Gut, 1991,32, 1386-1391

Morphological and immunohistochemical analysis ofthe human liver in chronic pancreatitis

R P Jalleh, J A Gilbertson, R C N Williamson, C S Foster

AbstractMorphological and immunohistologicalappearances of liver biopsy specimens aredescribed in a personal series of 52 patientsundergoing operation for chronic pancreatitis.The findings are compared with those in aseries of 10 histologically normal liver biopsyspecimens from patients without pancreatitis.Alcohol was the prime aetiological agent in 40ofthe 52 patients (77%). No obvious damage tohepatic parenchymal cells or biliary structureswas observed butminor morphological changesof alcohol associated liver disease were seen in42% of specimens. The most consistentfinding, present in 48 specimens (92%), was achronic inflammatory cell infiltration of portaltracts. In all but one case, T lymphocytespredominated, but a few B cells were present.In four biopsy specimens, T cells spilled overinto adjacent hepatic parenchyma, but therewas no evidence of T cell mediated cytotoxicdamage to the parenchymal cells or biliaryepithelium. It is suggested that these inflam-matory cells are in transit from the pancreasthrough the liver via the portal circulation andmay reflect the underlying pathogenesis ofchronic pancreatitis rather than alcoholic liverdisease.

Departments of Surgeryand Histopathology,Royal PostgraduateMedical School,Hammersmith Hospital,LondonR P JallehJ A GilbertsonR C N WilliamsonC S FosterCorrespondence to:Dr C S Foster, Department ofHistopathology, RoyalPostgraduate Medical School,Hammersmith Hospital, DuCane Road, London W12ONN.

Accepted for publication21 January 1991

Chronic pancreatitis affects approximately threeper 100 000 of the British population, althoughthe incidence may be increasing along with thenational consumption of alcohol.'2 The con-tinuing inflammatory process in the pancreas ischaracterised by irreversible morphologicalchange, especially fibrosis and loss of acini.Alcohol is the dominant aetiological factor inchronic pancreatitis and contributes to severalforms ofliver disease, notably alcoholic hepatitis,cirrhosis, and hepatocellular carcinoma. Yetclinical evidence of a definite associationbetween chronic pancreatitis and alcoholic liverdisease is lacking. 5 Indeed, in our ownexperience ofabout 100 alcoholic patients under-going operation for severe chronic pancreatitis,the liver has nearly always looked normal atlaparotomy.The observed histopathological appearances

of the liver in chronic pancreatitis includeminimal non-specific changes in architectureand are not characteristic of alcohol related liverdisease.36 Some of these changes have beenattributed to extrahepatic biliary obstruction,which is present in up to 27% of patients withchronic pancreatitis.7- Therefore, we haveemployed conventional histological techniques,immunohistochemistry, and lymphocytemorphometry to characterise liver disease in aseries of surgical patients with chronic pan-

creatitis of both alcoholic and non-alcoholicaetiology.

This study aimed to determine whether thelivers of patients with chronic pancreatitisexhibited morphological changes of ethanoltoxicity comparable with the severity of theirpancreatic disease. If not, the objective was toidentify morphological features, consistentlyexpressed, that might indicate alternative patho-genetic mechanisms in chronic pancreatitis.

Methods

PATIENTSLiver biopsy specimens were taken from anunselected personal series of 52 patients under-going operation for chronic pancreatitis betweenDecember 1985 and June 1990 at Bristol RoyalInfirmary or the Hammersmith Hospital,London. There were 40 men and 12 women witha mean age of41 2 years. Detailed social historiesindicated that alcohol was the cause in 40patients, with mean (SD) estimated daily ethanolconsumption of 216 (32) g over a period of 12.8(1.4) years. Causes in the other 12 patientsincluded previous acute pancreatitis in five andcholedochal cyst in one; no cause was identifiedin the remaining six patients. Liver functiontests were abnormal in only seven patients, andnone were jaundiced at the time of operation.Operations were performed by a single surgeon(RCNW) and comprised pancreatic resection(44), pancreaticojejunostomy (5), and sphinc-teroplasty (3). Liver biopsy specimens wereobtained intraoperatively using a Tru-cut biopsyneedle (Travenol Laboratories, Illinois, USA),and more than one specimen was obtained inmost cases.

CONTROL LIVER SPECIMENSTen histologically normal liver biopsy specimenstaken from patients without pancreatitis wereused as control tissues. Seven specimens werefrom patients with no known focal or generaliseddisease who were undergoing surgery aftertrauma. Two specimens were taken during distalpancreatectomy: one for a solitary neuroendo-crine tumour and the other for removal of abenign pseudocyst. One specimen was takenfrom the unaffected lobe of liver during excisionof a benign simple hepatic cyst.

HISTOLOGICAL EXAMINATIONSpecimens were fixed in buffered neutral formolsaline and embedded in paraffin wax. Tissuesections (2 lim) were routinely stained withhaematoxylin and eosin (H&E) to identify

1386

Morphological and immunohistochemical analysis ofthe human liver in chronic pancreatitis

TABLE I Histological and histochemicalfeatures of liverbiopsy specimens

Controls Pancreatitis(n=10) (n=52)

Feature No (%) No (%)

Parenchymal hyperplasia 0 7 (13)Mallory's hyaline 0 - 0Steatosis 0 18 (35)Pericentral fibrosis 0 4 (8)Cholestasis 0 5 (10)PAS 0 0 -Iron 0 1 1 (21)Copper 0 2 (4)Copper-associated protein 0 0 -Portal inflammation 3 (30) 48 (92)Marginal ductular proliferation 0 - 4 (8)Portal fibrosis 0 0Cirrhosis 0 0

PAS= periodic acid sehiff.

morphological features recognised to be associ-ated with ethanol toxicity (Table I). Whereverfeasible, these features were assessed usingnumerical grading systems which ranged from0-III according to whether none, 75% respectively of hepatic paren-chyma was affected. Parenchymal cell hyper-plasia was defined as hepatocyte plates two ormore cells thick and present in more than 15% ofeach specimen. Additional tissue sections werestained histochemically. Reticulin was identifiedby a modified Gordon and Sweet's technique.Periodic acid Schiff (PAS) staining before andafter diastase digestion detected glycogen orintracellular glycoproteins including al anti-trypsin. Perls' Prussian blue reaction was used toidentify iron, which was quantified according tothe criteria of Braillon et al. '° Copper and copperassociated protein were identified by rhodanineand orcein respectively.

IMMUNOHISTOCHEMISTRYMonoclonal antibody UCHLI (anti-human Tcell) was obtained as a gift from Dr PeterBeverley, University College Hospital, London.Monoclonal antibody L26 (anti-human B cell)was purchased from Dako Ltd (High Wycombe,UK). UCHL1 was diluted 1:10 and L26 1:100 in120 mmol/l sodium phosphate buffered saline(PBS, pH 7.4). Polyclonal rabbit antisera tolysozyme and to al antitrypsin (to identify macro-phages) were purchased from Dako Ltd. Bothwere diluted to 1:1000 in PBS. A conventionalindirect immunoperoxidase staining techniquewas used to detect bound monoclonal antibodies.Human tonsillar tissue was used as the positivecontrol. Polyclonal antisera were detected by theperoxidase antiperoxidase technique with humansplenic tissue as the positive control. These lattertissue sections were digested with trypsin (0-1%w/v in aqueous 0.1% CaC12 for seven minutes at37°C) after dewaxing. Endogenous peroxidasewas blocked with 0 3% H202 in distilled waterfor 30 minutes at room temperature. In all cases,sections were incubated with primary antibodiesovernight at 4°C. The resultant immuneperoxidase complexes were developed in 0.05%3,3'-diaminobenizidine hydrochloride (DAB,Aldrich, Gillingham, Dorset, UK) in PBS con-taining 0.03% (w/v) H202. Sections werecounterstained with haematoxylin and mountedin Pertex (Histolab, Hemel Hempstead, UK).

TABLE II Immunohistochemical staining ofmononuclearcells

Controls Pancreatitis(n= 10) (n=52)

Antibody No (%) No (%) p valueUCHL1(T lymphocytes) 3 30 47 90 p0.5

al Antitrypsin(macrophages) 7 70 34 65 p>0. 5

MORPHOMETRYB and T lymphocytes in portal areas werecounted using a lOx 10 mm graticule within themicroscope ocular and a x400 magnificationobjective. For each specimen, the number oflymphocytes in three representative portal areaswere counted and a mean value was calculated.

STATISTICSX2 and unpaired Student's t test were used forstatistical analysis of data.

Results

MORPHOLOGY AND HISTOCHEMISTRYThe morphological and histochemical appear-ances of 52 liver biopsy specimens from patientsundergoing operation for chronic pancreatitisand from the 10 control liver biopsy specimensare shown in Table I. Immunohistochemicalstaining patterns of T cells, B cells, and macro-phages are presented in Table II.

In all liver specimens examined, the hepaticarchitecture was intact and normal vascularrelations were preserved. Parenchymal cellhyperplasia was not identified in any of thecontrol specimens but was present in seven ofthespecimens (13%) from chronic pancreatitispatients. Mallory hyaline bodies were notobserved in any specimen.

In all biopsy specimens, the three portal tractelements were intact. No inflammation of bileducts was seen and no vasculitis was identified.There was no appreciable increase in fibrousconnective tissue, and cirrhosis was not presentin any of the specimens. The most strikingfeature was the relative absence of inflammationfrom portal tracts of the control specimenswhereas 47 of the 52 (90%) specimens frompatients with chronic pancreatitis containedportal tracts that had been infiltrated by increasednumbers of mononuclear cells. The nature ofthese cellular infiltrates is detailed below.

STEATOSISMacrovesicular steatosis of the type associatedwith alcohol induced liver disease was notobserved in any ofthe controls but was present in18 (35%) of biopsy specimens from patients withchronic pancreatitis. In 11 of these specimens,steatotic hepatocytes occupied less than 50% ofthe lobules (grade I). In the remaining sevenspecimens, fatty change occupied between 50and 75% ofthe hepatic lobules (grade II). In 15 of

1387

18alleh, Gilbertson, Williamson, Foster

the 18 (83%) specimens showing fatty change,hepatic parenchyma and portal tracts containedless than 10 polymorph neutrophils per highpower field.

PERICENTRAL FIBROSISA mild increase in pericentral venous fibrosiswas present around some, but not all, centralveins of the contained lobules in four (8%) liverbiopsy specimens from patients with chronicpancreatitis. No consistent association wasfound between perivenous fibrosis and any otherof the commonly accepted characteristics ofalcoholic liver disease.

CHOLESTASISCholestasis was present in only five of the 52 liverbiopsy specimens. In all ofthese, the distributionof bile was periportal and intracanalicular. Nopolymorph neutrophils were present, theappearances being those of non-inflammatorycholestasis. In only one of these five specimenswas marginal ductular proliferation seen. Noother morphological features of chronic biliaryobstruction were identified in any of the speci-mens. Of the five patients in whom morpho-logical features of cholestasis were identified,none were jaundiced and only two had twofoldincreases in serum bilirubin and alkalinephosphatase. However, all five had radiologicalevidence (endoscopic retrograde cholangio-pancreatograph (4) hepatic iminodiacetic acid(HIDA) scan (1)) of extrahepatic bile ductobstruction.

PAS STAININGHistochemical staining with PAS did not showintracellular PAS positive, diastase resistantglobules within hepatocytes of any of the liverspecimens examined. Nevertheless, in onebiopsy specimen, prominent staining for et,antitrypsin involving the hepatocytes indicated aprobable heterozygous al antitrypsin deficiency(an incidence of2% in this series).

SIDEROSISHepatocellular siderosis was noted in 11 patients(21%) and all were grade I. In three of thesepatients, Kupffer cells also contained stainableiron. Only two cases of siderosis were associatedwith steatosis, and in both the steatosis was mild.

Figure 1: Portal region ofliver biopsy specimen stainedusing monoclonal antibodyUCHLI . The portal tract isdensely infiltrated withlymphocytes ofT celllineage. There is spillage ofT lymphocytes into theadjacent parenchymawithout evidence ofcytolyticactivity to portal structures orto adjacent hepatocytes(original magnificationx480).

COPPER AND COPPER ASSOCIATED PROTEINIntracellular copper was not found in any of thecontrol biopsy specimens and was identified inonly two specimens from pancreatitis patientsusing a rhodanine technique. In both specimens,staining was scanty (grade I). The appearanceswere not those of a congenital copper storagedisease. Distribution was confined to a fewperiportal hepatocytes. No correlation wasapparent between this staining and the presenceof cholestasis, steatosis, or iron deposition.Orcein staining did not show stainable copperassociated protein in any of the control orpancreatitis specimens.

PORTAL INFLAMMATIONIn none of the 10 control liver biopsy specimenswere acute inflammatory cells or eosinophilsidentified. In three of these, chronic inflam-matory cells were seen. The phenotypes of thesecells were determined by immunohistochemis-try. However, 48 ofthe 52 specimens (92%) frompatients with chronic pancreatitis containedincreased numbers of inflammatory cells.Neutrophils were present in only 15 specimens.Several eosinophils were identified within theportal tracts ofone specimen in which cholestasiswas also present. Neither neutrophils noreosinophils made more than a scanty contributionto the inflammatory infiltrate within theirrespective specimens.

MARGINAL DUCTULAR PROLIFERATIONMarginal ductular proliferation was not presentin any of the control biopsy specimens but wasidentified in four specimens from the patientswith pancreatitis. In one of these specimens,marginal ductular proliferation was associatedwith cholestasis, while in the remaining three nocholestasis was evident.

PORTAL FIBROSISNo increased portal fibrosis or cirrhosis wereidentified in any specimens from controls orpancreatitis patients.

IMMUNOHISTOCHEMISTRYImmunohistochemistry was performed toidentify and phenotype the mononuclearinflammatory cells in the liver biopsy specimens.Using the monoclonal antibody UCHL1, Tlymphocytes were identified in only three of 10control specimens (30%) with a mean (SEM)density of 24 (4) cells per portal tract. T lympho-cytes, however, were identified in 47 of 52 (90%)specimens from pancreatitis patients. These cellswere almost entirely located in the portal areas(Fig 1). The mean (SEM) number ofT lympho-cytes per portal tract was 45 (5) cells. In fourspecimens, isolated T lymphocytes were seen insinusoidal and canalicular regions (Fig 2). Ineach of these specimens, there was an especiallydense portal infiltration of T cells: mean (SEM)130 (17) cells per portal tract. The relativedensity of T cell infiltration is shown in TableIII.

1388

Morphological and immunohistochemical analysis ofthe human liver in chronic pancreatitis

Figure 2: Focal collectionsofT lymphocytes withinhepatic parenchyma andwithout evidence ofcytolyticactivity to adjacenthepatocytes. No furtherorganisation ofthese smallcellular aggregates wasobserved and no granulomaswere identified (originalmagnification x480).

B lymphocytes were not identified in any ofthe control biopsy specimens. In contrast to theprevalence of T cells in the liver biopsy speci-mens from patients with chronic pancreatitis,however, B lymphocytes were identified in onlyeight specimens (15%). The pattern ofinfiltrationwas similar to that of T cells in that theselymphocytes were aggregated in portal regions.The density of this lymphoid infiltration wassubstantially less than that for T cells, with amean (SEM) of 10 (1) cells per portal area(p100 3 -

Figure 3: Liver biopsy specimen stained using polyclonalantiserum to lysozyme. The identified macrophages andhistiocytes were predominantly lining sinusoids ofparenchymal zone I. A few morphologically identicalsinusoidal lining cells remained unstained (originalmagnification x480).

pancreatitis and those with other causes arepresented in Table IV. Statistical analysis, wherepossible, has shown no differences between thetwo groups except that T lymphocytes weremore frequent in alcoholic patients (p

9Jalleh, Gilbertson, Williamson, Foster

TABLE IV Histologicalfeatures ofalcoholic and nonalcoholic patients with chronic pancreatitis

Alcoholic Non-alcoholic(n =40) (n= 12)No (%) No (%) p value

Portal fibrosis 0 (0) 0 (0) -Steatosis 12 (30) 6 (50) 05>p>01Cholestasis 3 (8) 2 (17) 05>p>0-1Iron 10 (25) 1 (8) 05>p>0-1Copper 2 (5) 0 (0) -UCHL1 (T cell) 38 (95) 9 (75) 005>p>0O02L26 (B cell) 8 (20) 0 (0)Lysozyme 34 (85) 9 (75) 05>p>01ctl Antitrypsin 25 (63) 9 (75) 0-5>p>0-1

increase in fibrous connective tissue withinportal tracts nor was there evidence of cirrhosis.The remarkable absence of severe liver damage,especially in the group of heavy drinkers, standsin sharp contrast to the severe changes of chronicpancreatitis observed in nearly every case.The commonest hepatic abnormality was

steatosis, which was present in 35% of patientsbut severe in only 13%. In only one patient wasthis associated with abnormal liver function.Hepatocellular fat is labile and may thereforereflect the patient's alcohol intake shortly beforeoperation rather than any significant liver celldysfunction. 11The 10% incidence of cholestasis is in keeping

with the recent Glasgow experience.6 Cholestasisprobably reflects extrahepatic biliary obstruc-tion, which was radiologically confirmed in allfive of our patients.7 Cholestasis was not seen infive other patients with biochemical, or radio-logical, or both, evidence of bile duct stenosis.Marginal ductular proliferation, anotherindicator ofrecent bile duct stenosis, was presentin only one of the five specimens showingcholestasis but was identified in an additionalthree patients. Intrahepatocyte copper, a featuresuggesting a chronic cholestatic disorder in theabsence of cirrhosis,'5 was not observed in any ofour cases with cholestasis.

In this present series, we have defined thecharacteristic features of inflammatory infil-tration of the liver in chronic pancreatitis.Lymphoid infiltration, seen predominantly inportal areas, comprised T lymphocytes ratherthan B lymphocytes. Only four patients had adetectable parenchymal distribution of Tlymphocytes. Since all four specimens containeda very dense portal infiltration, these paren-chymal lymphocytes probably reflect a 'spillover' mechanism. In the absence of featuresassociated with hepatocyte toxicity, our datasuggest that the presence of T lymphocyteswithin the parenchyma and portal tractsprobably reflects an upward migration of inflam-matory cells from the pancreas rather than adirect hepatotoxic process.The presence and density of macrophages

within liver biopsy specimens from patients withsuspected alcohol associated chronic pancreatitisis also ofnote. Several studies have shown cells ofthe monocyte macrophage system.to be pheno-typically heterogeneous. 16 l' Human hepaticmacrophages are known to originate fromcirculating monocytes of bone marrow origin.'8They form a self replicating population ofresident macrophages which remains stable

unless a major inflammatory stimulus affects theliver.'9 Immunohistological examination hasindicated the numbers of lysozyme positivehepatic sinusoidal macrophages to be decreasedin alcohol associated liver disease.20 During thatstudy, an undefined portion of sinusoidal liningcells was found to be negative for lysozyme. Asubsequent study confirmed these findings andshowed portal tract macrophages to be increasedin liver biopsy specimens exhibiting only steatosiswhereas lysozyme positive hepatic sinusoidalmacrophages were decreased.2' Heterogeneity ofliver macrophage phenotypes was confirmed bythe disparity of numbers of macrophagesidentified using three different immunohisto-chemical markers.

In this study, the numbers of macrophages incontrol liver biopsy specimens and in those frompatients with chronic pancreatitis were notsignificantly different. These observations aredistinct from the findings in alcoholic liverdisease20 21 and hence provide additional evidencethat alcoholic liver disease and alcohol associatedchronic pancreatitis do not involve exactly similarpathogenetic mechansims. Differences innumbers of macrophages identified using anti-sera to lysozyme and to a1 antitrypsin furthersubstantiate the phenotypic heterogeneity ofhepatic macrophages in both control and pan-creatitis biopsy specimens. Differences in aeti-ology and subsequent pathogenesis of chronicpancreatitis, thus differentially modulating thephenotype of hepatic macrophages, may accountfor the absence of lysozymal staining in nine ofthe specimens from patients with pancreatitis.

It is not understood why patients with alcoholicdamage to the pancreas so seldom have seriousliver disease. Although nutritional factors havebeen implicated,22 23 recent reports do notsupport this hypothesis.2"26 The original sugges-tion that a diet rich in fat and protein mightpredispose to alcoholic liver injury of thepancreas was contradicted by the subsequentdietary studies. Also, alcohol consumption isnotoriously difficult to quantify, particularly inrelation to nutritional status. If chronic pan-creatitis occurs earlier than alcoholic liverdisease, affected patients might not have time todevelop hepatic cirrhosis.3 Similarly, ethanolintake might change with the onset of pancreaticpain or the threat of operation. Yet evidence ofresidual hepatic disease secondary to alcoholwould be expected, but this was not found. Sincecertain human leukocyte antigen (HLA) haplo-types are associated with either alcoholic liverdisease or chronic pancreatitis, genetic factorscould help to determine which disease processdevelops; however, current data are inconclusiveand require further elaboration.27"29 Lastly, it isknown that in chronic pancreatitis liver andpancreatic damage proceed independently.30

Current evidence indicates that chronicpancreatitis is not a single defined entity, but adisease process comprising a group of differentaetiologies and pathogenetic mechanisms whichshare a few common morphological end points.If a single agent, such as alcohol, is the mostimportant aetiological factor in the genesis of thisdisease, then the exocrine pancreatic tissues ofaffected patients, when compared with hepatic

1390

Morphological and immunohistochemical analysis ofthe human liver in chronic pancreatitis 1391

parenchymal tissue, seem to be unduly sensitiveto its toxic effects. Such sensitivity is eithercongenital or acquired in origin. The finding ofraised numbers of T lymphocytes infiltratinghepatic portal tracts, particularly in the absenceof acute inflammatory cells, might indicate thata cell mediated immune component is involvedin human chronic pancreatitis. We suggest thatethanol is not acting as a simple toxin in a doserelated manner but rather as an adjunct or'trigger' that renders exocrine pancreatic tissuesensitive to cell mediated cytotoxicity.

We are grateful for the assistance of Professor J W B Bradfield,Department ofHistopathology, University of Bristol in permittingunreserved access to the archived specimens at the RoyalInfirmary.

1 Read G, Braganza JM, Howat HT. Pancreatitis - a retrospec-tive study. Gut 1976; 17: 945-52.

2 MacLaren IF. Observations and surgical management ofchronic pancreatitis in the British Isles. A review of thetwentieth century. WorldJ Surg 1990; 14: 19-27.

3 Angelini G, Merigo F, Degani G, et al. Association of chronicalcoholic liver and pancreatic disease: a prospective study.AmJ Gastroenterol 1985; 80: 998-1003.

4 Sarles H, Sarles JC, Camette R, et al. Observations in 205confirmed cases of acute pancreatitis, recurring pancreatitisand chronic pancreatitis. Gut 1965; 6: 545-52.

S Strum WB, Spiro HM. Chronic pancreatitis. Ann Intern Med1971; 74: 264-77.

6 Wilson C, Auld CD, Schlinkert R, et al. Hepatobiliarycomplications in chronic pancreatitis. Gut 1989; 30: 570-7.

7 Afroudakis A, Kaplowitz N. Liver histopathology in commonbile duct stenosis due to chronic alcoholic pancreatitis.Hepatology 1981; 1: 65-72.

8 Aranha GV, Prinz RA, Freaark RJ, Greenlee HB. Thespectrum of biliary tract obstruction from chronic pan-creatitis. Arch Surg 1984; 119: 595-600.

9 Scott J, Summerfeld JA, Elias E, Dick R, Sherlock S. Chronicpancreatitis: a cause of cholestasis. Gut 1977; 18: 196-201.

10 Braillon A, Capron JP, Herve MA, et al. Liver in obesity. Gut1985; 26: 133-9.

11 Dutta SK, Mobrahan S, Iber FL. Associated liver disease inchronic pancreatitis. Dig Dis 1978; 23: 618-22.

12 Scheuer PJ. Liver disease in the alcoholic. In: Scheuer PJ.

Liver biopsy interpretation. London: Bailliere Tindall, 1988:83-98.

13 Nokano M, Worner TM, Lieber CS. Perivenular fibrosis inalcoholic liver injury: ultrastructure and histologicprogression. Gastroenterology 1982; 83: 777-85.

14 Worner TM, Lieber CS. Perivenular fibrosis as precursorlesion of cirrhosis.JAMA 1985; 254: 627-30.

15 Guarascio P, Yentis F, Cevikbas U, Portmann B, Williams R.Value of copper-associated protein in diagnostic assessmentof liver biopsy. J Clin Pathol 1983; 36: 18-23.

16 Cline MJ, Lehrer RI, Territo MC, Golde DW. Monocytes andmacrophages: functions and diseases. Ann Intern Med 1978;88: 78-88.

17 Radzun HJ, Parwaresch MR. Differential immunohisto-chemical resolution of the human mononuclear phagocytesystem. Cell Immunol 1983; 82: 174-83.

18 Gale RP, Sparkes RS, Golde DW. Bone marrow origin ofhepatic macrophages (Kupffer cells) in humans. Science1978; 201: 937-8.

19 Summerfield JA, Jones EA. Further progress towards under-standing hepatic sinusoidal cells. J Hepatol 1986; 3: 413-8.

20 Mills LR, Scheuer PJ. Hepatic sinusoidal macrophages inalcoholic liver disease. J Pathol 1985; 147: 127-32.

21 Karakucuk I, Dilly SA, Maxwell JD. Portal tract macrophagesare increased in alcoholic liver disease. Histopathology 1989;14:245-53.

22 Sarles H. An international survey of nutrition and pancreatitis.Digestion 1973; 9: 389-403.

23 Uscanga L, Robles-Diaz G, Sarles H. Nutritional data andaetiology of chronic pancreatitis in Mexico. Dig Dis Sci1985;30: 110-3.

24 Pitchumoni CS, Sonnenschein M, Candido FM,Panchacharam P, Cooperman JM. Nutrition in the patho-genesis of alcoholic pancreatitis. Am J Clin Nutr 1980; 33:631-6.

25 Wilson JS, Bernstein L, MacDonald C, Tait A, McNeil D,Pirolla RC. Diet and drinking habits in relation to thedevelopment of alcoholic pancreatitis. Gut 1985; 26: 882-7.

26 Mezey E, Kolman CJ, Diehl AM, Mitchell MC, Herlong HF.Alcohol and dietary intake in the development of chronicpancreatitis and liver disease in alcoholism. Am J Clin Nutr1988;48: 148-51.

27 Eddleston ALWF, Davies M. Histocompatibility antigens inalcoholic liver disease. BrMed Bull 1982; 38: 13-6.

28 Saunders JB, Wodak AD, Williams R. What determinessusceptibility to liver damage from alcohol? J R Soc Med1984; 77: 204-16.

29 Forbes A, Schwarz G, Mirakian R, et al. HLA antigens inchronic pancreatitis. Tissue Antigens 1987; 30: 176-83.

30 Sandilands D, Jeffrey IJM, Haboubi NY, MacLennan IAM,Braganza JM. Abnormal drug metabolism in chronic pan-creatitis - treatment with antioxidants. Gastroenterology1990; 98: 766-72.