Between July 16, 2019, and December 9, 2019, there were 167 patients on mechanical ventilation in the ICU, and screening for full eligibility criteria was completed in 132 of them (Fig. 2). Among the 52 eligible patients, we excluded 15 that met the exclusion criteria, leaving 37 patients in the study. In all 37 patients, two EtCO2 measurements and PaCO2 were collected. The median age was 70 years (IQR, 60–79), 57% (21/37) were male, 76% (28/37) were emergency ICU admissions, and the median acute physiology and chronic health evaluation II score was 19 (IQR, 16–25) (Table 1). The major diagnoses were cardiovascular (19% [7/37]), respiratory (11% [4/37]), gastrointestinal (27% [10/37]), and neurological diseases (22% [8/37]).
At the time of EtCO2 measurements, the median number of days from ICU admission was 2 days (IQR, 2–2), and the median time from intubation was 21 hours (IQR, 14–24). Most of the patients had stable vital signs; the median systolic blood pressure was 129 mmHg (IQR, 109–144), body temperature was 36.9°C (IQR, 36.6–37.4), and respiratory rate was 16 breaths per minute (IQR, 14–19). The median partial pressure of arterial oxygen was 93.0 mmHg (IQR, 83.8–124.0), PaCO2 was 38.2 mmHg (IQR, 34.9–41.6), and the ratio of partial pressure arterial oxygen to fraction of inspired oxygen was 364 (IQR, 317–410). The ventilator mode was continuous spontaneous ventilation in 46% (17/37) of the patients. The median fraction of inspired oxygen was 0.25 (IQR, 0.21–0.35) and PEEP was 5 cmH2O (IQR, 5–5). The median ICU stay was 4 days (IQR, 2–7), and the hospital mortality rate was 16% (6/37).
The paired observations on the Y-piece side and the patient side of the HMEF are presented in Table 2. The means of EtCO2 were 37.4 mmHg (SD, 5.5) and 37.5 mmHg (SD, 6.1) on the Y-piece side and the patient side, respectively. The mean difference in EtCO2 (Y-piece side – patient side) was − 0.1 (SD, 1.6). The estimated accuracy difference (ΔY-piece side - Δpatient side) was − 0.14 mmHg (95% CI, -0.58–0.29). The upper limit of the CI for accuracy difference did not exceed the predefined margin of + 1 mmHg, establishing non-inferiority (P for non-inferiority < 0.001) (Fig. 3).
In the Bland-Altman analyses, the mean difference between PaCO2 and EtCO2 on the Y-piece side was 0.99 mmHg (95% CI, -0.66–2.63), and 95% limits of agreement was − 8.67 mmHg (95% CI, -11.51– -5.84) to 10.65 mmHg (95% CI, 7.82–13.49). The mean difference between PaCO2 and EtCO2 on the patient side was 0.87 mmHg (95% CI, -0.80–2.54), and 95% limits of agreement was − 8.93 mmHg (95% CI, -11.81– -6.06) to 10.67 mmHg (95% CI, 7.80–13.55) (Supplementary Information 1: Supplemental Table 1). The estimated 95% limits of agreement were almost identical, indicating that the degree of agreement between PaCO2 and EtCO2 was similar on the Y-piece side and the patient side (Fig. 4). The square root of within-patient variance of EtCO2 was 0.65 mmHg on the Y-piece side and 1.01 mmHg on the patient side, suggesting that the repeatability of the two EtCO2 were sufficient. Histograms of differences between PaCO2 and each of the two EtCO2 measurements showed roughly normal distributions. Graphical inspection revealed that the variations in the differences were constant across the range of measurements (Fig. 4).
The relationship between tidal volume and difference in Y-piece side EtCO2 and patient-side EtCO2 are assessed in Supplemental Fig. 1 (Supplementary Information 1). There was no significant linear relationship (regression coefficient, -0.001 [95% CI, -0.005–0.003], correlation coefficient, -0.096).