Objective: To determine the blood discard volume, as a multiple of deadspace, that is required for accurate arterial blood gas and electrolyte testing from arterial catheters.

Design: Prospective, controlled, crossover trial.

Setting: Eighteen-bed intensive care unit of a metropolitan teaching hospital.

Patients: A total of 84 critically ill patients with a 20-gauge, radial arterial cannulae, pressure monitoring transducer set, and stable oxygenation.

Interventions: System deadspace (priming volume from sampling port to catheter tip) was established. Patients had six 0.5-mL arterial blood samples taken sequentially in random order using discard volumes of 1, 1.5, 2, 2.3, and 3.6 times the deadspace (experimental values) and 5.5 times the deadspace (control).

Measurements and Main Results: Samples were analyzed for Pao2, Sao2, pH, Paco2, HCO3 −, Na+, and K+. We performed repeated-measures analysis of variance with post hoc linear contrasts and compared mean experimental and control values. The smallest discard volumes that provided measurements that were statistically equal to control were twice the deadspace (Pao2, p = .563; Sao2, p = .371) and 3.6 times the deadspace (pH, p = .107; Paco2, p = .519; HCO3 −, p = .10). All discard volumes tested provided results that were statistically different from control for Na+ (p < .003) and K+ (p < .001).

Conclusions: Many results were statistically different from control, although the actual discrepancies were very small. At clinically relevant levels of measurement, there was minimal variation between values obtained after a discard volume of twice the deadspace and control values. The level of error was clinically acceptable and within or close to the precision limits of the blood gas analyzer. Slight fluctuation in patient variables during sampling could also have contributed to the error. A blood discard volume of twice the deadspace is recommended for all variables. This will provide clinically accurate results and avoid the deleterious effects of unnecessary blood loss.