25, 9, 30
The recirculation measurement looks genuinely useful and is the more valuable half of this technique. The cardiac output is not usable: limits of agreement of roughly plus or minus 3 L/min around an output of 3 L/min describe a method that cannot distinguish a good heart from a failing one. Accuracy without precision is not a measurement, and the headline bias figure is far more reassuring than the interval.
High-precision ultrasonic flow probes, with Bland-Altman analysis
DOI 10.1097/ALN.0000000000004895
Cold bolus injected into the arterial ECMO limb with thermistors in both circuit limbs and the pulmonary artery, giving recirculation fraction by area under the curve and by bolus-volume partitioning, and cardiac output from the partition
In vitro only; no patients; performance degrades exactly as recirculation rises, which is when the measurement is most wanted. Abstract only. [VERIFICATION REQUIRED]
Recirculation fraction was accurate and precise: bias -5.4% with limits of agreement -18.6 to 7.9% by area under the curve, and -5.9% with -18.8 to 7.0% flow-based. Cardiac output was accurate but imprecise: bias 0.56 L/min with limits of agreement -2.27 to 3.4 L/min (area-under-the-curve method) and 0.48 L/min with -2.22 to 3.19 L/min (flow-based). A rising recirculation fraction increased bias and decreased precision.
In vitro simulation: two parallel ECMO circuits sharing a right atrium, with added limbs for recirculation and pulmonary shunt; ECMO flows 1 to 2.5 L/min, cardiac outputs 2.5 to 3.5 L/min, recirculation fractions 0 to 80%