Twenty-nine surgically resected specimens of pancreatic head cancer were examined from a clinicopathological stand point. Serial sections 3 to 5 mm in thickness were cut from each specimen to demonstrate the precise histopathological extent of cancer. Tumor size histologically determined was then divided into several groups as follows: t1 (less than 2 cm in diameter), t2a (2.1-3.0 cm), t2b (3.1-4.0 cm), t3 (4.1-6.0 cm) and t4 (greater than 6.1 cm). The number of cases in each group was 3 in t1, 10 in t2a, 9 in t2b, 6 in t3 and 1 in t4. Tumor size was not always correlated with prognosis and it was necessary to consider other important influential factors, namely, invasion to the serosa of the pancreas (s factor), retroperitoneal infiltration beyond the pancreas (rp factor), and regional lymph node metastasis (n factor). All patients with a positive s factor died less than 2 years after surgery, while the survival rate of s factor-negative patients was 41% at 2 years and 8% at 5 years. Survival rate of patients with positive n factor was 13% at 2 years and 6% at 5 years, and that of n factor-negative patients was 71% at 2 years and 33% at 5 years. Survival rate of patients with positive rp factor was 21% at 2 years, while that of rp factor-negative patients was 75% at 2 years and 50% at five years. Positive rp factor patients comprised 100% of t1, 80% of t2a, 89% of t2b and 83% of t3. This suggested that rp factor was the most important factor even in cases of t1. Furthermore, the frequency with which retroperitoneal surgical margins of specimens were infiltrated ranged from 64% of patients with extended retroperitoneal dissection by the translateral retroperitoneal approach, to 100% of those without such procedure. These results indicate that retroperitoneal radical dissection is basically necessary for pancreatic cancer and that a newer approach combined with some other modality also needs to be done.