Association of Dynamic Arterial Elastance With Fluid Responsiveness in Critically Ill Children

Abstract

OBJECTIVES:

Cardiac mechanical efficiency has been shown to be a predictor of fluid responsiveness (FR) in adults. Our goal was to assess the association between mechanical efficiency as measured by dynamic arterial elastance (Ea_dyn) and mean arterial pressure (MAP) after fluid bolus in children with MAP less than or equal to 50th percentile for age.

DESIGN:

This was a retrospective, observational cohort study.

SETTING/PATIENTS:

This studied IV crystalloid fluid boluses of greater than or equal to 10 mL/kg given to patients less than or equal to 18 years old within the first 72 hours of admission to an academic PICU.

INTERVENTIONS:

None.

MEASUREMENTS AND MAIN RESULTS:

Ea_dyn was calculated in 10-second intervals during the 20 minutes pre-bolus. FR was defined as an increase of greater than or equal to 10% in MAP from pre-bolus to the average MAP over 20 minutes post-bolus. Kruskal-Wallis test was used to assess associations. We analyzed 490 fluid boluses given to children with MAP less than or equal to 50th percentile for age across 365 PICU encounters. Pre-bolus Ea_dyn was not associated with FR (p > 0.1). This lack of association persisted in subgroup analysis among those mechanically ventilated or on vasoactive medication, and in stratification by MAP percentile for age and duration of time in MAP percentile. Additionally, mechanical efficiency was high (Ea_dyn > 0.7) for most children, even in the lowest MAP percentile for age cohorts.

CONCLUSIONS:

Further research is needed in children to understand the changing cardiac physiology of children as blood pressure decreases to develop more targeted, age-based shock management strategies.

KEY POINTS

Question: Will mechanical efficiency as measured by dynamic arterial elastance (Ea_dyn) predict blood pressure increase with fluid bolus administration in children?

Findings: In this retrospective observational study, Ea_dyn was not associated with an increase in mean arterial pressure (MAP) with IV crystalloid fluid boluses in patients less than or equal to 18 years old and MAP less than or equal to 50th percentile for age, despite stratification by MAP percentile for age and duration of time in MAP percentile. Unlike adult studies, children with MAP less than 10th percentile for age did not demonstrate mechanical inefficiency.

Meaning: Further research is needed to clarify potential pathophysiological differences between cardiovascular failure in children and adults with shock, to better target shock treatments for children.

Resuscitation guidelines for pediatric shock recommend fluid administration followed by vasoactive infusion (1–3), but the optimal volume of fluid administered before a child is considered refractory to fluids and vasoactive medication is initiated remains controversial (3). Children who do not respond to fluid are at risk for organ-damage and mortality from both prolonged time in shock (4–7) and fluid overload (8, 9). Anticipating which children will respond to fluid is challenging, as many tools used to predict fluid responsiveness in adults perform poorly in children (10, 11). This may be due to higher arterial compliance in children (10), more limited capacity to increase stroke volume during stress (12), and age-related changes in microvascular venous capacitance (13). Even if these tools did perform well in children, most only predict an increase in stroke volume with fluid administration, and not necessarily an increase in blood pressure. For those patients who are hypotensive, achieving a normal blood pressure must be the first step in shock treatment, as perfusion pressure to organs must be restored to maintain appropriate blood flow (14). Thus, predicting which children with lower blood pressure will have an increase in blood pressure with fluid administration could lead to more targeted resuscitation and potentially decrease morbidity and mortality.

Dynamic arterial elastance (Ea_dyn) is a measurement of ventriculo-arterial coupling that reflects cardiac mechanical efficiency. It is calculated as the ratio of pulse pressure variation (PPV) to stroke volume variation (SVV) brought about as preload varies with intrathoracic pressure changes due to breathing through one or more respiratory cycles. As Ea_dyn has been shown to vary by both cardiac contractility and systemic vascular resistance (15), it may also reflect cardiovascular receptivity to fluid administration. Indeed, Ea_dyn has been shown to predict a blood pressure increase with fluid bolus administration in adults with hypotension (16–25). On the other hand, early studies of Ea_dyn in children have not found this association (26–28). However, these studies are small and did not account for exposure to vasoactive medications, which may improve the predictive performance of Eadyn in adults (15, 25), or the degree and duration of lower blood pressures, which may show Eadyn to be predictive in children at different mean arterial pressure (MAP) thresholds than in adults.

We hypothesized that the ability of Ea_dyn to predict increase in blood pressure with fluid bolus administration in children would be modified by the degree and duration of lower blood pressures, as well as exposure to vasoactive infusion.

METHODS

Procedures were followed in accordance with the ethical standards of the Helsinki Declaration of 1975 and approved by the Institutional Review Board (IRB) of the Ann & Robert H. Lurie Children’s Hospital of Chicago (IRB No. 2018-1630) on March 1, 2020, with waiver of consent under the study title “Clinical Phenotypes, Organ Dysfunction, and Team Situational Awareness (CODE Aware) Study.”

Study Design and Population

We conducted a retrospective, observational study of IV crystalloid fluid boluses of at least 10 mL/kg given to patients 18 years old or less within the first 72 hours of admission to the Ann & Robert H. Lurie Children’s Hospital of Chicago’s PICU between August 2013 and August 2020 who had an arterial line. All PICU patients with arterial lines have archived arterial waveforms using BedMaster (Hillrom, Chicago, IL), a medical device integration software that stores the high-frequency signal data from the bedside monitor for each patient. As only bolus end times were documented, a 30-minute bolus duration was estimated, and we analyzed data in 20 minutes pre- and post-bolus periods (Fig. 1). At least 2 hours were allowed between boluses to avoid overlapping bolus periods (e.g., two 10 mL/kg boluses within 2 hr were excluded and not treated as one 20 mL/kg bolus, but two 10 mL/kg boluses > 2 hr apart were both included). A bolus volume of 10 mL/kg was selected due to frequency of institutional practice. As BedMaster only stored the data based on the monitor display settings, the top or bottom of the waveform could be lost if not optimized in real-time as the waveform was no longer in the captured display scale (i.e., “clipping”). We included boluses that had at least 30 seconds in the pre-bolus period when both: a) the arterial line MAPs were less than the 50th percentile for age (29) and b) there was no waveform “clipping.”

Figure 1.

Visualization of bolus period; including 30-min estimated bolus time and 20-min pre-bolus and post-bolus periods with associated physiologic variables extracted in each time period. Ea_dyn = dynamic arterial elastance, HR = heart rate, MAP = mean arterial pressure, PPV = pulse pressure variation, SVV = stroke volume variation.

Data Extraction

Pre-bolus age-normalized heart rate (HR), PPV, and SVV derived from stroke volume estimated using the Liljestrand-Zander formula (30, 31) were calculated in R (version 4.0.3, R Core Team, 2020). Beats were detected using a modified version of the WABP algorithm (32), operating on each observation beat-wise to extract the measures, then applying a windowing function to calculate the across-beat metrics. Metrics were then averaged in 10-second intervals during the pre-bolus period to allow for at least one respiratory cycle. HR was normalized for age using median and interquartile range (IQR) based on published ranges in a large PICU population (29). The ratio of 10-second averaged PPV-to-SVV ratio was used to calculate Ea_dyn. Post-bolus MAP was calculated as the average numeric value visualized on the monitor during the post-bolus period rather than the MAP calculated from the waveform to minimize potential loss of responders due to sudden blood pressure increases leading to “clipping.” Fluid responsiveness was defined as MAP increase of at least 10% between pre-bolus hypotensive MAP and averaged 20-minute post-bolus MAP (11). Clinical data, including demographic data and medical treatments, were extracted from the electronic health record using standard queries and data quality assurance checks.

Definitions

“Medical complexity” was defined by the pediatric complex chronic conditions classification system, Version 2 (33). As “hypotension” is poorly defined in children (34, 35), and the MAP threshold at which Ea_dyn would be associated with fluid response in children was unknown, children with less than or equal to 50th percentile for age were selected to start analysis, with the goal to allow for subgroup analysis of decreasing MAP percentile for age. “Suspected infection” was defined as any patient who received antibiotics and microbiological testing within 24 hours.

Statistical Analysis

The Kruskal-Wallis test was used to test for association between fluid responsiveness and pre-bolus Ea_dyn. Subgroup analysis was conducted for those receiving vasoactive medications, those who were mechanically ventilated, bolus volumes above (≥ 15 mL/kg) vs. below average (< 15 mL/kg) for our dataset, and stratification by MAP percentile for age and duration of time in MAP percentile. Analyses were performed using R (Version 4.0.3; R Core Team, 2020, Vienna, Austria).

RESULTS

There were 890 boluses administered to 565 children with stored arterial line waveform data across 539 encounters between August 2013 and August 2020. Fifty-five percent of those (490 boluses given to 356 children across 365 encounters) had at least 30 seconds of MAP less than or equal to 50th percentile for age, no other qualifying bolus within 2 hours, and adequate waveform quality to calculate Ea_dyn. The majority of boluses (81%, 324/400 boluses) that did not qualify were given to patients with MAP greater than 50th percentile for age. Children that met inclusion criteria had high disease severity, with median Pediatric Risk of Mortality III score of 11 (IQR, 7–18), 58% (n = 207) mechanically ventilated, 10% (n = 37) on vasoactive infusion at the time of bolus, and 13% (n = 48) mortality. The majority of these children were medically complex (88%, n = 313) and had suspected infection (82%, n = 291) (Table 1). Children with lower blood pressures were more likely to be fluid responsive (32–45%) (Fig. 2).

TABLE 1.

Demographics and Clinical Characteristics

Variable	All Patients (n = 356)
Age, yr, median (IQR)	6 (2–12)
Immunocompromised (%)	11 (3)
Medically complex (%)	313 (88)
Pediatric Risk of Mortality III score, median (IQR)	11 (7–18)
Suspected infection (%)	291 (82)
Mechanical ventilation (%)	207 (58)
Vasoactive infusion (%)	37 (10)
Length of stay, d, median (IQR)	14 (8–29)
Mortality (%)	48 (13)

IQR = interquartile range.

Data are presented as median (IQR) or total n and percentage (%). Only first bolus per patient represented.

Figure 2.

Patient flow diagram; describing number of included patients in each blood pressure cohort, with total cohort less than 50th percentile for age and subgroup analysis cohorts of less than 25th percentile and less than 10th percentile for age, as well as number of responders and nonresponses in each cohort. MAP = mean arterial pressure.

Pre-bolus Ea_dyn was not associated with fluid responsiveness. This association between pre-bolus Ea_dyn and fluid response was absent in the entire cohort (nonresponders [NR], 1.18 [1.00–1.35] vs. responders [R], 1.14 [0.99–1.36]; p = 0.72), when stratified by bolus volume (< 15 mL/kg: NR, 1.20 [1.01–1.35] vs. R, 1.13 [1.00–1.34]; p = 0.61); greater than or equal to 15 mL/kg (NR, 1.17 [0.99–1.33] vs. R, 1.16 [0.97–1.38]; p = 0.99), those undergoing mechanical ventilation (NR, 1.20 [1.01–1.36] vs. R, 1.13 [0.99–1.36]; p = 0.74), and those exposed to vasoactive infusion (NR, 1.20 [1.04–1.31] vs. R, 1.23 [1.08–1.35]; p = 0.45) (Table 2). Similarly, pre-bolus HR, SVV, and PPV were not associated with fluid responsiveness (Table 2).

TABLE 2.

Pre-Bolus Arterial Waveform Measures Stratified by Fluid Response

Variable	All Boluses	Nonresponders	Responders	p
All boluses (n)	490	332	158
Age-normalized HR,a (IQR)	0.60 (–0.05 to 1.20)	0.61 (–0.04 to 1.18)	0.52 (–0.06 to 1.23)	0.69
PPV, % (IQR)	0.14 (0.10–0.20)	0.14 (0.10–0.20)	0.14 (0.10–0.21)	0.90
SVV, % (IQR)	0.12 (0.09–0.17)	0.16 (0.12–0.23)	0.16 (0.12–0.23)	0.77
Eadyn (IQR)	1.18 (1.00–1.35)	1.18 (1.00–1.35)	1.14 (0.99–1.36)	0.72
On mechanical ventilation (n)	285	189	96
Age-normalized HR,a (IQR)	0.53 (–0.08 to 1.08)	0.53 (–0.11 to 1.11)	0.50 (0.02–1.08)	0.86
PPV, % (IQR)	0.13 (0.09–0.19)	0.12 (0.09–0.18)	0.14 (0.10–0.20)	0.28
SVV, % (IQR)	0.11 (0.08–0.16)	0.11 (0.07–0.16)	0.11 (0.09–0.16)	0.19
Eadyn (IQR)	1.18 (1.00–1.36)	1.20 (1.01–1.36)	1.13 (0.99–1.36)	0.74
On vasoactive infusion (n)	56	38	18
Age-normalized HR, (IQR)a	0.59 (0.16–1.30)	0.52 (0.13–1.32)	0.71 (0.31–1.08)	0.94
PPV, % (IQR)	0.15 (0.10–0.22)	0.14 (0.10–0.22)	0.19 (0.12–0.21)	0.38
SVV, % (IQR)	0.13 (0.08–0.19)	0.12 (0.07–0.20)	0.15 (0.11–0.19)	0.45
Eadyn (IQR)	1.21 (1.05–1.32)	1.20 (1.04–1.31)	1.23 (1.08–1.35)	0.45

Ea_dyn = dynamic arterial elastance, HR = heart rate, IQR = interquartile range, PPV = pulse pressure variation, SVV = stroke volume variation.

^aAge-normalized HR presented as number of IQRs above or below a normalized median of zero based on age.

Data are presented as median (IQR).

The subgroup analysis studying Ea_dyn and fluid response by MAP percentile for age and duration of time in MAP percentile also did not demonstrate a statistically significant association, although this analysis was limited by the sample size. Additionally, Ea_dyn measurements were consistent with high mechanical efficiency (> 0.7) across all degrees and durations of MAP percentile for age (Table 3).

TABLE 3.

Pre-Bolus Dynamic Arterial Elastance Stratified by Fluid Response, Degree and Duration of Hypotension

Duration of Time in MAP Percentile	< 5 min		5–15 min		> 15 min
	n	Eadyn	n	Eadyn	n	Eadyn
		Median (IQR)		Median (IQR)		Median (IQR)
MAP < 10th percentile
Nonresponder	32	1.25 (1.09–1.44)	51	1.23 (1.06–1.40)	55	1.36 (1.18–1.48)
Responder	41	1.22 (1.08–1.40)	40	1.15 (1.05–1.38)	30	1.27 (1.18–1.48)
MAP < 25th percentile
Nonresponder	48	1.26 (1.02–1.34)	73	1.17 (0.98–1.32)	100	1.26 (1.08–1.40)
Responder	46	1.10 (1.00–1.31)	56	1.16 (0.98–1.32)	41	1.25 (1.09–1.45)
MAP < 50th percentile
Nonresponder	138	1.14 (0.96–1.31)	102	1.17 (1.00–1.35)	165	1.21 (1.02–1.37)
Responder	111	1.20 (1.04–1.40)	64	1.12 (0.99–1.30)	58	1.20 (0.99–1.38)

Ea_dyn = dynamic arterial elastance, IQR = interquartile range, MAP = mean arterial pressure, n = number of boluses.

All p values comparing nonresponders to responders in each MAP percentile/time duration subset > 0.1.

DISCUSSION

In this retrospective observational cohort study, we found that pre-bolus Ea_dyn was not associated with MAP increase with fluid bolus in critically ill children with MAP less than or equal to 50th percentile for age. This lack of association persisted in subgroup analysis among those patients who were mechanically ventilated or on vasoactive medications, bolus volumes less than and greater than or equal to 15 mL/kg, and with stratification by MAP percentile for age and duration of time in MAP percentile. These results suggest that Ea_dyn may not be a helpful tool to predict fluid responsiveness in children. Additionally, Ea_dyn measurements in our cohort were high across all stratifications of MAP percentile for age and duration of time in MAP percentile—even when MAP was less than 10th percentile for age—suggesting that hypotension in pediatric shock may not be a state of mechanical inefficiency.

Ea_dyn is a measure of ventriculo-arterial coupling—reflecting the mechanical efficiency of the interaction between left ventricular contractility and afterload by describing how well the energy of the ventricle is transferred to the arterial system (15). Mechanical efficiency has been reported to be clinically useful in managing treatment in adults with hypotension, where adrenergic surge overwhelms the local metabolites that control blood flow to organs during homeostasis (14). As the body will seek to restore flow to the critical pressure-dependent organs first (i.e., brain, heart), physiologic response to treatment will be reflected in MAP (14). This mental framework has been applied successfully both in predicting response to fluid administration (16–25) and in titration of vasopressors (36–40) in adults with hypotension. Of note, in studies in which adults are not hypotensive, Ea_dyn has been shown to be less effective (15). This decrease in effectiveness may be due to blood pressure no longer being a true reflection of physiologic response as local mediators still control organ flow (14) or because vasopressor exposure is required to restore arterial compliance (25).

Despite promising results in adult studies, Ea_dyn has not been shown to be associated with fluid response in children (26–28). All previous pediatric studies, however, have exposures that may limit their applicability to treat shock in a PICU setting: they were all conducted in children who are not hypotensive for age and who were mechanically ventilated in an operating room (OR) setting. Practice location (especially when comparing ORs to ICUs) has been shown to change the predictive performance of both Ea_dyn in adults and general measures of fluid responsiveness in children (11, 25). Additionally, in none of the previous pediatric Ea_dyn studies were patients hypotensive for age. One study (27) did attempt to evaluate hypotension but defined hypotension cutoffs as values that have been found to be around the 50–75th percentile for age in the PICU population (29). Additionally, two of the three previous studies had no exposure to vasoactive medications (26, 28), which has been hypothesized to change the performance of Ea_dyn in adults (25). Our study adds to the body of literature in that it evaluates Ea_dyn in children with varying MAP percentiles for age and duration of time in MAP percentiles, as well as exposure to mechanical ventilation and vasoactive medications in a PICU setting.

The growing body of evidence suggests that mechanical efficiency as reflected in Ea_dyn might not be a useful tool to predict subsequent increase in MAP with fluid bolus in children. Two possible explanations for this include: a) difficulty measuring the components of Ea_dyn in children (41–43) and b) an inherent physiologic difference in shock states between children and adults. The possibility that shock states in children are less frequently a reflection of lack of mechanical efficiency than in adults is supported by our finding of high Ea_dyn values in children with lower blood pressures, consistent with prior findings that ventriculo-arterial coupling measures in children show improved mechanical efficiency as compared with adults (44). In fact, the adult cutoff values of Ea_dyn to predict fluid response in meta-analysis (0.7–0.8) are below the IQR for Ea_dyn in children in our study as well as in previous studies in children (26–28). As we continue to explore the underlying pathophysiology of shock in children, as well as how to leverage measures of shock physiology to inform treatment, we may need to consider that the pathophysiology of hypotension in pediatric shock is different than that of adults. Indeed, Ea_dyn would not be the first cardiovascular measure to suggest a difference between children and adults—only 20% of a cohort of children with fluid refractory shock being monitored with pulmonary artery catheters were found to have the low systemic vascular resistance state commonly associated with adult septic shock (45).

Although our results are compelling and extend prior findings in children, this study has several limitations. First, tidal volume and HR to respiratory rate ratios were unable to be estimated, although the potential effect of these on results are unclear. Second, as only bolus end times were documented, exact bolus timing and speed could not be verified, and it is unclear if the speed of bolus administration will change fluid response (46, 47). Third, the etiology of lower blood pressures for these patients was unknown. It is possible that either: a) there are hemodynamic or pathophysiologic responses that affect blood pressure fluid response to fluid administration that are not completely reflected in cardiac mechanical efficiency or b) cardiac efficiency is not completely reflected in Ea_dyn for children. Fourth, subgroup analysis of degree and duration, as well as on and off vasoactive medications, had small sample sizes, potentially introducing type II error, and therefore should be considered hypothesis generating. Fifth, we did not consider colloid administration. Children receiving colloid only resuscitation have the potential to be sicker (48). Additionally, some of this cohort may have been exposed to both colloid and crystalloid in the study period—changing both the volume and the type of fluid administered. Although recent meta-analyses of fluid responsiveness in children did not find a difference in either tool response (11) or clinical outcome (49) when comparing crystalloid vs. colloid administration, the effects of resuscitation with both types of fluid rather than one or the other is unknown. Sixth, we excluded smaller boluses within each bolus study window (e.g., two 10 mL/kg boluses within 2 hr were excluded and not treated as one 20 mL/kg bolus, but two 10 mL/kg boluses > 2 hr apart were both included). Although we felt this was necessary to avoid overlapping bolus effects, this could also have unintentionally excluded boluses given to patients with suspected cardiac or kidney dysfunction, for whom clinicians might have given smaller boluses more frequently.

CONCLUSIONS

In this retrospective observational cohort study, we found that pre-bolus Ea_dyn was not associated with blood pressure increase after fluid bolus in critically ill children with MAP less than or equal to 50th percentile for age. Additionally, we found that overall mechanically efficiency of critically ill children was high, even when MAP was less than 10th percentile for age. Further research is needed to clarify potential pathophysiological differences between critically ill children and adults with hypotension to develop more targeted, age-based shock management strategies.