AIMS: To determine the minimum number of calibrant plasmas required to assign an accurate System International Sensitivity Index (ISI). METHODS: Eighteen lyophilised adsorbed and eight frozen coumarin plasmas were despatched to 57 participating laboratories. Each centre determined the prothrombin time of all of the plasmas using their routine system of coagulometer and thromboplastin reagent and provided their system mean normal prothrombin time (MNPT). For each laboratory, the slope of the line relating local system results, including MNPT, to calibrated values obtained with BCT 441, RBT 1303 or RBT 90 was derived by orthogonal regression. The number of adsorbed calibrant plasmas included in the calculation was reduced, two at a time, from 18 to two. The validity of the System ISI was tested using the frozen coumarin plasmas. RESULTS: All systems tested gave higher International Normalised Ratios (INR) than those obtained manually with the reference thromboplastins. All systems therefore required correction. No changes were seen in the System ISI as the number of calibrant plasmas was reduced from 18 to 12. Below 12, inconsistent results were obtained with some systems and were dependent on plasma selection. CONCLUSIONS: Although System ISI could be assigned using as few as two calibrant plasmas, to avoid imprecise calibration, a minimum of 12 calibrant plasmas is recommended.