Acquired Gastrointestinal
Fistulas: Classification,
Etiologies, and Imaging
Evaluation1
Fistulas are abnormal communications between two epithelial-lined surfaces. Gastrointestinal
fistulas encompass all such connections that involve the alimentary
tract, and they can be congenital or acquired in nature. This review focuses on
acquired gastrointestinal fistulas. Development of an acquired gastrointestinal fistula
can greatly affect patient outcome, yet the clinical manifestations are often protean
in nature and the etiology, elusive. Imaging plays an important role in the detection
and management of acquired gastrointestinal fistulas. The more routine use of
cross-sectional imaging (especially computed tomography and magnetic resonance
imaging) has altered the standard sequence of radiologic evaluation for possible
fistulas, but fluoroscopic studies remain a valuable complement, especially for
confirming and defining the anomalous communications. In this review, a classification
scheme for gastrointestinal fistulas is provided, major causes are discussed,
and individual fistula types are elaborated with an emphasis on contemporary
imaging approaches.
Gastrointestinal (GI) fistulas represent abnormal ductlike communications between the
gut and another epithelial-lined surface, such as another organ system, the skin surface, or
elsewhere along the GI tract itself. A GI sinus tract, in comparison, is a similar ductlike
passage that communicates with the gut at one end but ends blindly at the other. The
development of a GI fistula can markedly increase patient morbidity and mortality,
rendering detection of the fistula critical. Imaging often plays a pivotal role in the
diagnosis and management of GI fistula, with fluoroscopic contrast agent–enhanced
studies serving as the traditional standard bearer. The emergence of cross-sectional imaging
techniques, however, has modified the radiologic approach to GI fistulas. Instead of
replacing fluoroscopic contrast-enhanced studies, cross-sectional methods complement
their conventional counterparts in the evaluation of GI fistulas.
In this review, we will provide an organ-system approach to classifying GI fistulas. A
brief discussion of the major causes of acquired GI fistulas will follow. Last, a systematic
review of GI fistulas according to our classification scheme will be provided, with an
emphasis on contemporary imaging evaluation. The relative contribution and effectiveness
of the various imaging modalities will be discussed for individual fistula types,
because many unique features and challenges exist. The salient clinical features of specific
GI fistulas, including management, will also be covered.
CLASSIFICATION OF GI FISTULAS
GI fistulas are generally named according to their participating anatomic components, and
virtually every imaginable combination has been reported in the medical literature. Rather
than recite all possible permutations, a more general approach is presented here (Fig 1).
Because the terminology can be somewhat variable, we have attempted to use fistula
names that prevail in the literature, regardless of underlying etiology. To begin, it is useful
to separate congenital and acquired causes, since their clinical settings and implications
obviously differ greatly. Congenital GI
fistulas are best understood by realizing
their embryologic origin and include
such entities as branchial, tracheoesophageal,
and omphalomesenteric fistulas.
Congenital fistulas, however, are beyond
the scope of this review and will not be
considered further.
Acquired GI fistulas can be categorized
as external or cutaneous if they communicate
with the skin surface or internal if
they connect to another internal organ
system or space, including elsewhere
along the GI tract itself. Internal GI fistulas
can be further divided into two types:
intestinal and extraintestinal. Intestinal
fistulas refer to a gut-to-gut connection
and may consist of any combination of
stomach, small bowel, and colon. An enteroenteric
fistula may refer to any intestinal
fistula in the generic sense, although
some may restrict this term to
small-bowel fistulas only. Extraintestinal
internal fistulas imply communication of
the GI tract with another organ system
such as the genitourinary system, biliary
tree, or respiratory tract. Complex fistulas
contain both internal and external components.
For the purposes of this review,
GI sinus tracts will not be covered in detail,
nor will intentional surgically created
fistulas.
CAUSES OF ACQUIRED GI
FISTULAS
The underlying causes of acquired GI fistulas
are diverse and can include virtually
any process resulting in bowel perforation
from within or bowel penetration
from an extraintestinal process (Fig 2).
The majority of external (cutaneous) fistulas
represent a complication of recent
abdominal surgery (1). The leading causes
of internal fistulas in the industrialized
world are Crohn disease, diverticulitis,
malignancy, or a complication of treatment
of these entities. Not surprisingly,
many cases are the result of multiple contributing
factors; common examples include
cancer patients who have undergone
radiation therapy and patients with
Crohn disease who have undergone prior
bowel surgery. The specific location and
type of fistula can often suggest certain
causes, as will be seen when individual GI
fistulas are covered more in depth. Some
general features of the more common inflammatory
causes will briefly discussed in
the following paragraphs. Most of the remaining
noninflammatory causes listed
in Figure 2 will be covered in more detail
in upcoming sections.
Fistula formation is a hallmark of Crohn
disease, occurring in up to 20%–40% of
patients described in surgical series (2). Sinus
tracts and fistulas often involve the
distal small bowel, and peritoneal abscess
or phlegmon may be an associated finding
(4). The clinical and radiologic manifestations
vary widely because these internal
fistulas can involve nearly any organ system,
but ileocolic and enterovesical fistulas
are the most common types (5). External
fistulas are also common, especially in the
perianal region (6). Fistula formation is
considerably less common in ulcerative colitis,
which, unlike Crohn disease, is not a
transmural process (3). Rectovaginal fistula
is the most frequent spontaneous GI fistula
that develops in ulcerative colitis, followed
by rectovesical fistula (7,8).
Diverticulitis is a common cause of colonic
fistula formation, with the fistula
most often communicating with the urinary
bladder (9). Colovaginal fistulas are
also relatively common in women with
sigmoid colon diverticulitis, particularly
after hysterectomy. Fistulas are seen in
up to 20% of cases of surgically treated
diverticular disease (9,10). Another relatively
common finding in diverticulitis is
a fistulous tract that parallels the colonic
lumen, representing a localized form of
colocolic fistula that has been termed
“double tracking.” Not surprisingly, fistula
formation of the sigmoid colon predominates
in diverticular disease, but
other colonic segments are occasionally
involved.
Other than Crohn disease and diverticulitis,
other less common inflammatory
causes of GI fistulas include atypical infections,
cholecystitis, pancreatitis, and
appendicitis (11,12). Among the various
atypical infectious causes that have been
reported are tuberculosis, histoplasmosis,
actinomycosis, xanthogranulomatous pyelonephritis,
amebiasis, echinococcosis,
and lymphogranuloma venereum (13–19).
OVERVIEW OF IMAGING
TECHNIQUES
Fluoroscopic contrast-enhanced studies
and conventional radiographic studies
have traditionally served as the cornerstone
for imaging of spontaneous GI fistulas.
However, technical advances and
the increased availability of cross-sectional
imaging modalities have challenged
this paradigm. The result has been
a more flexible hybrid approach that utilizes
the strengths of the various complementary
imaging modalities now available.
The preferred imaging approach
will vary according to fistula type and the
specific clinical scenario. Furthermore,
even individual fistula types often elude
generalization and must be treated on a
case-by-case basis. This underscores the
importance of the radiologist in determining
the most appropriate sequence of
imaging studies for a given case. Because
many fistulas may be detected incidentally
on cross-sectional images obtained
because of other indications, familiarity
with the direct and indirect signs of fistulas
is essential for this unsuspected diagnosis.
Despite this wide variability, some
broad comments can be made with regard
to the imaging approach. Once the
selection is made between conventional
and cross-sectional imaging as the initial
study, other technical considerations follow.
Contrast-enhanced fluoroscopic examinations
often remain the initial study
of choice and are generally superior to
endoscopy in demonstrating the presence
and extent of a GI fistula (4). Fistu-
Figure 1. Classification of GI fistulas.
Figure 2. Major causes of acquired GI fistulas.
10 _ Radiology _ July 2002 Pickhardt et al
Radiology
lography is adequate for diagnosis of
most external (cutaneous) fistulas and is
also useful for follow-up in these cases
(20). On occasion, enteric contrast-enhanced
studies, such as a small-bowel
study or enema, will provide as much or
more diagnostic information. For extraintestinal
internal fistulas, one must
decide between pursuing a primary bowel
study and a study that directly opacifies
the communicating organ system, such
as urography, vaginography, cholangiography,
and others. For intestinal (gut-togut)
fistulas, enteric contrast-enhanced
studies are superior and may be the only
noninvasive method able to demonstrate
these fistulas in some cases.
The choice of contrast agent is another
important factor in the performance of
conventional GI studies. A water-soluble
iodinated contrast agent is generally
used, at least initially, for abdominal fistulography
and enteric studies when
frank perforation is suspected or pneumoperitoneum
is present. This is predicated
on the potential for extravasated
barium to incite an inflammatory reaction
in the peritoneum, which can be
followed by the formation of dense fibrous
adhesions (21–23). The risk of clinically
important chemical peritonitis,
however, is minimal unless a relatively
large amount of barium has leaked, especially
with the newer barium preparations.
A similar but more localized and
less severe foreign body reaction can occur
with retroperitoneal and extraperitoneal
barium extravasation (24). Despite
these caveats, it is important to remember
that barium is more sensitive than
aqueous contrast agents for demonstrating
GI fistulas due to the tendency of the
latter to dilute, resulting in lower radiographic
opacity (1). This dilution of water-
soluble contrast agents is especially
limiting for small-bowel examination, and
initial evaluation with barium should be
strongly considered in patients without
pneumoperitoneum, particularly for intestinal
(gut-to-gut) fistulas (22). When a water-
soluble contrast agent is used initially, a
negative study should be followed by a barium
study when the index of suspicion
remains high.
For imaging of most internal GI fistulas
with extraintestinal communication, an
aqueous contrast agent is generally preferred.
This is obviously the case when
the extraenteric component is primarily
opacified, as with urographic and cholangiographic
contrast-enhanced studies. A
water-soluble agent should also be used
when there is a possibility of vascular
communication, due to the potentially
life-threatening complication of barium
embolization (25). An exception to this
rule of using an aqueous contrast agent
for imaging of extraintestinal involvement
involves GI fistulas communicating
with the tracheobronchial tree, where
barium is indicated and generally well
tolerated. A water-soluble agent in this
setting can lead to potentially lethal pulmonary
edema due to their relatively
high osmolarity, although the risk is
lower with nonionic agents (26,27). Furthermore,
unless a large esophageal leak
is suspected, barium remains the contrast
agent of choice for esophagography,
since the risk of clinically important mediastinitis
or granuloma formation appears
to be negligible with small amounts
of barium extravasation (27).
Cross-sectional imaging, particularly
computed tomography (CT), has further
strengthened the radiologist’s armamentarium
for evaluating GI fistulas. CT effectively
complements conventional radiography
with its ability to demonstrate
extraluminal disease, including associated
abscesses, tumor, or other coexisting
processes. Although CT may be less sensitive
for direct detection of some GI fistulas,
there are instances where it may be
more sensitive than conventional studies,
such as with enterovesical fistulas
(4,28,29).
Regardless of whether the fistula is directly
detected at CT, CT often yields
more valuable information overall with
respect to patient care. Furthermore, it is
important to at least consider the need
for obtaining a CT scan prior to performing
a conventional barium examination,
because residual barium can produce
troubling artifacts on CT. Technical advances
such as multi–detector row CT
allow for effective multiplanar reformations
and volume-rendering techniques
to directly display fistulas not oriented in
the traditional transverse plane. Often,
CT directly or indirectly demonstrates
the presence of a GI fistula, elucidates the
underlying cause, and obviates further
imaging. An additional advantage of CT
is its utility in guiding percutaneous
drainage of associated abscesses.
Magnetic resonance (MR) imaging
holds similar promise for the evaluation
of certain forms of possible GI fistulas,
but the application of MR imaging has
been most visible in the evaluation of
enterocutaneous fistulas, especially in
the perianal region (30–32). Faster imaging
sequences and the use of oral contrast
agents may further expand the role of MR
imaging in the future, but CT remains
the primary cross-sectional modality for
fistula evaluation because it is rapid, generally
available, and less costly than MR
imaging. Sonography plays a much more
limited role in the evaluation of most GI
fistulas and typically requires a corroborative
study for confirmation (28,33).
The remainder of this review will focus
on the specific forms of GI fistulas.
Figure 3. Enteroenteric and enterocolic fistulas. (a) Frontal radiograph from barium-enhanced
small-bowel study in a 25-year-old man with Crohn disease shows multiple fistulous tracts extending
from the terminal ileum (arrowheads), converging to a small mesenteric cavity (_), and communicating
with the cecum and more proximal ileum (arrows). (b) Transverse contrast-enhanced CT scan
in a 24-year-old man with Crohn disease shows irregular bowel wall thickening, mesenteric infiltration,
and contrast agent–filled extraluminal tracts (arrows) centered in the ileocecal region. This
complex enterocolic fistula involved distal ileum, cecum, ascending colon, and sigmoid colon.
Volume 224 _ Number 1 Acquired Gastrointestinal Fistulas _ 11
Radiology
INTERNAL GI FISTULAS
Internal GI fistulas include both intestinal
and extraintestinal types. Although
some of these fistulas may be suspected
on clinical grounds, their presence is often
first discovered at imaging, sometimes
quite unexpectedly. Complex GI
fistulas (Fig 1) may consist of virtually
any combination of internal and external
communication, but for the purpose of
this review each component will be considered
separately.
Intestinal Fistulas
Intestinal (gut-to-gut) fistulas may involve
any or all combinations of the
small bowel, colon, and stomach. The
clinical manifestation of this subset may
be subtle, since only the alimentary tract
is involved. Diarrhea, with or without abdominal
pain, is the most common
symptom overall (9). There are several
factors that influence which segments of
bowel are involved in the fistulous communication.
In cases where a primary
bowel abnormality is the underlying
cause, the segment of diseased bowel will
obviously be at highest risk. Proximity to
the pathologic process, be it intestinal or
extraintestinal, is also important. Finally,
a preexisting or preferred pathway between
certain portions of the gut, as with
a connecting ligament or mesentery, explains
the predisposition for some intestinal
fistulas to form (see discussion of
gastrocolic fistula later).
Enteroenteric and enterocolic fistulas
are common complications of Crohn disease,
where fistulas are often multiple
and favor the ileocecal region (Fig 3). Enterocolic
fistulas in Crohn disease are
usually due to primary small-bowel disease,
whereas the opposite is true for colonic
diverticulitis. Overall, coloenteric
fistulas constitute fewer than 10% of fistulas
complicating diverticulitis (9). A
more common form of intestinal fistula
from diverticulitis is the so-called doubletracking
colocolic fistula (Fig 4). This intraloop
form of intestinal fistula results
from localized perforation and paracolic
extension that parallels the bowel lumen.
A similar appearance can be seen with
Crohn disease and perforated adenocarcinoma
of the colon. Intestinal fistulas
can also be seen in cases of other abdominal
malignancies, radiation therapy, surgery,
and foreign bodies (Fig 5) (34–36).
In general, contrast-enhanced fluoroscopic
GI studies remain the most effective
means for help in diagnosing
intestinal fistulas. When the colon is
involved, a contrast agent enema examination
is the study of choice and will
demonstrate the fistulous communication
more often than an upper GI examination,
due to the increased intraluminal
pressure in the latter (Fig 5) (20). A
small-bowel follow-through examination
may be the only noninvasive means for
detecting some enteroenteric fistulas, but
a successful examination requires vigilance
and a high index of suspicion by
the radiologist. Enteroclysis may be more
sensitive for the detection of some enteroenteric
fistulas, but it is a more invasive
procedure that requires small-bowel
intubation. These fistulas tend to be subtle
on cross-sectional images but may be
detected as a serendipitous finding on
occasion (Fig 6b) (37). Although treatment
principles are similar in many respects
to the enterocutaneous fistulas discussed
later in this article, intestinal
fistulas rarely close spontaneously, and
surgical correction is generally required
(38).
Discussion of all types of intestinal fistulas
is not feasible in this review but one
specific type, the gastrocolic fistula, is
worthy of further attention. The gastrocolic
ligament allows for bidirectional
spread of pathologic processes between
the greater curve of the stomach and the
transverse colon. Although carcinomas
of the stomach and colon were once
thought to be the most common cause of
gastrocolic fistula, it now appears that
most cases are due to penetrating benign
gastric ulcers, particularly in the setting
of nonsteroidal antiinflammatory drug,
or NSAID, use (39). Not surprisingly, various
other neoplastic and inflammatory
causes have also been reported in the literature
(37). Unfortunately, the more
suggestive clinical symptoms of feculent
Figure 4. Colocolic (double-tracking) fistula. (a) Frontal radiograph from air-contrast barium
enema examination in a 50-year-old man 1 month after an episode of acute diverticulitis shows
a long-segment narrowing (arrowheads) involving the sigmoid colon. At the distal aspect of the
stricture, a second channel (arrow) parallels the colonic lumen, the so-called double-tracking
sign. Note additional scattered diverticula. (b) Transverse contrast-enhanced CT scan obtained 1
month earlier than a during an acute episode shows pericolonic inflammatory changes and a
small peridiverticular abscess (arrow). Adjacent large diverticulum (arrowhead) may represent the
point of eventual fistula reentry. Perforation with a localized fistula was confirmed at surgery and
pathologic examination.
Figure 5. Enterocolic fistula. Spot radiograph
obtained during air insufflation for air-contrast
barium enema examination in a 58-year-old
man shows unsuspected communication between
sigmoid colon and small bowel (arrowheads).
The patient had undergone successful
surgical removal of an infected abdominal aortic
graft 6 months earlier. Note also faint contrast
agent (arrow) extending along aortic region.
12 _ Radiology _ July 2002 Pickhardt et al
Radiology
vomiting and undigested food in the
stool are less common than nonspecific
findings such as abdominal pain. Contrast-
enhanced enema examination remains
the most reliable means for detection,
but its superiority over a contrastenhanced
upper GI study has likely been
overemphasized (Fig 6a) (39). Although
barium studies will more often demonstrate
the fistula directly and can usually
suggest a benign or malignant cause, CT
is likely more accurate for evaluation of
the gastrocolic region for the presence of
a bulky mass or alternate cause (Fig 6b).
Most gastrocolic fistulas are treated with
en bloc resection. However, gastrocolic
fistulas due to benign gastric ulcer disease
are the exception, because they may
spontaneously resolve after NSAIDs are
withdrawn (39).
Extraintestinal Fistulas
The extraintestinal fistulas constitute a
diverse and intriguing collection of acquired
GI fistulas since they can connect
the gut with virtually any other organ
system. Extraintestinal fistulas involving
the genitourinary, biliary, vascular, and
respiratory systems are considered below.
Genitourinary tract.—Communication
between the GI and genitourinary tracts
represents a major subset of extraintestinal
internal fistulas. The bladder and vagina
are most often affected, but involvement
of the upper collecting system,
urethra, or uterus is occasionally seen.
Available diagnostic modalities for evaluating
these lesions include urographic
studies, contrast-enhanced GI studies,
cross-sectional imaging, and endoscopic
procedures. The most appropriate initial
study varies according to fistula type, and
the management approach continues to
evolve. Both CT and MR imaging have
proved to be useful for noninvasive evaluation
of pelvic fistulas (30,31).
The term enterovesical fistula is often
generally applied for bladder communication
with the colon, small bowel, rectum,
or appendix (28,40). Sigmoid diverticulitis
is the single most common cause
of enterovesical (specifically, colovesical)
fistula (9). Furthermore, fistulas to the
urinary bladder account for over half of
all internal fistulas encountered in diverticular
disease (Fig 7). Crohn disease accounts
for most small-bowel–to-bladder
fistulas and may be present in up to
3%–4% of patients with this disease (Fig
(29). Pelvic malignancy, especially
colorectal adenocarcinoma, is the other
major cause of a GI fistula to the bladder,
followed by radiation- and surgically induced
fistulas. Approximately 20% of all
enterovesical fistulas are rectovesical, and
fewer than 5% are appendicovesical (Fig
9). Specific clinical symptoms (fecaluria
and pneumaturia) are present in 40%–
70% of patients, but nonspecific symptoms
such as cystitis are invariably
present (9,28,40–42).
Cystoscopy usually demonstrates inflammatory
changes in the bladder but is
nondiagnostic for fistula in the majority
of cases (9,28,40,41). Likewise, conventional
contrast-enhanced genitourinary
and GI studies such as cystography and
barium enema examination also yield
false-negative results in most cases, as do
Figure 6. Gastrocolic fistulas. (a) Frontal radiograph from solid-column barium enema examination
in a 57-year-old man shows fistulous communication between the transverse colon and
stomach via a large benign gastric ulcer (_) extending into the gastrocolic ligament. Note smooth
folds radiating from the ulcer crater and absence of a gastric or colonic mass. (b) Contiguous
transverse contrast-enhanced CT scans in a 59-year-old woman with abdominal pain and vomiting
show pericolonic inflammatory changes surrounding a large transverse colonic diverticulum
(arrows) in the gastrocolic region. The process blends imperceptibly with thickened gastric
antrum (arrowheads). (c) Image from contrast-enhanced enema examination in the same patient
as in b shows gastrocolic fistula (arrowhead), which proved to be secondary to diverticulitis at
surgery and pathologic examination.
Volume 224 _ Number 1 Acquired Gastrointestinal Fistulas _ 13
Radiology
sonography and GI endoscopy. CT, however,
has demonstrated 90%–100% sensitivity
for diagnosis (albeit with the use
mainly of indirect signs and not direct
demonstration of the fistula itself) and
has been advocated for initial evaluation
(28,41). Diagnostic CT findings that are
present with most enterovesical fistulas
include air in the bladder lumen (in the
absence of recent catheterization) associated
with focal bladder and/or bowel wall
thickening (Figs 7–9) (28). Conventional
studies are much less sensitive and specific
for the detection of intravesical air.
Although the fistula itself is often not
directly demonstrated at CT (Fig 9), its
location can generally be inferred from
the secondary findings (Figs 7, 8). Furthermore,
if an intravenous contrast
agent is not used, the presence of an enteric
contrast agent in the bladder on CT
scans is diagnostic of a fistula. Alternatively,
if an enteric contrast agent is not
used, the presence of an intravenous contrast
agent in the bowel is also diagnostic.
Sensitivity can be increased further with
direct rectal administration of a contrast
agent or with a CT cystographic technique.
Rescanning after active urination
and defecation may also be useful when a
suspected fistula is not demonstrated on
the initial scan. CT can also provide important
extraluminal information, such
as the presence of an offending tumor
(Fig 8). More recently, MR imaging has
shown similar success in facilitating diagnosis
(30). Treatment of enterovesical fistulas
consists of single-stage surgical resection
in the majority of cases (40,41).
GI fistulas to the kidney or upper urinary
tract are much less common than
bladder fistulas and are more often due to
urologic disease rather than a primary GI
process (17,43,44). The “retroperitonealized”
portions of the colon and duodenum
are most often involved. Communication
of the colon with the pelvicaliceal system
(renocolic fistula) or the ureter (ureterocolic
fistula) typically occurs secondary
to chronic suppurative renal infection in
the setting of urolithiasis and/or obstruction,
as seen with xanthogranulomatous
pyelonephritis (17,43,45). Less common
causes include tuberculosis, trauma, surgery,
radiation therapy, Crohn disease,
diverticulitis, and malignancy (Fig 10) (46–
48). Although fecaluria and pneumaturia
may be present, the clinical picture more
often is nonspecific and related to chronic
or recurrent urinary tract infection (44).
The combination of urography and CT
is a useful approach in the evaluation for
possible GI fistulas involving the kidneys
and upper urinary tract (Fig 10) (43). Urographic
studies are best for direct visualization
of the fistula, and CT can be used
to reliably assess the involved organs and
surrounding tissues for the cause and extent
of disease. It is important to note,
however, that an excretory urogram will
fail to opacify the fistula when the kidney
is nonfunctioning, and a retrograde
study may be nondiagnostic if an obstruction
is present distal to the fistula.
Direct antegrade pyelography can be useful
in these situations (49). Surgical excision
of the fistula, often with nephroureterectomy,
is necessary in most cases.
The acquired rectovaginal fistula is the
most common GI fistula involving the
genital tract in women. Most cases are
related to obstetric complications, inflammatory
bowel disease, or some combination
of gynecologic malignancy (particularly
cervical cancer), surgery, and
radiation therapy (7,50). Although the
clinical symptoms, particularly the passage
of feces through the vagina, usually
indicate the presence of a fistula, its detection
is often difficult on conventional
GI studies unless a relatively large communication
is present (Fig 11). Vaginography
may demonstrate the fistula more
clearly in subtle cases (51). More recently,
CT and MR imaging have been shown to
be useful for detection of rectovaginal
and enterovaginal fistulas, whereas endorectal
sonography appears to be relatively
insensitive (30,33,52). An enteric contrast
agent and/or air within the vagina
can be demonstrated on CT scans in the
majority of cases (52). Simple excision is
often not adequate in these complex
cases, and at least temporary colonic diversion
is usually necessary. GI fistulas
involving the uterine body and fallopian
tubes are rare, compared with vaginal fistulas,
and most often involve the left side
of the colon (53). Most colouterine and
colotubal (salpingocolic) fistulas result
from diverticulitis, but a variety of other
GI and genitourinary causes are possible,
especially in younger women (9,53). Actinomycosis
should be considered in the
setting of an intrauterine device. A combination
of hysterosalpingography and
CT will provide a comprehensive preoperative
assessment in most cases (Fig 12).
Acquired rectourethral fistulas in males
are also rare, with the majority of cases
related to treatment for prostate cancer
(Fig 13) (54).
Biliary tract.—Spontaneous internal biliary
fistulas represent a complication of
cholelithiasis or choledocholithiasis in
over 90% of cases (11,55). Infrequent
causes include peptic ulcer disease, malignancy,
and prior surgery. In most series,
cholecystoduodenal fistulas are the
most common type, followed by cholecystocolic
and choledochoduodenal fistulas
(11). The clinical manifestation of
enterobiliary fistulas is often nonspecific,
and most cases are diagnosed on the basis
of an unsuspected imaging finding (11).
Distal small-bowel obstruction from an
impacted ectopic gallstone, so-called
gallstone ileus, is an unusual complication
of chronic cholecystitis and affects
only a minority of patients with cholecystoduodenal
fistulas. Gallstones that
result in intestinal obstruction typically
exceed 2 cm in diameter (56). Obstruction
at the level of the gastric outlet or
duodenum represents a specific subset of
gallstone ileus that is referred to as Bouveret
syndrome (57). Surgery is indicated
to relieve the obstruction in cases of gall-
Figure 7. Colovesical fistula. Transverse contrast-enhanced CT scans in a 56-year-old-man with
pneumaturia and prior diverticulitis show air (arrowhead) in the bladder and the site of fistulous
communication (arrow) between sigmoid colon and bladder. Note diverticulosis of the sigmoid
colon.
14 _ Radiology _ July 2002 Pickhardt et al
Radiology
stone ileus, and surgical correction is required
for the biliary fistula, to prevent
future complications.
Pneumobilia seen on imaging studies
strongly suggests the presence of an internal
biliary fistula in the absence of prior
sphincterotomy, surgical bypass procedure,
recent endoscopic retrograde cholangiopancreatography,
or passed common
duct stone. The Rigler triad of small-bowel
obstruction, pneumobilia, and ectopic
gallstone(s) is virtually pathognomonic
for gallstone ileus but is present on conventional
radiographs in only 30%–35% of
cases (Fig 14a) (58). This triad of findings,
however, is more readily apparent on CT
scans (Fig 14b) (59). CT can also provide
important information on the degree of
bowel obstruction and suggest the likely
site of fistula formation. Endoscopic retrograde
cholangiopancreatography is a
sensitive technique for direct demonstration
of enterobiliary fistulas, especially
those of the choledochoduodenal type.
Conventional contrast-enhanced GI studies
are somewhat less direct for direct
demonstration but nonetheless are relatively
noninvasive and may help detect
an unsuspected communication with the
biliary tree (Fig 15). Compared with CT,
sonography is less accurate for detection
of cholecystoenteric fistulas, but suggestive
findings include an irregular contracted
gallbladder, nonvisualization of
the gallbladder, and pneumobilia (56).
Vascular system.—Enteric fistulas involving
the vascular system, whether arterial
or venous, are potentially lethal
and often require urgent correction. A
high index of clinical suspicion is necessary,
since a favorable outcome relies on
prompt diagnosis. Imaging studies, particularly
CT and contrast-enhanced GI
studies, play an important role in the preoperative
detection of these fistulas.
The aorta lies in proximity with the GI
tract for much of its thoracic and abdominal
course. Aortoenteric fistulas, therefore,
can potentially involve the gut anywhere
from the esophagus to the colon
(60–62). The majority of cases occur in
the presence of aortic aneurysm disease,
either as a primary event or a secondary
complication following surgical repair
(60). Aortic fistulas involving the duodenum
and esophagus warrant further consideration.
The duodenum participates in the majority
of aortoenteric fistulas, owing to the
proximity between its third portion and
the underlying abdominal aorta. Primary
aortoduodenal fistula is a rare life-threatening
cause of gastrointestinal bleeding that
results most commonly from an atherosclerotic
aortic aneurysm (60,63). Unusual
causes of a primary fistula include aortitis,
radiation therapy, malignancy, and peptic
ulcer disease (64,65). Most patients have
upper or lower GI bleeding, but the classic
triad of abdominal pain, GI bleeding, and
pulsatile mass is present in fewer than
25% of cases (60,66). A “herald bleed”
frequently precedes lethal exsanguination,
and patient survival hinges on
prompt diagnosis and emergent therapeutic
laparotomy. Unfortunately, a cor-
Figure 8. Enterovesical fistula. (a) Contiguous
transverse CT scans obtained with intravenous
and oral contrast agents in a 69-year-old
woman with longstanding Crohn disease
show a heterogeneous soft-tissue mass (M) associated
with thickened ileal loops and adjacent
bladder wall thickening (arrowhead). A
small gas bubble (arrow) is present in the bladder
lumen. (b) Fluoroscopic image shows contrast
agent injection through a communicating
enterocutaneous fistula and demonstrates
the fistula (arrowhead) between the ileal segment
and bladder. Small-bowel adenocarcinoma
complicating Crohn disease was proved at
surgery.
Figure 9. Rectovesical fistula. Transverse contrast-
enhanced CT scan in a 65-year-old-man
with ulcerative colitis shows air in a fistulous
tract (arrow) between inflamed rectum and
bladder. Note also air (arrowheads) in bladder
lumen.
Volume 224 _ Number 1 Acquired Gastrointestinal Fistulas _ 15
Radiology
rect preoperative diagnosis is determined
in only a minority of cases, underscoring
the importance of heightened clinical
suspicion (60). Endoscopy is often the
initial diagnostic study performed, but
blood pooling may impair luminal visibility,
and an alternate presumed source
of bleeding is frequently identified, acting
as a “red herring” (60). Conventional
upper GI study, sonography, aortography,
and tagged red blood cell scintigraphy
all have marked limitations for diagnosis
(67). CT, however, provides rapid
and effective evaluation in hemodynamically
stable patients suspected of having
an aortoenteric fistula. CT findings such
as perianeurysmal hematoma, pseudoaneurysm,
contrast agent extravasation,
periaortic or intraluminal gas, and focal
duodenal wall thickening are highly suggestive
of a fistula in the appropriate clinical
setting (Fig 16a) (63,67,68).
Secondary aortoduodenal fistulas develop
in fewer than 2% of aortic reconstructions
but are still more common
than primary fistulas (60,69). Clinical
suspicion remains the linchpin for diagnosis
of this condition, which must be
considered in any patient with a prosthetic
aortic graft and GI bleeding. As
with primary fistulas, endoscopy and CT
are the most useful diagnostic studies for
initial evaluation in hemodynamically
stable patients (69). However, although
endoscopy may reveal mucosal defects or
even graft eroding into the duodenum, it
is diagnostic in fewer than 25% of cases
(69). CT is a sensitive technique but its
specificity for the diagnosis of fistulas is
relatively low, especially in the early
postoperative period when perigraft fluid
and gas can be a normal finding. The CT
appearance of secondary aortoenteric fistula
overlaps substantially with that of
graft infection, although the presence of
extraintestinal air and associated duodenal
abnormality is less common in the
latter (Fig 16b) (70,71).
Aortoesophageal fistulas are rare; in
the majority of cases, they are caused by
localized rupture of a thoracic aortic aneurysm
(72,73). Unusual reported causes
include esophageal carcinoma, foreign
body ingestion, syphilis, and infected
aortic graft (73–76). The clinical manifestation
is fairly characteristic and is de-
Figure 10. Ureteroduodenal fistula. (a) Frontal radiograph obtained after retrograde contrast
agent injection of right upper urinary collecting system in a 67-year-old man shows contrast
agent within the duodenum (arrows) from an unsuspected fistula. Note wire (black arrowheads)
and small amount of retained contrast agent (white arrowhead) in the collecting system. (b) Contrast-
enhanced CT scan performed after a shows site of contact (white arrowhead) between
duodenum and right ureter. The fistula likely resulted from injury during previous aortofemoral
bypass surgery. Note vascular graft (black arrowhead) and right ureteral stent (arrow).
Figure 11. Rectovaginal fistula. Lateral radiograph
from air-contrast barium enema examination
in a 38-year-old woman with ulcerative
colitis shows air and contrast agent within the
vagina (V). The site of communication (arrow)
is visible inferiorly. The rectosigmoid region
appears somewhat foreshortened and featureless.
Figure 12. Salpingocolic (colotubal) fistula.
Frontal pelvic radiograph from hysterosalpingogram
in a 28-year-old woman with a history
of pelvic inflammatory disease shows left hydrosalpinx
and contrast agent filling a tuboovarian
abscess cavity (A), with extension superiorly
into the left side of the colon
(arrowheads).
Figure 13. Rectourethral fistula. Oblique radiograph
from retrograde urethrogram in a 64-
year-old man with a history of brachytherapy
for prostate cancer shows contrast agent in the
rectum (arrowheads). Contrast agent entered
the rectum via a large communication with the
prostatic urethra (arrow). As expected, contrast
agent is also present in the anterior urethra
and bladder (B). Note radiopaque brachytherapy
implants in prostatic region.
16 _ Radiology _ July 2002 Pickhardt et al
Radiology
scribed by the Chiari triad: midthoracic
pain or dysphagia, a sentinel episode of
hematemesis, and a symptom-free interval
that gives way to massive upper GI
bleeding (73,76). Diagnosis prior to exsanguination
is obviously imperative for
successful surgical repair. The chest radiograph
will typically demonstrate the
presence of an enlarged or tortuous thoracic
aorta (72). In stable patients, the
combination of CT and contrast-enhanced
esophagography will usually be
diagnostic. The esophagogram will usually
demonstrate deviation of the esophagus
due to the aneurysm, with or without
ulceration (Fig 17b) (72). The CT
findings are analogous to those seen with
aortoduodenal fistulas (Fig 17a). Aortography
was performed in the past but provides
less information than does chest CT
and only rarely will show the fistula directly
(72,77).
Other than typical portomesenteric venous
gas due to intestinal ischemia,
pneumatosis, and other causes, true enterovenous
and colovenous fistulas are
rare but potentially lethal entities. The
most common reported causes of duodenocaval
fistula include migration of caval
filters, right nephrectomy, peptic ulcer
disease, and ingestion of a foreign body
(78). Fistulas involving the mesenteric
small bowel and the colon are usually
secondary to Crohn disease and diverticulitis,
respectively (79,80). These fistulas
are often detected unexpectedly on barium
studies by observing intravasation of
the contrast agent (Fig 18). Substantial
barium intravasation reportedly carries a
high mortality rate (25,81). If the patient
survives the acute episode, diffusely increased
opacity of the spleen, greater
than that of the liver, can be seen on
both radiographs and CT scans due to
reticuloendothelial uptake of the barium
sulfate (81,82).
Respiratory tract.—Acquired esophagorespiratory
fistulas account for the majority
of intrathoracic GI fistulas and consist
of communication with either the tracheobronchial
tree or the pleura. Fistulas
that communicate between the respiratory
tract and the intraabdominal GI
tract (ie, gastrobronchial, enterobronchial,
and colobronchial fistulas) are rare
but may result from a penetrating subphrenic
abscess or a postsurgical complication
(83,84). Likewise, gastropleural
and colopleural fistulas are also rare and
are usually associated with diaphragmatic
herniation or prior pulmonary resection
(85–87). Of these GI-respiratory
fistulas, communication of the esophagus
with the tracheobronchial tree and
the pleura warrants further consideration.
Direct invasion by esophageal carcinoma
is the most common cause of acquired
tracheoesophageal and bronchoesophageal
fistulas, seen in approximately
5% of cases (88,89). Fistulas are especially
common following radiation therapy in
these patients. Other causes of tracheoand
bronchoesophageal fistulas include
primary lung and tracheal carcinoma,
Figure 14. Gallstone ileus from cholecystoduodenal fistula. (a) Supine radiograph in a 76-year-old woman shows bowel gas pattern suggestive of
small-bowel obstruction, two ectopic calcified gallstones (arrowheads), and air in the biliary tree (arrows). These findings constitute the Rigler triad.
(b) Transverse CT scans obtained without intravenous contrast agent in an 85-year-old woman show pneumobilia (arrowheads) and high-grade
small-bowel obstruction from an ectopic gallstone (short arrow). Note also orthotopic gallstones (long arrow) with a similar appearance.
Figure 15. Cholecystocolic fistula. Spot radiograph
from barium enema examination in
an 81-year-old man with nonspecific abdominal
complaints shows contrast agent within
the gallbladder (*) from communication with
the hepatic flexure. Air (arrowheads) is present
within the biliary tree.
Volume 224 _ Number 1 Acquired Gastrointestinal Fistulas _ 17
Radiology
esophageal instrumentation, tracheal
intubation, trauma, presence of foreign
bodies, and granulomatous infection
(89 –91). Patients typically present with
dysphagia and symptoms suggestive of
aspiration. The barium esophagogram
remains the study of choice, because it
effectively differentiates fistula from aspiration
(Fig 19a, 19b). The lateral projection
will generally best define tracheoesophageal
fistulas, whereas bronchoesophageal
fistulas may require a slightly different
obliquity. As previously mentioned, aqueous
contrast agents should be avoided because
of the risk of potentially lethal pulmonary
edema. CT can be a useful adjunct
in selected cases for evaluation of the
underlying cause or assessment of tumor
burden (Fig 19c). Treatment options for
malignant fistulas include palliation with
gastrostomy or jejunostomy, surgical bypass
or correction, and endoprosthetic
stent placement (88,92). For benign fistulas,
the goal of treatment is generally to
achieve definitive repair (89).
Esophagopleural fistulas are perhaps
best considered as a subset of esophageal
perforation and usually result from prior
surgery, endoscopic procedures, esophageal
carcinoma, or radiation therapy (93).
Clinical diagnosis is often difficult due to
inconstant and nonspecific symptoms,
especially in the absence of substantial
mediastinal involvement (93). Chest radiography
in patients with esophagopleural
fistulas will demonstrate either
pleural effusion or hydropneumothorax
on the affected side in essentially all
cases. Conventional esophagography is
indicated for confirmation and localization
of esophagopleural fistulas (Fig 20b).
CT can also demonstrate pleural air,
fluid, and/or contrast agent and can
sometimes demonstrate the fistula itself
(Fig 20a, 20c) (94).
Other fistulas.—Less common sites for
extraintestinal GI fistula formation include
the pericardium, pancreas, and skeletal system.
Fistula formation between the pericardial
space and the esophagus or stomach
should be considered in the setting of
nontraumatic spontaneous pneumopericardium.
Most esophagopericardial fistulas
are due to direct invasion from esophageal
cancer or a complication of treatment for
the cancer, whereas most gastropericardial
fistulas result from benign penetrating gastric
ulcers (95,96). In stable patients, CT
findings will help confirm the presence of
pericardial air and often suggest the underlying
cause (Fig 21a), while fluoroscopic
upper GI examination with a water-soluble
contrast agent is the simplest method for
directly demonstrating the pericardial communication
(Fig 21b). Endoscopy can also
be performed, but there is a potential risk
of inducing pericardial tamponade from
air insufflation (97).
Fistulas complicating surgical de´bridement
for severe necrotizing pancreatitis
are most often enterocutaneous and/or
pancreatricocutaneous, but internal pancreaticoenteric
communication is demonstrated
on rare occasions (12,98). Fistulas
may also form from spontaneous
rupture of a pseudocyst or peripancreatic
fluid collection into the stomach, colon,
or duodenum. In some cases, this transenteric
pseudocyst rupture will ameliorate
symptoms and serve a therapeutic
function. Depending on the situation,
conventional contrast-enhanced GI studies,
endoscopic retrograde cholangiopancreatography,
and/or CT can provide useful
diagnostic information (Fig 22).
Direct extension of GI inflammatory
and neoplastic processes to the abdominal
wall musculature, such as the psoas
and rectus abdominis, is common but
Figure 17. Aortoesophageal fistula. (a) Transverse contrast-enhanced CT scan in a 52-year-old
man with hematemesis and prior repair of thoracic aortic aneurysm with an endoluminal
stent-graft shows air (arrowhead) in the aortic lumen adjacent to the stent-graft. Irregular air
collection is also present in the expected region of the esophagus (arrow). (b) Oblique radiograph
from contrast-enhanced esophagogram directly demonstrates aortoesophageal fistula (arrows),
which was confirmed at surgery.
Figure 16. Primary and secondary aortoduodenal
fistulas. (a) Primary aortoduodenal fistula.
Transverse nonenhanced CT scan in an
80-year-old woman with GI bleeding shows a
calcified abdominal aortic aneurysm (A) with
intraluminal gas (arrow). Massive high-attenuation
retroperitoneal hemorrhage (H) surrounds
the aorta. (b) Secondary aortoduodenal
fistula. Contiguous transverse contrast-enhanced
CT scans in a 71-year-old man with GI
bleeding and history of aortic repair shows air
(black arrow) in the lumen of the aortic graft
and tethering (white arrow) of overlying duodenum
associated with periaortic inflammatory
changes.
18 _ Radiology _ July 2002 Pickhardt et al
Radiology
generally considered to represent a sinus
tract and not a true fistula when it is
contained. Rare acquired GI fistulas with
the skeletal system include colonic communication
with the hip (colocoxal) and
bowel communication with the spine
(enterospinal and colospinal) (99–102).
The mechanism for formation of these
fistulas is typically multifactorial, with
varying combinations of GI and orthopedic
surgery, radiation therapy, malignancy,
and inflammatory disease.
EXTERNAL (CUTANEOUS)
FISTULAS
Notwithstanding the deliberate creation
of a gastrostomy, jejunostomy, or colostomy,
the majority of unintended enterocutaneous
fistulas represent a complication
of prior surgery. Diverticulitis,
appendicitis, Crohn disease, and other
causes listed in Figure 2 may also manifest
with a spontaneous external fistula
on occasion (1). Perianal fistulas are
somewhat unusual in that most appear
to be idiopathic in nature or due to
Crohn disease; these will be considered
separately at the end of this section (103).
Factors that predispose to postoperative
enterocutaneous fistula formation
include anastomotic failure (eg, due to
inadequate blood supply, diseased bowel,
undue tension), adjacent abscess formation,
distal obstruction, and certain underlying
disease processes (12,104). Many
of these same factors will also influence
the likelihood of spontaneous closure of
such fistulas. Enterocutaneous fistulas are
further categorized according to their degree
of fluid production. High-output fistulas
drain more than 100–200 mL/day
and generally originate in the upper GI
tract, whereas low-output fistulas drain
less than this amount and are typically
more distal. Clinical management issues
beyond diagnosis and treatment of the
fistula itself include addressing potential
electrolyte imbalances, sepsis, malnutrition,
and wound care (105).
Enterocutaneous fistulas can be adequately
delineated and followed with fistulography
in the majority of cases (Fig
23) (20). In some instances, the fistulous
tract may be shown to equal or better
advantage with use of an intraluminal
enteric contrast agent (Fig 22). The most
common approach for fistulography is to
gently insert a soft-tipped catheter and
inject a water-soluble contrast agent with
fluoroscopic guidance to opacify the
tract. Contrast agent injection into an
existing surgical drain can also yield informative
findings. In addition to a variety
of occlusive adaptors and balloons to
obturate the skin site, it is often useful to
have the able patient hold the catheter
firmly in place during contrast agent in-
Figure 19. Tracheoesophageal and bronchoesophageal
fistulas. (a) Lateral radiograph from
barium esophagogram in a 61-year-old man
with esophageal cancer shows contrast agent
delineating tracheoesophageal communication
(arrowhead). Note widening of tracheoesophageal
stripe (_) and mass effect on the trachea
from tumor. (b) Lateral radiograph from barium
esophagogram in a 61-year-old man with
recurrent pneumonia shows fistula (arrow) between
esophagus and airway that was secondary
to histoplasmosis. (c) Reformatted oblique
transverse multi–detector row helical CT scan
in a 47-year-old man with bronchogenic carcinoma
shows irregular fistulous tract extending
from the left bronchial tree (black arrowhead)
to the esophagus (black arrow). Five standard
transverse CT images (not shown) were needed
to sequentially demonstrate the oblique course
displayed on this single reformatted image.
Note oral contrast agent (white arrow) in segmental
bronchus and peripheral airspace consolidation
(white arrowhead). The patient was
treated with a covered esophageal stent.
Figure 18. Colovenous fistula. Postevacuation
radiograph from barium enema examination
shows contrast agent throughout the inferior
mesenteric venous system (arrowheads).
Colovenous fistula was due to diverticulitis.
Note extensive sigmoid diverticulosis in this
region. (Case courtesy of Charles A. Rohrmann,
MD, Seattle, Wash.)
Volume 224 _ Number 1 Acquired Gastrointestinal Fistulas _ 19
Radiology
jection. Although fistulography will usually
demonstrate the bowel connection
and any large communicating abscess,
CT is usually performed in conjunction
not only to help identify all abscess cavities
but also to guide percutaneous drainage
of any abscesses found (106,107). In
current practice, external fistulas and associated
abscesses are often first discovered or
evaluated on CT scans, especially if the fistulous
tract extends only to the subcutaneous
tissue (Fig 24). Hydrogen peroxide–enhanced
sonography has been used recently
at some centers to delineate enterocutaneous
fistulas (108).
Conservative medical therapy, consisting
of total parenteral nutrition and somatostatin
analogues, constitutes the
standard initial approach to most postop-
Figure 20. Esophagopleural fistulas. (a) Transverse
nonenhanced CT scan in a 43-year-old
woman shows large air-fluid collection in left
pleural space (P) that encroaches on the esophagus
(arrow). (b) Subsequent esophagogram
shows contrast agent leak (arrowheads) from
the distal esophagus into empyema. (c) Transverse
contrast-enhanced CT scan in a 29-yearold
man with a history of radiation therapy for
Hodgkin disease shows communication (arrow)
between esophagus and apical hydropneumothorax
(P). L _ aerated right upper
lobe.
Figure 21. Gastropericardial and esophagopericardial fistulas. (a) Transverse contrast-enhanced
CT scan in an 82-year-old woman with spontaneous pneumopericardium (P) shows hiatal hernia
with inflammatory changes (arrowheads) adjacent to the pericardium. Gastropericardial communication
(arrow) is suggested and was confirmed at fluoroscopic examination with water-soluble
contrast agent (not shown). A penetrating benign gastric ulcer was the underlying cause.
(b) Oblique esophagogram in a 73-year-old man with spontaneous pneumopericardium after
distal esophagectomy for cancer shows contrast agent filling the pericardial space (arrowheads)
via the esophagopericardial fistula. (Case courtesy of Charles A. Rohrmann, MD, Seattle, Wash.)
Figure 22. Pancreaticocolocutaneous fistula.
Frontal radiograph from barium enema examination
in a 55-year-old man with severe pancreatitis
shows contrast agent filling an irregular
retroperitoneal collection (_) extending
from the region of the pancreatic tail. Communication
with the pancreatic ductal system is
not apparent. Note also colocutaneous fistula
(arrowheads).
Figure 23. Enterocutaneous fistula. Oblique
radiograph from pelvic fistulogram in a 29-
year-old man with abdominal tuberculosis
shows enterocutaneous fistula (arrowheads).
Note second cutaneous fistula (arrow) that
communicates with injection site.
20 _ Radiology _ July 2002 Pickhardt et al
Radiology
erative enterocutaneous fistulas (109,110).
Adjunctive measures include an appropriate
antibiotics regimen and correction of
electrolyte abnormalities. Surgical correction
is generally indicated when conservative
measures have failed or if peritonitis
develops (38,109,110). Failure of conservative
therapy is not a trivial matter, because
the mortality rate for surgical management
approaches 25% in some series (111). Percutaneous
catheter management by the radiologist
for enterocutaneous fistulas has
been shown to be a feasible alternative to
surgery in patients in whom medical treatment
has failed (111). This approach utilizes
fistulography to guide cannulation of
the fistulous tract by means of catheter and
guidewire manipulation, as well as CT for
detection and drainage of abscesses not
communicating with the main fistulous
tract. A 90% success rate for this nonsurgical
approach has been shown in high-output
fistulas in which conservative therapy
has failed (111). The closure rate for the
low-output fistulas (65%) is somewhat
lower, and these fistulas also require more
time to resolve than do high-output fistulas.
As mentioned earlier, perianal fistulas
represent a unique subset of external fistulas.
In the majority of cases, preoperative
imaging is not necessary and patients
generally do well with simple
fistulotomy (103). In a minority of complex
cases, however, preoperative imaging
will be necessary to provide a road
map for surgery. The primary goals of
imaging are to determine if the fistula
traverses the external sphincter (transsphincteric)
or levator ani muscles and to
identify any secondary fistulous tracts or
abscess cavities that could lead to treatment
failure (32). Most attempts at preoperative
imaging with conventional fistulography,
endoscopic sonography, and
CT have failed to surpass the accuracy of
the clinical examination (112–114). MR



