Pancreatic Carcinoma

Detailed images of upper abdominal vascular anatomy can now be obtained by experienced operators using modern real-time ultrasound scanners. To help determine the role of real-time ultrasound in preoperative evaluation of pancreatic carcinoma, ultrasound and angiographic results were compared in a series of 45 patients admitted with documented carcinoma of the pancreas. When surgical results were available, these were compared with the ultrasound and angiographic findings. The main tributaries of the portal system were evaluated with ultrasound and angiography for evidence of compression, displacement, or occlusion. A sensitivity of 90% and specificity of 95% were achieved for splenic vein abnormalities by ultrasound compared to angiography. For portal vein evaluation, sensitivity was 75%, and specificity was loo%, whereas for the superior mesenteric vein a sensitivity of 100% and a specificity of 94% were attained. Ultrasound can make angiography prior to laparotomy unnecessary in many cases if thorough evaluation of the portal venous system is emphasized during the initial ultrasound study. Indexing Words: Angiography * Pancreatic carcinoma * Portal vein * Veins, mesenteric * Vein, splenic * Ultrasound The incidence of pancreatic carcinoma has steadily increased since the 1 9 3 0 ~ . ~ , ~ Consequently, clinicians in many centers have faced management decisions in patients with this disease with increasing frequency. Advanced disease is clinically obvious in more than half of patients at the time of d i agn~s i s ,~ .~ and appropriate therapy is largely palliative in such circumstances. However, in early cases of pancreatic cancer, surgical resection by the Whipple procedure or by total pancreatectomy offers the only possibility of ~ ~ 1 8 . ~ 3 ~ Although pancreatic resection is a technically demanding and high-risk procedure, most surgical oncologists advocate pancreatic resection with curative intent whenever technically possible in medically fit paFrom the *Department of Radiology, Uniformed Services University of the Health Sciences, Bethesda, Maryland 20814, ?Diagnostic Radiology Department, The Clinical Center, and $Surgery Branch, National Cancer Institute, National Institutes of Health, Bethesda, Maryland 20892. For reprints contact Brian s. Garra, MD, Diagnostic Radiology Department, Building 10, Room 1C660, NIH, The Clinical Center, Bethesda, Maryland 20892.

The incidence of pancreatic carcinoma has steadily increased since the 1 9 3 0~.~,~ Consequently, clinicians in many centers have faced management decisions in patients with this disease with increasing frequency. Advanced disease is clinically obvious in more than half of patients at the time of d i a g n~s i s ,~.~ and appropriate therapy is largely palliative in such circumstances. However, in early cases of pancreatic cancer, surgical resection by the Whipple procedure or by total pancreatectomy offers the only possibility of ~~1 8 . ~3~ Although pancreatic resection is a technically demanding and high-risk procedure, most surgical oncologists advocate pancreatic resection with curative intent whenever technically possible in medically fit pa-t i e n t~.~,~ It is therefore important to attempt to define which patients are likely to be resectable before exploratory laparotomy . Local spread of tumor causing occlusion or encasement of the major veins in the portal system is a generally accepted criterion of surgical unrese~tability.~,~J~ Angiographic evaluation of the portal venous system has been shown to be an important predictor of operability when done either by direct injedionl1J2 or as part of an arteriographic study.13 It remains the only widely used method for portal system evaluation in pancreatic cancer patients. Prior studies on the use of ultrasound in pancreatic carcinoma patients have emphasized identification of tumor mass, biliary obstruction, and metastasislP16 but have not emphasized the value of ultrasound in the evaluation of the portal system. Modern real-time scanners allow accurate delineation of the upper abdominal vascular anatomy and are often capable of demonstrating tumor involvement of the portal system. 17 We recently compared our ultrasound results with angiographic and surgical findings in a series of patients admitted for treatment of pancreatic car- cinoma to determine what roles angiography and ultrasound should play in the preoperative evaluation of the portal system in these patients.

PATIENTS AND METHODS
All patients admitted to the National Institutes of Health between October 1981 and May 1984 with documented pancreatic carcinoma who received both an angiogram and an ultrasound evaluation were included in this study. For the 60 patients fulfilling these requirements, the ultrasound studies and angiograms were reviewed to determine in which patients the portal venous system was adequately demonstrated by both modalities. The portal system was considered adequately demonstrated by ultrasound when either two of the three main veins (portal vein, splenic vein, superior mesenteric vein) or the contiguous anatomic area was visualized. Adequate angiographic demonstration of the portal system was defined as opacification of two of the three major veins or opacification of associated abnormalities such as venous collaterals. The ultrasound films of each case were then examined without reference to the written report or angiogram, and each major vein was classed as normal, occluded, displacedkompressed, or not well visualized (Figs. 1-4). This film review was undertaken since normal venous structures are often seen but not always mentioned in written reports. No discrepancies between the film review results and the written reports with respect to abnormal findings were discovered. Thus, for comparison with the angiographic and surgical findings, the results of the film review and written reports were pooled.
The angiograms and angiogram reports were next reviewed in a similar fashion and were compared with the ultrasound results. Finally, the surgical reports for patients who had a laparotomy were correlated with the ultrasound and angiographic results. son. Of the 15 cases excluded, 14 had a good-quality angiogram but poor visualization of the area of the pancreas in the ultrasound examination. One patient had a good-quality ultrasound evaluation, but complications occurring during the angiogram resulted in poor opacification of the portal venous system.

Of the
Abdominal exploratory surgery was performed on 38 of the 45 patients in the comparison group, with 21 undergoing operative evaluation of the portal venous system. The remaining 17 patients received only a limited exploration because liver or peritoneal metastases were found (12 patients), severe pancreatitis or other medical problems were present (three patients), or the mass was quickly seen at laparotomy to be too large to be resectable (two patients).
The results of our comparison of ultrasound and angiography for the splenic vein (SV) are shown in Fig. 5. The splenic vein was well visualized by both modalities (unshaded cells in Fig. 5) in 30 patients. Grouping those patients with compressed veins together with those having an oc-cluded splenic vein in a single abnormal group results in a sensitivity of 90% and a specificity of 95% for ultrasound with respect to angiography. If only veins reported as occluded are counted as abnormal, the sensitivity and specificity were 57% and 96%, respectively. One also can see from Fig. 5 that the ultrasound and angiogram results generally agreed as to whether the abnormal veins were compressed or occluded. For cases in which the splenic vein was not visualized by ultrasound, slightly less than 42% were abnormal; this is only a little more than the percentage of abnormals in the entire group (33%). Table 1 shows the correlation of surgical results with ultrasound and angiography. Surgery and angiography agreed in 14 of 15 cases in which the splenic vein was successfully evaluated by all modalities. In two of these cases ultrasound overestimated the extent of involvement when compared to the other methods, whereas in one case ultrasound underestimated the extent of disease. In three other cases the surgical reports suggested portal obstruction but did not specifically describe JOURNAL OF CLINICAL ULTRASOUND the status of the splenic vein. One case not visualized by ultrasound showed SV occlusion by both angiography and surgical exploration. Figure 6 compares the ultrasound and angiographic results for the portal vein (PV). In 38 cases the portal vein was visualized by both modalities (unshaded cells in Fig. 6), and in 18 of these surgical correlations were available ( 97%. Ultrasound and angiography usually agreed regarding the extent of involvement, with ultrasound underestimating the extent in two cases and overestimating it in one case. In two of the three discrepant cases, the surgical findings (Table 1) supported the angiographic findings. In five cases surgical findings suggested portal vein obstruction when the angiogram showed a patent vessel, but the surgical diagnoses were based on finding dilated mesenteric veins rather than direct visualization of the portal vein.
A comparison of ultrasound and angiographic results for the superior mesenteric vein (SMV) is  shown in Fig. 7. The vessel was identified by both modalities in only 21 patients (unshaded cells in Fig. 7). Nonvisualization by ultrasound occurred in over half of the cases studied. The sensitivity and specificity of ultrasound with respect to angiography for normal versus abnormal were 100% and 94%, respectively. For occluded versus patent veins, the sensitivity was 50%, and the specificity was 95%. Surgical correlations were available in nine of the 21 cases (Table 1). Of the three cases in which the ultrasound and angiographic results disagreed, surgical correlation was available in two. The surgical results supported the angiographic finding in one case and the ultrasound finding in the other. In the cases in which angiography visualized the SMV but ultrasound did not, three more surgical correlations were available, all supporting the angiographic findings.

DISCUSSION
In pancreatic carcinoma patients, the term resectable is generally used by surgeons to indicate VOL. 15, NO. 2, FEBRUARY 1987 that all macroscopically evident tumor can be removed by partial or total pancreatectomy. Specific signs of unresectability used by surgeons at laparotomy vary somewhat but include (1) visceral metastases or peritoneal implants, (2) gross tumor invasion into the retroperitoneum, stomach, liver, or mesentery, (3) encasement or obstruction of the inferior vena cava, celiac axis, hepatic artery, or superior mesenteric artery, and (4) complete obstruction of the portal venous system.18 A partially obstructed portal system may or may not be resectable depending on how long a segment of the involved vein or veins must be removed to complete the resection. A number of the patients in our study were deemed unresectable because of liver or peritoneal metastases noted before laparotomy and therefore did not have extensive enough exploration to allow surgical evaluation of the portal system. Of the patients who did not have surgery, only one was excluded from operation because of extensive vascular involvement seen angiographically.
In our study, most of the operated patients were deemed unresectable at surgery (381, and of this group, 66% were unresectable because of metastases or implants, and 26% were unresectable because of vascular obstruction or invasion. The large number of patients with metastases accounts for the lack of surgical correlation in a number of patients, since the portal system was not explored when metastases were found. Our results seem to suggest that angiography is much more reliable than ultrasound at visualizing the portal system, as 14 of the 15 cases excluded from comparison because of poor visualization of the portal system represented ultrasound failures. This may not be true in view of the fact that most of the earlier ultrasound examinations did not include a careful search for portal system veins. Once the potential of ultrasound for abdominal vascular evaluation was recognized, more thorough ultrasonic examination and reporting of the portal system status was begun, and fewer ffnonvisualizations~' occurred. An analysis of the percentage of nonvisualized major portal system veins (PV, SV, SMV) by year showed a drop from 49% and 45% in 1982 and 1983, respectively, to 15% in 1984, when more thorough portal evaluation was emphasized. Since multiple radiologists of varying experience interpret and monitor ultrasound examinations at our institution, the 15% nonvisualization rate could probably be reduced further if experienced sonographers aggressively conducted all examinations. Examination of the "errors" made by ultrasound in splenic vein evaluation showed that in one case in which the splenic vein was reported as patent by ultrasound, angiography showed obstruction at the splenic hilus. Thus, obstruction of the splenic vein may not be identified ultrasonographically if the site of obstruction lies leftward away from the region best seen by ultrasound. Doppler examination of the visualized portion of the SV may help to avoid such errors by identi-fying abnormal flow. In another case in which ultrasound suggested SV compression, the angiogram showed no abnormality. This may be a result of the fact that the flattening of the vein seen ultrasonographically was in the anterior-posterior direction, which is difficult to appreciate on an AP angiogram if compression in the cephalad-caudal direction is not also present.
In one of the three cases in which the ultrasound disagreed with the angiogram regarding extent of venous involvement, the angiogram incorrectly suggested obstruction, whereas the ultrasound finding of compression was verified surgically. In the other two cases, one showed occlusion by ultrasound and compression by angiography, and the other vice versa; no cause for the discrepancies between the two studies could be identified. A possible explanation is that a venous collateral was mistaken for a narrowed main

JOURNAL OF CLINICAL ULTRASOUND
Doppler study would be helpful in distinguishing compression from occlusion if a patent venous channel is identified proximal or distal to the site of abnormality.
The results of this comparative study indicate that although angiography remains the more accurate modality, careful ultrasound evaluation can replace angiography for demonstration of the portal system in many cases. With the addition of pulsed Doppler, the accuracy of ultrasound could be increased further. With greater emphasis on examination of the portal system during the ultrasound study, the percentage of incomplete or inadequate studies of the portal system can be significantly reduced and could fall below the 15% value noted during the final phase of our series. For the detection of venous occlusion only, ultrasound was less sensitive than angiography but was still quite specific. Thus, when ultrasound shows occlusion of the portal vein or occlusion of both the superior mesenteric and splenic veins, angiography can be avoided, and the surgeon may plan on palliative surgery rather than total resection. In cases in which only compression is seen or where the surgeon feels that the occlusion seen still might allow resection, computed tomography or angiography may be used to help estimate the extent of vascular involvement. Computed tomography (CT), which is normally performed anyway, frequently adds useful information about the portal system, especially if done with bolus IV contrast enhancement.
We believe that the best approach is to perform ultrasound and CT first. If no other signs of unresectability are seen, and these modalities do not clearly demonstrate the extent of portal involvement or disagree, then angiography should be employed. If arteriography is performed early in the work-up but good opacification of the portal venous system is not achieved, ultrasound can be used to aid the venous evaluation. Since careful examination of the portal system requires additional time, busy ultrasound clinics may find it easier to reschedule those patients with a pancreatic mass seen by CT or ultrasound for a portal system evaluation rather than to try to carry out the examination at the time of the initial screening ultrasound evaluation. Wider acceptance of ultrasound as a major part of the preoperative evaluation of the portal system will depend on careful attention to portal venous structures during the ultrasound examination plus a willingness to report findings of tumor involvement or normality to the surgeons.