The eligible subjects were mechanically ventilated adults (age ≥ 18 years) with circulatory shock and without spontaneous respiratory efforts, for whom the attending physician decided to perform a fluid challenge, where circulatory shock was defined as the presence of one or more of the following signs: 1) systolic arterial pressure < 90 mmHg, mean arterial pressure < 65 mmHg, or requiring vasopressor administration; 2) skin mottling; 3) urine output < 0.5 mL/kg/h for ≥ 2 h; 4) lactate level > 2 mmol/L. Patients would be excluded if they met the following criteria: no indwelling arterial or central venous catheterization, aortic valve surgery, equipped with extracorporeal membrane oxygenation or a pacemaker, contraindications to fluid challenge, poor echogenicity, atrial fibrillation, refractory shock expected to die within 24 h, or decline to participate.

All eligible patients received invasive radial arterial monitoring and central venous catheterization, with the catheter tip positioned in the superior vena cava or the right atrium. Pressure calibration was performed in the supine position, with pressure transducers zeroed at the phlebostatic axis, a position corresponding to the right atrium's level (the midpoint of the fourth intercostal space at the midaxillary line) (13). A pressure-controlled ventilation mode was set, and sedative and analgesic medications were continuously administered to avoid spontaneous breathing efforts. Once enrollment was confirmed, a baseline set of hemodynamic variables was measured, and transthoracic echocardiography (TTE) was performed. Meanwhile, arterial and central venous blood gases were simultaneously sampled and analyzed using a GEM Premier 3,500 blood gas analyzer (Instrumentation Laboratory Company, Bedford, MA, USA). Immediately after that, a fluid challenge test was conducted by administering a pressurized bolus of 500 mL of Ringer's solution over 15 min in the 45° semi-recumbent position. Immediately after the fluid challenge, a second set of the above measurements was taken. During the study period, no adjustments were made to body position, ventilator settings, vasopressors, inotropes, or sedative and analgesic drugs. Figure 1 illustrates the detailed process of this study.

We collected demographic information (including age, gender, body mass index, and comorbidities), causes of shock, ventilator-related parameters (including tidal volume, positive end-expiratory pressure, driving pressure, respiratory rate, and fraction of inspired oxygen), acute physiology and chronic health evaluation II score, sequential organ failure assessment score, sedative and analgesic drugs, and vasoactive agents at the time of enrollment. We measured and recorded hemodynamic variables [including heart rate (HR), central venous pressure, systolic arterial pressure, diastolic arterial pressure, and mean arterial pressure], echocardiographic parameters [including stroke volume (SV), cardiac index, and velocity-time integral (VTI)], arterial and central venous blood gases parameters (including potential of hydrogen (PH), arterial lactate level, arterial partial pressure of carbon dioxide (PaCO₂), arterial oxygen saturation (SaO₂), central venous partial pressure of carbon dioxide (PcvCO₂), and ScvO₂), and oxygen-CO₂ derived variables (including DO₂, VO₂, and P(cv-a)CO₂) at baseline and after the fluid challenge. The hemoglobin (Hb) concentration was measured together with arterial blood gas analysis using the GEM Premier 3,500 blood gas analyzer (Instrumentation Laboratory Company, Bedford, MA, USA). Patients were followed up until ICU discharge.

SV was computed as: VTI × LVOT area, where VTI was measured via continuous Doppler transaortic flow on an apical five-chamber view, and LVOT area was calculated as π × (LVOT diameter/2)² (LVOT refers to the left ventricular outflow tract). Then, cardiac index was calculated as (SV × HR)/body surface area. TTE examination was conducted with the CX50 ultrasound system (Philips Medical System, Suresnes, France), which was performed by an independent ICU physician who was blinded to the study outcomes. The representative value of echocardiographic parameters was obtained by averaging three consecutive measurements, regardless of the respiratory cycle.

ΔScvO₂ and ΔP(cv–a)CO₂ induced by volume expansion were calculated as absolute changes, that is, subtracting the baseline value from the value after volume expansion. The fluid-induced changes in cardiac index and VO₂ were calculated as relative changes: (the value after fluid infusion – the baseline value)/ the baseline value×100%. Fluid responsiveness and VO₂ responsiveness were defined by a ≥10% increase in cardiac index and VO₂ in response to volume expansion, respectively. According to the Fick principle, DO₂ was calculated as (SV × HR) × (1.34 × Hb × SaO₂ + 0.003 × PaO₂), and VO₂ was calculated as (SV × HR) × [(1.34 × Hb × SaO₂ + 0.003 × PaO₂) − (1.34 × Hb × ScvO₂ + 0.003 × PcvO₂)], where PaO₂ is the arterial oxygen tension, and PcvO₂ is the central venous oxygen tension.

Statistical analyses were performed using SPSS version 17.0 (IBM, New York, USA). The normal distribution of continuous variables was assessed using the Kolmogorov–Smirnov test. Normally distributed variables are presented as means ± standard deviation (SD), and skewed variables are reported as medians with interquartile ranges (IQR). Categorical variables are described as frequencies and percentages. For the continuous data, either the Student's t-test or the Mann–Whitney test was used for inter-group comparison, depending on the data distribution, and the Student's paired t-test was applied for the intra-group comparisons. The Chi-squared test or Fisher's exact test was utilized to compare categorical variables. The Spearman's rank correlation coefficient (rho) was computed to evaluate the association between the fluid-induced changes in cardiac index and ΔScvO₂ and ΔP(cv–a)CO₂.

The MedCalc Statistical Software (MedCalc Software bvba, Ostend, Belgium) was employed to construct ROC curves for assessing the diagnostic accuracy of ΔScvO₂ and ΔP(cv–a)CO₂ for fluid responsiveness. The optimal cutoff value was determined by maximizing the Youden index, while taking into account the smallest detectable differences (SDD) of ScvO₂ (±3%) and P(cv–a)CO₂ (±2 mmHg), as previously reported (14). Additionally, the gray zone approach was used to avoid the binary constraint of a “black-or-white” decision of the optimal cutoff value (15). We calculated the gray zone for ΔScvO₂ and ΔP(cv–a)CO₂ based on values that did not allow for having 10% of diagnosis tolerance (i.e., a sensitivity of <90% or a specificity of <90%) (15). To identify potential factors affecting the diagnostic accuracy, we performed subgroup analyses based on baseline ScvO₂ value (≥70% or <70%), baseline P(cv–a)CO₂ value (≥6 mmHg or <6 mmHg), and fluid-induced VO₂ responsiveness (yes or no). The DeLong's test was used to determine the difference in AUC between subgroups with a minimum calculated sample size (16).

The Power Analysis and Sample Size software (NCSS, LLC, Kaysville, UT, USA) was utilized to determine the statistical power. Previous studies indicated that both ΔScvO₂ and ΔP(cv–a)CO₂ had an area under the receiver operating characteristic (ROC) curve (AUC) of approximately 0.8 (2, 3). To achieve a power of 80% with an alpha risk of 0.05, it was determined that 26 subjects would be sufficient. Therefore, at least 52 patients were required to ensure adequate statistical power for each arm in the subgroup analyses. In addition, we randomly selected 10 patients to calculate the coefficient of variation (CV) and the least significant change (LSC) to assess the intra-operator reproducibility for VTI. A two-tailed P-value of less than 0.05 was considered statistically significant.

Before the fluid challenge, no significant differences in the baseline hemodynamic variables were observed between the responders and non-responders. After the fluid challenge, SV, cardiac index, and DO₂ were remarkably increased in the responders, but not in the non-responders. Hemoglobin was significantly decreased after fluid expansion in both groups. In the responders, volume expansion led to an elevated ScvO₂ and a reduced P(cv–a)CO₂, whereas these values remained unchanged in the non-responders. Table 2 shows the fluid-induced changes in hemodynamic variables in detail.

Spearman correlation analyses revealed that the fluid-induced changes in cardiac index were positively correlated with ΔScvO₂ (rho = 0.36, P = 0.006) and were negatively correlated with ΔP(cv–a)CO₂ (rho = −0.35, P = 0.006) (Figure 2).

ΔScvO₂ and ΔP(cv-a)CO₂ defined fluid responsiveness with AUC values of 0.76 (95% CI: 0.63–0.86; P < 0.001) and 0.72 (95% CI: 0.59–0.83; P < 0.001) (Figure 3), respectively. Based on the Youden index, the optimal cutoff value of ΔScvO₂ was 2%, with a sensitivity of 76.7% and a specificity of 75.0%. However, considering the repeatability of ScvO₂ (a SDD of ±3%), the optimal cutoff value was 5%, yielding a sensitivity of 50%, a specificity of 92.9%, a positive predictive value (PPV) of 88%, and a negative predictive value (NPV) of 63%. The gray zone approach identified a ΔScvO₂ range of −3% to 4.6%, which included 51.7% of the patients (Figure 3). According to the Youden index, the optimal cutoff value of ΔP(cv-a)CO₂ was −2 mmHg, which exceeded the SDD of P(cv-a)CO₂ (±2 mmHg). Thus, the optimal cutoff value of −2 mmHg yielded a sensitivity of 50%, a specificity of 82.1%, a PPV of 75%, and an NPV of 61%. A range of −2.7 mmHg to 1 mmHg represented the gray zone for ΔP(cv-a)CO₂ that comprised 48.3% of patients (Figure 3).

Given the minimum calculated sample size, the findings from the subgroup analysis with a sample size of 26 cases or more were considered statistically valid. Subgroup analyses demonstrated that ΔScvO₂ potentially exhibited better accuracy for assessing fluid responsiveness in VO₂ non-responders and patients with a baseline ScvO₂ < 70%. Meanwhile, the diagnostic accuracy of ΔP(cv-a)CO₂ was slightly improved in VO₂ non-responders and patients with a baseline P(cv-a)CO₂ ≥ 6 mmHg (Table 3). The AUC of ΔScvO₂ in the subgroup with a baseline ScvO₂ < 70% was slightly higher than that in the subgroup with a baseline ScvO₂ ≥ 70%, despite no statistical significance (P = 0.219). However, the comparisons of AUC in other subgroups were not conducted because the minimum sample size in some subgroups was not reached.