Criteria for monitoring carcinoembryonic antigen: variability of sequential assays at elevated levels.

Fifty-five patients with advanced cancer and elevated plasma carcinoembryonic antigen (CEA) levels greater than 20 ng/mL (Roche assay) were monitored by clinical parameters for disease activity and a daily plasma specimen was obtained and stored frozen. In 45 patients with evaluable metastatic lesions, 35 were stable; five had progressive disease; and five had regressive disease. Plasma CEA in patients with stable disease showed an overall coefficient of variation of 13%. The CV did not differ according to various quantitative CEA levels from less than 100 ng/mL to greater than 1,000 ng/mL. The coefficient of variation in responding and progressive disease patients ranged from 13% to 63%. An analysis of CEA variability relative to the baseline CEA level was possible using the formula square root 2 times the variability about the mean; this yields a value of +/- 36% representing the range within which approximately 95% of sequential CEA levels would lie in the absence of a clinical change in disease. In 225 CEA determinations in stable disease patients, 6% demonstrated an increase beyond this level (36%) and none demonstrated a decrease of 36% or more from the baseline level. This study establishes guidelines for the boundaries of change in plasma CEA that may be applied as a criterion (in conjunction with standard objective disease parameters) for determination of tumor response to therapy.