Combined assessment of end-tidal CO₂ and pulse pressure variability during passive leg raising for non-invasive prediction of preload responsiveness in septic shock Authors: Xiao-Fen Zhou, Ding-Shen Qiu, Kun Cheng, Xiao-Yan Ding, Bing-Ying Chen, Han Chen Published: 2025 ABSTRACT AbstractBackgroundTraditional assessment of preload responsiveness during passive leg raising (PLR) relies on invasive cardiac output monitoring. We hypothesized that the physiological coupling between cardiac output, end-tidal carbon dioxide (EtCO2), and pulse pressure variability (PPV) could provide a reliable non-invasive alternative for predicting preload responsiveness.MethodsWe conducted a prospective observational study in 71 mechanically ventilated septic shock patients. Pulse index Contour Cardiac Output (PiCCO), arterial pressure, and EtCO₂ were monitored during standardized PLR maneuvers. Cardiac index (CI), arterial pressure, stroke volume variability (SVV), PPV, and EtCO₂ were measured before and after PLR. Patients with a PLR-induced CI increase ≥ 15% were classified as preload responders. Linear regression analysis and Pearson correlation coefficients were calculated to assess the relationships between PLR-induced changes in CI and other parameters. We evaluated the predictive performance of individual parameters and their combination using receiver operating characteristic (ROC) analysis, with the DeLong test used to compare the performance of the combined model to individual predictors.ResultsAmong 71 patients, 37 (52.11%) were preload responders. The absolute change in EtCO₂ (∆EtCO₂), relative change in EtCO₂ (∆EtCO₂%), and PPV were all significantly higher in responders compared to non-responders. A positive correlation was observed between PLR-induced changes in CI and both EtCO₂ (r = 0.538, p < 0.001) and PPV (r = 0.584, p < 0.001). The area under the ROC curve (AUC) for ∆EtCO₂% was 0.88 (95% CI: 0.79–0.95) with an optimal cutoff value of > 5.88% (sensitivity 91.89%, specificity 76.47%). The AUC for PPV was 0.85 (95% CI: 0.75–0.93) with a cutoff value of > 10% (sensitivity 81.1%, specificity 79.41%). Importantly, a novel logistic regression-derived combined PPV-∆EtCO₂% score demonstrated improved predictive performance (AUC 0.94, 95% CI: 0.86–0.98) with a cutoff value of > 0.67 (sensitivity 83.8%, specificity 94.1%).ConclusionThe combination of PPV and ∆EtCO₂% during PLR provides a promising non-invasive method for predicting preload responsiveness in mechanically ventilated septic shock patients, outperforming either parameter alone. This approach offers a promising alternative to invasive cardiac output monitoring in guiding fluid therapy decisions.Trial registrationThe study was registered on December 26th, 2019, at the Chinese Clinical Trial Registry (ChiCTR1900028210). URL of trial registry record:https://www.chictr.org.cn/bin/project/edit?pid=46568.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12871-025-03490-4.Keywords: Preload responsiveness, Passive leg raising, Cardiac output, Pulse pressure variability, End-tidal carbon dioxide Background Fluid administration is critical in managing septic shock, but only around 50% of hemodynamically unstable patients demonstrate improved cardiac output in response to volume expansion (fluid responsiveness) [1]. Inappropriate fluid administration carries risks, including pulmonary edema, prolonged mechanical ventilation, and increased mortality [2, 3]. Recent trials suggest that a restrictive approach to fluid therapy might be beneficial in certain populations [4, 5]. Therefore, accurate prediction of fluid responsiveness is essential for optimizing patient outcomes. The passive leg raising (PLR) test predicts preload responsiveness by inducing a transient increase in venous return, acting as a “self-volume challenge” [6]. Such PLR-induced changes in cardiac output have shown good predictive value across various clinical settings [7–9]. However, a recent international survey revealed that clinicians frequently avoid PLR testing because they perceive it as insufficiently accurate in certain situations, highlighting the need to incorporate additional parameters [10]. Our study aims to address this limitation by integrating two well-established hemodynamic principles: (1) End-tidal carbon dioxide (EtCO2) directly reflects pulmonary blood flow and cardiac output under stable ventilation and metabolism [11]. Changes in EtCO2 during PLR have been shown to correlate with changes in cardiac output (CO) [12, 13]. (2) Dynamic preload responsiveness, indicated by pulse pressure variability (PPV) during mechanical ventilation, which reflects the position on the Frank-Starling curve, with higher values suggesting greater preload responsiveness [14]. Recent consensus statements have reinforced the importance of dynamic parameters in guiding fluid therapy [15]. By integrating both PPV and PLR-induced changes in EtCO2, we propose an assessment approach that provides complementary physiological information: PPV reflects baseline preload dependency, while ΔEtCO2 captures the actual cardiac response to the PLR-induced preload change. We hypothesize that this combined approach will enhance the accuracy of non-invasive preload responsiveness prediction compared to either parameter alone. The aim of this study was to evaluate whether the combination of PPV and PLR-induced changes in EtCO₂ could provide a reliable non-invasive method for predicting preload responsiveness in mechanically ventilated patients with septic shock, potentially eliminating the need for invasive cardiac output monitoring. Methods Study population This prospective observational study was conducted in the intensive care unit (ICU) of Fujian Provincial Hospital from January 2021 to February 2023. The study protocol was approved by the Ethics Committee of Fujian Provincial Hospital (Approval # K2018-01–023) and registered at the Chinese Clinical Trial Registry (ChiCTR1900028210). Written informed consent was obtained from each patient’s next of kin. The research process was performed in accordance with the Declaration of Helsinki. We enrolled adult patients (≥ 18 years) with septic shock according to the Sepsis-3 definition [16] who required mechanical ventilation. To ensure stable baseline measurements, patients were included only if their blood pressure and heart rate fluctuated less than 10% within 15 min before enrollment. All patients were sedated with remifentanil and midazolam. Bolus of remifentanil was used to temporarily eliminate spontaneous breathing efforts during PLR test. Exclusion criteria included: anticipated death within 24 h, contraindications to PiCCO monitoring, use of intra-aortic balloon counter-pulsation or extracorporeal membrane oxygenation, and presence of arrhythmias that would interfere with PPV assessment. Before enrollment, demographic data including age and gender were recorded for all patients. The primary diagnosis and comorbidities were documented based on medical records and clinical evaluation, including conditions such as pneumonia, abdominal infections, and bloodstream infections. The Acute Physiology and Chronic Health Evaluation II (APACHE II) score and the Sequential Organ Failure Assessment (SOFA) score were calculated using the worst values obtained within 24 h of ICU admission. Additionally, we recorded comorbidities, including coronary atherosclerotic heart disease, hypertension, and diabetes. We also collected data on ventilator settings, including tidal volume, total positive end-expiratory pressure (PEEP), peak and plateau airway pressures, and the fraction of inspired oxygen. Ventilatory and hemodynamic measurements All patients were mechanically ventilated in volume-controlled mode with tidal volumes of at least 8 mL/kg predicted body weight. This tidal volume setting was chosen to optimize the reliability of PPV measurements, as previous research has demonstrated that PPV accuracy is highest with tidal volumes ≥ 8 mL/kg [17]. Respiratory rate was adjusted to maintain normocapnia (PaCO₂ 35–45 mmHg), and PEEP was set at 5–10 cmH₂O based on individual requirements. Arterial blood gas analysis was performed to measure pH, partial pressure of arterial oxygen (PaO2), partial pressure of arterial carbon dioxide (PaCO2), and arterial oxygen saturation (SaO2). Hemodynamic monitoring included: (1) Cardiac index (CI) and stroke volume variability (SVV) measured using the PiCCO system (PC 4000, Pulsion, Feldkirchen, Germany); (2) Continuous arterial pressure monitoring via a radial artery catheter, and heart rate (HR), systolic blood pressure (SBP), diastolic blood pressure (DBP), mean arterial pressure (MAP) were collected; (3) Central venous pressure (CVP) via a subclavian venous catheter; (4) End-tidal CO₂ (EtCO₂) measured using a sidestream infrared gas analyzer (M2741A, PHILIPS, Amsterdam, Netherlands). In addition, we also collected the dynamic changes of the indices: ΔHR, ΔCVP, ΔSBP, ΔDBP, ΔMAP, and ΔPP. Changes in end-tidal CO2 were calculated in mmHg (ΔEtCO2 mmHg), and in percentage (ΔEtCO₂%). These indices were calculated based on the measurements recorded at baseline and post-PLR. The PiCCO system was calibrated according to the manufacturer’s instructions before each measurement series. Calibration was performed using the transpulmonary thermodilution technique with three consecutive central venous injections of 15 mL of cold (< 8 °C) 0.9% saline solution. The arterial catheter was zeroed at the midaxillary line, and the arterial waveform quality was carefully checked before each measurement to ensure accurate pulse pressure recording. The system was recalibrated if hemodynamic conditions changed (> 10% change in heart rate or blood pressure) or at least every 8 h to maintain measurement accuracy. Pulse pressure variability (PPV) was calculated according to standard formula: PPV (%) = 100 × (PPmax - PPmin)/[(PPmax + PPmin)/2], where PPmax and PPmin represent the maximum and minimum pulse pressure over a single respiratory cycle. Passive leg raising protocol The PLR test was performed according to a standardized protocol [18]: Baseline measurements were obtained with patients in a semi-recumbent position (45° head-up). Baseline measurements were obtained with patients in a semi-recumbent position (45° head-up). The bed was then automatically adjusted to simultaneously lower the upper body to horizontal and elevate the legs to 45°. The bed was then automatically adjusted to simultaneously lower the upper body to horizontal and elevate the legs to 45°. Post-PLR measurements were recorded at the point of maximum hemodynamic effect (within 60–90 s). Post-PLR measurements were recorded at the point of maximum hemodynamic effect (within 60–90 s). All hemodynamic and respiratory parameters (CI, SVV, PPV, arterial pressures, CVP, and EtCO₂) were recorded at baseline and post-PLR. Patients were classified as preload responders if the PLR-induced increase in CI was ≥ 15%. This threshold was selected based on previous research indicating that a 15% increase in CI represents a clinically response associated with improved outcomes [1, 7]. Statistical analysis Sample size calculation was based on previous studies investigating the diagnostic accuracy of EtCO₂ changes during PLR [12, 13]. Assuming an area under the ROC curve of 0.85, a precision of 0.10, and a preload responder prevalence of 50%, we calculated a required sample size of 70 patients. Continuous variables were tested for normal distribution using the Kolmogorov-Smirnov test and presented as mean ± standard deviation or median (interquartile range) as appropriate. Between-group comparisons were performed using t-tests or Mann-Whitney U tests. Categorical variables were expressed as frequency (percentage) and compared using Chi-square or Fisher’s exact test. Linear regression analysis was performed to examine the relationships between continuous variables and Pearson correlation coefficients were calculated. For regression analyses, assumptions of linearity, normality of residuals, and homoscedasticity were assessed prior to analysis. To develop a combined predictor that integrates both PPV and ΔEtCO2% into a single score for dimensionality reduction, we constructed a logistic regression model with preload responsiveness (≥ 15% CI increase during PLR) as the dependent variable. This approach allows the combination of two independent hemodynamic parameters into one composite index. The logistic regression yielded the following equation: Z = −3.247 + 0.186×PPV + 0.312 × ΔEtCO2%, where 0.186 and 0.312 represent the regression coefficients (B values) for PPV and ΔEtCO2% respectively. The combined score for each patient was calculated as: combined score = 1/(1 + e− Z), providing a probability value between 0 and 1. The weighting in our model indicates that ΔEtCO2% has approximately 1.7-fold greater predictive contribution than PPV (coefficient ratio: 0.312/0.186 = 1.68). During the construction of the logistic regression model, we ensured that the underlying assumptions were met, including linearity of the log-odds, absence of multicollinearity, and independence of observations. Receiver operating characteristic (ROC) curves were constructed to evaluate the diagnostic performance of individual parameters and their combination. The area under the curve (AUC), sensitivity, specificity, positive predictive value, and negative predictive value were calculated. The optimal cutoff values for each predictor were determined using the Youden index, which was calculated as (sensitivity + specificity − 1). The point on the ROC curve with the maximum Youden index value was selected as the optimal threshold for each parameter. AUCs were compared using the DeLong method. Ten-fold cross-validation was performed to assess the robustness of the combined model. Statistical analyses were conducted using SPSS (ver. 25.0, IBM, NY, USA) and MedCalc (ver. 18.2.1, Mariakerke, Belgium). A two-sided p-value < 0.05 was considered statistically significant. Results Patient characteristics We enrolled 71 patients with septic shock who all successfully completed the PLR test. Based on PLR-induced changes in CI, 37 patients (52.11%) were classified as preload responders and 34 (47.88%) as non-responders. Baseline demographic and clinical characteristics were comparable between groups (Table 1). Table 1Characteristics and demographic data of study population (n = 71)Responders (n = 37)Non-responders (n = 34)p valueAge (yr)57 [44, 69]51 [45, 69.3]0.734Lactate (mmol/L)8 [5.3, 12]6 [5, 11]0.164APACHEII15 [9, 21.5]18 [10, 21]0.904SOFA8.1 ± 3.58.9 ± 3.50.364Male (%)25 (67.57)24 (70.59)0.783Infection site Pneumonia13140.123 Abdominal cavity17160.234 Bloodstream1090.512 Others210.432Underlying diseases CAD14100.215 Hypertension1080.139 Diabetes23220.458Respiratory parameters Tidal volume, mL/Kg predicted body weight8.2 ± 0.28.3 ± 0.30.137 Total PEEP, cmH2O6.2 ± 0.56.1 ± 0.40.398 Peak airway pressure, cmH2O21.2 ± 1.420.8 ± 1.00.267 Plateau airway pressure, cmH2O15.0 ± 1.014.6 ± 0.90.151 FiO2, %42.3 ± 6.241.3 ± 5.10.475 Driving Pressure, cmH2O8.8 ± 1.18.6 ± 1.00.317 Compliance, mL/cmH2O56.9 ± 8.159.0 ± 8.20.860Values are expressed mean ± standard deviation (SD) or as median [25–75% interquartile range, IQR], or absolute numbers (percentage), as appropriateAPACHE-II Acute Physiologic and Chronic Health Evaluation, SOFA Sequential Organ Failure Assessment, CAD Coronary Atherosclerotic Heart Disease Characteristics and demographic data of study population (n = 71) Values are expressed mean ± standard deviation (SD) or as median [25–75% interquartile range, IQR], or absolute numbers (percentage), as appropriate APACHE-II Acute Physiologic and Chronic Health Evaluation, SOFA Sequential Organ Failure Assessment, CAD Coronary Atherosclerotic Heart Disease Hemodynamic and EtCO₂ responses to PLR The PLR maneuver induced changes in hemodynamic parameters and EtCO₂ in both groups, but with significantly greater magnitudes in responders (Table 2). In preload responders, PLR resulted in significant increases in SBP, DBP, MAP, PP, and EtCO₂ (all p < 0.001), with no changes in HR (p = 0.869) or CVP (p = 0.092). Non-responders also showed statistically significant increases in these parameters, but of smaller magnitude. Table 2Effects of PLR on hemodynamic variables and end-tidal carbon dioxideParameterTime PointResponders (n = 37)Non-responders (n = 34)p-valueHR (b.p.m)Before PLR99.11 ± 19.4390.74 ± 14.780.043After PLR103 [82, 117.5]89.5 [76.75, 101]0.028CI (L/min/m²)Before PLR3.28 ± 0.673.7 ± 0.790.010After PLR4.3 ± 0.843.69 ± 0.790.008CVP (mmHg)Before PLR10 [8, 9]10 [7, 13.25]0.777After PLR11 [8.5, 11.5]11 [9.75, 14]0.118SBP (mmHg)Before PLR105 [99, 113.5]119 [105, 132.25]0.001After PLR120 [113, 132]128.5 [120.5, 136.5]0.052DBP (mmHg)Before PLR47.59 ± 9.4554.44 ± 9.790.004After PLR51 [45, 60]56.5 [51.5, 66.75]0.005MAP (mmHg)Before PLR65.67 [62.17, 69.5]79 [66, 83.08]< 0.01After PLR72.33 [70, 83]82.67 [75.33, 85.17]0.005PP (mmHg)Before PLR59.51 ± 10.5364.62 ± 14.120.112After PLR70.11 ± 11.2868.91 ± 14.50.968EtCO2 (mmHg)Before PLR35.22 ± 4.838.03 ± 5.070.015After PLR38.81 ± 4.9339.03 ± 4.510.548Dynamic indexSVV%16 [11, 20.5]9 [6, 12]< 0.001PPV%15.78 ± 6.877.71 ± 4.03< 0.001△HR%0.98 [−1.44, 3.39]0 [−3.36, 2.88]0.216△CVP%0 [0, 13.39]8.12 [0, 42.5]0.143△SBP%14.78 [10.1, 21.15]6.86 [−1.48, 21.07]0.030△DBP%8.51 [4.96, 17.7]7.29 [−2.32, 25.76]0.427△MAP%10.95 [6.35, 21.34]5.49 [−1.4, 18.67]0.076△PP%19.6 [8, 26.79]0 [−5.16, 15.47]0.001△EtCO2 mmHg4 [3, 4.5]1 [0, 2]< 0.001△EtCO 2 %10.39 ± 3.662.97 ± 5.08< 0.001Values are expressed mean ± standard deviation (SD) or as median [25–75% interquartile range, IQR], as appropriate. The p-values presented in this table indicate the results of intergroup comparisonsETCO2 end-tidal carbon dioxide, PLR passive leg raising test, HR heart rate, CI cardiac index, CVP central venous pressure, SBP systolic blood pressure, DBP diastolic blood pressure, MAP mean arterial pressure, PP pulse pressure, SVV stroke volume variability, PPV pulse pressure variability, △HR% the variations of HR presented as percentage change from baseline, △CVP% the variations of CVP presented as percentage change from baseline, △SBP% the variations of SBP presented as percentage change from baseline, △DBP% the variations of DBP presented as percentage change from baseline, △MAP% the variations of MAP presented as percentage change from baseline, △PP% the variations of PP presented as percentage change from baseline △EtCO2 absolute change in EtCO2 from baseline, △EtCO2% relative change in EtCO2 from baseline Effects of PLR on hemodynamic variables and end-tidal carbon dioxide Values are expressed mean ± standard deviation (SD) or as median [25–75% interquartile range, IQR], as appropriate. The p-values presented in this table indicate the results of intergroup comparisons ETCO2 end-tidal carbon dioxide, PLR passive leg raising test, HR heart rate, CI cardiac index, CVP central venous pressure, SBP systolic blood pressure, DBP diastolic blood pressure, MAP mean arterial pressure, PP pulse pressure, SVV stroke volume variability, PPV pulse pressure variability, △HR% the variations of HR presented as percentage change from baseline, △CVP% the variations of CVP presented as percentage change from baseline, △SBP% the variations of SBP presented as percentage change from baseline, △DBP% the variations of DBP presented as percentage change from baseline, △MAP% the variations of MAP presented as percentage change from baseline, △PP% the variations of PP presented as percentage change from baseline △EtCO2 absolute change in EtCO2 from baseline, △EtCO2% relative change in EtCO2 from baseline The relative change in EtCO₂ (ΔEtCO₂%) was significantly greater in responders compared to non-responders (10.39% ± 3.66% vs. 2.97% ± 5.08%, p < 0.001). Similarly, baseline PPV was higher in responders (15.78% ± 6.87% vs. 7.71% ± 4.03%, p < 0.01). Association between CI changes and other parameters Linear regression analysis revealed a positive association between PLR-induced changes in CI and changes in EtCO₂ (r = 0.54, p < 0.001, Fig. 1) and between CI changes and baseline PPV (r = 0.58, p < 0.001, Fig. 2). Linear regression analysis of the relationship between PLR-induced changes in CI and EtCO2CI cardiac index, EtCO2 end-tidal CO2, PLR passive leg raising Linear regression analysis of the relationship between PLR-induced changes in CI and baseline PPVCI cardiac index, PPV pulse pressure variability, PLR passive leg raising Prediction of preload responsiveness ROC curve analysis demonstrated that SVV, PPV, ΔEtCO₂, and ΔEtCO₂% all performed well in predicting preload responsiveness, with AUCs > 0.80 (Table 3; Fig. 3). The AUC for ΔEtCO₂% was 0.88 (95% CI: 0.79–0.95), with an optimal cutoff value of > 5.88%, yielding a sensitivity of 91.89% and specificity of 76.47%. The AUC for PPV was 0.85 (95% CI: 0.75–0.93), with a cutoff value of > 10%, yielding a sensitivity of 81.1% and specificity of 79.41%. Table 3Factors associated with response to PLRVariablesCut-off value%AUCp valueSensitivity (%)Specificity (%)PV+PV-LR+LR-SVV> 13 0.84 (0.73 to 0.91) < 0.001 72.97 (55.9 to 86.2) 82.35 (65.5 to 93.2) 81.8 (68 to 90.5) 73.7 (61.7 to 82.9) 4.14 (1.9 to 8.8) 0.33 (0.2 to 0.6)PPV> 10 0.85 (0.75 to 0.93) < 0.001 81.08 (64.8 to 92) 79.41 (62.1 to 91.3) 81.1 (68.5 to 89.4) 79.4 (66 to 88.5) 3.94 (2 to 7.8) 0.24 (0.1 to 0.5)△SBP> 9.73 0.65 (0.53 to 0.76) 0.030 78.38 (61.8 to 90.2) 61.76 (43.6 to 77.8) 69 (58.5 to 77.9) 72.4 (57.4 to 83.7) 2.05 (1.3 to 3.2) 0.53 (0.2 to 0.7)△PP> 4.48 0.72 (0.60 to 0.82) 0.001 83.78 (68 to 93.8) 64.71 (46.5 to 80.3) 72.1 (61.6 to 80.6) 78.6 (62.9 to 88.8) 2.37 (1.5 to 3.8) 0.25 (0.1 to 0.5)△EtCO 2> 2 0.89 (0.79 to 0.95) < 0.001 86.49 (71.2 to 95.5) 79.41 (62.1 to 91.3) 82.1 (70 to 90) 84.4 (70.1 to 92.5) 4.2 (2.1 to 8.2) 0.17 (0.07 to 0.4)△EtCO 2 %> 5.88 0.88 (0.79 to 0.95) < 0.001 91.89 (78.1 to 98.3) 76.47 (58.8 to 89.3) 81 (69.7 to 88.7) 89.7 (74.2 to 96.3) 3.91 (2.1 to 7.2) 0.11 (0.04 to 0.3)Combined PPV-ΔEtCO 2 % index> 0.67 0.94 (0.86 to 0.98) < 0.001 83.78 (68 to 93.8) 94.12 (80.3 to 99.3) 93.94 (87.8 to 90.92) 84.21 (79.21 to 82.93) 14.24 (3.7 to 55.1) 0.17 (0.08 to 0.4)AUC area under the curve, △EtCO2 end-tidal carbon dioxide variation with passive leg raising, PV + positive predictive value, PV- negative predictive value, LR + positivelikelihoodratio, LR- negative likelihood ratio, ETCO2 end-tidal carbon dioxide, SBP systolic blood pressure, PP pulse pressure, SVV stroke volume variability, PPV pulse pressure variability, △SBP% the variations of SBP presented as percentage change from baseline, △PP% the variations of PP presented as percentage change from baseline, △EtCO2 The variations of EtCO2 from baseline, △EtCO2% the variations of EtCO2 presented as percentage change from baseline Factors associated with response to PLR AUC area under the curve, △EtCO2 end-tidal carbon dioxide variation with passive leg raising, PV + positive predictive value, PV- negative predictive value, LR + positivelikelihoodratio, LR- negative likelihood ratio, ETCO2 end-tidal carbon dioxide, SBP systolic blood pressure, PP pulse pressure, SVV stroke volume variability, PPV pulse pressure variability, △SBP% the variations of SBP presented as percentage change from baseline, △PP% the variations of PP presented as percentage change from baseline, △EtCO2 The variations of EtCO2 from baseline, △EtCO2% the variations of EtCO2 presented as percentage change from baseline ROC curves showing the ability of the PLR-induced changes in SVV, PPV, △SBP, △PP, △EtCO2 mmHg, △EtCO2% and combined PPV-ΔEtCO2% index △EtCO2 absolute change in end-tidal carbon dioxide, △EtCO2% relative change in end-tidal carbon dioxide, △PP pulse pressure variation, △SBP systolic blood pressure variation, PPV pulse pressure variability, SVV stroke volume variability Notably, the combination of PPV and ΔEtCO2% improved predictive performance, with an AUC of 0.94 (95% CI: 0.86–0.98), which was significantly higher than either ΔEtCO₂% (p = 0.046) or PPV (p = 0.011) alone. The optimal cutoff value for the combined score was > 0.67, providing a sensitivity of 83.8% and, importantly, a much higher specificity of 94.1%. The cross-validation analysis confirmed the robustness of the combined model, with a mean cross-validation AUC of 0.94 ± 0.10 (Fig. 4). Cross-validation ROC curves of combined PPV-ΔEtCO2 % index Ten-fold cross-validation was conducted. Dash lines repsresents the curve of each fold and the solid line represents the mean cross-validation curve. cvAUC cross-validation area under curve, SD standard diviation Discussion This study demonstrates that the combination of PPV and PLR-induced changes in EtCO₂ provides improved accuracy in predicting preload responsiveness compared to either parameter alone in mechanically ventilated patients with septic shock. This finding has clinical implications as it offers a non-invasive alternative to direct cardiac output measurement during PLR testing. Physiological basis for the integrated approach Our findings can be explained by the complementary physiological information provided by PPV and ΔEtCO₂. PPV reflects the dynamic heart-lung interactions during mechanical ventilation, with higher values indicating greater preload dependency and position on the steep portion of the Frank-Starling curve [19]. Meanwhile, changes in EtCO₂ during PLR directly reflect the cardiac output response to the transient preload challenge, assuming stable ventilation and metabolism [20]. By integrating these two parameters, we capture both the baseline preload dependency status (PPV) and the actual cardiac response to a standardized preload challenge (ΔEtCO₂ during PLR). This provides a more comprehensive assessment of cardiovascular physiology than either parameter alone, explaining the improved diagnostic performance of the combined approach. The integration of complementary physiological signals aligns with recent recommendations for a more holistic approach to hemodynamic assessment in critical care [15, 21]. Comparison with previous studies Our findings regarding individual parameters are consistent with previous research. Monnet et al. [12] demonstrated that PLR-induced changes in EtCO₂ could predict preload responsiveness with an AUC of 0.93 (95% CI: 0.87–0.99), using a cutoff value of 5%. Similarly, Monge Garcia et al. [13] reported an AUC of 0.97 (95% CI: 0.84–0.99) with a cutoff value of 5.8%. Our slightly lower AUC for ΔEtCO₂% (0.884) may be attributed to differences in patient populations, as we specifically focused on septic shock patients who may have more complex pathophysiology. Regarding PPV, while our findings align with meta-analyses on PPV diagnostic performance [8, 9, 22], we observed notably lower PPV threshold despite using optimal ventilation conditions (TV ≥ 8 mL/kg). A recent large meta-analysis by Chaves et al. including 40 studies reported mean PPV threshold values of 11.5% with an AUC of 0.87 (0.84–0.90) [9]. This difference may reflect the specific pathophysiology of septic shock, where systemic vasodilation, altered vascular compliance, and concurrent vasopressor therapy may dampen respiratory-induced pressure variations. Despite the lower absolute value, PPV maintained discriminative ability in our study (AUC 0.852) using a 10% threshold rather than the conventional 11–12% cutoff value, suggesting population-specific optimization may be necessary. Most importantly, our study demonstrated the added value of combining PPV with ΔEtCO2 during PLR testing, offering a novel approach to non-invasive preload responsiveness assessment that overcomes the limitations of either parameter used alone. The concept of combining different physiological parameters to improve diagnostic accuracy in critical care has gained support in recent literature. Messina et al. [23] highlighted the value of integrating multiple parameters for a more comprehensive assessment of hemodynamic status, which aligns with our multiparameter approach. Moreover, our findings align with the comprehensive evidence presented in a recent meta-analysis by Alvarado Sánchez et al., which systematically evaluated variables influencing the prediction of preload responsiveness across different clinical settings [24]. The findings of this meta-analysis particularly support our rationale for combining multiple parameters, as technical variables like measurement methods and clinical factors influence individual predictor performance. Our study extends these insights by demonstrating that EtCO2-derived parameters, which reflect changes in pulmonary blood flow secondary to cardiac output variations, can serve as valuable complements to traditional dynamic indices in septic shock patients. Outliers Several notable outliers in our dataset warrant specific discussion. We observed two distinct patterns of extreme responses during PLR testing (Figs. 1 and 2). First, a few patients demonstrated paradoxical CI decreases (25–40% reduction) during PLR. This phenomenon may be explained by several mechanisms: ventricular interdependence in patients with right heart dysfunction, where increased venous return exacerbates right ventricular overload and impairs left ventricular filling through septal shift; severe diastolic dysfunction with restrictive filling patterns, where additional preload cannot improve and may worsen cardiac performance; or acute changes in ventricular compliance during the maneuver. Conversely, several patients exhibited dramatic cardiac index increases (> 50%, reaching up to 90%), suggesting they were operating on the steep portion of the Frank-Starling curve. These extreme responders likely represent patients with severe hypovolemia, cardiac reserve, or those with very low baseline preload states. While these outliers did not affect our overall statistical analysis, their presence highlights the complex pathophysiology underlying preload responsiveness and emphasizes the importance of considering individual patient characteristics when interpreting PLR results. Clinical implications While PLR testing is widely recognized as a valuable method for assessing preload responsiveness, its implementation has been hindered by the requirement for real-time cardiac output monitoring, which often necessitates invasive devices. By demonstrating that the combination of PPV and ΔEtCO2 provides improved diagnostic accuracy, our study may offer clinicians a promising non-invasive alternative. Both PPV and EtCO₂ can be monitored continuously and non-invasively in most ICU settings. PPV can be derived from the arterial pressure waveform obtained from a standard arterial catheter, which is routinely placed in critically ill patients. Similarly, EtCO2 monitoring is a standard component of care for mechanically ventilated patients. Therefore, implementing our combined approach would not require additional invasive procedures or specialized equipment beyond what is typically available in ICUs. The improved specificity of the combined approach (94.1% vs. 76.47% for ΔEtCO₂% and 79.41% for PPV) is particularly important, as it would reduce unnecessary fluid administration in non-responders, potentially preventing fluid overload and its associated complications. Recent evidence from the CLASSIC trial has demonstrated that restrictive fluid strategies may be beneficial in septic shock, further emphasizing the importance of accurate preload responsiveness prediction [4]. An important consideration for clinical translation of our combined score is its population-specific derivation. The logistic regression coefficients (PPV: 0.186, ΔEtCO2%: 0.312) reflect physiological relationships specific to our septic shock cohort, where systemic vasodilation, altered vascular compliance, and concurrent vasopressor therapy create unique cardiovascular conditions. Different patient populations will likely demonstrate different parameter weightings due to variations in: underlying cardiovascular reserve, myocardial function, degree of systemic vasodilation and vasopressor dependency, pulmonary mechanics, ventilation-perfusion relationships, and metabolic status affecting CO2 production and elimination. Future multicenter validation studies involving larger, heterogeneous cohorts will likely require model recalibration with population-specific coefficients. This underscores the need for either developing population-specific scoring systems or creating more generalizable algorithms that incorporate patient-specific physiological factors. Until comprehensive external validation is completed, our combined approach should be viewed as a proof-of-concept for the septic shock population, with the understanding that the diagnostic thresholds and parameter weightings may need adjustment for broader clinical application. To enhance the clinical applicability of our findings, we have developed an Excel spreadsheet that serves as a practical bedside tool (Supplementary File 1). This spreadsheet includes built-in formulas that allow clinicians to easily input the values for PPV and ΔEtCO2% and quickly obtain the assessment results, facilitating the application of our combined equation in practice. Limitations Several limitations should be considered when interpreting our results. First, our study was conducted in a specific population of deeply sedated, mechanically ventilated septic shock patients without spontaneous breathing efforts. The applicability of our findings to other populations, particularly spontaneously breathing patients or those with different shock etiologies, requires further investigation. Second, we used tidal volumes ≥ 8 mL/kg, which optimizes the reliability of PPV but may differ from lung-protective ventilation strategies (6 mL/kg) commonly employed in acute respiratory distress syndrome. Future studies should evaluate the performance of our combined approach across different ventilation strategies, as the diagnostic performance of PPV may be reduced with lower tidal volumes [17]. Third, the stability of ventilation and metabolism is a prerequisite for the reliable interpretation of EtCO₂ changes. While the short duration of the PLR test minimizes the likelihood of metabolic changes, this remains a theoretical limitation that clinicians should consider when implementing this approach. Fourth, intra-abdominal pressure was not measured in our study population. Elevated intra-abdominal pressure can impair venous return during PLR and reduce the accuracy of the test, potentially leading to false-negative results. Fifth, we did not perform a gold standard fluid challenge test with direct assessment of hemodynamic response to actual fluid administration. The correlation between PLR response and actual fluid challenge response, while generally good, is not perfect and may be influenced by various clinical factors. Moreover, from a terminological standpoint, PLR assesses preload dependence rather than definitive fluid responsiveness. Strictly speaking, only actual fluid administration can determine true “fluid responsiveness.” Therefore, we used the terminology of “preload responders” and “non-responders”. Finally, our combined score represents a population-specific predictive model derived from septic shock patients in a single-center setting. As discussed above, this limitation necessitates external validation. While we performed cross-validation to assess model robustness, external validation in an independent cohort would further strengthen our findings. Conclusions The combination of PPV and PLR-induced changes in EtCO₂ provides a promising non-invasive method for predicting preload responsiveness in mechanically ventilated septic shock patients compared to either parameter alone. This approach represents a promising alternative to invasive cardiac output monitoring in guiding fluid therapy decisions, potentially improving the risk-benefit ratio of volume expansion in critically ill patients. Future studies should focus on validating this approach in diverse patient populations and exploring its impact on clinical outcomes. Supplementary Information Supplementary Material 1. Supplementary File 1: Combined preload responsiveness score calculator. Acknowledgements None. Abbreviations cardiac index cardiac output central venous pressure change of end-tidal carbon dioxide partial pressure diastolic blood pressure end-tidal carbon dioxide heart rate intensive care unit interquartile range mean arterial pressure Pulse index Contour Cardiac Output passive leg raising pulse pressure pulse pressure variability Richmond agitation-sedation scale receiver-operating characteristic systolic blood pressure standard deviation stroke volume stroke volume variability Authors’ contributions The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. Study concept and design: XFZ, HC, DSQ and KC. Acquisition of data: XFZ, DSQ, KC, XYD, BYC and HC. Analysis and interpretation of data: XFZ, DSQ, KC and HC. Drafting of the manuscript: XFZ, DSQ, KC and HC. Critical revision of the manuscript for important intellectual content: XFZ, DSQ, KC, XYD, BYC and HC. Qualitative analysis: XFZ, DSQ, KC, XYD, BYC and HC.Obtained funding: HC and XFZ. Funding HC is supported by the Youth Top Talent Project of Fujian Provincial Foal Eagle Program. XFZ is supported by the General Program of Natural Science Foundation of Fujian Province (Grant Number: 2023J011179). The funders had no role in study design, data collection, and analysis, the decision to publish, or manuscript preparation. Data availability De-identified data are available at the discretion of the corresponding author. The datasets used and/or analysed during the current study available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The formal ethical Committee studied the prospective study and agreed to its ethical approval on January 23rd, 2018 (Approval # K2018-01-023). The study protocol and consent forms were approved by the Ethics Committee of Fujian Provincial Hospital. The study was registered on December 26th, 2019, at the Chinese Clinical Trial Registry (ChiCTR1900028210). The research process was in line with the Helsinki Declaration. Informed consent was obtained from all subjects and/or their legal guardians. All experiments were performed in accordance with relevant guidelines and regulations. Consent for publication All authors approved the final manuscript and submission to this journal. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Xiao-Fen Zhou, Ding-Shen Qiu and Kun Cheng contributed equally to this work. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary File 1: Combined preload responsiveness score calculator. Data Availability Statement De-identified data are available at the discretion of the corresponding author. The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.