Optimizing CO by increasing preload is based on the assumption that increased blood volume may subsequently promote stroke volume (SV) as described by the Frank-Starling law. Predicting fluid responsiveness is to test whether the patient was in the ascending part of the Frank-Starling curve (preload dependence state) and may benefit from fluid treatment[24, 27–30]. Although numerous measurements can predict FR, the fluid challenge is a gold standard for assessing FR and is widely used. The new concept of mini-fluid challenge is to use a minimun amount of fluid to avoid overload risk, whereas a median volume of 500 mL has been used before.[14, 31, 32]. A metanalysis demonstrated that a mini volume of 100 mL could predict preload responsiveness, the pooled AUC was 0.91 (95%CI 0.85–0.97), with a pooled sensitivity of 0.82 (95%CI 0.76–0.88) and specificity of 0.83 (95%CI 0.77–0.89)[33]. No similar findings were found in our study: the predictive power of 100 mL is disappointing with an AUROC of 0.69.
The main reason for these differences may be the difference in the study population. Our study focused on septic shock patients, whose cardio-vascular system has undergone major perturbations and profound alterations to the endothelium. Capillary leakage caused by epithelial barrier dysfunction occurs, as a consequence, the intravascular volume is insufficient and unstressed volume substantially reduced[34]. The adjustment and shift between unstressed volume and stressed volume introduce a role for volume expansion that is not simply to increase CO but rather to ensure reserves[35]. In these patients, a larger dose may be required for an effective test, whereas 100 mL of fluid seemed too little to predict the FR. Among the previous 7 studies, only 1 mini-fluid study focused on septic shock. Wang’s study showed a similar result to ours that the predictive power with 100 mL was less satisfying. In other mini-fluid challenge investigation, the study population was perioperative patients in a stable hemodynamic state. In this study, we chose colloid of gelatin to perform the fluid challenge, which can stay in vascular longer and may increase CO more than crystalloid. The marked change, however, showed in 200 mL.
In spite of the effect of heterogeneity of study population, Aya et al[22]. and Smoerenberg et al[16]. were unable to reproduce similar results as well, who showed the predicted minimal volume is 4 mL/kg to defined FC (CO measured by LiDCOplus monitor), or minimal volume of 150 mL CO measured by ModelflowR COm and 200 mL by PulseCORCOli. These two studies implied a negative predictive value of a minimum dose of 100 mL to reliably assess FR even in a fairly homogeneous sample of postoperative cardiac surgery patients. It suggests that the specificity and sensitivity values of mini fluid challenge may change depending on the device used to measure the CO. To the best of our knowledge, CO measurements including thermodilution (PAC), pulse index continuous CO, Doppler echocardiography (TTE or TEE) and Fick techniques, have been most commonly used to assess FR in recent decades. In the mini fluid challenge, pulse contour CO was used in 4 out of the 7 cases, and TTE was used in 2 out of the 7 cases, while PAC, as a standard clinical reference method for CO monitoring was not used in the above studies.
However, the CO measurements by Doppler echocardiography (we have discussed only the VTi method, which is the most popular and presented in recent studies) have been validated for a long time for its noninvasive, but its accuracy is limited by the devices and technicians. In method of echocardiography, SV is affected by VTi and the aortic valve area. When estimating the changes in SV, the value of VTi is the main component of the SV calculation whereas the aortic valve area is supposed to be constant. It is well known that either the quality and sharpness of image or the process of technician may have an influence on the value of VTi. Moreover, Gorassi et al.[36] demonstrated that pulsed Doppler has limitations in detecting high cardiac output values when the blood flow velocity is greater than 2 m/sm, especially in septic shock patients who showed the greatest variability in CO. What is more, the CO measured by PAC integrated over several heartbeats, while the VTi was measured on a beat-to-beat basis, and then calculates CO by the products of SV and HR, which may increase the potential error of CO estimation. VTi seemed to be more likely influenced by heart-lung interaction than PAC, and previous studies have shown that CO is more reliable on an average of serial measurements[37]. Based on the published studies evidences, the accuracy of PAC is higher than that of the transthoracic echocardiography[23, 36, 38].
At the same time, the precision of new generation CO measurement, especially methods coupled with thermodilution and pulse contour analysis, such as Pulse index Continuous Cardiac Output combination (PiCCO), is also unclear, because more variables will affect the estimation of CO. Pulse contour analysis measures CO indirectly by integrating a variety of characteristics of the pressure waveform to calculate stroke volume. In the equation of CO = SV × HR, SV is calculated from area under systolic portion of arterial waveform trace and aortic impedance, in which the arterial pressure waveform is complex and aortic impedance varies between individuals and within individuals, especially in septic shock patients whose hemodynamic state is characterized by distributive shock and the vascular impedance is unpredictable. Although the pulse contour analysis has been calibrated by trans-cardiopulmonary thermodilution to overcome some limitations, the analysis remains potential inaccuracy due to multitude of variables in the equation of CO calculation. A great deal of studies and meta-analysis have been done to compare the precision of new generation measurements of CO with PAC, but why not perform a mini-fluid challenge in classic thermodilution by PAC.
In addition to taking the mini-volume into fluid challenge, the threshold of predictive value is also a controversial topic. In most clinical practice, and even in most studies, the definition of FR is based on the assumption that thermodilution is the only method validated to detect a 10–15% increase in CO[23]. This consensus was originally derived from the understanding of the sources of errors of PAC, that is, three measurements are sufficient if the CO differs by 10% or less. On the contrary, if the CO difference exceeds 10%, the measurements are considered unreliable[39]. It is worth noting that these technique errors were mostly introduced by the measurement of injecting iced water while CO was measured[38–41]. Our application for CO measurement by PAC (Swan-Ganz CCOmbo CCO/SvO2, Vigilance II™ monitor, Edwards Lifesciences, Irvine, CA, USA) in this study is automatic heating, and iced water injection is no longer needed. The sources of error including temperature of iced water, volume of injection and even injection speed will not affect the CO measurement accuracy during this protocol. In order to avoid measurement errors and enhance the reproducibly of the results, we took an average CO value of 3 measurements at each point in time as a CO determination. In this study, the best cutoff value of 1.9% was an inferior threshold to the interobserver variability of PAC. Thus, considering both clinical feasibility and reproducibility, the cutoff value was 5.2%, with a specificity of 83.3% and 90.9%, respectively. An increase in CO greater than 5.2% after 200 mL may be more clinically relevant. The correlation between ΔCO200mL and ΔCO500mL suggests that the greater the increase in ΔCO200mL, the more we can expect a similar increase in ΔCO500mL. If we accepted CO > 10% in mini-FC, both the sensitivity and specificity were too low (60.7% and 73.7%) to predict FR, and up to 32% of R may possibly be misclassified as NR (Fig. 9).
In the design of this study, we attempted to find a surrogate of CO, which can evaluate FR simply and practically when there is no CO monitoring available or inconvenient. SvO2 and ScvO2 are a pair of parameters wildly used in clinical practice to assess whether CO and oxygen delivery (DO2) are sufficient to meet the patient’s need and to guide fluid resuscitation[42]. The well-known study by Rivers et al.[43] indicated that targeting a ScvO2 > 70% in the early stage of resuscitation may improve outcomes. In our study, both in R and NR, the mean value of ScvO2 were higher than 70% while that of SvO2 were higher than 65%. Neither the actual value of ScvO2 and SvO2 nor the ΔScvO2 and ΔSvO2 were meaningless to predict FR. In fact, ScvO2 and SvO2 can be influenced by oxygen extraction. We confirmed the result of Velissaris et al[42] that a high level of ScvO2 and SvO2 levels did not exclude FR. In this study, there were 17 cases out of 37 (45.9%) with ScvO2 were greater than 70%, and 20 cases out of 37 (54.1%) with ScvO2 greater than 65% in FR. Based on the same theory, Pcv-aCO2 are considered as alternative markers of tissue hypoperfusion and are used to guide treatment for shock. However, there was no difference between R and NR before and after FC. For the indicators we simply used in clinical practice, MAP < 61 mmHg before FC is a strong indicator for FR, with a sensitivity and specificity of 100% and 100%. Although ΔDO2I exhibited predictive power in this study, it correlated to the changes of CO and could not be used as a surrogate of CO.
Several limitations of this study need to be discussed. First, as we discussed before, the accuracy varies with different hemodynamic techniques. With the development of techniques in assessing the circulatory function, the new generation measurements noted by less or non-invasive, such as TEE, bio-impedance and TTE are more likely to be applied in recent decade[36, 44]. Our findings depend on PAC to monitor CO, and the results may not be extrapolated to other techniques used to monitor CO. Second, 77% of the patients responded to fluid administration after a total volume of 500 mL gelatin, and the proportion of responders (PR) was higher than in other studies. Our FC was completed within 40 minutes from the first bolus. Theoretically, PR decreases with a long infusion time, as described by Toscani et al.[45], but our results are just the opposite. We compared the similar results of high PR, which appeared in Smorenberg’s study, with a PR of 71% at the end of FR. These trials shared two same points, one was that patients subsequently received a 500 mL volume expansion by several intravenous boluses, and the other was that both used the colloid in FC. So, a reasonable question is whether the high PR may be due to the accumulative of fluid or the colloid, which supposed to remain in the intravascular compartment longer. To answer this question, we need to further study the pharmacodynamic outcomes and its effect on different approaches of FC. Third, we did not find a meaningful surrogate of CO that can predict FR and guide fluid therapy. In future study, a feasible and simple clinical indicator for predicting FR needs to be found. Above all, future research may be conducted to investigate the diagnostic value of the mini-FC using crystalloids, and the accuracy grey zone to detect CO changing by using different CO monitor. Furthermore, pharmacodynamic and pathophysiology mechanisms FC need to be studied.