Accuracy of Combining End-Tidal Carbon Dioxide and Pulse Pressure Variability Measurements in Predicting Fluid Responsiveness During Passive Leg Raising Test in Septic Shock Patient

It is essential to assess the patients’ volume responsiveness before uid infusion in patients with circulatory failure. The hypothesis is that the combining of end-tidal carbon dioxide (EtCO 2 ) and pulse pressure variability (PPV) measurements can non-invasively predict uid responsiveness during passive leg raising (PLR) test. Pulse indicates continuous cardiac output, right radial artery blood pressure, and EtCO 2 were monitored in 71 septic shock patients with mechanical ventilation. A standard PLR test was performed; cardiac index (CI), arterial pressure, stroke volume variability (SVV), PPV, and EtCO 2 were measured before and after the PLR test. Patients with an increase in CI greater than 15% after the PLR test were dened as uid responders. Receiver-operating characteristics (ROC) curve analysis was carried out to assess the predictive performance of the measured parameters.


Background
It is essential to assess the patients' volume responsiveness before uid infusion in patients with circulatory failure. The hypothesis is that the combining of end-tidal carbon dioxide (EtCO 2 ) and pulse pressure variability (PPV) measurements can non-invasively predict uid responsiveness during passive leg raising (PLR) test.

Methods
Pulse indicates continuous cardiac output, right radial artery blood pressure, and EtCO 2 were monitored in 71 septic shock patients with mechanical ventilation. A standard PLR test was performed; cardiac index (CI), arterial pressure, stroke volume variability (SVV), PPV, and EtCO 2 were measured before and after the PLR test.
Patients with an increase in CI greater than 15% after the PLR test were de ned as uid responders. Receiveroperating characteristics (ROC) curve analysis was carried out to assess the predictive performance of the measured parameters.

Conclusion
The combination of PPV and △EtCO 2 can better predict uid responsiveness during PLR.

Background
The rst-line treatment for patients with circulatory failure is volume therapy. Optimized adjustment of intravascular volume and cardiac preload is essential for improving cardiac output (CO) and tissue perfusion [1] . However, only 50% of hemodynamically unstable critically ill patients in the intensive care unit (ICU) are volume-responsive [2] . It is essential to assess the patients' volume responsiveness before uid administration, since excessive uid load may lead to heart failure, prolonged mechanical ventilation duration and ICU stay, reduced oxygen delivery, and increased risk of mortality [3] .

Page 3/16
The passive leg raising (PLR) test can transfer blood from the legs and abdomen to the heart cavity via posture changes, and its effect is similar to endogenous uid supplementation [4] . The advantage of the PLR test is that it is still valuable for detecting preload responsiveness in the situations that the application of dynamic hemodynamic indicators is limited (e.g., strong spontaneous breathing, low tidal volume ventilation, arrhythmia, etc.) [5][6][7][8][9][10][11][12][13][14] . However, the disadvantage of the PLR test is that it needs to measure CI directly. At present, the gold standard for assessing volume responsiveness is still to observe changes in cardiac index (CI) or stroke volume (SV) before and after volume loading tests, which requires a thermodilution catheter in place. Another way to measure CI or SV is bedside echocardiography, where the non-invasive nature of echocardiography is also attractive. However, echocardiography requires skilled operators, and continuous estimation for it cannot be made [15,16] .
Studies have shown that the end-tidal carbon dioxide (EtCO 2 ) depends on cellular metabolism and the CO 2 content transported by venous blood ow (which equals to CO at steady status), and is also related to the ability of the lungs to remove CO 2 . Therefore, in patients with stable breathing, The EtCO 2 mainly depends on cardiac output [17][18][19] . It has been shown that a change of end-tidal carbon dioxide partial pressure (△PetCO 2 ) ≥ 2 mmHg during PLR determines whether there is volume reactivity [20,21] . However, the absolute value of this number is too small, which means it is prone to errors. Studies have shown that pulse pressure variability (PPV) can be used to accurately predict volume responsiveness. We can obtain the PPV value by continuously monitoring the radial artery blood pressure. We hypothesis that the combination of PPV and △EtCO 2 % can non-invasively predict volume responsiveness in septic shock patients.

Patients
This prospective study was approved on January 23rd, 2018, by the Ethics Committee of Fujian Provincial Hospital (Approval # K2018-01-23). The study was registered on December 26th, 2019, at the Chinese Clinical Trial Registry (ChiCTR1900028210). Written informed consent was obtained from each patient's next of kin.
The inclusion criteria were: 1) Patients diagnosed septic shock according to the Sepsis 3.0 de nition [21] ; 2) Age ≥ 18; 3) blood pressure and heart rate of the enrolled patients uctuated less than 10% within 15 min before the start of the study, with relatively stable hemodynamics. The exclusion criteria were: 1) preliminary judgment at the end of the disease, may die within 24 hours; 2) patients with contraindications for the use of pulse indicating continuous cardiac output (PiCCO) monitoring; 3) patients with intra-aortic balloon counterpulsation or arti cial membrane pulmonary oxygenation.

Study design and measurements
All patients with septic shock were treated in accordance with the relevant guidelines [21] . Characteristics and demographic data of the study population were recorded. Patients were intubated and treated with sufentanil and midazolam, with a Richmond agitation-sedation scale (RASS) maintained at -3. Patients were kept with no spontaneous breathing, and with a tidal volume greater than 8 ml/kg, and PEEP was at a constant value.
Hemodynamic monitoring was performed with the PiCCO catheter (PC 4000, PULSION, Feldkirchen, Germany). Invasive blood pressure was also monitored with catheterization of the right radial artery. Heart rate (HR) was measured via electrocardiograph monitor; central venous pressure (CVP) was measured via the right subclavian venous catheter; CI, stroke volume variability (SVV) was measured via the PiCCO catheter; pulse pressure variability (PPV), systolic blood pressure (SBP), diastolic blood pressure (DBP), mean arterial pressure (MAP) and pulse pressure (PP) was measured via the right radial artery catheter. EtCO 2 was continuously measured by using a sidestream infrared gas analyzer connected to the patient's monitor (M2741A, PHILIPS, Amsterdam, Netherlands).
For the PLR test, the rst step was to raise the head of the bed by 45 degrees; patients were kept in the semirecumbent position for one minute. At baseline, we obtained the rst set of measurements, including HR, CVP, CI, SVV, PPV, SBP, DBP, MAP, PP, and EtCO 2 . The second step was to elevate both legs at 45 degrees, and the trunk was placed in a horizontal position. The postural change was performed by using the automatic motion of the bed [4] . The second set of measurements was obtained.

Statistical analysis
Patients with an increase in CI ≥ 15% after PLR test were de ned as uid responders (R). Patients with a change of < 15% were de ned as non-responders (NR).
Statistical analysis was performed using SPSS statistical software (ver. 25.0, IBM, NY, USA) and MedCalc software (ver. 18.2.1, Mariakerke, Belgium). Continuous variables were tested for normal distribution (Kolmogorov-Smirnov test) and were expressed as mean ± standard deviation (SD) or median (interquartile range, IQR), as appropriate. Comparison between the two groups was performed by t-test or Mann-Whitney U test, as appropriate. Categorical variables were expressed as frequency (percentage), and the comparison between the two groups was performed by Chi-square test or Fisher exact test.
The combination of EtCO 2 and PPV was performed by using a logistic regression model which involved both EtCO 2 and PPV as co-variables (use "enter" as the method of variable selection), and the software would then output a predict value. We used this predict value as a "score" of a combination of EtCO 2 and PPV. Receiveroperating characteristic (ROC) curves were constructed to determine the performance of each variable in predicting volume responsiveness. The Hanley-McNeil test was used to compare the areas under the ROC curves [22] . Correlations were tested by the Pearson method. All reported p values are two-sided, and a p < 0.05 was considered signi cant.

Patients' characteristics
From January to May 2020, 71 patients were included. All patients successfully completed the PLR test. There were 37 uid responders (52.11%) and 34 non-responders (47.88%). Patients' characteristics were summarized in Table 1. Values are expressed as mean ± SD, median (25th to 75th percentile) or absolute numbers, as appropriate.

Effects of the PLR maneuver
The SVV, PPV, △SBP, △PP, △EtCO 2 and △EtCO 2 variation ratio (△EtCO 2 %) of responder were all greater than that in the non-respond patients ( Table 2). The responders had a greater △EtCO2% than the non-responders (10.39% ± 3.66% vs. 2.97% ± 5.08%, p < 0.001). The responders also had a greater PPV than the nonresponders (15.78% ± 6.87% vs. 7.71% ± 4.03%, p < 0.01). All responders had higher SBP, DBP, MAP, PP, and EtCO 2 than their baseline after PLR (all p < 0.001). In the uid responsiveness group, there was no statistical difference in the changes of HR before and after PLR (p = 0.869). In the uid responsiveness group, there was no statistical difference in the changes of CVP before and after PLR (p = 0.092). All uid non-responders showed higher CVP, SBP, DBP, MAP, PP and EtCO 2 than their individual baseline (p < 0.01, p < 0.01, p = 0.002, p = 0.001, p = 0.023, p = 0.003; respectively). In the uid nonresponsiveness group, there was no statistical difference in the changes of HR before and after PLR (p = 0.369).
A positive relationship was observed between the change of CI and the change of EtCO 2 (Fig. 1, r = 0.538; p < 0.001). A positive relationship was also observed between the change of CI and PPV (Fig. 2, r = 0.584; p < 0.001).

Prediction of uid responsiveness
Dynamic hemodynamic indexes were selected to evaluate uid responsiveness. Table 3 Table 3). There was no statistical difference between the AUC of PPV and △EtCO 2 % (0.852 vs. 0.884, p = 0.5843, Table 3). The combination of PPV and △EtCO 2 % was a better predictor of uid responsiveness than each variable independently.

Discussion
The main nding of this study is that EtCO 2 variation was correlated with changes in cardiac output induced by a simpli ed PLR maneuver. PPV, △EtCO 2, and △EtCO2% of uid-responders were signi cantly greater than those in the non-responders. The AUC of PPV-△EtCO 2 % combination was signi cantly greater than either △EtCO 2 % or PPV.
CO 2 is the product of tissue cell's metabolism. It simply diffuses to the circulating blood, which is then transported to the lung and diffuses from the lungs into the exhaled gas, which is measured by the EtCO 2 monitoring technology [23] . The normal value of EtCO 2 is 30 to 43 mmHg, which is 2 to 5 mmHg lower than the CO 2 content of arterial blood. The EtCO 2 value is affected by the following three parts: the amount of CO 2 generated by cell metabolism, the ability of the lungs to remove CO 2 from the veins, the cardiac output. The absolute value of EtCO 2 cannot be simply used to judge the cardiac output due to the different ventilation and cell metabolic states of patients at different times. Within a few minutes of the PLR test, assuming that the volume of alveolar ventilation and body metabolism are constant, when the output volume drops, the blood ow in the lungs decreases signi cantly, and EtCO 2 decreases accordingly, and vice versa. The change of EtCO 2 can sensitively re ect the change of CI, and thus can determine the uid responsiveness. In the PLR test, the uid responsiveness could be predicted by observing the change of EtCO 2 .
Numerous studies have proved that dynamic hemodynamic indicators such as SVV and PPV can also be used to determine whether there is uid responsiveness [24] . Our data showed that both △EtCO 2 % and PPV positively correlate with CI. This is similar to the results of the study of Toupin and colleagues. They conducted a PLR study on 90 cardiac surgical patients who were receiving mechanical ventilation and found that △EtCO 2 % was positively correlated with △CI [20] .
We found that the AUC of SVV, PPV, and △EtCO 2 % were greater than 0.8, showing that △EtCO 2 % and PPV could predict uid responsiveness well with high speci city and sensitivity. The results of several studies show that the cut-off value of EtCO 2 % during PLR is 5% [18,19,25] . Yao et al. conducted a PLR study on 41 postcardiac shock patients who were receiving mechanical ventilation. They used PiCCO to monitor cardiac output and mainstream sampling equipment to monitor EtCO 2 . The result showed that the AUC of △EtCO 2 % was 0.87, its cut-off value was > 5.8%, its sensitivity was 75.2%, and its speci city was 90% [26] . Toupin et al. found that during PLR, the AUC of △EtCO 2 was 0.8, its cut-off value was ≥ 2 mmHg, its sensitivity was 75%, and its speci city was 70% [20] . These were also con rmed by our data in patients with sepsis.
Our data suggested that the combination of △EtCO 2 and PPV can predict uid responsiveness better than the use of only △EtCO 2 or PPV. We combined these two methods to overcome their shortcomings. Because the cut-off value of △EtCO 2 was so small (2 mmHg), the application of this threshold value is prone to errors, which means the change may come from measurement error rather than a response to a changed CI. On the other hand, although PPV can be easily obtained by minimally invasive radial artery monitoring without a PiCCO catheter, the accuracy of PPV in predicting is not satisfactory. Therefore, we combined △EtCO 2 % with PPV to make a new predictor. The reason we chose △EtCO 2 % plus PPV was that one could monitor PPV without invasive CI monitoring (e.g., PiCCO). Meanwhile, one could obtain △EtCO 2 % by using a sidestream infrared gas analyzer connected to the patient's monitor, which is also non-invasive. By doing so, one can predict uid responsiveness accurately, easily, and non-invasively.

Conclusions
It was effective to predict the uid responsiveness of patients with septic shock by using △EtCO 2 and PPV in the PLR test. The combination of PPV and △EtCO 2 % was a better predictor of uid responsiveness than solely the use of each variable.

Declarations
Ethics approval and consent to participate This prospective study was approved on January 23rd, 2018, by the Ethics Committee of Fujian Provincial Hospital (Approval # K2018-01-23). The study was registered on December 26th, 2019, at the Chinese Clinical Trial Registry (ChiCTR1900028210). Written informed consent was obtained from each patient's next of kin.

Consent for publication
Not applicable Availability of data and material The datasets used and/or analyzed during the present study are available from the corresponding author on reasonable request.  Linear regression analysis of the relationship between PLR-induced changes in CI and PPV CI cardiac index, PPV pulse pressure variability, PLR passive leg raising