How accurate is helical computed tomography for clinical staging of pancreatic cancer?

BACKGROUND: Our goal was to determine if findings on an index computed tomo graphy (CT) scan would correlate with survival in patients with pancreatic adenocarcinoma. We know that as this tumor extends out of the gland, surviv al decreases. Are there any CT findings that assess tumor extension suffici ently that also correlate with survival? Once identified, these CT areas wo uld be the best factors to clinically stage patients. METHODS: Between 1993 and 1997, 160 patients with biopsy-proven adenocarcin oma of the pancreatic head were included if an index helical CT scan and cl inical follow-up were available. All CT scans were reviewed by the same rad iologist blinded for outcomes. CT scans were interpreted using a graded ext ension of tumor out of the pancreatic head in four areas: retroperitoneum ( RP); anterior pancreatic capsule (S); portal/superior mesenteric veins (PV) ; and celiac/superior mesenteric arteries (A), Extension of tumor was grade d as follows: Grade 0 (negative margin); 1 (suspicious); 2 (positive); or 3 (extensively involved). Also recorded and graded were signs of metastases: nodal enlargement greater than or equal to 1.5 cm (N); and lesions consist ent with hepatic metastases (H), Survival was compared between grades for e ach CT area using the methods of Kaplan and Meier and relative risk estimat es of death (Cox regression models). RESULTS: Compared with grade 0, the following CT areas had significantly de creased survival curves: grade 1 (only S and A), grade 2 and 3 (RP, PV, S, A). N and H did not correlate with survival unless greater than or equal to 1.5 cm nodes were in the liver or splenic hilum or there were multiple liv er nodules. CONCLUSION: Although postoperative microscopic H or N involvement is a reli able prognostic sign, only extensive CT involvement of H or N predicts surv ival preoperatively. A better CT finding that predicts decreased survival p reoperatively was extension out of the pancreatic head (especially S or A). Clinical methods of staging should use CT areas such as S, A, PV, and RP, and not H and N. Am J Surg. 1999;177:428-432. (C) 1999 by Excerpta Medica, Inc.