We compared determinations of cardiac output using various combinations of injectate volumes and temperatures to results obtained with 10 ml of iced (0 degrees C) injectate (standard technique) in 33 critically ill patients. The use of a 10-ml injectate at room temperature resulted in comparable reproducibility (12.7 vs 10.8 percent; not significant) and a small and nonsignificant error (-0.013 +/- 0.543 L/min). Five milliliters at room temperature resulted in markedly decreased reproducibility (17.9 vs 8.9 percent; p less than 0.05); however, the error associated with the technique was still not significant (0.136 +/- 0.829 L/min). When a 5-ml injectate at 0 degrees C was used, a reproducibility comparable to that of 10 ml at 0 degrees C was obtained (12.3 vs 7.5 percent; not significant). The results underestimated values obtained with 10 ml at 0 degrees C (-0.360 +/- 0.857 L/minute); however, the difference did not reach statistical significance. The use of 3 ml at 0 degrees C was associated with a substantial increase in variability, with a coefficient of variation of 32.0 percent (10.4 percent for 10 ml of iced injectate; p less than 0.01); however, the differences between the average value of cardiac output obtained with this technique and the standard technique were only minimal (error, -0.063 +/- 0.455; not significant). We reached the following conclusions: (1) the use of 10 ml at room temperature and 5 ml at 0 degrees C as the indicator for thermodilution determinations of cardiac output results in small and insignificant differences in reproducibility and accuracy from the standard technique; (2) five milliliters at room temperature and 3 ml at 0 degrees C are associated with markedly decreased reproducibility; however, the error in values for cardiac output obtained with these techniques is not statistically significant; and (3) the ability to use injectate at room temperature and in small volume should substantially simplify the technique, lowering its cost, and should prevent volume overloading.