The value of clinical assessment for preoperative prediction of liver metastases in suspected gastrointestinal cancer. A prospective, consecutive study.

Surgeons' ability to predict liver metastases preoperatively was studied in 185 operable, nonicteric patients strongly suspected to have gastrointestinal cancer. The prediction was based on a clinical assessment comprising medical history, physical examination and low-cost laboratory tests. A pathoanatomic verification procedure was used. The observed predictive value of a positive test was 0.67 and there was a false-negative ratio of 0.79. At the clinical assessment, presence of metastases was believed to be more common in gastric than in colorectal disease (6/43 vs. 3/142), but the actual incidence of metastases was similar in the two groups. Serum alkaline phosphatase (AP) was a priori believed to be of value in the clinical assessment. But the clinical assessment with inclusion of AP was inferior to AP alone as a predictor of metastases, due to undervaluation of the importance of elevated AP in cases of colorectal disease. The difficulties in standardization of the clinical assessment are underlined.