Fluid administration is the first line of treatment for critically ill children who are admitted to the pediatric intensive care unit (PICU) with unstable hemodynamics. However, only 40% to 69% of these children show a response to fluid administration (1). Fluid responsiveness is defined as an increase in cardiac output of more than 10% to 15% after an intravenous fluid challenge (1–3). Early administration of fluid in patients who are responsive improves survival. However, fluid administration to those who are unresponsive may cause fluid overload, leading to longer ventilator days and higher morbidity and mortality rates (4–6).

Many hemodynamic parameters have been used to predict fluid responsiveness in critically ill children. These parameters can be divided into static and dynamic parameters (Supplementary Table S1). Static parameters are measured at a specific time point during observation. Dynamic parameters are measured by monitoring changes in physiological responses based on cardiopulmonary interaction (e.g., variability change in preload during mechanical ventilation). Most studies have suggested that dynamic parameters are more accurate than static parameters for predicting fluid responsiveness (1, 7–9).

Dynamic parameters can be measured in an invasive or non-invasive manner. Ultrasonic cardiac output monitoring and electrical cardiometry are non-invasive methods that are commonly used to assess dynamic parameters in the intensive care unit (ICU) setting.

Previous studies of dynamic parameters were conducted in different circumstances and populations (10–36). To date, there are no standard parameters that can be used across all critically ill children, especially in mechanically ventilated children, who are prone to fluid overload. This systematic review aimed to compare the diagnostic performance of dynamic parameters for predicting fluid responsiveness in mechanically ventilated children.

This study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) reporting guideline (37). The protocol was registered and approved by the international prospective register of systematic reviews PROSPERO (CRD42020206400) on 1 October 2020. Inclusion criteria included the following: (i) children aged 1 month to 18 years old who were under mechanical ventilation and required an intravenous fluid challenge; (ii) diagnostic accuracy studies of dynamic parameters for predicting fluid responsiveness compared with the gold standard definition of fluid responsiveness (10%–15% increase in cardiac output after a fluid challenge as measured by the pressure recording analytic method, an echocardiogram, or non-invasive cardiac output monitoring), and the measurements needed to be performed before and after a fluid challenge; and (iii) the diagnostic performance included the cutoff value, sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve. Meta-analyses, systematic reviews, narrative reviews, clinical practice guidelines, conference proceedings, case series and case reports with a sample size < 10, and non-English articles were excluded.

The identification and selection of studies are shown in Figure 1. A total of 27 studies were included in the final systematic review (10–36), which comprised 1,005 participants and 1,138 intravenous fluid challenges. A total of 77% (21/27) of studies were published after the last systematic review (1). Twenty-five studies were conducted as prospective observational cohorts (10–16, 18–36), and only 1 study was retrospective cohort study (17). There were 4 major groups of patients in different clinical settings as follows: (i) the congenital heart surgery group in 14 studies; (ii) the general surgery group in 5 studies; (iii) the neurological surgery group in 4 studies; and (iv) the general PICU group in 4 studies. Among the subgroups of participants, different fluid types and volumes were administered. Patients with congenital heart surgery mostly received colloid or blood components; only 2 of 14 studies used isotonic crystalloids. The other 3 groups of participants mostly received crystalloids with larger bolus volumes.

Table 1 shows the diagnostic performance of dynamic parameters compared with the gold standard measurement of fluid responsiveness. The gold standard measurement was an increase in cardiac output of 10%–15% after fluid administration, which was represented by multiple parameters as follows: the stroke volume index in 15 studies, stroke volume in 5 studies, the cardiac index in 4 studies, and the velocity–time integral in 2 studies. Eleven dynamic parameters (see Supplementary Table S2 with equations) were investigated in the 27 included studies.

The respiratory variation in aortic peak velocity (ΔVpeak) was the most common dynamic parameter examined (12/27 studies). Moreover, ΔVpeak provided a reliable diagnostic performance. All studies of ΔVpeak showed that the area under the ROC curve ranged from 0.71 to 1.00, and the cutoff value of ΔVpeak for determining fluid responsiveness ranged from 7% to 20%.

Because patients with congenital heart surgery were included in approximately half of all studies, we allocated participants to 2 new subgroups as follows: the congenital heart surgery subgroup (10–23) and the non-cardiac surgery subgroup (general surgery, neurological surgery, and general PICU patients) (24–36). In congenital heart surgery subgroup, ΔVpeak showed the best sensitivity of 100% at the cutoff value of 7% when performed by transesophageal echocardiogram (TEE) (11). The best specificity of ΔVpeak was 92% at the cutoff values 13%–14% by TEE (13, 21). Another reliable dynamic was the pulse pressure variation (PPV), with the sensitivity of 94% (at the cutoff value of 18%) and the specificity of 100% (at the cutoff value of 30%) (17). In the non-cardiac surgery subgroup, ΔVpeak performed by transthoracic echocardiogram (TTE) showed the best sensitivity of 100% (at the cutoff values 10% and 12.2%) (25, 29) with the best specificity of of 100% (at the cutoff value 10%) (29). Note that plethysmographic variability index (PVI) measured by the transflectance adhesive forehead sensor exhibited the second-best sensitivity of 94.1% (at the cutoff value of 6%) (26), while stroke volume variation (SVV) provided the second-best specificity of 93.3% (at cutoff values 16.5%) (33).

The risk of bias assessment of all included studies is shown in Table 2. The reference standard domain was judged to have a high risk of bias in 9 studies because the interpretation of the reference standard test was made with knowledge of index test results. The flow and timing domain were also judged to have a high risk of bias in 15 studies because all included patients were not in the final analysis (per-protocol analysis).

In 2013, Gan et al. (1) studied static and dynamic parameters, and found that dynamic parameters were more reliable in predicting fluid responsiveness in children. Several new dynamic parameters have since been introduced and studied in the pediatric population during the last 10 years. Therefore, we conducted this review to extend the work of Gan et al. (1) on dynamic parameters and to provide an update with newly examined parameters.

New dynamic parameters from non-invasive ultrasonic cardiac output monitoring, electrical cardiometry, and ultrasound are easily accessible and widely used in the PICU. These new parameters are reliable and can be measured by non-experienced physicians in a few minutes (40, 41). Therefore, they could be useful tools for clinicians to determine whether patients should undergo a fluid challenge.

This systematic review showed that ΔVpeak had a promising diagnostic performance across all populations. The ΔVpeak was studied as a single parameter or together with other dynamic parameters. The cutoff values for predicting fluid responsiveness ranged from 7% to 20%, while the average values ranged from 12% to 14%. In group of congenital heart surgery, the echocardiogram performed by transesophageal technique but in other groups, mostly performed by transthoracic technique. A major disadvantage of ΔVpeak is that this parameter requires an experienced operator of echocardiography.

The highest sensitivity of ΔVpeak in patients who had congenital heart surgery is advantage because fluid overload can increase the risk of acute kidney injury and poor postoperative outcomes in patients with congenital heart disease (42, 43). Therefore, a parameter with high specificity, such as ΔVpeak, could reduce such adverse events and complications by decreasing an unnecessary fluid challenge in this patient subgroup. When ΔVpeak is not accessible, new dynamic parameters from non-invasive methods such as ultrasonic cardiac output monitoring, electrical cardiometry, and arterial line variable parameters should be considered, because of easy accessibility and mostly non-operator dependent methods. Pulse pressure variation could be used as alternative because it also had a high specificity. Patients in the non-cardiac subgroup are most likely to benefit from early fluid resuscitation. The ΔVpeak and PVI should be considered in this context because they have a high sensitivity.

Each study with patients in the congenital heart surgery group reported inotropic and vasopressor administration in various forms, including the percentage of inotrope use in the population and the Vasoactive Inotropic Score, and some studies did not report inotropic or vasopressor data. Therefore, we did not perform analysis for specific dynamic parameters based on inotropic status.

There are some limitations to our study. First, our search strategy was limited to the last 10 years. The reason for his limitation was to focus on new dynamic parameters that appeared after the systematic review in 2013 by Gan et al. (1) Second, there was heterogeneity of the study design, including multiple participant groups in different clinical settings, different fluid types, varying amounts of volume (5–20 ml/kg), and the definition of fluid responsiveness using different parameters across the studies.

The findings from this systematic review suggest some future research opportunities. The ΔVpeak, which is the most reliable parameter for predicting fluid responsiveness in mechanically ventilated children, has not been investigated in children with spontaneous breathing. Preload challenge maneuvers (e.g., calibrated abdominal compression, mini-fluid bolus, the passive leg raising test, and the end-expiratory occlusion test) have been extensively studied in the adult population for predicting fluid responsiveness (44). However, these maneuvers have not been well investigated in pediatric population.

The ΔVpeak exhibited a promising diagnostic performance in predicting fluid responsiveness in mechanically ventilated children. The sensitivity of ΔVpeak is advantageous in non-cardiac surgical patients and the PICU setting because early fluid resuscitation improves survival in these patients. Furthermore, the specificity of ΔVpeak is beneficial in congenital heart surgery because fluid overload is particularly detrimental in this group of patients.