GASTROENTEROLOGY 2001;121:1064 –1072
A Comparison of Endoscopic Ultrasound, Magnetic Resonance Imaging, and Exam Under Anesthesia for Evaluation of Crohn’s Perianal Fistulas DAVID A. SCHWARTZ,* MAURITS J. WIERSEMA,* KIKA M. DUDIAK,‡ J. G. FLETCHER,‡ JONATHAN E. CLAIN,* WILLIAM J. TREMAINE,* ALAN R. ZINSMEISTER,§ IAN D. NORTON,* LISA A. BOARDMAN,* RICHARD M. DEVINE,㛳 BRUCE G. WOLFF,㛳 TONIA M. YOUNG-FADOK,㛳 NANCY N. DIEHL,§ JOHN H. PEMBERTON,㛳 and WILLIAM J. SANDBORN* *The Division of Gastroenterology and Hepatology, the ‡Department of Radiology, the §Section of Biostatistics, and the 㛳Division of Colorectal Surgery, Mayo Clinic, Rochester, Minnesota
Background & Aims: To determine accuracy of endoscopic ultrasound (EUS) and magnetic resonance imaging (MRI) for evaluation of Crohn’s disease perianal fistulas. Methods: Thirty-four patients with suspected Crohn’s disease perianal fistulas were prospectively enrolled in a blinded study comparing EUS, MRI, and examination under anesthesia (EUA). Fistulas were classified according to Parks’ criteria, and a consensus gold standard was determined for each patient. Acceptable accuracy was defined as agreement with the consensus gold standard for >85% of patients. Results: Three patients did not undergo MRI; 1 did not undergo EUS or EUA; and consensus could not be reached for 1. Thirtytwo patients had 39 fistulas (20 trans-sphincteric, 5 extra-sphincteric, 6 recto-vaginal, 8 others) and 13 abscesses. The accuracy of all 3 modalities was >85%: EUS 29 of 32 (91%, confidence interval [CI] 75%–98%), MRI 26 of 30 (87%, CI 69%–96%), and EUA 29 of 32 (91%, CI 75%–98%). Accuracy was 100% when any 2 tests were combined. Conclusions: EUS, MRI, and EUA are accurate tests for determining fistula anatomy in patients with perianal Crohn’s disease. The optimal approach may be combining any 2 of the 3 methods.
rohn’s disease is a chronic inflammatory condition of unknown etiology, which may involve the small bowel, the colon, or both. Subgroups of patients may have predominantly fistulizing, stricturing, or inflammatory disease. The lifetime risk for developing a fistula in patients with Crohn’s disease is 20%– 40%.1– 4 Patients with Crohn’s disease perianal fistulas experience significant morbidity and are at risk for incontinence either directly through the fistulas themselves or as a result of damage to the anal sphincters from progressively destructive fistulization or from inadvertent damage during surgical exploration and therapy. Examination under an-
C
esthesia (EUA) by a colorectal surgeon is critical for the accurate diagnosis and treatment of many and perhaps most patients with Crohn’s disease perianal fistulas. Surgical principles to be followed in the care of these patients include adequate incision and drainage of all abscesses, fistulotomy of superficial fistulas, and low trans-sphincteric or inter-sphincteric fistulas in the absence of active rectal disease and in some cases, placement of non-cutting setons in patients with active rectal disease or high fistulas.5 Failure to identify perianal abscesses, and performing fistulotomy in patients with high fistulas, may lead to suboptimal patient outcomes. Diagnostic modalities that could be combined with EUA to improve accuracy or used in place of EUA to diagnose and classify perianal abscesses and fistulas would be of use to clinicians, and could ultimately improve patient outcomes. Imaging of perianal fistulas with fistulography in patients with fistula-in-ano6,7 or with pelvic computed tomography in patients with Crohn’s disease8,9 is not sufficiently accurate to be clinically useful. In contrast, imaging of perianal fistulas with rectal ultrasound probes or endoscopic ultrasound (EUS)8,10 –13 or with pelvic magnetic resonance imaging (MRI)10,14 –16 has been shown to be accurate in patients with fistula-in-ano. This prospective triple-blinded study was designed to assess the accuracy of rectal EUS, pelvic MRI, and surgical EUA for evaluation of Crohn’s disease perianal fistulas. Abbreviations used in this paper: CI, confidence interval; EAS, external anal sphincter; EUA, examination under anesthesia; EUS, endoscopic ultrasound; MRI, magnetic resonance imaging. © 2001 by the American Gastroenterological Association 0016-5085/01/$35.00 doi:10.1053/gast.2001.28676
November 2001
IMAGING OF PERIANAL CROHN’S DISEASE
1065
Figure 1. Fistula Anatomy Recording Form. Modified with permission from Choen S, Burnett S, Bartram CI, Nicholls RJ. Comparison between anal endosonography and digital examination in the evaluation of anal fistulae. Br J Surg 1991;78: 445– 447.13
Methods Selection of Patients The study was performed between July 1, 1999 and September 1, 2000. Eligible patients had Crohn’s disease with suspected perianal fistulas. The following patients were not eligible: those who were pregnant; those with contraindications to pelvic MRI (implanted metal devices); those in need of immediate abdominal surgery for active gastrointestinal bleeding, obstruction, or intra-abdominal abscess; and those with severe anal stenosis that would preclude endoscopy. The Institutional Review Board of the Mayo Clinic approved the study, and all patients gave written informed consent.
Study Design The study was a prospective, triple blind comparison of rectal EUS, pelvic MRI, and EUA. A fistula anatomy recording form indicating the number and location of fistulas, abscesses, and other findings was completed by the physician or surgeon performing each diagnostic modality evaluation (Figure 1). Fistulas were classified according to the Parks’ criteria17 as superficial, or inter-, trans-, supra-, or extra-sphincteric (Figure 2). Fistulas that tracked between the rectum and vagina were defined as “recto-vaginal.” In addition, the presence and location (inter-sphincteric, infralevator, and supralevator) of any horseshoeing were also recorded. The presence and location
1066
SCHWARTZ ET AL.
GASTROENTEROLOGY Vol. 121, No. 5
The examiner completed the fistula anatomy recording form and sealed it in an opaque envelope.
MRI
Figure 2. The Parks’ classification. (A) A superificial fistula tracks below both the internal anal sphincter and EAS complexes. (B) An inter-sphincteric fistula tracks between the internal anal sphincter and the EAS in the inter-sphincteric space. (C) A trans-sphincteric fistula tracks from the inter-sphincteric space through the EAS. (D) A suprasphincteric fistula leaves the inter-sphincteric space over the top of the puborectalis and penetrates the levator muscle before tracking down to the skin. (E) An extra-sphincteric fistula tracks outside of the internal and external anal sphincters and penetrates the levator muscle into the rectum. Modified with permission from Parks AG, Gordon PH, Hardcastle JD. A classification of fistula-in-ano. Br J Surg 1976; 63:1–12.17
(superficial or supralevator) of any abscesses were also documented. Rectal EUS, pelvic MRI, and EUA were all performed within the same week to ensure that no changes in fistula anatomy occurred between the assessment of the various diagnostic modalities.
Rectal EUS The rectal EUS was performed by 1 of 3 gastroenterologists with experience performing EUS including anorectal examination. All 3 endosonographers were blinded to results of all other imaging studies. Rectal EUS examinations were performed after flexible sigmoidoscopy or colonoscopy, and additional sedation with meperidine and/or midazolam was administered if necessary for patient comfort. Patients were examined in the left lateral position using radial scanning echoendoscope (GFUM30; Olympus Corporation, Melville, NY) and a 7.5 MHz biplane probe (Acoustics Imaging, Phoenix, AZ). The biplane probe has a variable focus that is adjusted electronically across the depth of imaging. Fistulas were identified as hyperechoic tracks or beads within a larger hypoechoic tract corresponding to air within an active fistula.
The images were interpreted by 1 of 2 radiologists experienced in reading pelvic MRIs. Both radiologists were blinded to the results of all other imaging studies. The pelvic MRI was performed using a Signa 1.5 T superconducting magnet (GE Medical Systems, Milwaukee, WI) and coupled with a torso phased-array coil. Images of the rectum were obtained using a fat-saturation fast-spin echo pulse sequence in 3 orthogonal imaging planes (axial, coronal, and sagittal). Specific imaging parameters included: a 22-cm field of view, 5-mm slice thickness with 0 mm skip, in addition to other standard parameters (echo time [TE] ⫽ 120 ms, repetition time [TR] ⫽ 4000 ms, echo train length ⫽ 8, 256 ⫻ 192 matrix, and 2 number of excitations [NEX]). Fast spin-echo images were also acquired without fat saturation pulses in the axial plane. These were used to define sphincter anatomy and fistula location. Corresponding 5-mm interleaved T1– weighted spin-echo images were acquired in the axial and coronal planes, before and after gadolinium enhancement (gadoteridol; ProHance, Bracco Diagnostics, Inc., Princeton, NJ), using similar parameters (22 field of view, TE minimum value full, TR ⬍ 700 ms, 256 ⫻ 256 matrix, and 1–2 NEX; fat saturation for the gadolinium-enhanced images). The examiners completed the fistula anatomy recording form and sealed it in an opaque envelope.
Surgical EUA One of 7 experienced colorectal surgeons blinded to the results of the previous rectal EUS and pelvic MRI studies, evaluated the patient and performed an EUA if clinically indicated. Standard probing techniques were used, aided by methylene blue instillation where necessary to assist in identifying internal fistulous openings. After the surgical evaluation was finished (while the patient was still under anesthesia) and the fistula anatomy recording form had been completed, the results of the rectal EUS and pelvic MRI were immediately revealed to the surgeon. At that time, the colorectal surgeon had the opportunity to reassess any discordant results to determine the correct anatomy. If the surgeon had elected not to perform an EUA before being shown the results of the 2 imaging studies, that decision could be altered based on the additional data.
Outcomes and Statistical Analysis After the results of the 3 studies (EUA, EUS, MRI) were recorded for each patient, the findings of all of the studies were reviewed periodically by investigators representing each diagnostic modality to establish a consensus of the “correct anatomy,” which we refer to as the gold standard for each patient. This necessitated a change of opinion of 1 or more of the investigators about the initial interpretation of the findings of 1 or more studies when there was discordance between the 3 studies. More specifically, the investigators representing 1 of
November 2001
the diagnostic modalities had to concede that a mistake had been made in the initial interpretation of the findings with their modality. After the modified readings were used to establish the gold standard, the original interpretation of each study was used for analysis of the accuracy of the tests. Consensus opinion was used as the gold standard to allow for the possibility that rectal EUS and/or pelvic MRI were actually superior to the historic gold standard of EUA. The accuracy of each modality was judged as to its agreement with the results of the consensus gold standard. Before initiation of the study, we defined agreement for a modality with the consensus gold standard as complete agreement for fistula classification; the presence (and location) of any horseshoeing of the fistula; and the presence (and location) of any abscess. The results of each diagnostic modality were categorized as accurate or not accurate for the evaluation of Crohn’s disease perianal fistulas. The primary endpoints for the study were the results of the various diagnostic modalities. The primary analyses were to test whether the accuracy for each modality, i.e., the proportions of patients diagnosed with perianal fistulas using each diagnostic modality whose perianal fistulas were accurately classified based on the consensus diagnosis, was high enough to be clinically useful. Each diagnostic modality was considered to be clinically useful if its accuracy for correctly classifying perianal fistulas was at least 85% compared with the consensus classification. The accuracy (and 95% confidence intervals [CIs]) of EUS, MRI, and EUA for Crohn’s disease perianal fistulas were determined. In addition, the agreement of the 3 diagnostic modalities with each other was estimated using Kappa (a chance adjusted measure of agreement, 1.0 ⫽ perfect agreement). Considering both accurate and inaccurate results together, McNemar’s test was used to test whether the discordance between the 2 results was statistically significant. The results for all pairwise comparisons of agreement between the diagnostic modalities were summarized in tabular form. The accuracy for all pairwise combinations of EUS, MRI, and EUA was also evaluated.
IMAGING OF PERIANAL CROHN’S DISEASE
1067
time only a cursory exam of the patient’s perianal disease was performed. Pelvic MRI and rectal EUS differed in their classification of the patient’s fistula track; thus, a consensus could not be reached, and this patient was not included in the analysis. A consensus gold standard was established for 32 patients. The disposition of all 34 patients is shown in Figure 3. Of the 34 patients, 17 (50%) were male, 13 (38%) were current smokers, and 10 (29%) were former smokers. The median age was 36 years (range, 18 –70 years). The anatomic distribution of the patients’ luminal Crohn’s disease was ileal in 3 (9%), ileocolonic in 15 (44%), and colonic in 13 (38%). In 3 patients (9%), the diagnosis of Crohn’s disease was eventually disproved. During the year before study enrollment, 22 (65%) patients had received immunosuppressive therapy (azathioprine or 6-mercaptopurine), 14 (41%) had received infliximab or the humanized anti–tumor necrosis factor antibody CDP-571, and 19 (56%) had received antibiotics (metronidazole or ciprofloxacin). Eighteen (53%) patients had undergone a previous surgical procedure for their perianal Crohn’s disease (5 had 1 procedure, 6 had 2 procedures, 6 had ⬎2 procedures, 1 had 6 procedures). The diagnostic evaluations with EUS, MRI, and EUA documented a total of 40 fistulas and 13 abscesses in 32 patients (Table 1).
Sample Size For the sample size of 32 patients who could be evaluated, the study had approximately 80% power to detect a departure from the (null) hypothesis value of 85% accuracy for any one of the 3 diagnostic modalities (EUS, MRI, EUA) corresponding to 67% accuracy or less. This is based on a one-sample test for proportions using a one-sided alpha level of 0.05.
Results A total of 34 patients enrolled in the study. One patient withdrew before any diagnostic evaluations being performed and was not included in the analysis. The study co-investigators could not reach a consensus opinion regarding the correct anatomy for another patient who required a proctocolectomy because of the severe and refractory nature of his perianal Crohn’s disease, at which
Figure 3. A graphic representation of patient movement throughout the course of the study.
1068
SCHWARTZ ET AL.
GASTROENTEROLOGY Vol. 121, No. 5
Table 1. Fistula Classification Fistula type Superficial Inter-sphincteric Trans-sphincteric Supra-sphincteric Extra-sphincteric Recto-vaginal Other (vaginal-cutaneous, blind sinus, anal fissure) Totala No fistulaa Abscess Superficial Supralevator Total Horseshoe Inter-sphincteric Infralevator Supralevator Total
Number of patients
Number of fistulas
2 2 15 1 4 5
2 2 20 1 5 6
3 32 4
3 39 0
Number of patients
Number of abscesses
9 2 11
11 2 13
Number of patients
Number of fistulas
2 2 1 5
2 3 1 6
aIndicates that this table includes the number of patients with each type of fistula. Seven patients had more than 1 fistula and are thus counted more than once.
Rectal EUS Rectal EUS correctly classified 29 of 32 (91%; 95% CI 75%–98%) patients and thus misclassified only 3 patients. One patient was reported on rectal EUS to have a low trans-sphincteric fistula because rectal EUS showed partial involvement of the medial aspect of the external anal sphincter (EAS). Neither pelvic MRI nor EUA showed any EAS involvement, and the consensus diagnosis was an inter-sphincteric fistula. One patient was reported to have a recto-vaginal fistula on rectal EUS. This patient had an endorectal advancement flap procedure performed 2 years earlier for a previous rectovaginal fistula. This caused distortion of the patient’s anatomy. Pelvic MRI and EUA classified the fistula as trans-sphincteric because they did not show any connection of the fistula with the vagina. The third patient was reported to have an extra-sphincteric fistula on rectal EUS; both pelvic MRI and EUA classified this fistula as a high trans-sphincteric fistula. This patient had undergone multiple perianal surgeries including several incision and drainage procedures and seton placement that may have led to the formation of scar tissue that was misinterpreted as part of the fistula track. The agreement of rectal EUS with pelvic MRI and EUA is shown in Table 2. A representative image from a rectal EUS
examination of a patient with Crohn’s disease perianal fistulas is shown in Figure 4A and B. The median examination length was 29 minutes (range, 10 –96 minutes). Using rectal EUS to study patients with Crohn’s disease and suspected perianal fistulas is a recent development at our institution. Over the course of the study, the endosonographers became more efficient in their examinations. The median time for the last 13 exams was 20 minutes (range, 18 – 40 minutes) versus a median time of 44 minutes (range, 25–96 minutes) for the first 10 patients studied. None of the examinations had to be terminated because of patient discomfort. No patient had to be excluded because of anal stenosis. Pelvic MRI Pelvic MRI correctly classified 26 of 30 (87%; 95% CI 69%–96%) patients, and thus misclassified a total of 4 patients. Three of the 4 pelvic MRI errors involved superficial lesions. One patient who presented with posterior perianal pain was reported to have a posterior superficial fistula based on the information obtained from pelvic MRI. Rectal EUS and EUA did not reveal any evidence of a fistula. This patient had had a posterior fissure that had healed 1 year before. When co-investigators reviewed the films at the consensus meeting, the track previously called a superficial fistula was reported to be more representative of scar tissue. A second patient’s fistula was misclassified as trans-sphincteric according to the data available from MRI. Rectal EUS and EUA did not find any evidence of sphincter involvement and classified the fistula as superficial. This Table 2. Agreement of EUS, MRI, and EUA Evaluations for (⫹) Result
Comparison
Kappa (95% CI)
P value
EUS & EUA EUS & MRI MRI & EUA
⫺0.103 (⫺0.186 TO ⫺0.021) ⫺0.129 (⫺0.229 TO ⫺0.030) ⫺0.129 (⫺0.229 TO ⫺0.030)
1.000 0.705 0.705
# of discordant pairs 6 7 7
NOTE. Table 2 provides the estimated Kappa (a chance adjusted measure of agreement, 1.0 ⫽ perfect agreement) and 95% confidence intervals for Kappa. The marginal totals are very unbalanced. Thus, even though there is good “overall agreement” (71%– 81%) among the pairs of diagnostic modalities, the chance-adjusted measure of agreement (Kappa) indicates poor (actually negative) agreement between the diagnostic approaches. Specifically, the marginal (univariate) agreement of each diagnostic modality with the consensus ranged from 87%–91%, and thus a high overall agreement just because chance alone was expected. In fact, these were higher than the observed agreement, thus yielding negative Kappa estimates. In addition, the P value from McNemar’s test for the difference between the 2 diagnostic modalities is listed along with the number of discordant pairs.
November 2001
IMAGING OF PERIANAL CROHN’S DISEASE
1069
Figure 4. Example of a study patient illustrating the use of all 3 diagnostic modalities. The patient presented with recurrent perianal pain and drainage. (A ) The radial EUS image shows a trans-sphincteric fistula with horseshoeing. (B) The linear ultrasound image shows a long fistula tract with air within it. (C ) The MRI image illustrates a complex horseshoe fistula. (D) The patient then had an EUA that confirmed a horseshoe fistula.
patient had had a fistulotomy for a perianal fistula 9 months before he presented with new perianal pain and discharge. The misdiagnosis was likely the result of misinterpretation of scar tissue from the previous surgery that extended above the superficial tract as an active fistula. The third patient misclassified by the radiologist interpreting the pelvic MRI presented with a large perianal abscess. All 3 modalities clearly identified the abscess, but the associated fistula was misinterpreted as a superficial fistula on the images available from MRI. Rectal EUS and EUA found the associated fistula tract to be trans-sphincteric in course. At the consensus meet-
ing, after reviewing the images, involvement of the EAS was also appreciated on images from the pelvic MRI. One patient’s fistula was described as supra-sphincteric on pelvic MRI, but the endosonographer and surgeon showed a high extra-sphincteric fistula on rectal EUS and EUA. This patient had undergone multiple perianal surgeries in the year before entry into the study, including a fistulotomy and seton placement that may have led to scar tissue formation that was misinterpreted as part of the fistula track. The agreement of pelvic MRI with rectal EUS and EUA is shown in Table 2. A representative image from a pelvic MRI examination of a
1070
SCHWARTZ ET AL.
patient with Crohn’s perianal fistulas is shown in Figure 4C. Image acquisition took less than 50 minutes. The median image interpretation time was 20 minutes (range, 15– 40 minutes). One patient required sedation using midazolam for mild anxiety. Surgical Evaluation Surgical evaluation correctly classified 29 of 32 (91%; 95% CI 75%–98%) patients. Two of the 3 patients misclassified by surgical evaluation had EUA performed. The 1 patient who did not undergo EUA was reported to have an extra-sphincteric fistula based on probing of the primary fistula tract in the office. Rectal EUS and pelvic MRI showed a complex trans-sphincteric fistula with an associated supralevator horseshoe component to the fistula. This patient continued to have perianal pain and drainage and 2 months later underwent an EUA that showed a complex trans-sphincteric fistula with an associated supralevator horseshoe component as seen on the 2 imaging studies. Technically, this is a failure of the surgical evaluation rather than a failure of EUA. However, to mimic clinical practice as closely as possible, we did not separate patients who underwent EUA from those who did not undergo EUA. In this case, failure to perform an EUA was an error in surgical judgment. The second surgical misclassification occurred in a patient who was classified at the time of EUA as having a superficial fistula. This patient presented with abdominal and perianal pain. During his evaluation, he was found to have an intra-abdominal abscess. Before surgery, it was decided to focus on the abdominal abscess and to save the perianal manifestations for a later date. As a result, only a precursory perianal exam was performed to rule out any occult abscess. Pelvic MRI and rectal EUS showed an extra-sphincteric fistula in this patient. The third surgical misclassification occurred as the result of a missed superficial abscess. This patient had had an active fistula for 8 years. The associated scarring and inflammation may have made palpation of the abscess difficult. The agreement of EUA with rectal EUS and pelvic MRI is shown in Table 2. A representative image from an EUA in a patient with Crohn’s disease perianal fistulas is shown in Figure 4D. In addition to the diagnostic examination, therapeutic surgical procedures were performed in 18 of 32 (56%) patients (incision ⫹ drainage of perianal abscess only in 3 patients, placement of non-cutting setons with or without incision ⫹ drainage in 12 patients, and EUA with other surgical procedures in 3 patients). There were no patients in whom the surgeon initially decided an EUA was not warranted, who then required
GASTROENTEROLOGY Vol. 121, No. 5
an EUA after being shown the results of the imaging studies (aside from the misclassification described in the preceding paragraph). The surgeon made the decision of whether an EUA was warranted based on their office evaluation. If the surgeon decided an EUA was not necessary, the fistula anatomy form was completed by the colorectal surgeon, and the results of the EUS and MRI were revealed. After viewing the results of the imaging studies, the decision not to perform an EUA could have been changed, and for the purpose of the surgical assessment arm of the study, the initial opinion regarding fistula anatomy would have been used. However, for the purpose of determining the “true” or consensus anatomy, the findings at EUA would have been used. These results are analogous to clinical practice in which the colorectal surgeon has the opportunity while the patient is still under anesthesia to reassess any discordant results to determine the correct anatomy. Only one colorectal surgeon (J.H.P.) participated in the periodic meetings in which investigators representing each diagnostic modality met to establish a consensus of the “correct anatomy” (the gold standard result) for each patient. None of the surgical misclassifications involved patients he had evaluated. Combination of Diagnostic Modalities The combination of EUS and MRI correctly classified 30 of 30 (100%; 95% CI 88%–100%) patients. The combination of EUS and EUA correctly classified 32 of 32 (100%; 95% CI 89%–100%) patients. The combination of MRI and EUA correctly classified 30 of 30 (100%; 95% CI 88%–100%) patients. Either of the 2 imaging modalities (EUS, MRI) correctly characterized the 3 patients misclassified during surgical evaluation. Thus, 100% accuracy would have been obtained by combining any 2 of the diagnostic modalities.
Discussion We found that the accuracy of rectal EUS, pelvic MRI, and surgical EUA all exceeded the prespecified threshold of 85% accuracy compared with the consensus gold standard, which we believed to be clinically useful. Although the possibility that these diagnostic modalities are not accurate cannot be excluded (given that the 95% CIs for each modality overlapped 85%), the lack of statistical significance for these trends is more likely to result from sample size issues than from a true lack of accuracy. The 3 diagnostic modalities all had similar accuracy, ranging from 87%–91%, and the differences between the modalities were not significant (the 95% CIs for the modalities overlapped each other). More impor-
November 2001
tantly, none of the modalities in isolation, including surgical evaluation, were ⬎91% accurate. When any 2 of the diagnostic modalities were combined, the accuracy increased to 100%. Both rectal ultrasound and pelvic MRI were well tolerated. The assessment of the patient with perianal Crohn’s disease can be challenging even for experienced colorectal surgeons. The difficulty arises from the inability to directly visualize fistulas and/or abscesses. The surgeon must discern the anatomy through touch. This can be difficult because of the induration and inflammation that is often present in these patients. The 91% accuracy of surgical evaluation in our study is amongst the highest reported in the literature; however, it still resulted in misclassification in 9% of patients, all of whom required additional surgical procedures. We used the consensus opinion of the investigators who performed the diagnostic studies to establish the “correct diagnosis,” which we termed the gold standard for each patient. This strategy is similar to the process used in clinical practice to resolve conflicting findings from different diagnostic tests to arrive at a final diagnosis. For example, mucosal biopsies of the colon may show changes compatible with Crohn’s disease from a region of the colon with a normal colonoscopic appearance, and the clinical diagnosis would be that there was Crohn’s colitis present in the region of the colon in question. In a similar fashion, in this study, the surgeon, ultrasonographer, or the radiologist who interpreted the MRI yielded to the findings of the other studies if the preponderance of the data pointed to a different finding than the original interpretation. This strategy did not constitute a circular argument because the original interpretation of the study was used for determination of the accuracy of the test, not the final interpretation that was used for the gold standard. A recent study from Beets-Tan et al.18 showed that images obtained at pelvic MRI provided important additional structural information when compared with EUA in 21% of the 56 patients who were enrolled in the study.18 For the 15 patients in the Beets-Tan study with Crohn’s disease, pelvic MRI provided important additional information regarding perianal anatomy in 6 patients (40%). Several other studies have also evaluated the use of pelvic MRI in Crohn’s patients with perianal fistulas.19,20,21 In the largest of these studies, Makowiec et al.21 studied 54 patients with suspected perianal Crohn’s disease. He found a total of 90 fistulas and 83 abscesses in these patients. The authors of this study reviewed the proctological, pelvic MRI, and intraoperative findings to determine a consensus gold standard that
IMAGING OF PERIANAL CROHN’S DISEASE
1071
they used as their benchmark. The sensitivity of pelvic MRI for correctly determining fistula anatomy was 82%. Similar to our study, pelvic MRI was shown to have a tendency to miss the short or superficial fistula tracks. These findings confirm our decision to use consensus opinion as the gold standard, and highlight the potential for the routine addition of a second diagnostic modality to the evaluation of patients with perianal Crohn’s disease to improve patient outcomes. There has been only 1 previous prospective study comparing rectal EUS, pelvic MRI, and EUA in patients with Crohn’s perianal fistulas.22 This pilot study by Orsoni et al.22 that included 22 patients and used surgical EUA as the gold standard, found rectal EUS to be the most sensitive modality for imaging Crohn’s disease perianal fistulas. The agreement for fistulas with rectal endosonography and pelvic MRI when compared with the surgical findings was 82% and 50%, respectively. Rectal ultrasound was performed with a 7 MHz linear scanning probe similar to the linear instrument used in our series. The use of both radial and linear scanning instruments in our study may account for our improved accuracy (91%) versus 82% in this study. In our experience, the linear scanning probe is the most useful of the ultrasound instruments for imaging perianal fistulas. However, radial scanning provides complementary information and in rare cases revealed fistulas not apparent with the linear probe. The poor accuracy of pelvic MRI in the Orsoni study (50%) may result from relying on a body coil for imaging instead of adding local coils such as the pelvic phased array coil used in our study. Because the pelvic phased array coil is a receive-only coil, it provides better spatial resolution than is available with the body coil alone, allowing thinner slices and smaller field of views. Because the coil is positioned closer to the anus, the pelvic phased-array coil also results in improved signal to noise ratios. In addition, the use of surgical examination as the benchmark may have led to an underestimation of each modality’s accuracy in the Orsoni study. The results of our study show that the accuracy of the diagnostic evaluation of patients with Crohn’s disease perianal fistulas can be improved from between 87% and 91% up to 100% by combining any 2 of the diagnostic studies. However, in our study, 56% of patients required a surgical therapeutic intervention (incision ⫹ drainage of a perianal abscess and/or placement of draining setons or other surgical procedures). An additional consideration is cost. The Medicare 2001 reimbursement for the 3 diagnostic procedures are: rectal EUS (includes both physician component and facility fee for hospital outpa-
1072
SCHWARTZ ET AL.
tients)—$550; pelvic MRI with and without contrast (hospital outpatients)—$536; and EUA—$1400. Thus, in most instances, surgical evaluation combined with EUS or MRI is the most logical combination. This strategy avoids the situation in which the patient has both rectal EUS and pelvic MRI, and then requires EUA with incision ⫹ drainage and/or placement of a noncutting seton. The costs for the 2 strategies are similar overall because the cost savings associated with EUS ⫹ MRI for patients who do not require surgical intervention are offset by the costs of doing an EUA after EUS ⫹ MRI in the 56% of patients who require a surgical intervention. The choice between EUS and MRI depends on local institutional expertise. In selected patients in whom the suspicion of a perianal lesion is low, combining rectal EUS and pelvic MRI may prevent an unnecessary and more invasive EUA. It is reasonable to assume that increased diagnostic accuracy will improve treatment outcomes in patients with Crohn’s disease perianal fistulas. However, this hypothesis should be prospectively tested in future studies. The potential for rectal EUS or pelvic MRI to predict response to medical therapy, and to guide decisions regarding the need for ongoing biologic therapy with infliximab to maintain fistula closure, etc., should also be further explored. Similarly, studies are needed to determine whether the optimal therapy for draining fistulas is surgical placement of a non-cutting seton, medical therapy (i.e., infliximab), or a combination of both surgical and medical therapy. In conclusion, pelvic MRI, rectal EUS, and EUA are all reasonably accurate ways of classifying perianal Crohn’s disease. The combination of either pelvic MRI or rectal EUS and surgical evaluation may be the optimal approach to patients with Crohn’s disease suspected of having a perianal fistula or abscess. The exact imaging modality used should depend on the local expertise available.
GASTROENTEROLOGY Vol. 121, No. 5
6. 7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17. 18.
19.
20.
References 1. Hellers G, Bergstrand O, Ewerth S, Holmstrom B. Occurrence and outcome after primary treatment of anal fistulae in Crohn’s disease. Gut 1980;21:525–527. 2. Farmer R, Hawk W, Turnbull RJ. Clinical patterns in Crohn’s disease. A statistical study of 615 cases. Gastroenterology 1975;68:627– 635. 3. Rankin GB, Watts HD, Melnyk CS, Kelley ML Jr. National Cooperative Crohn’s Disease Study: extraintestinal manifestations and perianal complications. Gastroenterology 1979;77:914 –920. 4. Schwartz D, Loftus E, Tremaine W, Panaccione R, Sandborn W. The natural history of fistulizing Crohn’s disease: a population based study (abstr). Gastroenterology 2000;118:A337. 5. Schwartz D, Pemberton J, Sandborn W. Diagnosis and treatment
21.
22.
of perianal fistulas in Crohn’s disease: a review. Ann Intern Med 2001;(in press). Kuijpers HC, Schulpen T. Fistulography for fistula-in-ano. Is it useful? Dis Colon Rectum 1985;28:103–104. Weisman RI, Orsay CP, Pearl RK, Abcarian H. The role of fistulography in fistula-in-ano. Report of five cases. Dis Colon Rectum 1991;34:181–184. Schratter-Sehn AU, Lochs H, Vogelsang H, Schurawitzki H, Herold C, Schratter M. Endoscopic ultrasonography versus computed tomography in the differential diagnosis of perianorectal complications in Crohn’s disease [see comments]. Endoscopy 1993;25:582–586. Berliner L, Redmond P, Purow E, Megna D, Sottile V. Computed tomography in Crohn’s disease. Am J Gastroenterol 1982;77: 548 –553. Lunniss PJ, Barker PG, Sultan AH, Armstrong P, Reznek RH, Bartram CI, Cottam KS, Phillips RK. Magnetic resonance imaging of fistula-in-ano [see comments]. Dis Colon Rectum 1994;37: 708 –718. Tio TL, Mulder CJ, Wijers OB, Sars PR, Tytgat GN. Endosonography of peri-anal and peri-colorectal fistula and/or abscess in Crohn’s disease. Gastro Endosc 1990;36:331–336. Hussain SM, Stoker J, Schouten WR, Hop WC, Lameris JS. Fistula in ano: endoanal sonography versus endoanal MR imaging in classification. Radiology 1996;200:475– 481. Choen S, Burnett S, Bartram CI, Nicholls RJ. Comparison between anal endosonography and digital examination in the evaluation of anal fistulae. Br J Surg 1991;78:445– 447. Spencer JA, Chapple K, Wilson D, Ward J, Windsor AC, Ambrose NS. Outcome after surgery for perianal fistula: predictive value of MR imaging. AJR Am J Roentgenol 1998;171:403– 406. Chapple KS, Spencer JA, Windsor AC, Wilson D, Ward J, Ambrose NS. Prognostic value of magnetic resonance imaging in the management of fistula-in-ano. Dis Colon Rectum 2000;43:511–516. Beckingham IJ, Spencer JA, Ward J, Dyke GW, Adams C, Ambrose NS. Prospective evaluation of dynamic contrast enhanced magnetic resonance imaging in the evaluation of fistula in ano. Br J Surg 1996;83:1396 –1398. Parks AG, Gordon PH, Hardcastle JD. A classification of fistulain-ano. Br J Surg 1976;63:1–12. Beets-Tan R, Beets G, van der Hoop A, Kessels AG, Vliegen RF, Baeten CG, van Engelshoven JM. Preoperative MR imaging of anal fistulas: does it really help the surgeon? Radiology 2001; 218:75– 84. Koelbel G, Schmiedl U, Majer MC, Weber P, Jenss H, Kueper K, Hess CF. Diagnosis of fistulae and sinus tracts in patients with Crohn disease: value of MR imaging. AJR Am J Roentgenol 1989; 152:999 –1003. Skalej M, Makowiec F, Weinlich M, Jenss H, Lanaido M, Satrlinger M. Kernspintographie bei perianalem morbus Crohn. Dtsch Med Wochenschr 1993;118:1791–1796. Makowiec F, Laniado M, Jehle E, Claussen C, Starlinger M. Magnetic resonance imaging in perianal Crohn’s disease. Inflamm Bowel Dis 1995;1:256 –265. Orsoni P, Barthet M, Portier F, Panuel M, Desjeux A, Grimaud JC. Prospective comparison of endosonography, magnetic resonance imaging and surgical findings in anorectal fistula and abscess complicating Crohn’s disease [see comments]. Br J Surg 1999;86:360 –364.
Received April 11, 2001. Accepted July 12, 2001. Address requests for reprints to: William J. Sandborn, M.D., Mayo Clinic, 200 First Street SW, Rochester, Minnesota 55905. e-mail:
[email protected]; fax: (507) 266-0335.
