Precision of the transpulmonary thermodilution measurements

This prospective study was conducted in the 15-bed intensive care unit of a university hospital. As approved by the Institutional Review Board of our institution, patients were included according to an emergency procedure. A deferred informed consent was asked from the patient's surrogate as soon as possible. As he/she recovered consciousness, a deferred informed consent was asked from the patient. If the patient or his/her next of kin refused to consent, the patient's data were not entered into analysis.

Patients were included if they had a femoral arterial catheter (Pulsiocath PV2015L20N, Pulsion Medical Systems, Munich, Germany) and an internal jugular catheter in place and were routinely monitored by a PiCCO2 device (Pulsion Medical Systems, Munich, Germany). Patients were excluded if they were less than 18 years old and if haemodynamic instability did not allow the mean arterial pressure to remain stable (changes by more than 10%) during at least five minutes before starting the study. Patients with cardiac arrhythmias were not excluded. No patient had a pacemaker.

Immediately after inclusion, one of the investigators (RP) injected five successive cold boluses, each according to the manufacturer's recommendation [7]. For each bolus, we injected 15 mL 0.9% saline at 6°C through the distal port of the internal jugular catheter. The injection was performed as rapidly as possible, irrespective of the respiratory cycle. The injectate temperature was carefully checked to be <6°C for all boluses, as displayed by the PiCCO device. For ensuring that boluses were <6°C, we used two packs of saline, one frozen and one at 6°C. For each bolus, we sampled 20 mL from the 6°C saline pack, injected it into the iced pack and re-sampled 15 mL from this saline that had been cooled by the contact with ice. These 15 mL were used for performing the bolus. The thermodilution curve recorded by the arterial thermistance was automatically analyzed by the PiCCO2 device, allowing obtaining the value of cardiac output, of GEDV indexed for body surface (GEDVi) and of EVLW indexed for predicted body weight (EVLWi). The five boluses were performed one after another, as soon as blood temperature had returned to its baseline value, as indicated by the device. The values of CI, GEDVi and EVLWi obtained from each thermodilution were collected. No thermodilution curve was rejected from analysis. Treatments were kept unchanged and patients were not mobilized during the study period. All measurements were performed by the same operator (RP).

In each patient, we calculated the coefficient of variation (CV) of the TPTD variables (CI, GEDVi and EVLWi). The CV is a normalized measure of dispersion of a probability distribution. It was calculated as being the standard deviation divided by the mean of the five measurements. This relatively large number of measurements thus allowed obtaining a reliable value of CV. The coefficient of error (CE) was obtained by using the formula CE = CV/√n, were n was the number of replicates of measurements in each patient. The precision was calculated as being two CV for a single measurement and two CE for averaged measurements. It is usually considered that a measurement precision level ≤10% is desirable [8]. The least significant change (LSC) is the minimum change that needs to be measured by a device in order to recognize a real change of measurement [9]. The LSC was calculated using the following equation: LSC = CE × 1.96 × √2.

A hundred different patients were initially included in the study. Their characteristics are detailed in Table 1. In nine patients, the TPTD could not measure any variable, likely due to low CI, as suggested by the values observed during the preceding hours. The PiCCO device was instituted due to septic shock in 80% of patients in all patients and it was in place since 28 (5 to 50) hours. Atrial fibrillation was observed in 20% of patients (Table 1). Eighty-three percent of patients were under mechanical ventilation and 74% were sedated. A spontaneous breathing activity was observed in 23 (26%) patients. Eight patients were under continuous veno-venous hemofiltration for renal replacement therapy. The dialysis catheter was inserted in the jugular vein and the blood pump flow was 300 mL/minute.

Considering five boluses, the average CI was 3.5 (2.7 to 4.3) L/minute/m², the GEDVi was 812 (708 to 932) mL/m² and the EVLWi was 9 (8 to 14) mL/kg (Table 1). The precision for one single measurement was 14 (10 to 21)% for CI, 15 (10 to 24)% for GEDVi and 15 (11 to 24)% for EVLWi and the precision for CI was ≥10% in 66 patients. If two boluses were used for TPTD, the precision was reduced to 10 (7 to 15)% for CI, 10 (7 to 17)% for GEDVi, 10 (8 to 17)% for EVLWi (Figure 1) and the precision for CI was ≥10% in 47 patients. If three boluses were used for TPTD, the precision was reduced to 8 (6 to 12)% for CI, 8 (6 to 14)% for GEDVi, 8 (7 to 14)% for EVLWi (Figure 1) and the precision for CI was ≥10% in 33 patients.

The LSC (for one single measurement) was 20 (14 to 29)% for CI, 20 (14 to 34)% for GEDVi and 20 (16 to 33)% for EVLWi. If two boluses were used for TPTD, the LSC was reduced to 14 (10 to 21)% for CI, 14 (10 to 24)% for GEDVi and 14 (11 to 23)% for EVLWi (Figure 2). If three boluses were used for TPTD, the LSC was 12 (8 to 17)% for CI, 12 (8 to 19)% for GEDVi and 12 (9 to 19)% for EVLWi (Figure 2).

Excluding the eight patients under continuous veno-venous hemofiltration did not significantly change CV, neither for CI (8 (6 to 12) vs. (8 (6 to 12) for three measurements) nor for GEDVi (8 (5 to 13) vs. 8 (5 to 13) for three measurements) and EVLWi (8 (6 to 14) vs. 8 (6 to 14) for three measurements). Excluding the 23 patients with spontaneous breathing activity did not significantly change CV, neither for CI (8 (5 to 12) vs. (8 (6 to 12) for three measurements) nor for GEDVi (8 (5 to 13) vs. 8 (5 to 13) for three measurements) and EVLWi (8 (6 to 13) vs. 8 (6 to 14) for three measurements). There was no significant correlation between the dose of norepinephrine and the precision of CItd (P = 0.06), GEDVi (P = 0.10) and EVLWi (P = 0.91).