The authors reviewed their institutional experience with liver resection for metastatic colorectal carcinoma to (1) determine whether perioperative blood transfusion affects survival; (2) identify prognostic determinants; and (3) estimate the patient requirement for a prospective randomized trial designed to demonstrate efficacy of liver resection. Two hundred eighty consecutive patients treated by potentially curative liver resection between 1960 and 1987 were included. Data were obtained for all but 10 patients for at least 5 years after operation or through 1990. Actuarial survival curves related to potential prognostic determinants were analyzed with the log-rank test. Overall, survival was 47 +/- 3% at 3 years and 25 +/- 3% at 5 years, including 4% 60-day operative mortality rate. Eighty-one patients who did not receive blood 7 days before to 14 days after operation had 60 +/- 6% 3-year and 32 +/- 6% 5-year survival compared with 40 +/- 4% and 21 +/- 3% survival rates for 183 patients who received at least one unit (p = 0.03, operative deaths excluded). Extrahepatic disease (p = 0.015), extrahepatic lymph node involvement (p = 0.002), satellite configuration of multiple metastases (p = 0.0052), and initial detection by abnormal liver enzymes (p = 0.0005) were associated with poor survival rates. Synchronous presentation of metastatic and stage B primary disease was associated with a favorable prognosis (p = 0.003). The requirement for a prospective randomized trial estimated by an exponential survival model would be 36, 74, 168, or 428 patients if 5-year survival without resection were 1, 5, 10, or 15%. We conclude that (1) perioperative blood transfusion may be adversely associated with survival; (2) extrahepatic disease, extrahepatic lymph node involvement, satellite configuration, and initial detection by clinical examination or a liver enzyme abnormality portend a poor prognosis; and (3) a prospective randomized trial of liver resection is impractical because of the large patient requirement, at least by a single institution.

To assess the frequency and significance of complete resolution of inflammatory activity following corticosteroid therapy, 115 patients with severe HBsAg-negative chronic active hepatitis were followed regularly for 84 +/- 5 months. Of 83 patients eligible to revert to normal liver tissue, 18 did so after 56 +/- 8 months. Five of the 18 relapsed after treatment withdrawal. Only patients who improved spontaneously after cessation of treatment from histologic features of chronic persistent hepatitis to normal invariably sustained the improvement. Of 32 patients with cirrhosis at presentation, 17 reverted to inactive cirrhosis after 48 +/- 10 months, but only 3 remained inactive after discontinuation of treatment. Mortality was similar in patients with and without reversion to normal tissue (0 vs. 14%, p greater than 0.2), but the frequency of relapse was less after complete resolution (28 vs. 76%, p less than 0.001). Reversion to inactive cirrhosis did not improve survival or reduce relapse frequency after remission and treatment withdrawal. Findings prior to therapy did not predict outcome. We conclude that complete resolution of inflammatory activity is possible, but that it occurs slowly, infrequently and unpredictably after therapy. In patients without cirrhosis, reversion to normal liver tissue decreases the likelihood of relapse and the requirement for retreatment. In patients with cirrhosis at presentation, elimination of inflammatory activity is rarely sustained and does not improve prognosis after remission and treatment withdrawal.

Incidence rates of blood transfusion for "causal" planning of blood collections are presented here for the first time. The probability of receiving a transfusion of RBCs in any year rises by 20-fold from the rate in those less than 40 years old to that in those more than 65 years old, who receive 53.3 percent of the red cell units transfused.

Receipt of a blood transfusion can be used as a descriptive epidemiologic index of morbidity in the general population, as it is independently predictive of mortality, adding to the predictive value of age, gender, and previous hospitalization. There is a dose-response relationship between the amount of blood components received and a reduction in the subsequent length of survival. However, when a county's entire population is studied, posttransfusion mortality due to underlying disease is substantially lower than that previously reported in look-back investigations.