Two automatic coagulometers, ACL 810 (Instrumentation Laboratory), a laser-nephelometric centrifugal analyzer, and KoaguLab 40 A (Ortho Diagnostics), an optical automatic coagulometer, were compared with the manual tilt-tube method for the performance of activated partial thromboplastin time (APTT). Seven commercial APTT reagents were used for duplicate determinations in 30 normal controls, 26 patients with liver disease, and 33 patients on full-dose heparin treatment. Clotting times were longer with the manual method than with ACL 810 and, to a lesser extent, with KoaguLab 40 A. Average imprecision of duplicate determinations (coefficient of variation [CV]) was less with ACL 810 (less than 1.5%) than with KoaguLab 40 A (2.9%) and the manual method (2.4%). Differences in slope of the regression curves of clotting times obtained with the coagulometers over the tilt-tube method were observed with all the reagents tested (P less than 0.01). Transformation of clotting times of controls, patients with liver disease, and patients on heparin therapy to APTT ratios did not eliminate the bias resulting from the different reagents (P less than 0.001) and clot-detection methods (P less than 0.001); in controls, significant (P less than 0.001) reagent-method interaction was also observed. The in vitro heparin sensitivity differed with the APTT reagents evaluated and was influenced by the clot-detection method used. Transformation of APTT ratios of anticoagulated patients to apparent plasma heparin levels--as derived from in vitro dose-response curves--effectively eliminated the bias resulting from the different clot-detection methods but had no effect on the bias resulting from the different APTT reagents. In vitro heparin activity curves thus have little, if any, relevance for the ex vivo monitoring of heparin treatment.