You are here
Magnetic resonance imaging and clinical assessments for perianal Crohn's disease: Gain and limits
Digestive and Liver Disease, Volume 46, Issue 12, December 2014, Pages 1072–1076
Assessment of perianal Crohn's disease remains challenging. European Crohn's and Colitis Organisation (ECCO) recommend magnetic resonance imaging (MRI) as a gold standard, but both accuracy and advantages remain limited compared to systematic clinical assessment. The aim of this study was to define their actual diagnostic value.
We performed a retrospective analysis of a prospective database of consecutive patients with perianal Crohn's disease assessed by magnetic resonance imaging and clinical examination from 2006 to 2012. At each outpatient visit, perianal activity (Perianal Disease Activity Index) and perianal phenotype (Cardiff-Hughes classification) were noted. MRI was interpreted according to Cardiff-Hughes and Van Assche classifications.
Overall, 122 combined evaluations were assessed in 70 patients. Radiological imaging failed to show superficial ulcerations in 20/21 patients (95%) and severe ulcerations in 13/15 patients (87%). It consistently failed to diagnose anal stenosis (n = 21, 100%). For fistulising lesions, the global agreement between the two methods was 71/122 (58%) in assessing complex fistulas. Clinical assessment underestimated 44/68 (65%) of multiple or ramified fistula tracts. Clinical examination failed to diagnose half of the radiological abscesses.
Current ECCO guidelines should be applied with some caution because of the low sensitivity of MRI for the diagnosis of non-fistulising perianal disease. Combining clinical and MRI assessments improves diagnostic accuracy.
Keywords: Diagnostic accuracy, Pelvic magnetic resonance imaging, Perianal Crohn's disease, Perianal examination, Sensitivity.
Perianal lesions were not mentioned as a feature of Crohn's disease (CD) in the initial disease description and were only recognised as such 30 years later  . However, perianal CD (PCD) is one of the most unfortunate complications of CD and leads to anal pain, discharge and incontinence, in addition to emotional symptoms and impaired social functioning  . Moreover, PCD independently denotes a more aggressive and disabling CD phenotype  . Data suggested that findings of rectal inflammation or stenosis had prognostic implications and were relevant in determining the treatment strategy. According to European guidelines, use of the Montreal classification to categorise the CD phenotype is advocated to better define and treat CD  . In this classification, PCD is summarised with a “p”, whereas prognosis and significance may vary according to the type of perianal lesions. Cardiff-Hughes classification allows more precise description of PCD  with ulcerations, fistulas and strictures.
Recent European Crohn's and Colitis Organisation (ECCO) guideline recommendations state: “Magnetic resonance imaging (MRI) is the most accurate diagnostic imaging test for perianal CD with accuracy surpassing examination under anaesthesia, and is recommended during the initial diagnosis unless there is a need for immediate drainage of sepsis”  .
Interestingly, MRI accuracy in diagnosing fistulising PCD allows almost perfect anatomical classification of fistulising lesions, but the predictive value of clinical examination, including the determination of the presence of perianal indurations for the diagnosis of inflammation of such lesions, remains unknown. Conversely, MRI diagnosis of non-fistulising lesions is unknown and underreported, while non-evidence based experience favours clinical examination. Finally, the accuracy of a clinical assessment in an out-patient visit is underreported in the literature as compared to the evaluation under anaesthesia of surgical series. This study aimed, therefore, to define the diagnostic value of both methods.
2.1. Study population
Between January 2006 and April 2012, a tertiary referral centre database recorded prospectively the main events of consecutive patients with an established diagnosis of PCD based on clinical, biological, radiological, endoscopic and/or histological evidence. The following data were prospectively recorded in a secure database: sex, age at diagnosis, height, weight, smoking habits, luminal CD phenotype according to the Montreal classification  at diagnosis, treatment (including steroids, 5-aminosalicylates [ASA], immunosuppressants, tumour necrosis factor [TNF] antagonists) and past surgical history. At each visit, CD activity was assessed using the Harvey-Bradshaw Index (HBI)  , anatomical classification of PCD was described according to the Cardiff-Hughes classification (including a digital examination)  , and PCD activity was evaluated with the Perianal Disease Activity Index (PDAI)  . Rectal inflammation was assessed by endoscopy. According to the HBI, CD was considered in clinical remission with an HBI score below 4, mild to moderately active with an HBI score ranging from 4 to 12 and severely active with an HBI score above 12.
Inclusion criteria were the diagnosis of fistulising PCD assessed by MRI. All patients with fistulising PCD were extracted from the database and compared with the radiology records of all patients assessed by pelvic MRI. The clinical assessment immediately before MRI evaluation within 1 month was taken into account for analyses to obtain a blind assessment.
2.2. MRI data
Pelvic MRI was performed in patients suffering from severe LAP where abscesses or complex fistulas were suspected and/or to assess the effectiveness of surgical drainage. The pelvic MRI was performed using 2 different MRI scanners: one manufactured by Achieva (3.0 tesla; Philips Medical Systems) and one manufactured by Verio (3.0 tesla; Siemens Medical Systems). MRI used surface coil Torso 16 elements on the Achieva instrument and cardiac 32 elements on the Verio instrument. The acquisition protocol used high-resolution T2-weighted fat-saturation fast-spin echo pulse sequences and T1-weighted fat-saturation fast-spin echo or three dimensional gradient echo sequences before and after intra-venous injection of gadolinium (Dotarem®, Guerbet). T2 sequences were acquired in 3 orthogonal planes depending on the anal canal axis. T1 sequences were acquired at least in the anal canal axis. An anti-spasmodic was administered to all patients before acquisition. All records were reviewed by one experienced radiologist (TR) blinded to the previous interpretation form, the clinical assessment and the history of CD. MRI data were collected according to Van Assche classification  , using anatomic descriptions of fistulising lesions and the intensity of the T2 signal. The Cardiff-Hughes classification was used to describe the 3 elementary lesions of PCD: fistula, stricture and ulceration.
2.3. Statistical analysis
Quantitative variables were described as the median and interquartile range [IQR 25–75]. Categorical variables were presented as counts and percentage of the cohort. Concordance rates between clinical and MRI assessments were quantified. The sensitivity and the specificity were determined according to a reference standard: clinical for ulcerations and stenosis and MRI for fistulising lesions and inflammation  . Likelihood ratios (LR), global concordance and Youden index (J) offered a global assessment of diagnostic performance. These performance tests were calculated using the following equations: (1) Positive LR, [Sensitivity/(1 − Specificity)]; (2) negative LR, [(1 − Sensitivity)/Specificity]; (3) global concordance, [(true positives + true negatives)/N]; (4) Youden index J, [(sensitivity + specificity) − 1]. The LR is the probability that a test result would be expected in a patient with the target disorder compared to the probability that the same result would be expected in a patient without the target disorder. A likelihood ratio greater than 1 indicates a test result associated with the disease, and a likelihood ratio greater than 10 indicates a powerful association. The Youden index is a summary measure of the characteristics of a test and assesses its global accuracy, including its specificity and sensitivity  . Statistical analyses were performed using JMP®Pro 10.0.0 software.
2.4. Ethical considerations
This study was approved by the Hospital Ethics Committee of Rennes and the “Commission Nationale Informatique et Liberté” (CNIL n°1412467). Patients were informed that their blinded data were being collected for scientific or teaching purposes (University Hospital) and provided consent for this use.
3.1. Study population
Seventy CD patients with anal fistulas and/or abscesses were identified from the prospective database, because they underwent at least one and some patients up to 5 concomitant MRIs and clinical assessments. Overall, 122 events of both clinical examination and pelvic MRI were assessed. Clinical examination was performed without general anaesthesia in 105 (85%) patients. The characteristics of the population are summarised in Table 1 .
|N (%) or median [IQR 25–75]|
|Total||N = 70|
|Male gender||26 (37%)|
|Age (years)||31.5 [24–40.3]|
|Former smoker||8 (14%)|
|Duration of CD (months)||78 [33–139.5]|
|Duration of perianal episode at baseline (months)||3 [0–7]|
|Previous medical treatment|
|Anti-TNF therapy||42 (60%)|
|Previous surgical treatment|
|Drainage of suppuration||52 (74%)|
|Intestinal resection||15 (21%)|
|Definitive stoma||1 (1%)|
|Luminal disease – Montreal classification|
|A1/A2/A3||1 (1%)/58 (83%)/11 (16%)|
|L0/L1/L2/L3||2 (3%)/4 (6%)/27 (39%)/37 (53%)|
|B1/B2/B3||53 (76%)/11 (16%)/3 (4%)|
|Luminal disease – activity|
|Rectal involvement||43 (65%)|
|Perianal disease – Cardiff-Hughes classification|
|Perianal disease – PDAI score||8.5 [5–12.25]|
CD, Crohn's disease; PDAI, Perianal Disease Activity Index; IQR InterQuartile Range; TNF, Tumour Necrosis Factor.
Table 2 summarises the occurrence of anal ulcerations as assessed by clinical examination or by MRI. MRI failed to show superficial ulcerations (U1) in 20 of 21 patients and severe ulcerations (U2) in 13 of 15 patients.
|Ulceration as assessed by MRI and clinical examination according to Cardiff-Hughes classification|
|Fistula as assessed by MRI and clinical examination according to Cardiff-Hughes classification|
|Fistula as assessed by T2 signal on MRI and clinical induration|
|MRI assessment||No T2 a||6||1||0||7|
|Moderate or intense T2 a||50||22||0||72|
|No tract||One tract||Multiple tracts||Total|
|Fistula as assessed by MRI and clinical examination according to number of tracts identified|
|MRI assessment||No tract||2||8||3||13|
a T2 hyper-signal.
Abbreviations: U, ulceration F, fistula; MRI, magnetic resonance imaging.
MRI consistently failed to diagnose both reversible stenosis (S1,n = 12) and severe irreversible anorectal stricture (S2,n = 9) as assessed by clinical evaluation.
3.4. Fistulas and abscesses
Table 2 summarises anal fistula and abscess assessments according to clinical evaluation and pelvic MRI. Notably, examination under anaesthesia (n = 18) was more accurate and missed only 6% of fistulas described on pelvic MRI. The global agreement between the assessments was 60% for assessing complex fistulas. Complex fistulas diagnosed by MRI were underestimated by the clinical examination in 17% of the patients. Moreover, clinical evaluation overestimated the presence of complex fistulas in 76% of the patients.
Clinical indurations and abscesses were compared to inflammation and abscesses as judged by T2 enhancement. Abscesses on MRI were noted in 25% of patients with no clinical induration or abscess. By contrast, clinical induration or abscesses were associated with inflammation or abscesses on MRI in 98% of cases. Examination under anaesthesia did not improve clinical accuracy, with only one third of abscesses described in both assessments.
Clinical evaluation underestimated 59% of multiple or ramified tracts ( Table 3 ). Clinically multiple or ramified tracts were confirmed by MRI in 67% of patients. Examination under anaesthesia did not improved clinical accuracy. There were agreements between MRI and examination under anaesthesia in 6/18 (33%) according to the number of the fistula tract, 11/18 (61%) according to the inflamed (induration) area, 12/18 (67%) according to the occurrence of an abscess. However, 6/9 (67%) MRI abscesses were missed by the clinical assessment under anaesthesia.
|Items||Test||G-S||Se||Sp||Youden index b||Pos LR||Neg LR||C|
a Cardiff-Hughes classification.
b Youden index J = Se + Sp − 1.
MRI, magnetic resonance imaging; G-S, gold standard; Se, sensitivity; Sp, specificity; Pos LR, positive likelihood ratio [Se/(1 − Sp)]; Neg LR, negative likelihood ratio [(1 − Se)/Sp]; C, concordance [(VP + VN)/total]
3.5. Overall test accuracy
Global concordance was acceptable for both assessments, but the accuracy of MRI was insufficient in evaluating non-fistulising lesions, and clinical assessment was not accurate in defining fistulising lesions ( Table 3 ).
The present study demonstrates the low sensitivity of MRI for the diagnosis of non-fistulising PCD (stenosis and ulcerations). Clinical examination for fistulising PCD remained consistent with high sensitivity in diagnosing fistulas but it had a low sensitivity for abscesses and inflammation.
This prospective recorded database has several strong points. Data were assessed using recommended classifications and validated scales, , and . Both fistulising and non-fistulising PCD were carefully recorded with equal interest. All MRIs were systematically reviewed by an investigator who was blinded to CD history and clinical data. The correlation between examination under anaesthesia and MRI for fistulas is well described, and we aimed to assess the value of the perianal examination in daily practice.
Conversely, there are limitations in the present study. The study design was a retrospective analysis of a prospective database. MRI was not performed for all patients with CD and anal fistulas: patients with severe or recurrent suppurations were preferentially enrolled in MRI evaluation. This may induce a bias, especially in the sensitivity of the clinical assessment of fistulas. The aim and design were defineda posteriori, which resulted in a lack of power. Moreover, the presence of fistulas was a confounder as both part of the inclusion criteria and a study endpoint. Only a few patients were examined under anaesthesia (n = 18), so it is not possible to define its accuracy compared to pelvic MRI.
The question of choice of the gold standard was a crucial point. ECCO guidelines  recommend the use of MRI as the most accurate tool for diagnosing PCD. These guidelines are based on studies that demonstrated the accuracy of MRI for fistulas  and are in line with our results. The sensitivity of the Cardiff-Hughes classification for fistulas remained very good (0.94), while the sensitivity for abscesses and indurations in active PCD is lower compared to T2 hyper enhancement (0.14 and 0.4, respectively). Notably, cutaneous opening of the fistula tract and mucous discharge as described by the PDAI is usually combined with induration. On the other hand, clinical evaluation overestimated the presence of a complex or high fistula. This may be explained by the level of induration by digital palpation. We could speculate that inflammation is pronounced in the Crohn's fistula thus leading to an overestimate the extension of the infection where MRI may differentiate inflammation from fistula tract. However, the better positive LR (2.86) was found for the presence of MRI inflammation where clinical induration was reported. The good specificity for clinical diagnosis of abscess led to advising urgent surgical drainage without MRI in this case. By contrast, MRI should systematically complement the absence of induration or abscess. Clinical and MRI concordance for assessment of the number of tracts remained weak (0.45), but we could not draw conclusions about the value of EUA (n = 18).
There is no reference to assess non-fistulising PCD in these guidelines  . To our knowledge, no study has yet analysed MRI performance for ulceration and stenosis. Our results suggest the weak accuracy of MRI for diagnosis of ulceration and stenosis (0.02 for ulceration and 0 for stenosis). Moreover, 5 patients (4%) with ulceration on MRI were reclassified as fistula by perianal exam. Schwartz et al.  had similar observations for one of their patients who presented with posterior perianal pain and was reported to have a posterior superficial fistula based on pelvic 1-tesla MRI that was not shown by rectal ultrasound and clinical assessment. This patient had a posterior ulceration that had healed 1 year before. Schwartz et al.  previously concluded that a combination of 2 out of 3 tools (EUA, anal ultrasound, MRI) provided the optimal approach to assessing PCD. A total of 34 patients were enrolled in their study, and each of the three methods exceeded the pre-specified threshold of 85% accuracy. This was improved from between 87% and 91% up to 100% by combining any 2 of the diagnostic studies. We reinforced these conclusions by restriction of the modalities to MRI and perianal examination for an overall (with or without anaesthesia) assessment of PCD.
In conclusion, ECCO guidelines for the assessment of PCD should be applied with some caution. Although this study confirmed the clear superiority of MRI for diagnosing fistulas, this imaging modality is unable to diagnose non-fistulising lesions in PCD. Thus, both evaluations are required in the assessment of PCD. Finally, examination under anaesthesia did not improve clinical accuracy, thus limiting its diagnosis value in the MRI era.
Conflict of interest
-  B. Crohn, L. Ginzburg, G. Oppenheimer. Regional ileitis: a pathologic and clinical entity. Journal of the American Medical Association. 1932;99:1323-1329 Crossref
-  S. Mahadev, J.M. Young, W. Selby, et al. Quality of life in perianal Crohn's disease: what do patients consider important?. Diseases of the Colon and Rectum. 2011;54:579-585 Crossref
-  L. Beaugerie, P. Seksik, I. Nion-Larmurier, et al. Predictors of Crohn's disease. Gastroenterology. 2006;130:650-656 Crossref
-  J. Satsangi, M.S. Silverberg, S. Vermeire, et al. The Montreal classification of inflammatory bowel disease: controversies, consensus, and implications. Gut. 2006;55:749-753
-  L.E. Hughes. Clinical classification of perianal Crohn's disease. Diseases of the Colon and Rectum. 1992;35:928-932 Crossref
-  J. Panes, Y. Bouhnik, W. Reinisch, et al. Imaging techniques for assessment of inflammatory bowel disease: joint ECCO and ESGAR evidence-based consensus guidelines. Journal of Crohn's and Colitis. 2013;7:556-585 Crossref
-  R.F. Harvey, J.M. Bradshaw. A simple index of Crohn's-disease activity. Lancet. 1980;1:514 Crossref
-  E.J. Irvine. Usual therapy improves perianal Crohn's disease as measured by a new disease activity index. McMaster IBD Study Group. Journal of Clinical Gastroenterology. 1995;20:27-32
-  G. Van Assche, D. Vanbeckevoort, D. Bielen, et al. Magnetic resonance imaging of the effects of infliximab on perianal fistulizing Crohn's disease. American Journal of Gastroenterology. 2003;98:332-339 Crossref
-  J. Yin, L. Tian. Optimal linear combinations of multiple diagnostic biomarkers based on Youden index. Statistics in Medicine. 2013;:6
-  J.M. Enríquez-Navascúes, J. Leal, E. Tobaruela, et al. The value of Hughes-Cardiff classification in the management of perianal Crohn's disease. Revista Española de Enfermedades Digestivas. 1997;89:583-590
-  K. Horsthuis, M.L.W. Ziech, S. Bipat, et al. Evaluation of an MRI-based score of disease activity in perianal fistulizing Crohn's disease. Clinical Imaging. 2011;35:360-365 Crossref
-  G.R. Lichtenstein, S.B. Hanauer, W.J. Sandborn, et al. Management of Crohn's disease in adults. American Journal of Gastroenterology. 2009;104:465-483 Crossref
-  D.A. Schwartz, M.J. Wiersema, K.M. Dudiak, et al. A comparison of endoscopic ultrasound, magnetic resonance imaging, and exam under anesthesia for evaluation of Crohn's perianal fistulas. Gastroenterology. 2001;121:1064-1072 Crossref
a Department of Hepato-Gastroenterology, University Hospital of Rennes, Pontchaillou, France
b INSERM U991, University of Rennes 1, Rennes, France
c Department of Radiology, University Hospital of Rennes, Pontchaillou, France
© 2014 Editrice Gastroenterologica Italiana S.r.l., Published by Elsevier B.V.