The Importance of Diagnostic Accuracyin Colonic Inflammatory Bowel DiseaseMartin Farmer, F.R.C.S., Robert E. Petras, M.D., Louise E. Hunt, F.R.C.S.(Ed), Janine E. Janosky, Ph.D.,and Susan Galandiuk, M.D.Price Institute of Surgical Research, Department of Surgery, University of Louisville School of Medicine,Louisville, Kentucky; Department of Anatomic Pathology, Cleveland Clinic Foundation, Cleveland, Ohio; andDepartment of Family Medicine and Clinical Epidemiology, School of Medicine, University of Pittsburgh,Pittsburgh, Pennsylvania

OBJECTIVE: Crohn’s disease (CD) and ulcerative colitis(UC) may both affect the colon. However, in approximately10–20% of these cases, it is impossible to distinguish be-tween these two entities either clinically or histologically,and a diagnosis of indeterminate colitis (IC) is made. Cor-rect diagnosis is important because surgical treatment andlong-term prognosis differ for UC and CD. The purpose ofthis study was to determine the extent of interobserveragreement among board-certified pathologists and a special-ist gastrointestinal (GI) pathologist regarding the histologi-cal diagnosis of colonic inflammatory bowel disease (IBD).

METHODS: A total of 24 university medical center patholo-gists from eight institutions evaluated 84 colectomy speci-mens and 35 sets of biopsy specimens from 119 consecutivepatients with colonic IBD. A specialist GI pathologist sub-sequently reviewed all cases without knowledge of clinicaldata and prior diagnosis.

RESULTS: The GI pathologist’s diagnoses differed from theinitial diagnoses in 45% of surgical specimens and 54% ofbiopsy specimens. Of 70 cases initially diagnosed as UC, 30(43%) were changed to CD or IC, whereas 4 of 23 cases(17%) initially diagnosed as CD were changed to UC or IC.The k coefficient for the overall agreement of initial diag-noses with the specialist GI pathologist’s diagnoses was20.01 (p5 0.98).

CONCLUSIONS: There is significant interobserver variationin the histological diagnosis of colonic IBD. This may havea profound effect on clinical patient care and, especially, onthe choice of operation. More accurate diagnostic criteriaare needed to facilitate patient care and to optimize treat-ment outcome. (Am J Gastroenterol 2000;95:3184–3188.© 2000 by Am. Coll. of Gastroenterology)

INTRODUCTION

Accurate histological diagnosis of colonic inflammatorybowel disease (IBD) remains an important clinical problem.Many factors may obscure the correct diagnosis of colonic

IBD, including drug therapy, disease severity, and the ex-perience of clinicians and pathologists involved in patientcare. Most hospitals in the United States do not have des-ignated specialist gastrointestinal (GI) pathologists on theirstaff, and exposure to IBD cases may vary greatly from oneinstitution to another. Depending on their experience withtreating these disorders, gastroenterologists and surgeonsmay rely heavily on histological diagnosis to choose themost appropriate types of therapy for their patients. Wewished to determine the extent of interobserver agreementbetween board-certified, university-affiliated, general pa-thologists and a specialist GI pathologist regarding the his-tological diagnosis of colonic IBD. We believe that thisexperience is broadly representative of a real clinical prob-lem faced by clinicians caring for IBD patients in small- tomid-sized communities. Furthermore, we recognize that dif-ferentiation between these disorders is very difficult and cantax the most skilled pathologists and clinicians.

There is recognized interobserver disagreement in thediagnosis of colonic IBD. Theodossiet al. (1) studied themagnitude of variation among 10 observers with a specialistinterest in GI pathology in examining the same slides of“true” Crohn’s disease (CD) or ulcerative colitis (UC). Therange of agreement between the observers was 65–76%,with the best agreement when discriminating between nor-mal slides and cases of confirmed IBD. Ulcerative colitiswas well diagnosed; however, true CD was often thought tobe UC.

Clinicians and pathologists are unable to distinguish be-tween CD affecting the colon and UC in 10–20% of casesof colonic IBD. This difficulty arises because a single de-fining histological characteristic is not present or absent ineach case of the disease (2). These indistinguishable casesare referred to as indeterminate colitis (IC). The frequencyof this diagnosis varies among pathologists.

MATERIALS AND METHODS

The histology of 119 consecutive patients with colonic IBDtreated in a single surgical practice was independently re-

THE AMERICAN JOURNAL OF GASTROENTEROLOGY Vol. 95, No. 11, 2000© 2000 by Am. Coll. of Gastroenterology ISSN 0002-9270/00/$20.00Published by Elsevier Science Inc. PII S0002-9270(00)01992-4

viewed by one specialist GI pathologist. Patients with grossmacroscopic terminal ileal CD (identified at either surgeryor endoscopy) in addition to colonic disease were excluded.A total of 24 university medical center–affiliated, board-certified general pathologists from eight institutions initiallyreviewed 84 surgical resection specimens and 35 sets ofbiopsy specimens as part of the patients’ routine clinicalcare. All surgical resection specimens were total colectomyspecimens with transection of the terminal ileum just prox-imal to the ileocecal valve. Biopsy specimens were takenduring surveillance colonoscopy. They included biopsiesfrom each anatomic area within the colon and rectum as wellas terminal ileal mucosa in the majority of cases. A special-ist GI pathologist (R.E.P.) then independently reviewed allcases without knowledge of prior histological diagnosis orclinical data. These results were then compared to the initialpathological diagnoses.

Histological Diagnostic Criteria inSurgical Resection SpecimensThe specialist GI pathologist required the following featuresfor the definitive histological diagnosis of UC (3): diffusedisease limited to the large intestine; involvement of therectum; more proximal colonic disease occurring in conti-nuity with an involved rectum (i.e., no gross or histologicalskip lesions); no deep fissural ulcers; no mural sinus tracts;no transmural lymphoid aggregates in an area not deeplyulcerated; and the absence of non-necrotizing granulomas(4).

The presence of nonnecrotizing granulomas or the pres-ence of transmural lymphoid aggregates in an area notdeeply ulcerated were required features for the diagnosis ofCD. In the absence of diagnostic criteria of UC or CD, caseswith ambiguous pathological features were classified as IC(4–8). These cases were often accompanied by fulminantclinical colitis, a setting in which the transmural inflamma-tion and fissural ulcers usually associated with CD can alsobe found in UC. Mucin granulomas were recognized butwere not considered in the diagnosis, as they are nonspecificand can be seen in other forms of colitis.

Diagnostic criteria for CD by the general pathologistincluded the presence of nonnecrotizing granulomas, trans-mural inflammation, and skip lesions. In the absence ofnonnecrotizing granulomas, general pathologists were re-luctant to make the diagnosis of Crohn’s colitis. The generalpathologists followed no standard diagnostic protocol, as weintended to examine current practice rather than institutenew protocols.

Histological Diagnostic Criteria in Biopsy SpecimensThe diagnosis of IBD in biopsy specimens is more difficultfor the pathologist. Diagnostic criteria used by the specialistGI pathologist in evaluating colonic biopsy specimens arelisted below (7). Patients were considered to have UC if thehistology of the biopsy specimens showed the following:diffuse chronic and active colitis; distortion; basal plasma-

cytosis (9, 10) in a regional distribution consistent with UC(e.g., disease limited to the large intestine, proximal colonicinflammatory changes found in continuity with an involvedrectum). Patients were considered as having CD if histologyof the biopsy specimens showed changes of primary inflam-matory disease with intervening histologically normal areas(e.g., histological skip lesion), contained nonnecrotizinggranulomas, or showed the focal active colitis pattern ofinjury associated with biopsy-proven terminal ileal inflam-mation.

Patients were considered as having IC if the histology ofbiopsy specimens showed deep obscuring ulceration,changes of IBD without regional distribution information, orspecimens that showed UC-like colonic changes with aninflamed terminal ileum. As with resection specimens, gen-eral pathologists were reluctant to make a diagnosis of CDin the absence of nonnecrotizing granulomas.

Statistical AnalysisThe overall proportion of agreement of diagnosis was cal-culated and corrected for chance agreement. Results wereanalyzed using ak coefficient to measure agreement be-tween the initial diagnosis and the GI pathologist’s diagno-sis (11–13).

RESULTS

The histology of 119 consecutive patients who clinicallywere believed to have colonic IBD was reviewed. Thisincluded 84 surgical resection specimens (71%) and 35 setsof biopsy specimens (29%). The number of paraffin blocksper surgical specimen varied greatly (range, 2–14). A totalof 70 patients (59%) had an initial diagnosis of UC; 23patients (19%) were diagnosed with CD; and in 26 patients(22%), the initial pathology review failed to confirm adiagnosis of IBD. In all, 92% of examinations were per-formed by three of eight university-affiliated hospitals, andnine of the 24 general pathologists examined eight or morespecimens (mean5 5, range 1–13).

An initial diagnosis of IBD was made in 75 of 84 surgicalspecimens and in 19 of 35 biopsies. Of the remaining 26patients in whom the diagnosis of IBD could not be con-firmed initially, 21 were believed to have IBD on review bythe specialist GI pathologist (Table 1). Nonnecrotizing gran-ulomas were absent in five of 10 newly diagnosed cases ofCD.

Of the 93 patients initially diagnosed with IBD (70 UC,23 CD), the diagnoses changed on review by the GI pathol-ogist in 34 patients (37%; Table 2). Four of 23 cases (17%)initially diagnosed as CD were changed to UC or IC (threepatients and one patient, respectively). A total of 30 (43%)of the 70 cases initially diagnosed as UC were changed toCD or IC (10 and 20 patients, respectively). The 23 newlydiagnosed IC patients consisted of 20 patients previouslydiagnosed as having UC, two patients with nonspecific

3185AJG – November, 2000 Importance of Diagnostic Accuracy in Colonic IBD

inflammation, and one patient initially believed to have CD.No diagnosis of IC was made by the general pathologists.

The initial diagnoses and the specialist GI pathologist’sdiagnoses were compared. Agreement was measured usingthek coefficient. Thek coefficient for the overall agreementbetween initial and GI pathologists’ diagnoses was20.01(p 5 0.98), indicating very poor agreement. There was nosignificant difference in overall agreement among the threemain institutions and the other university-affiliated hospi-tals.

DISCUSSION

Diagnostic disagreement between pathologists is by nomeans rare, nor is it limited to IBD. Many studies havedocumented diagnostic variation in pathology and have re-ported on “ambiguous” areas. Central pathology review forquality assurance in cancer trials has been an acceptedpractice for more than a decade (14). The Pathology Panelof the Dutch Melanoma Working Party (15) reviewed 1069consecutive referral cases. Of 158 lesions classified as in-vasive melanoma by the referring pathologists, 22 wereconsidered “benign” by the expert panel. There were also108 invasive melanomas newly diagnosed by the panel inwhich the melanomas had originally been considered to bebenign lesions, dysplastic nevus, or melanomain situ. In-terobserver differences in the diagnosis of superficial blad-

der cancer have also been documented (16). Abtet al. (17)reported a 9.1% incidence of discrepant diagnoses among777 patients as a result of interinstitutional anatomic pathol-ogy consultation. Our study aims to highlight this phenom-enon in the pathological diagnosis of colonic IBD.

Despite several discriminating features, there is signifi-cant overlap in the pathological features of colonic IBD.This study confirms that disparity exists between patholo-gists when reviewing cases of colonic IBD.

The definitive diagnosis of IBD—a complex process in-volving the patient, clinician, radiologist, and pathologist—often requires sequential investigations over time (18). Drugtherapy may alter histological features as well (19, 20). Onemust remember that histological diagnosis is but one pieceof information used in formulating a clinicopathologicaldiagnosis. In the present study, the specialist GI pathologistwas blinded to all clinical and endoscopic information. Thismay not always be the case in practice. There was greatvariation in the number of paraffin blocks per colectomyspecimen. Two are probably too few, and the cutting ofmore blocks might improve the accuracy of diagnosis. Nineto 12 paraffin blocks should probably be obtained per co-lectomy specimen, with sections from the proximal anddistal margins, as well as two sections for every 10-cmsection of colon. For diagnostic endoscopy, especially incases in which the diagnosis of colonic IBD is unclear, it ishelpful for the pathologist to have at least one biopsy fromeach anatomic area of the colon (rectum, sigmoid, descend-ing colon, transverse colon, both flexures, ascending colon,cecum, and terminal ileum). In the long term, it is not knownwhether the specialist GI pathologist’s histological diag-noses will be any more accurate or useful than the generalpathologist’s diagnoses. Correct diagnosis does, however,have important implications for the patient, clinician, andresearcher.

Treatment options differ with diagnosis. Crohn’s diseasemay recur, and surgery is reserved for failure of medicaltherapy and for disease complications such as obstruction orabscess. Certain medications (e.g., imfliximab) are used totreat CD but not UC (21). Surgery is usually curative withUC but not with CD. An operation such as colectomy andileal pouch–anal anastomosis may, however, have disas-trous consequences when inadvertently performed for CD.In general, this procedure is contraindicated in patients withCD because of increased morbidity (e.g., fistula and ab-scess) (22, 23). We have deliberately excluded patients withgross macroscopic terminal ileal disease from our studybecause, in the case of these patients, it is obvious to thesurgeon at the time of operation that Crohn’s disease ispresent, and ileal pouch–anal anastomosis is not performed.Complications requiring pouch removal can result in theloss of considerable lengths of small bowel. Virtually allreports of ileal pouch–anal anastomosis for presumed UCcontain 2–7% of patients in whom the actual diagnosisultimately proved to be CD (23–28). Although several stud-ies have concluded that IC clinically acts similar to UC, the

Table 1. Comparison of General Pathologist’s and Specialist GIPathologist’s Diagnosis When No Specific Inflammatory BowelDisease Was Initially Diagnosed

General Pathologist’sInitial Diagnosis

(no. patients)GI Pathologist’s Diagnosis

(no. patients)

Diverticulitis (1) Diverticulitis (1)

Benign normal mucosawith no inflammation (6)

No inflammatory boweldisease (4)

Crohn’s disease (2)

Inflammatory bowel diseasenot otherwise specified (2)

Crohn’s disease (1)Ulcerative colitis (1)

Nonspecific inflammation (17) Crohn’s disease (7)Indeterminate colitis (2)Ulcerative colitis (8)

Table 2. Initial Diagnosis of Inflammatory Bowel DiseaseCompared With Review by a Specialist GI Pathologist

General Pathologist’sInitial Diagnosis

(no. patients)GI Pathologist’s Diagnosis

(no. patients)

Crohn’s disease (23) Crohn’s disease (19)Ulcerative colitis (3)Indeterminate colitis (1)

Ulcerative colitis (70) Ulcerative colitis (40)Crohn’s disease (10)Indeterminate colitis (20)

3186 Farmer et al. AJG – Vol. 95, No. 11, 2000

rate of pouch failure after colectomy and ileal pouch–analanastomosis is higher in cases of IC compared to UC (29,30). In a report by McIntyreet al. (30), pouch failure washigher in patients with indeterminate ulcerative colitis(19%) compared to patients with ulcerative colitis (8%). Incontrast, the rate of pouch failure after ileal pouch–analanastomosis for CD is approximately 34% (31). The diag-nosis of IC, therefore, requires a different level and intensityof preoperative counseling regarding possible future se-quelae, compared with that for the patient with “straightfor-ward” UC.

Correct disease diagnosis also has implications for patientfollow-up. There is a greater risk for developing coloncancer in UC after 8–10 yr of disease duration. Beginningat 8 yr, annual surveillance colonoscopies with biopsy areperformed to screen for dysplasia. If this is identified, co-lectomy is recommended. The risk for colon cancer in CDis increased compared to that in the general population.Although some believe that the risk is still much less thanthat seen in UC, recent reports indicate that this risk mayactually be considerably higher (32).

The marked interobserver variation in diagnosis of co-lonic IBD may also have profound implications for research.Many studies using germline DNA from IBD patients aresearching for genetic associations with specific types ofIBD. Misdiagnosis can adversely affect the results of link-age studies and could obscure associations between variousmarkers or susceptibility loci and CD or UC. Improvementis needed in diagnostic criteria and in the use of more exactdefinitions of morphological features that constitute thesecriteria in colonic IBD (1). Scoring systems have beendesigned in an attempt to diagnose cases of IBD and also tofurther differentiate cases of IC (33, 34).

Serological testing has also been proposed in an attemptto facilitate correct diagnosis. Immunoglobulin (Ig) A andIgG anti-saccharomyces cerevisiaeantibodies (ASCA) andperinuclear antineutrophil cytoplasmic antibodies (pANCA)have been proposed as diagnostic tools. The former is morecommon with CD and the latter with UC. In one study, allANCA-positive CD patients had UC-like colonic disease(35). In some studies,.20% of CD patients have testedpositive for ANCA (36). The sensitivity and specificity ofASCA IgA and IgG as a marker for CD have been reportedto be 67% and 92%, respectively. High IgA and IgG ASCAlevels are thought to correlate with small bowel CD (37).

This study is in no way intended as a critique of pathol-ogists; rather, we wish to highlight the extreme diagnosticdifficulties of colonic IBD. Just as a clinician’s exposure toIBD will vary with practice and geographic location, so doesa pathologist’s. There are many smaller to mid-sized com-munities in which both clinicians and pathologists will haveless exposure to IBD. Especially in this scenario, develop-ment of genetic or immunological testing methods to makethe diagnosis of IBD more objective and accurate is neededto facilitate proper patient care and to optimize treatmentoutcome.

This study is supported in part by a gift from Mrs. BarbaraGreenberg.

Reprint requests and correspondence to:Susan Galandiuk,M.D., Department of Surgery, University of Louisville, Louisville,KY 40292.

Received Feb. 11, 2000; accepted May 15, 2000.

REFERENCES

1. Theodossi A, Spiegelhalter DJ, Jass J, et al. Observer variationand discriminatory value of biopsy features in inflammatorybowel disease. Gut 1994;35:961–8.

2. Jones JH, Lennard-Jones JE, Morson BC, et al. Numericaltaxonomy and discriminant analysis applied to non-specificcolitis. Q J Med 1973;42:715–32.

3. Goldman H. Ulcerative colitis and Crohn’s disease. In: MingS, Goldman H, eds. Pathology of the gastrointestinal tract.Philadelphia: Saunders, 1992:643–88.

4. Lee KS, Medline A, Shockey S. Indeterminate colitis in thespectrum of inflammatory bowel disease. Arch Pathol LabMed 1979;103:173–6.

5. Price AB. Overlap in the spectrum of non-specific inflamma-tory bowel disease–“colitis indeterminate.” J Clin Pathol1978;31:567–77.

6. Petras RE, Oakley JR. Intestinal complications of inflamma-tory bowel disease: Pathologic aspects. Semin Colon RectalSurg 1992;3:160–72.

7. Petras RE. Nonneoplastic intestinal diseases. In: Sternberg SS,ed. Diagnostic surgical pathology, 2nd ed. Philadelphia: Lip-pincott-Raven, 1994:1311–70.

8. Glotzer DJ, Gardner RC, Goldman H, et al. Comparativefeatures and course of ulcerative and granulomatous colitis.N Engl J Med 1970;282:582–9.

9. Nostrant TT, Kumar NB, Appelman HD. Histopathology dif-ferentiates acute self-limited colitis from ulcerative colitis.Gastroenterology 1987;92:318–28.

10. Seldenrijk CA, Morson BC, Meuwissen SG, et al. Histopatho-logical evaluation of colonic mucosal biopsy specimens inchronic inflammatory bowel disease: Diagnostic implications.Gut 1991;32:1514–20.

11. Fleiss JL, Spitzer RL, Endicott J, et al. Quantification agree-ment in multiple psychiatric diagnosis. Arch Gen Psychiatry1972;26:168–71.

12. Maclure M, Willett WC. Misinterpretation and misuse of thekappa statistic. Am J Epidemiol 1987;126:161–9.

13. Buckley NA, O’Connell DL, Whyte IM, et al. Interrater agree-ment in the measurement of QRS interval in tricyclic antide-pressant overdose: Implications for monitoring and research.Ann Emerg Med 1996;28:515–9.

14. Gilchrist KW, Harrington DP, Wolf BC, et al. Statistical andempirical evaluation of histopathologic reviews for qualityassurance in the Eastern Cooperative Oncology Group. Cancer1988;62:861–8.

15. Veenhuizen KCW, DeWit PEJ, Mooi WJ, et al. Quality as-sessment by expert opinion in melanoma pathology: Experi-ence of the Pathology Panel of the Dutch Melanoma WorkingParty. J Pathol 1997;182:266–72.

16. Witjes JA, Kiemeney LALM, Schaafsma HE, et al. The in-fluence of review pathology on study outcome of a random-ized multicentre superficial bladder cancer trial. Br J Urol1994;73:172–6.

17. Abt AB, Abt LG, Olt GJ. The effect of interinstitution ana-tomic pathology consultation on patient care. Arch Pathol LabMed 1995;119:514–7.

3187AJG – November, 2000 Importance of Diagnostic Accuracy in Colonic IBD

18. Kleer CG, Appleman HD. Ulcerative colitis: Patterns of in-volvement in rectal biopsies and changes with time. Am J SurgPathol 1998;22:983–9.

19. Bernstein CN, Shanahan F, Anton PA, et al. Patchiness ofmucosal inflammation in treated ulcerative colitis: A prospec-tive study. Gastrointest Endosc 1995;42:232–7.

20. Odze R, Antonioli D, Peppercorn M, et al. Effect of topical5-aminosalicylic acid (5-ASA) therapy on rectal mucosal bi-opsy morphology in chronic ulcerative colitis. Am J SurgPathol 1993;17:869–75.

21. Sands BE, Bank S, Sninsky CA, et al. Preliminary evaluationof safety and activity of recombinant human interleukin 11 inpatients with active Crohn’s disease. Gastroenterology 1999;117:58–64.

22. Schrock TR. Surgery for Crohn’s colitis. In. Bayless TM, ed.Current management of inflammatory bowel disease.Philadelphia: Decker,1989:290–4.

23. Hyman NH, Fazio VW, Tuckson WB, et al. Consequences ofileal pouch–anal anastomosis for Crohn’s colitis. Dis ColonRectum 1991;34:653–7.

24. Koltun WA, Schoetz DJ Jr, Roberts PL, et al. Indeterminatecolitis predisposes to perianal complications after ileal pouch–anal anastomosis. Dis Colon Rectum 1991;34:857–60.

25. Dozois RR, Kelly KA, Welling DR, et al. Ileal pouch–analanastomosis: Comparison of results in familial adenomatouspolyposis and chronic ulcerative colitis. Ann Surg 1989;210:268–73.

26. Galandiuk S, Scott NA, Dozois RR, et al. Ileal pouch–analanastomosis: Reoperation for pouch-related complications.Ann Surg 1990;212:446–54.

27. Deutsch AA, McLeod RS, Cullen J, et al. Results of the

pelvic-pouch procedure in patients with Crohn’s disease. DisColon Rectum 1991;34:475–7.

28. Poppen B, Svenberg T, Bark T, et al. Colectomy-proctomu-cosectomy with S-pouch. Operative procedures, complica-tions, and functional outcome in 69 consecutive patients. DisColon Rectum 1992;35:40–7.

29. Pezim ME, Pemberton JH, Beart RW Jr, et al. Outcome ofindeterminate colitis following ileal pouch-anal anastomosis.Dis Colon Rectum 1989;32:653–8.

30. McIntyre PB, Pemberton JH, Wolf BG, et al. Indeterminatecolitis. Long-term outcome in patients after ileal pouch-analanastomosis. Dis Colon Rectum 1995;38:51–4.

31. Phillips RK. Ileal pouch–anal anastomosis for Crohn’s dis-ease. Gut 1998;43:303–4.

32. Gillen CD, Walmsley RS, Prior P, et al. Ulcerative colitis andCrohn’s disease: A comparison of the colorectal cancer risk inextensive colitis. Gut 1994;35:1590–2.

33. Lennard-Jones JE, Lockhart-Mummery HE, et al. Clinical andpathological differentiation of Crohn’s disease and proctoco-litis. Gastroenterology 1968;54:1162–70.

34. Clamp SE, Myren J, Bouchier IA, et al. Diagnosis of inflam-matory bowel disease: An international scoring system. BrMed J 1982;284:91–5.

35. Ruemmele FM, Targan SR, Levy G, et al. Diagnostic accuracyof serological assays in pediatric inflammatory bowel disease.Gastroenterology 1998;115:822–9.

36. MacDermott RP. Lack of current clinical value of serologicaltesting in the evaluation of patients with IBD. Inflamm BowelDis 1999;5:64–5.

37. Targan SR. The utility of ANCA and ASCA in inflammatorybowel disease. Inflamm Bowel Dis 1999;5:61–3.

3188 Farmer et al. AJG – Vol. 95, No. 11, 2000