Correspondence to: Professor M. A. Kamm, St. Mark's Hospital, Watford Road., Harrow HA1 3UJ, UK. E-mail: firstname.lastname@example.org
Aims : To assess fistula track healing after infliximab treatment using magnetic resonance imaging.
Methods : Magnetic resonance imaging and clinical evaluation were performed before and after three infliximab infusions given over a 6-week period. Magnetic resonance images were evaluated for abscesses and fistula tracks. Paired magnetic resonance image examinations were rated ‘better’, ‘unchanged’ or ‘worse’. Magnetic resonance imaging and clinical outcomes were then compared.
Results : Of the 12 referred patients, pre-treatment magnetic resonance imaging detected abscesses in three (two not treated). Of the 10 treated patients, seven had peri-anal fistulas, two of whom also had recto-vaginal fistulas, and three had abdominal wall entero-cutaneous fistulas. After infliximab, four were in remission, one had a response and five were non-responders. One developed a peri-anal abscess. Magnetic resonance imaging improved in six, was unchanged in two and was worse in two. In four of the six with improvement in magnetic resonance imaging, the fistula track resolved, but two of these had clinically persistent entero-cutaneous fistulas. The clinical outcome and magnetic resonance imaging correlated in seven of the 10 patients; in three (two entero-cutaneous and one peri-anal), there was discordance.
Conclusions : Magnetic resonance imaging identifies clinically silent sepsis. Fistulas may persist despite clinical remission. Clinical response to infliximab and clinical correlation with magnetic resonance imaging were poor in patients with abdominal entero-cutaneous fistulas.
Crohn's disease is complicated by the development of fistulas in 20–40% of patients.1–4 The most common site is the perineum (peri-anal and recto-vaginal), although abdominal entero-cutaneous, entero-enteric, entero-vesical and other internal fistulas also occur. Treatment has traditionally been surgical. Fistula healing often requires multiple procedures over an extended period of time, indicating that relapse after treatment is common.5 The majority of simple fistulas that do not cross the external sphincter heal after drainage, fistulotomy or seton insertion. More complex fistulas (see Table 1)6 often require more aggressive surgery, combining the above approaches with bowel de-functioning or resection. Proctectomy is sometimes required to achieve healing. Despite an aggressive surgical approach, including de-functioning, resection or proctectomy, one-third of complex fistulas remain open.5
Table 1. Magnetic resonance imaging classification of fistulas
Superficial, linear or inter-sphincteric
Superficial, linear or inter-sphincteric with abscess or secondary tract
Trans-sphincteric with abscess or secondary tract
Supra-levator, trans-levator or extra-sphincteric
Treatment with the anti-tumour necrosis factor-α antibody, infliximab, has recently been shown to result in rapid fistula closure in two-thirds of treated patients. However, peri-anal abscess formation occurs in 10% of patients with treatment, suggesting the possible closure of the cutaneous opening prior to track healing.7 The effect of infliximab on healing of the fistula track is poorly understood, with preliminary studies suggesting that fistula tracks persist despite healing of the external orifice.8,9 An understanding of what happens to the fistula tract with infliximab treatment may help to identify those in whom recurrence is likely.
Both endo-anal ultrasound and magnetic resonance imaging (MRI) have been shown to be accurate in demonstrating fistula tracks in the peri-anal region.10 Magnetic resonance imaging has been shown to predict the long-term outcome of peri-anal fistulas better than surgical evaluation.11–13 The major advantages of MRI over endo-anal ultrasound are the multiplanar capacity, which allows examination in surgically relevant planes, and the high inherent contrast resolution, which highlights tracks and abscesses. Furthermore, MRI does not suffer the field-of-view limitations afflicting endo-anal ultrasound and affords excellent views of the supra-levator, retro-rectal and antero-anal spaces, where occult sepsis can be missed clinically due to extensive scarring or a remote location.6
This study aimed to identify those patients with undrained sepsis who may be at particular risk of abscess formation. It also aimed to assess the apparent healing of fistula tracks on MRI after treatment with anti-tumour necrosis factor-α antibody, and to correlate such healing with the short-term clinical outcome.
Patients with fistulas related to Crohn's disease were referred for consideration of inclusion in a study of infliximab treatment. Patients with Crohn's disease and draining fistulas (peri-anal, recto-vaginal or entero-cutaneous fistula to the abdominal wall), who were on stable medication and did not have clinically apparent active sepsis (abscess) or a stricture, were eligible.
Patients with a seton in situ for a peri-anal fistula were eligible for this study. Patients had their seton removed if the fistula appeared to be healing after infliximab treatment, two weeks after the first infliximab dose.
All patients underwent MRI and clinical evaluation in the 2 weeks prior to treatment. All eligible patients received three doses of infliximab, 5 mg/kg, at weeks 0, 2 and 6 of the trial. At week 14, patients were assessed clinically and by repeat MRI scan.
Demographic data were collected for each patient, together with the following information: date of onset of Crohn's disease, date of onset of fistula, site of each fistula, the presence of active luminal disease and previous surgical and/or medical treatments for the fistula (Table 2).
Drainage × 4 Partial fistulotomy Setons Azathioprine 7 years
Small and large bowel
Drainage × 5 Partial fistulotomy Seton last 2 years Azathioprine 1 year
For the purposes of this study, a fistula was defined as closed if drainage had ceased on history and no drainage could be demonstrated on gentle compression of the fistula site. Clinical response was defined as closure of at least 50% of draining fistulas at two consecutive time points 4 weeks apart, whilst remission was defined as closure of all draining fistulas also at these time points.
MRI was performed according to a standardized fistula protocol.14 The usual contraindications to MRI applied (implanted metal devices, severe claustrophobia, weight greater than 150 kg). MRI was performed using a 1.0-T static magnet (Gyroscan T10-NT, Philips Medical Systems, London, UK). Patients were scanned supine using a phased-array surface coil. The long axis of the anal canal was identified using a midline sagittal-localizing scan, and then axial and coronal short tau inversion recovery sequences were planned with respect to the anal canal axis using the following scan parameters: TR/TE, 1500/15; field of view, 375 mm; 256 × 256 matrix; slice thickness, 4 mm; interslice gap, 1 mm; four excitations. Sagittal T2-weighted images were sometimes employed at the discretion of the supervising radiologist. Scans were extended into the upper abdomen where a fistula or abscess at this site was a possibility. Images were assessed for the presence of sepsis or fistula.14
The pre-treatment MRI scan was examined primarily to determine whether there were any clinically occult undrained collections that might predispose patients to the development of a clinical abscess after infliximab treatment. Patients in whom a collection was seen on MRI were referred for examination under anaesthetic prior to reconsideration for trial entry. Patients with abdominal entero-cutaneous fistulas and underlying intra-abdominal collections requiring surgery or radiological drainage were deemed unsuitable for the trial. Peri-anal fistulas were classified as simple (superficial or inter-sphincteric) or complex (trans-sphincteric, supra-levator or extra-sphincteric) according to the MRI appearance (Table 1).6
Magnetic resonance imaging was repeated at week 14, 8 weeks after the last of the three infliximab infusions.
Pre- and post-treatment MRI scans were assessed by a single radiologist blind to the date of the examination in relation to treatment and the clinical outcome. The number and complexity of fistula tracks, their anatomical classification and the presence of a cavity or abscess were defined. The two examinations were then compared to determine whether there was a change in the fistula track, the volume of inflammation and the presence of any collections. Paired examinations were rated as ‘better’, ‘unchanged’ or ‘worse’. Improvement was defined as a disappearance or reduction in the number of fistula tracks and/or draining cavities. Deterioration was defined as the development of new tracks or collections, or an increase in the size or number of cavities or fistula tracks.
Twelve patients with fistulas due to Crohn's disease were referred for consideration of inclusion. Pre-treatment MRI detected an abscess in three patients: two intra-abdominal and one peri-anal. The first two were excluded from infliximab treatment and one was treated after drainage and seton insertion, resulting in 10 treated patients.
The demographic and previous treatment details of the patients are outlined in Table 2. Seven had peri-anal fistulas; two of these patients also had a recto-vaginal fistula. Three patients had abdominal wall fistulas. Seven patients had more than one fistula. Five of the seven patients with peri-anal fistulas had setons in situ in one or more fistulas at the start of treatment. These were removed 2 weeks after the first treatment to allow the fistula orifice to heal.
Results of anti-tumour necrosis factor-α treatment
At week 14, four patients were in remission (all fistulas closed), five were non-responders and one showed a response (> 50% of fistulas closed). The latter patient, who had three peri-anal fistulas, all with setons in situ pre-treatment, showed closure of two of the fistulas, but developed a peri-anal abscess at week 10 in the most complex fistula after the seton had been removed. This required drainage and re-insertion of a seton.
All four patients who achieved remission had peri-anal fistulas. Although a reduction in fistula drainage and an improvement in symptoms were seen in all three patients with abdominal wall fistulas, none showed closure of their fistulas (Table 3).
Table 3. Clinical outcomes at week 14
Site of fistula
Response (> 50% fistulas healed) (n = 1)
Remission (all fistulas healed) (n = 4)
Non-responders (n = 5)
Outcomes following treatment with infliximab at weeks 0, 2 and 6. Outcome described by site of fistula. PA, peri-anal; RVF, recto-vaginal fistula.
Peri-anal (n = 7)
4 (1 PA + RVF)
2 (1 PA + RVF)
Abdominal (n = 3)
MRI pre-treatment showed fistula tracks with active inflammation in all patients. The post-treatment MRI scans improved in six patients (three peri-anal, one peri-anal + recto-vaginal and two abdominal wall), with resolution of the fistula track(s) in four of these patients (Figure 1). In the two patients with residual tracks, there was a marked reduction in the track volume and residual inflammation. The MRI scan was unchanged after treatment in two patients (one peri-anal and one abdominal wall) (Figure 2) and was worse in two patients (one peri-anal + recto-vaginal and one peri-anal) (Table 4).
Table 4. Magnetic resonance imaging assessment of response to treatment
Site of fistula
Peri-anal (n = 7)
4 (track resolved in 2)
2 (1 had persistent collection and 1 had new collection)
Abdominal (n = 3)
Recto-vaginal (n = 2)
Correlation between MRI and clinical outcome
The clinical outcome and MRI correlated in seven of the 10 patients (Table 5). The post-treatment MRI scan improved in four patients who achieved clinical remission or response. In two of these, the fistula track was no longer visible. Three patients who did not respond to infliximab had no change or a worse appearance on MRI scan.
Table 5. Correlation between magnetic resonance imaging (MRI) and clinical outcome
Paired MRI comparison
In the remaining three patients (two abdominal wall and one peri-anal), there was discordance between imaging and clinical outcome. Two of these patients with persistent entero-cutaneous fistulas after treatment had MRI improvement with fistula track resolution on imaging. In contrast, MRI revealed a persistent horseshoe trans-sphincteric fistula in a patient in whom the external orifice had closed.
In this small group of patients with complicated, long-standing, fistulating Crohn's disease, unresponsive to multiple medical and surgical treatments, infliximab produced a good short-term clinical outcome in one-half of patients. Three of the five non-responders were patients with abdominal wall fistulas, suggesting that these patients are more difficult to treat. Only a small number of abdominal wall fistulas were included in the original published study of infliximab treatment for fistulating disease, and their results were not reported separately.7 Subgroup analysis of this group in subsequent trials is required to determine the role of infliximab and how it should be used in these patients.
Pre-treatment MRI identified collections related to fistulas in one-quarter of the patients initially referred for the study, two of whom had entero-cutaneous fistulas to the abdominal wall. The third patient, with a past history of recurrent horseshoe abscesses with a single fistula opening, had significant fibrosis, making digital examination unreliable. These results highlight that sepsis may not be identified by clinical examination in patients with complex fistulas, and confirm the findings from previous studies that MRI may reveal an abscess or secondary track which is not apparent on digital rectal examination.6 These results suggest that pre-treatment MRI is helpful to identify those patients with collections who need drainage or alternative therapy, particularly those with abdominal fistulas.
Healing of the external fistula orifice was associated with MRI resolution of the track in two of four patients who showed closure of all fistulas. Whether track healing can predict a prolonged remission or permanent healing is unknown. Fistula tracks persisted on MRI despite clinical remission after infliximab in the remaining two patients. Failure of complete resolution of the fistula track may be due to an inadequate time for this to occur, particularly in our group of patients with severe fistulating disease who had failed all other treatments. Alternatively, the apparent persistence of the fistula track on MRI may reflect the difficulty in distinguishing between healing granulation tissue and persistent unhealing Crohn's fistula; both will return a high signal on the sequences used. The use of intravenous gadolinium contrast agent is also likely to prove unhelpful for this distinction because areas of active sepsis and granulation tissue are both equally likely to enhance.6,11 Lastly, persistent tracks may represent a fistula that will recur despite temporary closure of the external orifice. Longer term follow-up will determine whether persistence of a fistula track predicts clinical recurrence.
Discordance between MRI findings and clinical outcome was seen in three patients: two with an ileostomy and peristomal fistulas, and one with a trans-sphincteric peri-anal fistula with horseshoe granulation tissue at examination under anaesthetic. This suggests that MRI identification of entero-cutaneous tracks in areas of previous surgery associated with scarring is difficult. Further study of larger numbers of patients with abdominal fistulas is required to determine the optimal approach in this group of patients. In patients with substantial peri-anal scarring, it is likely that a combination of MRI and examination under anaesthetic may be required.
In conclusion, MRI is a useful non-invasive method of assessing patients with fistulating Crohn's disease before and after treatment. Complete resolution of fistula tracks following infliximab was seen in one-half of patients who responded. MRI was unable to reliably identify fistula tracks in patients with abdominal entero-cutaneous fistulas or significant surgical scarring. Longer follow-up is required to determine the relationship between track healing on MRI and clinical outcome.