Despite technical progress in liver radioisotope scintigraphy, the overall accuracy in detecting metastases has not improved. It has been suggested that a fixed scintigram interpretation in reply to changing clinical problems is partly responsible. An evaluation of the possibilities of improving the clinical value of the test by using six cutoff levels in the scintigraphic interpretation is presented. The test population is a prospectively planned consecutive series of 185 patients strongly suspected of having gastric or colo-rectal malignancies. The incidence of liver metastases was 15%. The relationship between the predictive value of a positive test and the false negative ratio at different incidences of metastases is demonstrated graphically. To improve the clinical value of liver scintigraphy, a close cooperation between clinicians and physicians in nuclear medicine to define the clinical problem and the appropriate scintigraphic cutoff level is necessary.