Abstract Introduction We wanted to determine the number of cold bolus injections that are necessary for achieving an acceptable level of precision for measuring cardiac index (CI), indexed global end-diastolic volume (GEDVi) and indexed extravascular lung water (EVLWi) by transpulmonary thermodilution. Methods We included 91 hemodynamically stable patients (age 59 (25% to 75% interquartile range: 39 to 79) years, simplified acute physiologic score (SAPS)II 59 (53 to 65), 56% under norepinephrine) who were monitored by a PiCCO2 device. We performed five successive cold saline (15 mL, 6°C) injections and recorded the measurements of CI, GEDVi and EVLWi. Results Considering five boluses, the coefficient of variation (CV, calculated as standard deviation divided by the mean of the five measurements) was 7 (5 to 11)%, 7 (5 to 12)% and 7 (6 to 12)% for CI, GEDVi and EVLWi, respectively. If the results of two bolus injections were averaged, the precision (2 × CV/√ number of boluses) was 10 (7 to 15)%, 10 (7 to 17)% and 8 (7 to 14)% for CI, GEDVi and EVLWi, respectively. If the results of three bolus injections were averaged, the precision dropped below 10%, that is, the cut-off that is generally considered as acceptable (8 (6 to 12)%, 8 (6 to 14)% and 8 (7 to 14)% for CI, GEDVi and EVLWi, respectively). If two injections were performed, the least significant change, that is, the minimal change in value that could be trusted to be significant, was 14 (10 to 21)%, 14 (10 to 24)% and 14 (11 to 23)% for CI, GEDVi and EVLWi, respectively. If three injections were performed, the least significant change was 12 (8 to 17)%, 12 (8 to 19)% and 12 (9 to 19)% for CI, GEDVi and EVLWi, respectively, that is, below the 15% cut-off that is usually considered as clinically relevant. Conclusions These results support the injection of at least three cold boluses for obtaining an acceptable precision when transpulmonary thermodilution is used for measuring CI, GEDVi and EVLWi.