Do standardization and quantification of histopathological criteria improve the diagnosis of inflammatory bowel disease ?

First submission on 25/09/13; last submission on 05/03/14; accepted for publication on 27/03/14; published on 20/06/14 1. Master of Science in Microbiology, Parasitology and Pathology at Universidade Federal do Paraná (UFPR); pathologist at Hospital de Clínicas at UFPR; assistant professor at UFPR-Medical Pathology Department. 2. Doctor in Therapeutic and Internal Medicine at Universidade Federal de São Paulo (UNIFESP); pathologist at Hospital de Clínicas (UFPR); adjunct professor at UFPR-Medical Pathology Department. 3. Doctor in Surgery at UFPR; head of Department of Surgery of UFPR. 4. Master of Medicine by UFPR; assistant professor at UFPR-Internal Medicine Department. 5. Master of Science in Microbiology, Parasitology and Pathology at UFPR; surgeon at Hospital de Clínicas at UFPR. ABStrACt


introDuCtion
Inflammatory bowel disease (IBD) is the generic term for a specific group of chronic inflammatory disorders of unknown origin affecting the gastrointestinal tract, with outbreaks of acute exacerbation (10) .For not presenting specific pathognomonic signs the diagnosis is made by the correlation of clinical symptoms with various signals detected on endoscopic, radiological and histopathological findings, which are add together to a more specific conclusion.Colonoscopy is essential to make possible the diagnosis and evaluate the extent, severity and distribution of IBD.Information obtained at colonoscopy, associated histopathological criteria, providing data for classify IBD in one of its two major subgroups: ulcerative rectocolitis (UC) and Crohn's disease (CD).About 10%-20% of cases are designated as indeterminate IBD, which diagnosis cannot be defined as UC or CD (8) .The specific diagnosis is important to evaluate surgical treatment as well as long-term follow up since the UC and DC medical management is different as regards the evolution of the disease and the indication for surgery.The relative risk of colorectal cancer is highest in patients with IBD than in the general population (5,7) .

oBJECtivES
Choose and apply standardized and quantified histopathological diagnosis method, and compare the results and quality index with the original diagnosis.

MAtEriALS AnD MEthoDS
This study was approved by the Research Ethics Committee at Hospital de Clínicas-Universidade Federal do Paraná (UFPR), under protocol number: 718.137/2003-09, and was recorded in Research Database System at UFPR (BANPESQ) under code number: 2003005726.Retrospective study was performed on IBD of 255 outpatient registers at Hospital de Clínicas at UFPR, in July 2004.There were 130 patients diagnosed with CD and 125 UC.We found 445 histopathological examinations, colonic endoscopic biopsies of these patients in files of Department of Pathology.Among these, only 43 examinations of 37 patients had all the inclusion criteria (some had made more than one exam that met the inclusion criteria).Each test consisted of a series of 5-8 slides, and each slide assembly (relative to one exam) there was 9-39 fragments of bowel mucosa.All these material totaled 264 slides and 846 fragments.For sample selection we established the following criteria: 1) patients with active colitis at the time of biopsy; 2) clinical diagnosis established; 3) clinical segment of at least 12 months; 4) at least five different samples of the colon, one of which necessarily rectum.The material was previously set in aqueous 10% formalin and submitted for processing.The capsules were placed in an automatic tissue processor, which held dehydration, diaphanization, impregnation, and embedment in paraffin.The specimens were included in paraffin for holding samples microtome cuts of 4-5 mm.The material was stained with hematoxylin-eosin (HE) and examined under light microscopy (2,19,22) .For histopathological evaluation, the method developed by Tanaka et al. (25) was chosen.It was reproduced in Table 1 and 2 just like elaborated by the authors.The data in these tables defined criteria H1 to H9, which were marked as present (1) or absent (0).From these values were made the calculations to IBD and CD, which scores were used to classify the cases into the categories described in Table 1 and 2. The original histopathological diagnoses of 37 patients included in this study (43 exams) were classified as UC or CD when these hypotheses were suggested or indicated compatibility with them, and classified as inconclusive when these possibilities were not expected.Source: Tanaka et al. (25) ..CD: Crohn's disease; UC: ulcerative colitis; EG: epithelioid granuloma; NB*: number of biopsies; MMNII: monomorfonuclear inflammatory infiltrate.

rESuLtS
We selected 253 patients − 107 men (42%) and 146 women (58%) − from IBD outpatient, classified into 129 UC (51.0%) and 124 CD (49.0%), which had a total of 445 subjected to histopathological colon biopsies examination, with an average of 1.7 per patient.After surveying the results of 445 biopsies, only 43 (9.6%) met the inclusion criteria for the present study.5.For UC, the test result indicated the non-rejection of the null hypothesis at a significance level of 5% (p = 0.1250).Thus, it can be stated that there is no statistically significant difference between OD and RSD in relation to the percentage of samples with correct diagnoses for UC.For CD, the test result indicated rejection of the null hypothesis at a significance level of 5% (p = 0.0313).Thus, it can be stated that there is a statistically significant difference between OD and RSD in relation to the percentage of samples with correct diagnoses for CD.Table 5 shows that the percentage of correct cases by RSD (92.3%) is higher than the percentage for OD (46.1%).The results of the quality rates for the conclusive diagnoses obtained by OD and RSD are shown in Table 6.

DiSCuSSion
Inconclusive diagnoses such as "chronic inflammation" and "nonspecific inflammation" are of limited value for both the physician and the pathologist (11,28) , and despite the clinicopathological correlation be the goal of histopathology, it is not always possible because the information is not available, or clinical data provided contradict the histopathological findings.In these cases, the pathologist needs to extract as much information as possible from the biopsy.For an appropriate evaluation of biopsy in IBD, the pathologist should be informed about the duration of symptoms and what kind of treatment was given to the patient because the treatment can produce patchiness or discontinuity of mucosal inflammation in UC making a differential diagnosis more difficult (10) .Another study refers that the treatment of UC can lead to partial cure, which results in focal lesions simulating distribution of CD (29) .This fact has already been reported (14) highlighting the focal distribution of the lesions in the UC treated.The treatment can produce a variety of changes including normalization of the mucosa.Some authors emphasize that the information about therapeutics should be given to the pathologist avoid misdiagnosis (17,18,27) .Other authors state that only the epithelioid granulomas without ruptured crypts and or chronic active ileitis are features highly suggestive of CD on mucosal biopsy analysis and that some UC exams may show discontinuous disease, no rectal disease, inflammation in ileum, extracolonic involvement, granulomatous inflammation, aphthous ulcers or mural inflammation.In these cases, the differential diagnose with CD can be almost impossible (30) .In the appropriate clinical presentation the presence of granuloma inflammation in gastrointestinal biopsy specimens confirms the diagnosis of CD (16,32) , and when the granuloma is not found, the diagnoses of CD should be suggested emphasizing the necessity of correlation with others clinical, endoscopic and imaging features (9) .In a study the evaluation of mucosal biopsy in patients with CD demonstrated increased numbers of macrophages and microgranulomas with the help of the technique of immunohistochemistry for CD68+ (31) .Another difficulty is the presence of granulomas related with crypts ruptures.To distinguish between epithelioid granuloma and a granulomatous lesion by rupture of crypt, cuts in multiple tissue levels may be necessary to find the presence of neutrophils, lymphocytes, and foamy macrophages that are absents in the former (30) .
In the present study, the histopathological diagnosis was correct and conclusive in OD at 74.4% of patients, and in 97.7% RSD, demonstrating that the last one increased correlation with the gold standard for diagnosis.This percentage was also higher than those found by several studies (6,15,20,21,27,28) in which the correlation was 94%, 80%, 75%, 72%, 73%, and 73% respectively.Our hypothesis that the standardized histopathological diagnosis increases the correlation with the gold standard in IBD was supported on two other studies (3,13) .There is several studies on determining the histopathological criteria more discriminately, and on how better applying them to the diagnosis of IBD.In some studies, the criteria are repeated, consolidating its importance in the diagnosis of IBD, as occurred, for example, with the alteration in cryptic architecture, which is present in seven reviewed studies (6,13,15,21,23,24,27) .The authors of the study that was used as a model for the standardization review of cases evaluated (24) observed that basal plasmacytosis (Figure 1) associated with severe inflammatory infiltrate is more discriminative than plasmocytosis considered individually.They also observed that the diffuse crypt atrophy (Figure 2B) is characteristic of UC, whereas segmental atrophy is related with CD. Figure 2A shows a comparison between a normal mucosa and a mucosa with crypt atrophy.However, they did not consider cryptic abscesses and cryptitis discriminative for the differential diagnosis between UC and CD, contradicting what other authors have stated (13, 15, 21,    , 28) .Paneth metaplasia (Figure 3) was not analyzed in some studies (13,20,21) , but was analyzed and considered important in six other studies (6,(24)(25)(26)(27)(28) .In the studies reviewed, the histopathological findings were evaluated taking into account different criteria.The study we have chosen as model for standardized review, assessed 70 criteria (the largest number of criteria among the studies reviewed), which were tested and selected by statistical calculations (24) .The series of three studies of these authors (24)(25)(26) consisted of 431, 726, and 60 cases respectively.Together, it is the largest study that used the same criteria for IBD biopsies histopathological evaluation.In the present study, only one case of CD (number 25 in Table 3) was diagnosed as UC by RSD.In this case, we observed that rectal mucosa was compromised, there was diffuse distribution of atrophy and crypt distortion (Figure 4) (criterion H5), segmental distribution of mucin depletion (criterion H6) and preservation of mucin in areas of acute inflammation (criterion H7).A conflict was observed in the histopathological criteria in this case: the diffuse and continuous distribution of cryptic changes from the rectum is characteristic of UC according to some authors observations (23,24) .Another unusual feature found in this CD case was the commitment of the rectum, which is usually spared in CD (12) .The case inconclusive by RSD (number 34 in Table 3) was defined by the gold standard as UC, but it presented: segmental distribution of crypt architectural changes (criterion H5), segmental distribution of depletion of mucus (criterion H6), and focal inflammation, which are histopathological characteristics of CD (23,24) .On this exam request there was no information about the patient being in treatment at the time of biopsies, even though it has endoscopic appearance of UC, but with segmental distribution.In biopsies of children with UC, architectural distortion of crypts occurs in 32.1%, while it happens in 57.9% of adults (29) .The authors correlate this difference with the shorter duration of disease before biopsy in pediatric cases.There is a contradiction between the observations of some studies (23,29) about the time for the appearance of architectural changes promoted by IBD.A study argues (23) that they are precocious and can be detected in biopsies from patients with seven days of onset of disease activity, and moreover the architectural changes are the most important in the diagnosis of IBD histopathology, once it differ in each type of IBD: diffuse for UC, and focal for CD.These authors also note that these changes are not exclusive to IBD as they can be found in intestinal amebiasis and shigellosis cases, depending on the strain and duration of infection (greater than one week).The quality scores obtained by OD and RSD in the present research were not consistent with one study (4) .Here RSD presented specificity, positive predictive value, and accuracy for UC slightly larger than OD.RSD also showed sensitivity, negative predictive value, and accuracy for CD slightly larger than OD.In conclusive cases the specificity for UC and sensitivity to CD (both 92.3%) were higher in RSD than in OD (both 85.7%), agreeing with the results of several studies (1,3,13,(24)(25)(26)(27) that consider the standardization of histopathological criteria a necessary instrument for improving the quality indices of histopathological diagnosis.

ConCLuSion
RSD diagnosis had a higher percentage of correct and conclusive diagnoses than those presented in OD, especially statistically significant difference for CD exams.RSD's conclusive cases obtained slightly higher specificity rates, positive predictive value, and accuracy for UC; slightly higher sensitivity rates, negative predictive value, and accuracy for CD, when compared to OD.

tABLE 2 -
Simple criteria for differentiate CD and UC Source: Tanaka et al.

Table 4 .
To compare OD and RSD results in percentages of assertiveness and conclusiveness of diagnosis, binomial test was applied, since each sample was evaluated by two diagnostic methods (OD and RSD), determining the result classified into two categories.Values of p < 0.05 were considered statistically significant.The binomial test results indicated rejection of the null hypothesis at a significance level of 5% (p = 0.0117), this allows us to affirm that there is a statistically significant difference between OD and RSD in relation to the percentage of exams with inconclusive diagnoses.Table4shows that the percentage of tests considered inconclusive by OD (23.3%) is higher than the well regarded by RSD (2.3%).The assessment of the proportion of cases with correct diagnosis by OD and RSD for each disease (UC and CD) is presented in Table

tABLE 6 -
Quality indices and their estimated 95% interval of confidence obtained from OD and RSD OD: original diagnosis; RSD: reviewed standardized diagnosis.