Surgical and radiological decompression in malignant biliary obstruction: a retrospective study using multivariate...

We performed a retrospective computer-aided statistical study of 228 patients who received treatment for malignant biliary obstruction to compare survival rates after surgical or radiological biliary decompression. To adjust for the selection bias produced by the clinical choice of surgical vs. catheter drainage, we examined differences in survival rates after controlling for various clinical and laboratory factors, which are predictive of short and long-term survival. These factors were selected by multivariate analysis. Short-term survival was significantly correlated with preoperative metabolic status as reflected in serum albumin, and blood urea nitrogen (BUN) levels. Elevated serum bilirubin values had no additional negative effect on survival. Long-term survival was influenced mainly by the extent of the primary cancer and the preoperative presence of leukocytosis. Although there was a difference in the raw data for hospital mortality between patients who underwent surgical drainage (11%) and those who underwent radiological drainage (30%), we found no real effect produced by the form of therapy after controlling for prognosis variables, i.e., eliminating selection bias. We propose a simple statistical model to predict short-term hospital survival. The application of objective risk factor analysis according to accepted statistical methods should permit the assessment of new radiologic therapeutic techniques with greater clinical validity.