Key words
pancreas - CT - MR imaging - resectability - cancer - staging
Introduction
Although notable progress in diagnosis and therapy has been made in recent years,
the 5-year survival of patients diagnosed with ductal adenocarcinoma of the pancreas
is still less than 5 % [1]. To date, surgery is the only curative option for patients suffering from this aggressive
entity. Nowadays in high-volume centers, this procedure can be performed with a mortality
of 1 – 3 % and morbidity of about 40 % [2]. However, only a minority of patients is resectable and 5-year survival rates of
more than 20 % after resection are rare [2]. Radical resection of the tumor reaching negative resection margins is one of the
strongest predictors for long-term survival [3]. Surgical exploration is the gold standard to determine unresectability. However,
it is crucial to avoid passing the point of no return during explorative surgery.
Otherwise, the resection has to be completed with the result of an R2-resection which
implicates perioperative risk and does not improve prognosis [4]. Therefore, radiological assessment of resectability is an indispensible standard
procedure before surgery with the goal to plan surgery by displaying the critical
zones or to identify clearly unresectable patients.
Surgical techniques have evolved significantly over the recent decades with attempts
of more aggressive and radical resection. This is paralleled by technical advances
in imaging technology regarding computed tomography (CT) and magnetic resonance imaging
(MRI), which are the radiological modalities most frequently requested by surgeons
for preoperative diagnostics and resectability assessment [5]
[6]
[7]. Therefore, a continuous reappraisal of radiological methods and criteria in resectability
assessment and exact knowledge of current surgical techniques are necessary for radiologists
to give the correct advice to surgeons.
How to resect
Pancreatic cancer is located in 70 % of cases in the pancreatic head, in 20 % of cases
in the body and 10 % are located in the pancreatic tail. The oncologically radical
types of resection for pancreatic head cancer are en-bloc resection of the pancreatic
head and duodenum with partial gastrectomy (Kausch-Whipple procedure) or, today’s
standard, pylorus preserving without gastric resection (Traverso-Longmire) [2]. For cancer of the pancreatic body and tail, a left pancreatectomy with splenectomy
is performed [2]. Depending on the dimension of a left-sided carcinoma, the resection has to be extended
towards the pancreatic head to reach tumor-free resection margins (subtotal left-sided
pancreatectomy or total pancreatectomy). If the carcinoma is found to be of multilocular
origin, a total pancreatectomy plus splenectomy is indicated to ensure complete removal
of the tumor. Due to the absence of symptoms such as jaundice or gastric outlet stenosis,
it is not unusual to diagnose left-sided pancreatic tumors at an advanced stage with
involvement of adjacent organs such as the colon or stomach. In this case, multivisceral
resection with complete removal of the tumor remains an option because the median
survival is comparable to that of patients after standard pancreatic resections [8].
It can be deduced from the aforementioned characteristics of the different surgical
options that location, intrapancreatic extent, and extension into adjacent tissues
are the key information for surgical decision making. Regarding the extrapancreatic
extent, both contrast-enhanced multiplanar CT and MRI are capable of displaying the
tumor as well as the surrounding retroperitoneal and mesenteric fat, stomach, and
bowel wall. Here, in case of absent or depleted separating fat planes, blurred or
abolished boundary surfaces of tumor-adjacent tissues may indicate infiltration.
Inside the pancreas, the highest possible contrast of the tumor to the surrounding
pancreatic tissue is needed. For CT, it has been shown that after intravenous application
of contrast media, pancreatic tissue enhances early, while ductal adenocarcinoma of
the pancreas typically shows slow contrast accumulation [9]. The highest difference in radiodensity between tumor and pancreas is reached with
a scan delay of approximately 40 seconds after the start of contrast administration
[9] ([Fig. 1]). This contrast phase is therefore indispensable for preoperative CT imaging of
pancreatic cancer, and the results can be transferred to dynamic contrast-enhanced
MRI as well, which is equally suited to detect and delineate tumors [10]. The use of bolus tracking can improve the correct timing of this pancreatic parenchymal
phase in the individual patient by using an interval of approximately 25 seconds after
bolus arrival in the abdominal aorta [11].
Fig. 1 Dynamic computed tomography (a 20 sec. delay; b 45 sec. delay; c 70 sec. delay) demonstrating contrast intensification of pancreatic ductal adenocarcinoma
(asterisk). The best delineation of the tumor against the pancreatic parenchyma (arrow
head) is seen with a delay of 45 sec. because of quick contrast uptake in the pancreas
and slow accumulation in the tumor. This allows detailed visualization of tumor/vessel
contact (block arrow) towards the superior mesenteric vein (arrow). The superior mesenteric
artery (empty arrow) is surrounded by a fat plane indicating non-involvement. Note
the stent in the choledochal duct.
Abb. 1 Dynamische Computertomografie (a 20 s Verzögerung, b 45 s Verzögerung, c 70 s Verzögerung) zeigt die Kontrastverstärkung bei duktalem Adenokarzinom des Pankreas
(Stern). Die beste Abgrenzbarkeit gegenüber dem Pankreasparenchym (Pfeilspitze) besteht
bei 45 s Verzögerung wegen der raschen Kontrastmittelaufnahme des Pankreasgewebes
und der langsamen Akkumulation im Tumor. Dies ermöglicht eine detaillierte Darstellung
des Tumor/Gefäß-Kontaktes (Blockpfeil) zur V. mesenterica sup. (Pfeil). Die A. mesenterica
sup. (leerer Pfeil) ist von einer Fettschicht umgeben, demnach ist die Arterie hier
sicher tumorfrei. Anmerkung: Stent im D. choledochus.
Vessel invasion
The most common finding when determining unresectability besides the presence of metastases
is the local invasion of major vascular structures by the primary tumor. In the case
of pancreatic left resection, splenic vessels, even though their invasion is a negative
prognostic factor [12], are dispensable and are included in the en-bloc resection of the pancreatic tail
and spleen. However, in the case of the most frequent location of pancreatic cancer,
i. e., the pancreatic head, the surrounding vascular structures are complex and the
arterial supply of the liver, stomach, and bowel, as well as the mesentericoportal
venous axis has to be maintained in any kind of pancreatic resection. Many efforts
have been made to facilitate complete tumor resection despite vessel involvement and
technically, it is indeed possible to perform vessel resection and reconstruction.
However, the surgical capabilities have to be set in relation to the achievable clinical
outcome. With respect to the clinical relevance, a differentiation must be made between
venous and arterial involvement [13]
[14]. This is reflected by the current TNM staging system: the T3 category, representing
the locally invasive but potentially resectable tumor, includes also focal invasion
of the mesentericoportal venous axis, which can be handled surgically, while the T4
category, referring to the locally advanced unresectable tumor, includes invasion
of the superior mesenteric artery (SMA) or the celiac axis [15]. The basis for discussion of the surgical options and prognosis related to the local
tumor extent is the preoperative imaging with distinct analysis of the local tumor
situation.
Veins
Tumor infiltration of the mesentericoportal axis is common in pancreatic head carcinoma.
If the portal vein (PV) resection results in negative resection margins, survival
rates similar to those of patients without PV resection can be achieved [16]. For that, tumor involvement of the PV is not a general contraindication for curative
pancreatic head resection [13]. To maintain the venous drainage of the bowel after venous resection, several surgical
procedures are available for venous reconstruction. These are segmental resection
with reanastomosis (with or without interposition of a graft) and wall excision with
patch plastic. Exact analysis of the portal venous confluence anatomy and its normal
variants in this context is important and can be done with CT [17]. Extensive invasion of the mesenteric vein with separation of its branches as depicted
by CT or MRI can even result in unresectability ([Fig. 2, ]
[Table 1]). The intraoperative decision regarding how to manage venous involvement is made
to a great part after the point of no return of a pancreatic head resection. This
is why distinct preoperative knowledge of the extent of contact between the tumor
and the mesentericoportal venous axis is so important.
Fig. 2 Carcinoma of the pancreatic head invading the superior mesenteric vein (CT, coronal
MIP). The tumor (asterisk) occludes several main branches of the mesenteric vein (arrows).
In this case, spontaneous collateralization of all venous territories after resection
cannot be expected to be sufficient and surgical reconstruction of all relevant branches
is problematic resulting in unresectability upon explorative surgery.
Abb. 2 Pankreaskopfkarzinom mit Infiltration der V. mesenterica superior (CT, koronare MIP).
Der Tumor (Stern) verschließt mehrere Hauptäste der Mesenterialvene (Pfeile). In diesem
Fall kann eine ausreichende Kollateralisierung nach Resektion nicht erwartet werden
und die chirurgische Rekonstruktion von allen relevanten Ästen ist problematisch mit
dem Ergebnis der Irresektabilität nach explorativer Laparotomie.
Table 1
Signs of unresectability on computed tomography and magnetic resonance imaging examinations.
Tab. 1 Zeichen der Irresektabilität in der Computertomografie und der Magnet-Resonanz-Tomografie
|
radiological signs of unresectability
|
exceptions
|
|
vessel involvement
|
arteries
|
celiac trunk
|
left pancreatectomy with resection of the celiac axis for cancers of the body and
tail can be considered if superior mesenteric, gastroduodenal, and proper hepatic
arteries are free
|
|
common hepatic artery
|
|
proper regular or aberrant hepatic artery
|
|
|
superior mesenteric artery
|
|
|
veins
|
superior mesenteric vein at its branching
|
if surgical reconstruction is possible
|
|
inferior vena cava
|
|
|
renal vein
|
|
|
metastases
|
hepatic or other hematogenous metastases
|
|
|
peritoneal carcinomatosis
|
|
|
invasion of adjacent organs
|
spleen, colon, small bowel, stomach, adrenal gland
|
individual consideration if R0-resection seems possible
|
The radiological appearance of the tumor/vein interface can be with a fat plane between
the tumor and the vein securely indicating non-involvement and dissectability, while
complete encasement and vessel occlusion are a reliable sign for profound vessel invasion
[18]. Any other direct tumor contact to the vessel may indicate invasion or a non-dissectable
adhesion and has to be further characterized ([Fig. 1]). A recent study used simple descriptive criteria of the tumor/vein contact on CT
images to predict the actual infiltration depth in the wall of the mesentericoportal
veins and prognosis retrospectively in 358 patients resected for pancreatic cancer
[19]. An increasing number and increasing depth of vessel wall invasions were found with
increasing tumor-related narrowing of the vessel lumen (invasion into tunica media
or deeper: no narrowing, 0 %; unilateral narrowing, 27 %; bilateral narrowing, 42 %;
obstruction with collaterals, 63 %), paralleled by a significantly worse prognosis
for patients with bilateral narrowing or obstruction compared with no or unilateral
narrowing of the vein [19]. It has been shown that the use of multiplanar image reconstruction enhances accuracy
when determining vein invasion [20]. Reporting the findings not only by description and axial images but also by multiplanar
or three-dimensional reconstructed images helps the surgeon to plan the procedure
regarding complexity, duration, and material needed.
Arteries
Visceral arteries that are commonly affected by tumor growth are the SMA, the common
hepatic artery (CHA) and the celiac artery (celiac trunk). Additional arterial resection
during pancreatic head resection is associated with significantly increased perioperative
mortality and poor survival compared to patients without necessity for additional
arterial resection [21]. Arterial invasion is therefore considered an unresectability criterion ([Table 1]).
CT angiography (CTA) and MR angiography (MRA) are the methods of choice as they display
both the tumor and the vessel at high image quality [22]
[23]. An early study on CT in the preoperative evaluation of 25 patients with pancreatic
cancer showed that the extent of circumferential tumor growth around the artery correlates
with vessel wall invasion [24]. Tumor contact of more than half of the artery circumference and/or vessel constriction
indicated unresectability at this artery segment with a sensitivity and specificity
of 100 %. In case of lesser tumor contact without constriction, the tumor was still
surgically unresectable at half of the evaluated arteries [24]. A contemporary study used a distinct analysis of the tumor/vessel interface describing
it as a convex or concave contact [18]. Again, in the case of contact to the vessel, it was difficult to predict vessel
wall invasion. This underlines that between separating the fat plane ([Fig. 1]) and vessel encasement, there is a gray zone of tumor contact to the vessel without
a clear-cut differentiation of cases with and without arterial wall infiltration.
In contrast to the devices used in the studies cited above, new multidetector CT scanners
with fewer movement artifacts and higher resolution enable high-quality three-dimensional
multiplanar and curved image reconstruction. This facilitates improved depiction of
arterial constrictions as well as distinct assessment of circumferential and longitudinal
tumor contact independent of the vessel orientation. In a recent study evaluating
the involvement of the SMA, the CHA, and the celiac trunk with multidetector CT using
the traditional criteria of circumferential tumor contact and vessel constriction
in 70 pancreatic cancer patients, the sensitivity and specificity were increased from
88 % and 94 % for axial images to 100 % and 93 % for additional multiplanar reconstructions
[20]. Other recent studies showed similar results with high accuracy for the assessment
of arterial invasion for CT and even for MRI using latest generation devices [10]
[25]
[26]. These improvements help to increase the rate of correct indications for and successful
completions of open surgery with curative intent in pancreatic cancer patients [27]. However, despite the improvements of imaging in this respect, its remaining inaccuracy
regarding arterial involvement demands surgical exploration, at least in equivocal
cases, as this is still the gold standard in determining resectability.
Arterial variants
In pancreatic head resection, it is decisive to know about the individual arterial
anatomy. Variants of hepatic arteries are common and can be relevant because tumor
contact can occur at important arteries, which are not where the radiologist and the
surgeon expect them to be [26]. Conditions which may lead to impaired arterial blood supply of the liver after
pancreatic head resection are, e. g. preexisting severe stenosis of the celiac axis
which becomes relevant after cutting off the collateral flow from the gastroduodenal
artery (GDA) or accidental injury of an aberrant hepatic artery arising from the SMA
behind or – rarely – inside the pancreatic head ([Fig. 3]). Tumor infiltration of such an artery is a contraindication for resection just
as of the SMA or a regular CHA ([Table 1]).
Fig. 3 Rare arterial variant with intrapancreatic aberrant right hepatic artery (arrows)
arising from the superior mesenteric artery through the pancreatic head in close proximity
to the pancreatic head tumor (asterisk) to the liver (CT, axial MIP).
Abb. 3 Seltene arterielle Normvariante mit intrapankreatischer aberranter rechter Leberarterie
(Pfeile) die aus der A. mesenterica superior durch den Pankreaskopf in enger Lagebeziehung
zu einem Pankreaskopfkarzinom (⌘) zur Leber zieht (CT, axiale MIP).
As MRA and even more so CTA are highly accurate in displaying the arterial anatomy
in the upper abdomen preoperatively, it is obligatory to report the relevant arterial
anatomy in detail to the surgeon [22]
[26]
[28].
Metastases
According to the German S3 guideline, pancreatic adenocarcinoma with hematogenous
metastases or peritoneal spread is unresectable regardless of the local tumor extent
since it does not improve the overall prognosis [13]. Palliative resection despite present metastases is currently being debated but
is still not recommended outside studies. Thus, the exclusion of metastatic spread
remains an important part of resectability assessment by imaging and surgical exploration
([Table 1]).
Peritoneum
Regarding the peritoneal spread of pancreatic cancer, preoperative detection is problematic
as at an early stage these lesions may appear in the form of small flat spots on the
peritoneal surface, without sufficient dimensions to become visible on endoscopic
ultrasound (EUS), CT, or MRI. Imaging is limited to the depiction of nodular lesions
large enough to pass partial volume effects and to be differentiated from the adjacent
structures, such as bowel wall or liver surface [29]. A recent study using mostly CT (95 %) for preoperative staging showed an overall
rate of unanticipated peritoneal spread discovered during surgery of 5 % in 487 patients
[30]. Positron emission tomography with F18-Fluorodeoxyglucose (FDG-PET), initially promising
to overcome these problems, did not perform better than CT for the same reasons [31]. Often, only indirect findings, such as small amounts of free peritoneal fluid,
can be depicted to raise the suspicion for peritoneal spread. Such findings and other
risk factors for the presence of unanticipated peritoneal spread, like large primary
tumors and tumor location in the pancreatic body or tail, should prompt the surgeon
to start with laparoscopy instead of laparotomy in order to keep the trauma minimal
in case of termination of the exploration [32].
Liver
While CT has not shown any major improvement of the detection rates of liver metastases
over the recent decade, MRI has experienced innovations like fast 3 D sequences and
diffusion-weighted imaging (DWI) resulting in an increase of sensitivity from 70.2 %
until 2004 to 84.9 % until 2010, which is superior to CT (75 %, metaanalysis on colorectal
cancer metastases) [33]. For pancreatic cancer, a recent study reported a rate of unanticipated liver metastases
of 12 % after negative preoperative imaging (mostly CT, 95 %) [30]. Furthermore, the rate of unanticipated hepatic and peritoneal metastases increased
significantly from 10 % in patients who had been operated within 3 weeks after imaging
to 20 %, 25 %, and 35 % in patients who had undergone surgery during the fourth, fifth,
and sixth week after imaging, respectively [30]. Therefore, high-quality imaging of the liver and abdomen is needed shortly before
surgery to avoid false-negative results. High-quality imaging means contrast-enhanced
dynamic scanning using CT and/or MRI and, in the case of MRI, DWI besides the standard
sequences [34]
[35]. For both CT and MRI these prerequisites for liver imaging are compatible with pancreatic
imaging optimized for resectability assessment. FDG-PET has the strength of a high
specificity but a striking lack of sensitivity is reported regarding small liver metastases
[35]. Therefore, FDG-PET is currently not the method of choice for resectability assessment
of pancreatic cancer [13]
[36].
Lymph nodes
Pancreatic cancer resection with a standard lymphadenectomy includes the removal of
lymph nodes in the peripancreatic region, along the hepatoduodenal ligament, the celiac
trunk, and the SMA. Lymphatic metastases occur frequently and early with lymph node
involvement in 75 % of pT3 tumors [37]
[38]. In a large population-based study, N1 disease (n = 1,507) was associated with a
significantly worse 5-year survival rate of 4.3 % compared with 11.3 % in N0 disease
(n = 1,971), irrespective of other factors like grading, local extension, and number
of assessed lymph nodes [39]. Despite this negative prognostic impact, locoregional lymph node metastases of
pancreatic adenocarcinoma in general do not preclude a patient from the attempt of
curative surgery [13]
[39].
Given the high prevalence of lymphatic metastases, the minor role for unresectability,
and the well-known problems of diagnostic imaging in the detection of involved lymph
nodes in any kind of cancer, preoperative lymph node staging plays a minor role for
surgical decision-making. The accuracy for the detection of lymph node metastasis
by diagnostic imaging is limited as shown in a study on the detection of paraaortic
lymph node metastases in 69 pancreatic cancer patients with all six lymph node positive
cases being negative on CT, MRI, and FDG-PET [39]. Nevertheless, reporting enlarged, spherical, irregular-shaped, centrally necrotic, or otherwise
suspicious nodes can be of importance for the surgeon especially when occurring outside
the regions of standard lymphadenectomy because extended lymphadenectomy remains an
option for these patients [41].
Innovative treatment concepts
Innovative treatment concepts
Neoadjuvant therapy
If radiological signs of locally advanced disease with arterial infiltration are present,
the conversion of advanced pancreatic carcinoma from “non-resectable” to resectable
seems to be a promising concept [42]. It remains to be seen what significance neoadjuvant treatment (chemotherapy and/or
radiotherapy) will have in the future but already now radiologists are faced with
the reassessment of resectability after the completion of neoadjuvant therapy.
It is well known from other tumor entities that the accuracy of CT and MRI can be
impaired after neoadjuvant treatment [43]
[44]
[45]. Only a few studies have addressed this issue for pancreatic cancer with variable
results. An early study concluded that the prediction of resectability by CT after
neoadjuvant radiochemotherapy is comparable to cases without preceding therapy [46]. In contrast, a later study from the same group using CT to reassess initially borderline
resectable tumors after neoadjuvant chemo- or chemoradiation therapy showed radiographic
reduction of vascular involvement in only less than 1 % but an R0-resection rate as
high as 80 % [47]. Another recent study confirmed this tendency of overestimation of the local tumor
extent by CT and MRI after neoadjuvant therapy [48], while Kim et al. observed some inaccuracy in T-staging by CT after neoadjuvant
treatment but only minor effects on the assessment of resectability [49].
Left pancreatic plus celiac trunk resection
Carcinoma of the body of the pancreas with involvement of the celiac trunk and/or
the CHA is considered to be unresectable [13]. A radical distal pancreatectomy with splenectomy and en-bloc resection of the celiac
trunk without reconstruction of the celiac axis aims at providing a curative approach
for these cases ([Table 1]). A recent metaanalysis of this approach showed survival rates equal to regular
R0-resections [50]
[51]. The Achilles heel of this procedure is the interruption of the direct arterial
blood supply to the liver, bile ducts, and stomach. Despite collateral pathways via
the SMA, pancreaticoduodenal arcades, and the GDA, arterial perfusion of critical
organs can be compromised after this procedure, causing severe complications such
as liver failure, biliary duct necrosis, perforation of the stomach, and ischemic
ulcer [52]
[53].
In order to avoid these complications, an interdisciplinary approach was introduced.
Preoperative digital subtraction angiography with two catheters intubating the celiac
trunk and the SMA provides important information about the status of the relevant
vessels [53]. In the case of preexisting celiac trunk stenosis, spontaneous collateral flow can
be seen from the SMA over pancreaticoduodenal arcades to the GDA supplying the hepatic
artery. If no stenosis is present, a test occlusion of the celiac trunk with a balloon
catheter can be employed to provoke collateral flow and to confirm thereby the existence
of sufficient collateral arteries ([Fig. 4]) [53]. To enhance the collateral flow already before the operation, which follows typically
one week later, embolization of the celiac trunk is recommended in most reports on
this approach as this “training” of collateralization reliably prevents ischemic complications
[53]
[54].
Fig. 4 T2-weighted MRI a of a carcinoma of the pancreatic corpus (asterisk) with infiltration around the celiac
trunk (arrow). Selective celiaco- b and mesentericography c show no spontaneous collateral flow from the superior mesenteric artery to the hepatic
arteries. With test occlusion of the celiac trunk d with a balloon (empty arrow) presence of sufficient collaterals from the superior
mesenteric artery over pancreaticoduodenal arcades (arrow heads) to the gastroduodenal
artery and the proper hepatic artery (block arrow) can be confirmed before definite
coil embolization and subsequent en-bloc pancreatic left and celiac trunk resection.
Abb. 4 T2-gewichtete MRT a eines Pankreaskorpuskarzinoms (Stern) mit Infiltration um den Tr. coeliacus (Pfeil).
Die selective Coeliaco- b und Mesentericographie c zeigt keinen spontanen Kollateralfluss von der A. mesenterica sup. zu den Leberarterien.
Mit einer Testokklusion des Tr. coeliacus d mit einem Ballon (leerer Pfeil) kann die Existenz suffizienter Kollateralen von der
A. mesenterica sup. über pancreatikoduodenale Arkaden (Pfeilspitzen) zur A. gastroduodenalis
und A. hepatica propria vor einer definitiven Coilembolisation der zöliakalen Achse
und Pankreaslinksresektion mit Trunkusresektion geprüft werden.
Conclusion
Surgical exploration is the gold standard for the determination of tumor resectability
of pancreatic ductal adenocarcinoma. Preoperative imaging by CT or MRI can identify
clearly unresectable tumors and is essential for surgical planning in resectable and
borderline resectable patients.
Radiologic resectability assessment comprises location and extent of the primary tumor
including contact to adjacent vessels, vessel anatomy, presence of liver metastases,
signs of peritoneal carcinomatosis, and presence of lymph node metastases.
Even with modern imaging technologies, false-positive signs of unresectability may
occur. Therefore, the indication for surgical exploration should be made broadly to
not preclude any patient from the chance for complete tumor resection.
