Clin Pathol 1993;46:1 18-122

Importance of apoptosis in the histopathology ofdrug related lesions in the large intestine

F D Lee

AbstractAim: To investigate the possibility thatthe incidence of apoptotic bodies in thecryptal epithelium might help to identifycolonic lesions due to drugs, especiallynon-steroidal anti-inflammatory drugs(NSAIDs).Methods: The apoptotic count (AC) thenumber of apoptotic bodies per 100crypts was calculated in a series of colo-rectal biopsy specimens, stained withhaematoxylin and eosin from patientswith (a) known or suspected druginduced colitis and (b) inflammatorybowel disease before or after treatmentwith salazopyrine or corticosteroids.These specimens were compared withnormal biopsy specimens from a controlgroup of comparable age and sex distrib-ution.Results: Under normal conditions apop-totic bodies were seldom seen at all andthe mean apoptotic count was less than1'0. In untreated inflammatory boweldisease the mean apoptotic count wasmarginally increased (2.4), but whenthere was a partial response to drugtreatment the apoptotic count rose to13-1 (p 0.003). In colonic lesions directlyattributable to drugs the apoptotic countwas always increased, reaching its high-est level (106) with 5-fluorouracil. In coli-tis related to NSAIDs apoptoses wereassociated with inflammation, mostnotably an increase in lymphocytes inboth lamina propria and epithelium.Conclusion: The presence of crypt apop-toses in substantial numbers (with anapoptotic count in excess of 5) shouldalways raise the possibility of drug effect.The mechanisms involved are not clearbut with NSAIDs the changes might wellbe immunologically mediated.

(C Clin Pathol 1993;46:118-122)

incriminated in such diverse conditions asileal ulceration,' diaphragm disease of thesmall bowel,'3 non-specific colitis,4 colonicperforation5 and jejunal villous atrophy withmalabsorption.6 This list would no doubt beexpanded if some kind of objective histologi-cal marker for drug effects could be identified,particularly with regard to colorectal biopsyspecimens.The peculiar form of cell dissolution some-

times referred to especially in its physiologicalcontext as apoptosis' is well recognised totake place in the epithelium of the intestinalcrypts and gastric pits during the graft versushost reaction which may complicate bonemarrow transplantation.8 9 Similar changeshave also been seen in the colonic lesions inassociation with AIDS'°IO and in inflammato-ry bowel disease.'0 It has long been known,however, that apoptosis of the colonic epithe-lium may follow the administration of certaincytotoxic agents, most notably 5-fluoro-uracil."2 There is also experimental evidencethat melanosis coli may be related to anthro-quinone induced apoptosis of colonic epithe-lial cells."3The possibility that the presence of apopto-

sis especially in the colonic crypts might beindicative of exposure to other drugs, and inparticular to NSAIDs, was raised by twogroups of personal observations. First, apop-tosis was found to be a conspicuous feature incases of colitis related beyond reasonabledoubt to the administration of mefenamicacid and diclofenac (voltarol). Secondly dur-ing a review of biopsy specimens frompatients with putative inflammatory boweldisease (IBD) the impression was gained thatapoptosis was more often seen in treated thanuntreated cases. It was thus consideredappropriate to amplify these observations toassess the possible value of apoptosis in theidentification of drug induced intestinal dis-ease.

Department ofPathology, GlasgowRoyal Infirmary,Glasgow G4 3SFF D LeeCorrespondence to:F D LeeAccepted for publication12 August 1992

The aetiological importance of therapeuticagents in intestinal disease distal to the duo-denum is becoming increasingly apparent.With certain drugs, most notably the cytotox-ic agents, intestinal disturbance is an acceptedhazard and the pathological changes havebeen reasonably well documented.' In othercontexts, however, it is often difficult toestablish beyond reasonable doubt thatintestinal lesions are caused by drugs. This isespecially true of non-steroidal anti-inflamma-tory drugs (NSAIDs) which have been

MethodsAll the colorectal biopsy specimens from casesof colitis considered on clinical grounds to bepredominantly or exclusively drug inducedwere retrieved from files of the PathologyDepartment, Glasgow Royal Infirmary, cover-ing the years 1981 to 1988. Clinical suspicionwas based on either a clear associationbetween drug administration and the onset ofcolonic symptoms or clinical and pathologicalevidence that the lesions remitted followingwithdrawal of the drug. There were numerous

118

on Decem

ber 6, 2021 by guest. Protected by copyright.

http://jcp.bmj.com

/J C

lin Pathol: first published as 10.1136/jcp.46.2.118 on 1 F

ebruary 1993. Dow

nloaded from

Importance of apoptosis in the histopathology ofdrug related lesions in the large intestine

Table 1 Clinical and pathologicalfeatures observed infive cases ofdrug related colonic disease

HistologicalfeaturesCase Drug implicatedNo Agelsex Clinicalfeatures AC Inflammation Otherfeatures

1 71 M Rectal carcinoma Diarrhoea 5-fluorouracil 106clearly related to chemotherapy

2 51 M Gastric carcinoma Diarrhoea and 5-fluorouracil 8-7 Plasmacytosis in Nuclear atypiarectal bleeding clearly related to Adriamycin lamina propria in epitheliumchemotherapy Mitomycin Margination of Crypt thinning

neutrophils3 40 F Migraine Diarrhoea, abdominal pain Mefenamic 7-8 Lymphocytosis in IEL focally

anorexia, weight loss for 6 months acid lamina propria: increased inSymptoms improved and biopsy specimen also neutrophils cryptsreturned to normal on drug withdrawalNo pathogens isolated

4 64 F Diverticular disease: rectal bleeding Diclofenac 23-9 Plasma cells, lymphocytes IEL focallyNo pathogens isolated Symptoms sodium increased in increased in

improved and biopsy specimen returned (voltarol lamina propria, crypts alsoto normal on withdrawal of drug occasional neutrophils intraepithelial

eosinophilsCrypt abscessformation

5 70 F Chronic osteoarthritis Diarrhoea, Mefenamic 5-7 Plasma cells, lymphocytes, IEL focallywhich improved on withdrawal of drug acid eosinophils increased increased in

No pathogens isolated in lamina propria crypts: alsointraepithelialeosinophils

AC = apoptosis countIEL = intra-epithelial lymphocytes

cases of colitis in which drug effects were sus-pected, but only five cases fulfilled these strictcriteria (table 1).

All the colorectal biopsy reports fromcases of non-specific inflammatory boweldisease for the years 1986 to 1987 were alsoreviewed. These cases were divided into twomain groups. The first consisted of patientsundergoing initial investigation for colonicsymptoms and whose biopsy specimens wereconsistent with those of IBD in an activephase, as revealed by the presence of cryptitis-that is, neutrophil infiltration of the cryptepithelium-in addition to plasmacytic infil-tration in the lamina propria. This was desig-nated the presumptive untreated group andcomprised 32 cases (male female ratio 13:19,mean age 47 years, range 21-78 years). Thesecond consisted of patients with IBD whowere being treated with either salazopyrine (5cases), or steroids (9 cases), or both (6 cases).In all of these cases the diagnosis had beenreasonably well established from previousbiopsy analysis and clinical investigation. Theduration of the disease varied from a few

Figure I Rectal mucosa during treatment with S-fluoroouracil. Apoptosis bodies areclearly seen (haematoxylin and eosin).

weeks to 16 years and several patients hadexperienced multiple relapses. This treatedgroup (20 cases in all; male female ratio 3:13,mean age 39 years, range 11 to 59 years) wasfurther subdivided into those cases in whichthe disease was deemed, on histologicalgrounds, to be in remission by virtue of anabsence of leucocytic infiltration beyond thatobserved normally (9 cases) and those casesshowing persistent active or chronic inflam-matory reaction (11 cases). The latter wereregarded as being in a "partial response".There was no significant difference in themean duration of treatment between thosecases in remission (15-6 months) and thoseshowing a partial response (1 2- 1 months).

For control purposes a random series of 20histologically normal rectal biopsy specimensfrom patients with non-specific colonic symp-toms, such as unexplained diarrhoea or con-stipation, rectal bleeding, abdominal pain,was extracted from the 1986 files. This nor-mal group included seven males and 13females, the mean age being 44 years (range13 to 74 years). A group of biopsy specimensfrom other colonic diseases were alsoassessed; these included melanosis coli (5cases) infective colitis (4 cases), and radiationcolitis (6 cases). The last condition was divid-ed into a chronic type-presenting 18 monthsor more following irradiation (4 cases)-andan acute type-observed 10-14 days followingirradiation in preparation for surgical treat-ment for rectal carcinoma (2 cases).

All the specimens had been fixed in neutralbuffered formalin and processed by standardmethods. In each case three or more 5 pmsections stained with haematoxylin and eosinwere available, the best orientated beingselected for calculation of the apoptosiscount. The procedure for this is simple butrelies on positive identification of the struc-ture known as an apoptotic body within theepithelium lining the lower half of the cryptsof Lieberkiihn. Apoptotic bodies vary inappearance, but for the purposes of this studywere defined as round or ovoid structures

119

on Decem

ber 6, 2021 by guest. Protected by copyright.

http://jcp.bmj.com

/J C

lin Pathol: first published as 10.1136/jcp.46.2.118 on 1 F

ebruary 1993. Dow

nloaded from

120

*..W? .8 ......

AW VP$8* ' t*X

Figure 2 Rectal mucosa following exposure to mefenamic acid Note the increase inintraepithelial lymphocytes and the goblet cell depletion, albeit of minor degree. Anapoptotic body can also be seen (haematoxylin and eosin).

I. ..

. I

a. -'.

apparently bounded by a discrete membraneand enclosing aggregates of pyknotic intense-ly basophilic fragments of nuclear chromatinoften surrounded by a thin mantle of cyto-plasm (fig 1). Isolated fragments of pyknoticbasophilic material are quite often seen inbiopsy specimens but are not included in thecounts: they often indicated, however, thatapoptotic bodies might be present in thevicinity. Almost all biopsy specimens had evi-dence of nuclear fragments (usually withinhistocytes) beneath the surface epithelium.These are particularly prominent in colono-scopic biopsy specimens and are not includedin the study. Another problem encounteredwas the distinction between apoptotic bodiesand intraepithelial neutrophils; this did notusually cause great difficulty but in cases ofdoubt the lesion in question was omittedfrom the count. All the apoptotic bodiesfound within the location designated abovewere counted in each biopsy specimen. Thetotal numbers of crypts were also countedand the number of apoptotic bodies per 100

~* vm ^s4

*~~~~~~~~~~~~1

.. *:

*.. ;.

<.Z..v ....F * ..,.. * .: . . ...... :

t: @.:t:

.s s ...s b.^

- w"_...... .. .**r 4 f ;

14. ,.,zX^4!

*Ke 5W '-1tE

:..*~~~~~~~~~~~~r

Figure 3 Rectal mucosa in colitis associated with diclofenac treatment. There are

numerous apoptotic bodies. Intra-epithelial lymphocytes are also easily found and accountfor 10% of the cells in the epithelial surface (haematoxylin and eosin).

crypts was calculated. This was referred to asthe apoptotic count. The number of cryptscounted varied between 17 and 325. Most ofthe low counts were found in cases ofuntreated IBD in which there was extensivecrypt damage or dissolution. It is possiblethat this may have biased the results, but inuntreated IBD the mean apoptosis count percase was almost identical with that obtainedby aggregating the measurements made in allcases.The results were analysed using the Mann-

Whitney Wilcoxon two sample rank test.

ResultsThese are summarised in tables 1 and 2. Inthe normal group (table 2) the mean apopto-sis count was 0-94 (mean SE 0.7); in onlyfour of the 20 (25%) cases were apoptoticbodies found at all, the highest count record-ed being 7-8.

In the cases of drug induced colitis (table1) the apoptosis count was equal to or abovethe normal range with the exception of onecase of mefenamic acid induced colitis inwhich the apoptosis count was 5x7. The his-tological features of the rectal mucosa inthese cases were variable. As far as cytotoxicagents are concerned the biopsy specimenexposed to 5-fluoro-uracil alone (case 1)showed little deviation from normal despitethe high apoptosis count (106; fig 1). Withthe combination of 5-FU with other agents(case 2) there is plain evidence of epithelialdamage as revealed by etiolation of thecrypts, and flattening of the surface epitheli-um and nuclear atypia. There is also a mildplasmacytosis of the lamina propria with pro-nounced margination of neutrophils accom-panied by oedema and vascular congestion.With the other drugs implicated, all NSAIDs,the inflammatory changes are more pro-nounced. Mefenamic acid (cases 3 and 5)produces a sharp increase in lymphocytes inthe lamina propria and there is also a distinctincrease in intraepithelial lymphocytes(accounting for, respectively, 18% and 10%of the total number of cells in the epitheliallayer). Normally, intraepithelial lymphocytesaccount for less than 10% of cells in theepithelial surface, the mean count being4.7%.I6 Some neutrophils are also seen in theepithelium but there are no crypt abscesses.Even so the crypts show substantial gobletcell depletion (fig 2). With diclofenac(voltarol; case 5) the inflammatory changes inwhich both plasma cells and lymphocytesparticipate are accompanied by more severe

Table 2 Incidence of crypt apoptoses in normal rectalbiopsy specimens and in those from untreated and treatedIBD

N Apoptosis count Percentage ofcases with

Mean Range apoptosis

Normal 20 0 94 (SE 0 5) 0-7 8 20IBD, untreated 32 2-4 (SE 0 7) 0-14-3 56IBD, partial

responsephase 11 13-1 (SE 3 3) 0-34-3 91IBD, remission 9 1-4 (SE 1 1) 0-97 44

Lee

on Decem

ber 6, 2021 by guest. Protected by copyright.

http://jcp.bmj.com

/J C

lin Pathol: first published as 10.1136/jcp.46.2.118 on 1 F

ebruary 1993. Dow

nloaded from

Importance of apoptosis in the histopathology ofdrug related lesions in the large intestine

tion and even crypt dissolution. Again anincrease in intra-epithelial lymphocytesoccurred (fig 3). In cases 4 and 5 eosinophilswere quite prominent and were also foundwithin the crypt epithelium. In none of thecases, however, was there the kind of cryptirregularity seen in IBD.

In untreated IBD (table 2) the mean apop-tosis count was 2-4 (SE 0 658) which wasonly just significantly greater than that in thenormal group (p = 0085). Apoptosis wasobserved in 56% of the cases. There was nosignificant difference between ulcerative coli-tis and Crohn's disease in this regard, themean apoptosis count counts being 2-6 and2-2, respectively. In the "partial response"phase of IBD the mean apoptosis count risesto 13 1 (SE 3 34) which was significantlygreater than that found in untreated cases (p= 0-01 1). The small numbers available madeit impossible to measure exactly the influenceof specific forms of treatment: in generalterms, however, treatment schedules includ-ing salazopyrine gave the highest counts(mean apoptosis count 18-8). With steroidsalone the apoptosis count was 6-2, and in thesole case in which only salazopyrine was giventhe apoptosis count was 29-6. When the tis-sue enters remission, however, the apoptosiscount falls to almost normal levels (1 -4;SEl15) and in over 50% of cases apoptoticbodies could not be found at all. The numberof cases in each of the miscellaneous groupswas too small to produce significant resultsbut apart from acute radiation colitis thesewere not dramatic.

In melanosis coli the apoptosis count in thefive cases examined varies considerably (6-2,2-3, 0, 0, 3 0). This was also the case withinfective colitis; the apoptosis count was zeroin two cases of amoebic colitis and in the soli-tary case of bacillary dysentery. In one case ofcampylobacter colitis, however, an apoptosiscount of 9-5 was recorded. The biopsy speci-mens in the four cases of chronic radiationcolitis were too small to measure with anyconviction: two gave zero counts but theapoptosis count was 5 in each of the twoother cases. In contrast, the apoptosis countin the two cases of acute radiation colitis was20-5 and 14-0, respectively. In both instancesthere was extensive crypt damage associatedwith pronounced nuclear atypia, cytomegaly,and abnormal mitotic activity in the epithelialcells. Eosinophilic infiltration into the cryptepithelium and lamina propria (and some-times deeper) was another notable feature.

DiscussionIn circumstances which might be regarded asphysiological the phenomenon of apoptosis issometimes referred to as programmed celldeath.14 In the colonic mucosa it might beanticipated that under normal conditionsapoptotic bodies would be restricted to thesurface epithelium as epithelial cells reach theend of their natural lifespan; and this indeedwas usually the case. It should also be added,however, that in almost any colorectal biopsy

specimen (especially when taken duringcolonoscopy) nuclear fragments can be foundbeneath the surface epithelium usually withinhistiocytes which may be abundant in thislocation. This phenomenon can be so strik-ing as to resemble granuloma formation forwhich it must not be mistaken. The presenceof apoptosis in the crypts, however, cannot beregarded as truly physiological and shouldpersuade the observer to seek a cause even ifthe biopsy specimen otherwise appears nor-mal.The main purpose of this study was to

explore the possibility that the detection ofapoptotic bodies in the colorectal cryptsmight be of diagnostic relevance particularlywith regard to drug induced colonic distur-bance. It was appreciated from the outset ofcourse that the presence of apoptotic bodieshas been reported in disease states in theapparent absence of drug treatment, AIDSperhaps being the most notable example.'011In the present study it has not only been con-firmed that apoptotic bodies may also befound in untreated IBD" but that they maybe seen sporadically in infective colitis andeven in 20% of apparently normal biopsyspecimens. With some notable exceptions,however, only small numbers of apoptoticbodies can be detected in untreated colorectaldisease and do not seem to appear under nor-mal conditions.

It is against this background that the effectsof drug treatment on the colonic mucosa haveto be assessed. Apoptotic bodies have readilybeen found in all the reasonably well estab-lished cases of drug induced colonic distur-bance described in the study. Only in acuteradiation colitis does the apoptosis countapproach that seen in association with drugtreatment. As expected the highest apoptosiscount was recorded after treatment with 5-fluoro-uracil,12 but apoptotic bodies were pre-sent in substantial numbers in colitisassociated with diclofenac (voltarol) and to alesser extent in association with lesionsinduced by mefenamic acid. Moreover incases of IBD showing only a partial responseto treatment the apoptotic count in thecolonic mucosa becomes significantlyincreased over baseline levels, this being espe-cially noticeable with drug regimens whichinclude salazopyrine. It can therefore beasserted that even if the finding of apoptoticbodies in the colorectal mucosal crypt doesnot have any specific diagnostic meaning, thepresence of apoptotic bodies in substantialnumbers-in more than five out of 100crypts-should always raise the possibility ofdrug effects. This observation may be of valuenot only in the primary investigation of unex-plained colonic inflammation but also in theassessment of biopsy specimens in cases ofputative IBD in which there is a possibilitythat the histological appearance may havebeen modified by drug treatment.The mechanisms by which drugs bring

about apoptotic change are far from clear. Inthe cases of colitis associated with NSAIDsand reported in this study cryptal apoptoses

121

on Decem

ber 6, 2021 by guest. Protected by copyright.

http://jcp.bmj.com

/J C

lin Pathol: first published as 10.1136/jcp.46.2.118 on 1 F

ebruary 1993. Dow

nloaded from

122

were frequently accompanied by other histo-logical abnormalities, most notably anincrease in lymphocytes in the lamina propriaas well as an increase in intraepithelial lym-phocytes mainly in the crypts themselves. Insome instances there was evidence of moreacute crypt damage with incipient or evenfrank crypt abscess formation. This combina-tion of changes is a little unusual in colorectalbiopsy specimens. An increase in intraepithe-lial lymphocytes has only been recorded in afew conditions such as the lymphocytic vari-ant of microscopic colitis,'5-'7 collagenous col-itiS,'6 and coeliac disease following rectalgluten challenge.'8 An increase in intraepithe-lial lymphocytes and has also been observedin some cases of inflammatory bowel dis-ease.'6 The presence of eosinophils as well aslymphocytes in the crypts in some cases ofNSAID related colitis is of interest because itraises the possibility that in these instancesthe drug induced cell damage may well beimmunologically mediated and thus repre-sents a hypersensitivity reaction. Indeed, thechanges were analogous to those observed inlichenoid drug eruptions in the skin. Whethera similar mechanism is operating in acuteradiation colitis, in which eosinophilic infiltra-tion of the crypt epithelium may also be con-spicuous, is more doubtful. With cytotoxicagents such as 5-fluoro-uracil apoptosis ismore likely to be the result of a direct toxiceffect. The appearance of apoptosis duringthe treatment of IBD is not easy to explain byeither of these mechanisms. It may, however,be a "post-hyperplastic" effect'9 related totemporary overproduction of epithelial cellsin the crypts at a time when cell damage isbeing brought under control by drug treat-ment. This would explain the fall in the num-ber of apoptoses which takes place when thedisease goes into remission even though drugtreatment continues.

In summary, it has been shown in thestudy that although the detection of the occa-sional apoptotic body in a colorectal cryptmay not have any specific diagnostic connota-tion, the presence of apoptotic bodies in sub-stantial numbers (in more than five of 100crypts) should always raise the possibility of adrug effect. The mechanisms whereby drugsproduce these effects are poorly understood.

In the case of NSAID associated lesions,however, apoptosis is accompanied by inflam-matory changes and in particular by a focalincrease in intraepithelial lymphocytes in thecrypts and may well be immunologicallymediated.

I am most grateful to Dr Alfred Pauson for his help in the sta-tistical analysis and to Dr AG Howatson for his advice in thepreparation of this paper.

1 Cunningham D, Morgan RJ, Mills PR, et al. Functionaland structural changes of the human proximal smallintestine after cytotoxic therapy. J Clin Pathol 1985: 38:265-70.

2 Madhok R, McKenzie JF, Lee FD, Bruckner FE, TerryTR, Sturrock RD. Small bowel ulceration in patientsreceiving non-steroidal anti-inflammatory drugs forrheumatoid arthritis. QJ Med 1986; 58: 53-8.

3 Lang J, Price AB, Levi AJ, et al. Diaphragm disease:pathology of disease of the small intestine induced byNSAIDS.JClinPathol 1988: 41: 516-26.

4 Hall RI, Petty AH, Cobden I, Lendowin R. Enteritis andcolitis associated with mefenamic acid. Br Med J 1983;287: 1182.

5 Day TK. Intestinal performation association with osmoticslow release indomethacin capsules. Br Med J 1983;287: 1671-2.

6 Isaacs PET, Sladen GE, Filipe I. Mefenamic acidenteropathy. J Clin Pathol 1987: 40: 1221-7.

7 Kerr JFR, Bishop CJ, Searle J. Apoptosis. In: AnthonyPP. MacSween RNM, EDs. Recent advances inpathology. No 12. London: Churchill Livingstone, 1984:L. 1-16.

8 Epstein RJ, McDonald GB, Sale GE, Shulman HM,Thomas ED. The diagnostic accuracy of the rectal biop-sy in acute graft-versus-host disease. A prospectivestudy of thirteen patients. Gastroenterology 1980; 78:764-71.

9 Gallucci BB, Epstein R, Sale GE, Shulman HM,McDonald GB, Thomas ED. The fine structure ofhuman rectal epithelium in acute graft-versus-host dis-ease. AmJ Surg Pathol 1982; 6: 293-305.

10 Kotler DP, Gaetz HP, Lange M, Klein EB, Holt PR.Enteropathy associated with the acquired immunodefi-ciency syndrome. Ann Intern Med 1984; 101: 421-8.

11 Koder DP, Weaver SC, Terzakis JA. Ultrastructural fea-tures of epithelial cell degeneration in rectal crypts ofpatients with AIDS. AmJSurgPathol 1986: 10: 531-8.

12 Floch MH, Hellman L. The effect of 5-fluorouracil onrectal mucosa. Gastroenterology 1965; 48: 430-7.

13 Walker NI, Bennett RE, Axelsen RA. Melanosis coli. Aconsequence of anthroquinone induced apoptosis ofcolonic epithelial cells. AmJ Pathol 1988; 131: 465-76.

14 Wyllie AH. Commentary: What is apoptosis?Histopathology 1986, 10: 995-8.

15 Ong G, Price AB. Microscopic (lymphocytic) colitis: char-acteristic rectal biopsy features. J Pathol 1990; 160:170A.

16 Lazenby AJ, Yardley JH, Giardiello FM, Jessurun J,Bayless TM. Lymphocytic (microscopic) colitis. HumPathol 1989; 20: 18-28.

17 Giardiello FM, Lazenby AJ, Bayless TM, et al.Lymphocytic (microscopic) colitis. Dig Dis Sci 1989; 34:1730-8.

18 Austin LL, Dobbins WO. Studies of the rectal mucosa incoeliac sprue: the intraepithelial lymphocyte. Gut 1988;29: 200-5.

19 Ledda-Columbano GM, Columbano A, Coni P, Faa G,Pani P. Cell deletion by apoptosis during regression ofrenal hyperplasia. AmJ Pathol 1989; 135: 657-62.

Lee

on Decem

ber 6, 2021 by guest. Protected by copyright.

http://jcp.bmj.com

/J C

lin Pathol: first published as 10.1136/jcp.46.2.118 on 1 F

ebruary 1993. Dow

nloaded from