biliary bypass surgery and pancreatectomy(Gloucester).

between numbers of registrations and numbersof deaths, leading.to confidence in the estimatesof incidence for these cancers.

For patients with stones in the biliary tract, itis difficult to obtain an accurate figure forincidence, as there is no register or longitudinalrecord available. Apart from any special localrecord systems which may have been developed,statistics can only be provided from HospitalActivity Analysis (HAA) and kindred systems -forexample, Hospital In-Patient Enquiry (HIPE)which describes hospital In-Patient episodes andwhich does not readily distinguish betweenseparate patients, readmissions, and transfers.All these factors tend to inflate simple estimatesof incidence based on these sources.

Procedures laid down for data extraction andcoding also affect results in the HAA. Forexample, the codes of the ICD 9 revision attemptto distinguish stones in the bile duct (codes574 3-574 5) from calculus in the gall bladder.However, when coding, it is likely that in someinstances the stone in the duct is overlooked orspecificity is lost through precedence being givento the stone in the gall bladder. Similarly, theremay be some failures to code such apparentlysubsidiary operations as choledocholithotomy(code 511) or exploration of the common bileduct (510), where cholecystectomy is moreimmediately perceived and coded (code 522).The combined effect of these factors has dis-torted the picture provided by national statisticsin the past, particularly in the apparent ratiobetween types of operations performed.'3Earliest impressions of the fourth revision of theOffice ofPopulation Census and Surveys surgicalcodes, which were brought into use in Gloucesterin 1988, suggest that many of these codingdifficulties may not arise in the future.

Method. A preliminary and rigorous analysisof the regional HAA file to include subsidiaryHAA diagnosis and operations was carried out.Special searches of theatre registers in two dis-tricts showed that exploration of the commonbile duct appeared less frequently in HAA thanin theatre registers because of negative findingson exploration (Fig 2). Choledocholithotomywas mentioned more often in HAA especiallywhere calculus in the duct appears as one of thecoded diagnoses. In contrast, there was closecorrelation between theatre register and HAAdata for biliary bypass surgery (Fig 6).Thus, for the purpose of our study, data for

therapeutic events for 1977-88 were extracted byone of us (NAD) from local theatre and endo-

scopy records in the Gloucester and Portsmouthhealth districts. Additionally, in PortsmQuth, acomputer system operated in the division ofsurgery provided further information for onehospital from 1983-8, which confirmed the find-ings from some theatre registers.As theatre registers do not provide a record of

diagnosis, disease incidence is derived fromHAA and the National Registration Schemethroughout and is based on a whole region.Because the number ofpatients involved tends tobe overestimated in HAA due to readmission orhospital transfer, an attempt was made to estab-lish incidence by matching successive HAAadmissions with any mention of duct stone,within district, by unit number, date of birth,sex, and diagnosis.

In determining age specific rates, data forseveral years were compared and combined forthe two study districts and related to relevantOffice of Population Census and Surveysdenominators.'4 This provided a population ofapproximately 812 000 in 1981, the midpoint ofthe study, with a broad age composition similarto that for England as a whole (Table II).The data sources, together with brief details of

the problems encountered, broad populationnumbers, and the ICD 9 diagnostic codes used inthis study, are set out in Table I.

This study was undertaken on behalf of the Clinical ServicesCommittee of the British Society of Gastroenterology.We are most grateful to the Gloucester District Health

Authority and the research fund of the British Society ofGastroenterology for providing finance for this project. We aregrateful to the surgeons in Gloucester and Portsmouth for access tothe theatre records and to Mr W G Prout for the use of thePortsmouth computer system. We also thank Mr R Moody of theRegional Health Authority for assistance with the HospitalActivity Analysis system. We thank Mrs Gillian Kellett for all herhard work in the preparation of the paper and Miss Jane Price forthe diagrams.

1 Classen M, Demling L. Endoscopische Sphincterotomie derPapilla Vater. Dtsch Med Wochenschr 1974; 99: 496-7.

2 Huibregtse K, Havercamp HJ, Tytgat GN. Trans-papillarypositioning of a large bore 3-2 mm biliary endoprosthesis.Endoscopy 1981; 13: 217-9.

3 Report of a working party of the British Society of Gastro-enterology. The staffing ofendoscopy units. British Society ofGastroenterology, 3 St Andrew's Place, London NW1 4LB:1987.

4 Neoptolemos JP, London NJ, James D, Carr-Locke DL,Bailey IA, Fossard DP. Controlled trial of urgent endoscopicretrograde cholangiopancreatography and endoscopicsphincterotomy versus conservative treatment for acutepancreatitis due to gall-stones. Lancet 1988; 2: 979-83.

5 Neoptolemos JP, Carr-Locke DL, Fossard DP. Prospectiverandomized study of pre-operative endoscopic sphinctero-tomy versus surgery alone for common bile duct stones.BrMedj 1987; 294: 470-4.

6 Cotton PB. Endoscopic management of bile duct stones:(apples and oranges). Gut 1984; 25: 587-97.

7 Speer AG, Cotton PB, Russell RCG, et al. Randomized trial ofendoscopic versus percutaneous stent insertion in malignantobstructive jaundice. Lancet 1987; 2: 57-62.

8 Little JM. A prospective evaluation of computerized estimatesof risk in the management of obstructive jaundice. Surgery1987; 102: 473-6.

9 Shepherd HA, Royle G, Ross APR, Diba A, Arthur M, Colin-Jones DG. Endoscopic biliary endoprosthesis in thepalliation of malignant obstruction of the distal common bileduct: a randomised trial. BrJ Surg 1988; 75: 1166-8.

10 South West Regional Health Authority. Report of the CancerRegistry. SWRHA Bristol, 1987.

11 Summerfield JA. Biliary obstruction is best managed byendoscopists. Gut 1988; 29: 741-5.

12 Robertson DAF, Ayres R, Hacking CN, Shepherd H, Birch 5,Wright R. Experience with a combined percutaneous endo-scopic approach to stent insertion in malignant obstructingjaundice. Lancet 1987; 2: 1449.

13 Office of Population Census and Surveys. Hospital inpatientenquiry main tables. Series MB4, No 21, 23, 25, 27. London:HMSO, 1982-5.

14 Office of Population Census and Surveys. Population estimates.SeriesPPI No4, No. 6, No8. London: HMSO, 1979, 1981,1983.

on June 23, 2021 by guest. Protected by copyright.

http://gut.bmj.com

/G

ut: first published as 10.1136/gut.31.10.1150 on 1 October 1990. D

ownloaded from

http://gut.bmj.com/