Alveolar dead space fraction is not associated with early RV systolic dysfunction in pediatric ARDS

Abstract

Primary Hypothesis:

We hypothesized that higher alveolar dead space fraction (AVDSf) at pediatric acute respiratory distress syndrome (PARDS) onset would be associated with right ventricular (RV) systolic dysfunction within the first 24 h of PARDS.

Study Design and Methods:

We performed a retrospective single-center cohort study of PARDS patients with clinically obtained echocardiograms within 24 h. Primary exposure was AVDSf at PARDS onset. Primary outcome was RV systolic dysfunction as defined by RV global longitudinal strain (GLS) (>−18%). Secondary outcomes included pulmonary hypertension (PH) and RV systolic dysfunction as defined by other echocardiogram parameters, and measures of oxygenation. Unadjusted and adjusted logistic and linear regression were used to investigate AVDSf associations with outcomes.

Results:

Ninety-one patients were included: median age 6.2 years, 46% female, and 65% with moderate or severe PARDS. Median AVDSf was 0.2 (interquartile range [IQR] 0.0–0.3), 33% had RV dysfunction, and 21% had PH. Unadjusted and adjusted logistic regression showed no association between AVDSf and RV systolic dysfunction or PH by any echocardiographic measure, but unadjusted and adjusted linear regression did show an association between AVDSf and PaO₂/FiO₂.

Conclusion:

AVDSf at PARDS onset was not associated with RV systolic dysfunction or PH within 24 h but was associated with PaO₂/FiO₂ ratio and may be more reflective of pulmonary causes of ventilation-perfusion mismatch. Future investigations should focus on clarifying the clinical utility of AVDSf in relation to existing metrics throughout the course of PARDS.

1 |. INTRODUCTION

Pediatric acute respiratory distress syndrome (PARDS) has a high morbidity and mortality rate. While mortality rates are decreasing over time and vary across regions, overall mortality continues to be between 15% and 20%^1–3 with multiorgan dysfunction syndrome and refractory shock accounting for up to 60% of mortality.^4,5 Right ventricular (RV) systolic dysfunction and pulmonary hypertension (PH) have been independently associated with worse outcomes in PARDS, presumably by further impairing oxygenation and increasing the risk of multiple organ dysfunction syndrome.^6,7 Early detection and treatment of RV dysfunction and/or PH in PARDS may improve patient outcomes.

The lung physiologic dead space ratio (VD/VT) can be estimated through the Enghoff modification of the Bohr formula.⁸ The alveolar dead space fraction (AVDSf) is a measure of ventilation-perfusion mismatch that represents the proportion of inhaled air not participating in gas exchange and is a surrogate for alveolar dead space. It is an easily obtained clinical variable at the bedside, requiring only measurement of end-tidal carbon dioxide (PetCO₂) and arterial partial pressure of CO₂ (PaCO₂). AVDSf is low in healthy children and increases in the setting of pulmonary injury and low cardiac output.⁹ Higher AVDSf at the onset of PARDS is associated with mortality.^10,11 As it is affected by perfusion abnormalities from intrinsic pulmonary vascular disease, AVDSf may be a marker for risk of developing RV systolic dysfunction or acute cor pulmonale in adults.¹² However, the prognostic utility of AVDSf in PARDS may also be due to lung overdistension, iatrogenic generation of dead space, or reflect the risks associated with unsafe ventilator pressures. Whether AVDSf is primarily a marker of decreased pulmonary blood flow in the setting of RV dysfunction or PH or is primarily a marker of pulmonary causes of ventilation-perfusion mismatch in children with PARDS is unknown.

The primary objective of this study was to determine the association between AVDSf at PARDS onset with evidence of RV systolic dysfunction by 2D speckle tracking echocardiography. Our hypothesis was that higher AVDSf at PARDS onset would be associated with evidence of RV systolic dysfunction as measured by RV global longitudinal strain on clinically obtained echocardiograms within 24 h of PARDS onset.

2 |. MATERIALS AND METHODS

2.1 |. Study design and data source

This was a retrospective analysis of a single-center prospective cohort study as previously described¹³ who were admitted with PARDS between July 1, 2012, and April 30, 2018, requiring conventional invasive mechanical ventilation, and who had a clinically indicated echocardiogram performed within 24 h of PARDS onset. Other inclusion criteria were availability of paired end tidal PetCO₂ and PaCO₂ measurements as well as echocardiogram images of sufficient quality to perform RV strain analyses. Patients were excluded for PARDS onset outside of the study center PICU, chronic invasive mechanical ventilation, invasive mechanical ventilation >7 days, and for respiratory failure primarily secondary to congestive heart failure, fluid overload, or exacerbation of chronic lung disease. The study was approved by the Children’s Hospital of Philadelphia’s (CHOP) Institutional Review Board with a waiver of the need for informed consent.

2.2 |. Definitions

The first qualifying arterial blood gas for the diagnosis of PARDS was used for the AVDSf calculation and was considered the onset of PARDS for purposes of this study. AVDSf was calculated using the following equation: AVDSf = (PaCO₂ − PetCO₂)/PaCO₂. PaCO₂ and PetCO₂ values were excluded if they were not within 60 min. For patients with a PaCO₂ before PetCO₂ in the medical record (n = 13), chart review confirmed that there were no ventilator adjustments made between values. During the study time period, cuffed endotracheal tubes were exclusively used with PICU respiratory therapists adjusting the cuff using a minimal leak technique (MLT) once per 12-h shift. The chart documentation includes that the MLT was used to adjust the cuff and at what pressure a leak is heard (measured via manometer) but very few instances of an actual percent leak. Although this was inconsistently recorded in the medical records for this cohort, a prior publication out of our institution documented a median endotracheal tube leak of 0% (IQR 0-%), with 95% of patients have a leak less than or equal to 5%.¹¹ Patients with paired PaCO₂ and PetCO₂ values that yielded a negative AVDSf were normalized to a value of 0.

During the study period, there were no protocols in place with respect to the performance of echocardiograms for children with PARDS at our institution and performance was at the discretion of the clinical team. Echocardiograms were analyzed as previously described.⁶ Echocardiographic measures of RV systolic function included: qualitative assessment (normal or mildly, moderately, or severely diminished), fractional area change (FAC), tricuspid annular plan systolic excursion (TAPSE), RV peak systolic global longitudinal strain (RV GLS), and RV peak systolic strain (also known as free wall strain). Consistent with prior publications in adult and pediatric echocardiography, abnormal FAC was defined as less than 35%.¹⁴ Abnormal TAPSE was defined as a value less than −2 standard deviations below published pediatric normal values for age¹⁵ and calculated through an online calculator (parameterz.blogspot.com). Abnormal RV GLS and strain were defined as values greater than −18% and greater than −21%, respectively, as less negative values reflect worse function.^6,7 As pulmonary arterial catheters are rarely used in clinical practice, echocardiographic surrogates of pulmonary hypertension were defined as flattening or bowing of the septal position in systole in the parasternal short-axis view, tricuspid regurgitant velocity > 2.8 m/s², or RV systolic pressure estimate greater than ½ systemic systolic pressure at the time of echocardiogram.

2.3 |. Primary exposure and outcomes:

The primary exposure was AVDSf at PARDS onset analyzed as a continuous variable. The primary outcome was RV systolic dysfunction as defined by abnormal RV GLS (dichotomous as defined above). Secondary outcomes included echocardiogram evidence of pulmonary hypertension and RV systolic dysfunction as defined by RV strain, TAPSE z-score, FAC, and qualitative assessments as well as measures of oxygenation including oxygenation index (OI = mPaw × FIO₂ × 100/PaO₂) and PaO₂/FiO₂ ratio.

2.4 |. Data analysis and power calculations

Statistical analyses were performed with Stata 16.1 (StataCorp). Summary data for continuous variables were reported as median with interquartile ranges (IQR) and for categorical variables as numbers with percentages. Continuous data were compared using the Wilcoxon Rank-Sum test and categorical data were compared by Fisher exact test. The associations between AVDSf at onset and echocardiographic findings within the first 24 h as well as oxygenation metrics were assessed with univariable and multivariable logistic regression. Given the relatively small number of patients with echocardiographic abnormalities, potential confounders were included into the logistic regression model one variable at a time to not overfit the model. Additional analyses included the use of AVDSf as a binary variable based on various cutoffs and the use of RV dysfunction measures as continuous variables. p-values < 0.05 were considered significant for all analyses.

Initial power analysis using the assumptions of abnormal RV GLS rate of 35% in the population and a difference in AVDSf between groups of 0.15 with a standard deviation of 0.2 resulted in a total population of 65 patients required for the study (80% power, α = 0.05).

3 |. RESULTS

During the study timeframe, 115 patients had a clinically performed echocardiogram completed within 24 h of PARDS onset with 91 patients meeting full inclusion criteria (Figure 1). Median age was 6.2 years (IQR 1.9–9.9), 54 were male, and 28 patients (31%) were immunocompromised (Table 1). The most common diagnoses were infectious pneumonia (46%) and sepsis (27%) and the distribution of PARDS severity was: 35% mild, 33% moderate, and 32% severe (Table 1). Median time between PaCO₂ and EtCO₂ values was 15 min (IQR 0–30) and median time from ARDS diagnosis to echocardiogram was 0.9 h (IQR −4.1 to 4.1). 25/91 (27%) of patients were on inhaled nitric oxide (iNO) at the time of AVDSf measurement and 29/91 (32%) of patients were on inhaled nitric oxide (iNO) at the time of echocardiogram. Mortality was 32%.

FIGURE 1.

Study flowchart. PARDS, pediatric acute respiratory distress syndrome

TABLE 1.

Cohort characteristics and outcomes

Characteristic	Total
	N = 91
Age (years)	6.2 (1.9–9.9)
Male/Female	49 (54%)/42 (46%)
Race
White	46 (51%)
Black	25 (27%)
Asian	1 (1%)
Other	19 (11%)
Ethnicity
Hispanic	9 (10%)
Not Hispanic	82 (90%)
Weight (kg)	22.0 (10.6–35.0)
Diagnosis
Infectious Pneumonia	42 (46%)
Aspiration Pneumonia	7 (8%)
Sepsis	25 (27%)
Drowning	5 (5%)
Trauma	2 (2%)
Cardiac Arrest	2 (2%)
Other	8 (9%)
Stem Cell Transplant	12 (13%)
Immunocompromised	28 (31%)
PRISM III Score	14 (8–23)
Oxygenation Index at onset	10.3 (6.2–17.6)
PaO2/FiO2 at Onset	165 (98–246)
PARDS Severity at onset
Mild	32 (35%)
Moderate	30 (33%)
Severe	29 (32%)
AVDSf at Onset	0.2 (0.0–0.3)
RV GLS (%)	−20.3 (−24.2 to −16.7)
Abnormal RV GLS	30 (33%)
Pulmonary Hypertension	19 (21%)
ICU days	14 (8–23)
Ventilator-free days at 28 days	11 (0–19)
Death	29 (32%)

Note: Data are presented as median (IQR) for continuous measures, and n (%) for categorical measures.

Abbreviations: AVDSf, alveolar dead space fraction; kg, kilograms; PARDS, pediatric acute respiratory distress syndrome; PRISM III, Pediatric Risk of Mortality III; RV GLS, right ventricular global longitudinal strain.

Median RV GLS value for the cohort was −20.3% (IQR −24.2 to −16.7) with 30 patients (33%) having abnormal RV GLS (> −18%). 19 patients (21%) were found to have PH. The median values (IQRs) for other echocardiographic measures of RV systolic function are displayed in Table S1. There were no differences in the distribution of AVDSf between those with and without abnormal RV GLS or those without or without echocardiographic evidence of PH (Figure 2). There was no difference in RV dysfunction as defined by abnormal GLS between those with and without echocardiographic evidence of PH (7/19 [37%] and 23/72 [32%], Fisher’s exact p = 0.785).

FIGURE 2.

Box plots of alveolar dead space fraction (AVDSf) values in patients with normal and abnormal right ight ventricular (RV) global longitudinal strain (GLS) as well as with and without echocardiographic evidence of pulmonary hypertension (PH).

Unadjusted logistic regression showed no association of AVDSf at onset with the primary outcome of echocardiographic evidence of RV systolic dysfunction defined as RV GLS > −18% nor with the secondary outcome of echocardiographic evidence of PH within 24 h of PARDS diagnosis (Table 2). This remained true when adjusting individually for other clinically important covariates including immunocompromised status, PRISM III score, OI at onset, iNO use at the time of echocardiogram, and age (Table 2). Similar analyses showed no association of AVDSf at onset with other secondary outcomes of echocardiographic measures of RV systolic function (Table S2). Sensitivity analysis was performed with a cohort that excluded patients with PetCO₂ value before PaCO₂ (n = 13) with no change in the results for the primary analyses (Supplemental Table 3). The use of AVDSf as a binary variable based on various cutoffs (0.15, 0.25, and 0.35) and the use of RV dysfunction measures as continuous variables with the linear regression models showing no association between AVDSf at the onset with measures of RV systolic dysfunction or PH (data not shown).

TABLE 2.

Logistic regression models for the primary outcome of echocardiographic evidence of RV systolic dysfunction defined as RV GLS > −18% and secondary outcome of echocardiographic evidence of pulmonary hypertension within 24 h of PARDS diagnosis with the primary predictor of AVDSf at PARDS onset

Covariate	RV GLSOdds ratio (95% CI)	P	PHOdds ratio (95% CI)	p
AVDSf at onset (unadjusted)	0.56 (0.03–9.72)	0.691	5.25 (0.23–118.6)	0.297
AVDSf at onset adjusted for:
Immunocompromised	0.31 (0.01–5.94)	0.437	4.75 (0.19–118.5)	0.341
PRISM III	0.53 (0.03–9.28)	0.667	7.66 (0.27–221.4)	0.236
OI at onset	0.50 (0.03–8.70)	0.638	14.3 (0.43–475.4)	0.136
iNO at echo	0.56 (0.03–9.65)	0.689	5.27 (0.23–119.4)	0.297
Age	0.57 (0.03–10.2)	0.701	4.60 (0.19–109.3)	0.345

Abbreviations: AVDSf, alveolar dead space fraction; iNO, inhaled Nitric Oxide; PH, pulmonary hypertension; PRISM III, pediatric risk of mortality III; OI, oxygenation index; RV GLS, right ventricular global longitudinal strain.

AVDSf at onset was associated with PaO₂/FiO₂ ratio but not OI in both unadjusted and adjusted linear regression models (Table 3). There was no difference in AVDSf at onset between different ARDS severities (Kruskal-Wallis p = 0.7481, Cuzick non-parametric test of trend p = 0.454, Figure S1).

TABLE 3.

Linear regression models for secondary outcomes of measures of oxygenation with the primary predictor of AVDSf at PARDS onset.

Covariate	OI at onsetβ coefficient (95% CI)	P
AVDSfa at onset (unadjusted)	0.81 (−0.95 to 2.57)	0.360
AVDSfa at onset adjusted for:
Immunocompromised	0.82 (−1.00 to 2.64)	0.375
PRISM III	0.73 (−1.03 to 2.49)	0.412
iNO at time of echo	0.76 (−0.91 to 2.43)	0.368
Age	0.79 (−1.00 to 2.58)	0.380
	PaO2/FIO2 at onset β coefficient (95% CI)	p
AVDSfa at onset (unadjusted)	−13 (−23 to −3)	0.014
AVDSfa at onset adjusted for:
Immunocompromised	−13 (−23 to −2)	0.019
PRISM3	−13 (−23 to −2)	0.017
iNO at time of echo	−13 (−23 to −3)	0.013
Age	−13 (−24 to −3)	0.013

^aper every 0.1 increase in AVDSf.

Abbreviations: AVDSf, alveolar dead space fraction; FiO2, fraction of inspired oxygen; iNO, inhaled nitric oxide; OI, oxygenation index; PaO₂, arterial partial pressure of oxygen; PRISM III, pediatric risk of mortality III.

4 |. DISCUSSION

In this retrospective cohort study of PARDS patients with clinically obtained echocardiograms, there was no association between AVDSf at PARDS onset and RV systolic dysfunction or pulmonary hypertension within 24 h of PARDS diagnosis. AVDSf was, however, associated with the PaO₂/FiO₂ ratio at PARDS onset, suggesting that the primary determinant of AVDSf early during PARDS was pulmonary causes of ventilation-perfusion mismatch rather than reduced pulmonary blood flow from increased pulmonary vascular resistance.

In pediatrics, elevated AVDSf, RV systolic dysfunction, and PH have all been independently associated with mortality and worse outcomes in PARDS.^6,7,10,11 As alveolar dead space is increased in low cardiac output states and in intrapulmonary shunt, elevated AVDSf likely reflects a combination of these physiologies. Our primary hypothesis was that higher AVDSf would be associated with echocardiographic evidence of RV systolic dysfunction and PH due to an increase in pulmonary vascular resistance. Although we did not demonstrate a statistically significant association, the estimated effect size for PH was > 4 for all models (Table 2) and the wide confidence intervals suggest that we were likely limited in power for our outcomes. As pulmonary arterial catheters are rarely used in modern PARDS management, prospective and protocolized echocardiograms may also allow for other noninvasive surrogates of elevated pulmonary vascular resistance or RV-pulmonary arterial coupling to be more fully investigated. Alternatively, it is possible that pulmonary, rather than vascular, physiology ultimately determines AVDSf early in PARDS. In either case, the clinical utility of AVDSf requires clarification in future larger studies. Furthermore, our data highlight the heterogeneity of factors influencing AVDSf, some of which will require either adjustment or stratification in larger studies.

We did find an association between AVDSf and the secondary outcome of PaO₂/FiO₂ ratio, but not OI, at PARDS onset. It may be that very early during PARDS higher AVDSf could be more reflective of intrapulmonary shunt as positive pressure is being initiated and titrated to maximize lung recruitment. In one study of sheep models of early ARDS, increases in estimated AVDSf by PET scans were attributable primarily to redistribution of perfusion likely related to mechanisms including hypoxic pulmonary vasoconstriction and gravitational effects with redistribution of perfusion away from nondependent lung regions with no change in mean pulmonary artery pressures.¹⁶ Further, compared to other PARDS cohorts, our study cohort had a relatively high AVDSf at onset (median 0.2, IQR 0.0–0.3) and had a higher severity of illness. This may have created selection bias in our sample that could have biased toward the null hypothesis.

One recent adult study demonstrated that increases in AVDSf were associated with relative increases in echocardiographic surrogates for pulmonary vascular resistance over the course of ARDS,¹² suggesting that a changing and increasing AVDSf over time may be an indicator of those who may be at risk for RV systolic dysfunction. As new or persistent RV systolic dysfunction during the first 8 days of PARDS is associated with worse outcomes,⁷ future studies in pediatric patients should investigate the longitudinal association of changing AVDSf as a marker or potential predictor of PARDS-related pulmonary hypertension or subsequent RV dysfunction. Similar to studies showing that oxygenation metrics and pulmonary mechanics early in PARDS are not associated with pulmonary injury and outcomes,^4,13,17 it is possible that using AVDSf at later time points, the trajectory of AVDSf, or the tracking of AVDSf in populations at known risk for PH (e.g., stem cell transplant recipients or chronic lung disease) will have more value, particularly as PARDS pathophysiology evolves into worsening hypoxemia and/or multiple organ dysfunction syndrome. Finally, the fact that we did not observe differences in AVDSf between categories of PARDS severity highlights the need for future studies to determine if the primary reason for elevated AVDSf (pulmonary vascular dysfunction vs. primary pulmonary cause) may be influenced by other factors, such as age.

Our study has limitations including its retrospective design and single-center cohort. We selected PARDS patients who had a clinically obtained echocardiogram and available PetCO₂ data, which restricted our cohort to approximately 26% of the total patients from the PARDS database during the study timeframe. The clinically performed echocardiograms analyzed for this study were performed close to the time of the research-defined PARDS onset time and may have been too early in the course of PARDS to detect RV dysfunction from increased afterload. As discussed above, this resulted in a cohort with a higher severity of illness with bias toward patients more likely to have hemodynamic instability or specific diagnoses, such as sepsis. Therefore, generalizability to all PARDS patients. Another limitation from our retrospective design was the quality of ETCO₂ data available for review. ETCO₂ values were directly recorded from the electronic medical record using the closest value within an hour of the first qualifying blood gas at PARDS onset and had a median time from blood gas of 15 min. We did verify that no ventilator changes were made between the measurements, however, it is possible that the ETCO₂ value may have changed at the time of the blood gas. We attempted to control for iNO exposure in the statistical models and, while there were 4 patients (4.3% of the total cohort) who were on iNO at the time of echo who were not on iNO at the time of AVDSf determination, it was not possible from the medical record to discern the intent of iNO for all patients. Finally, ideally alveolar dead space as measured by volumetric capnography would also be included as the correlation of alveolar dead space estimated with that method compared to AVDSf is reduced in patients with hypoxemia.^8,9

5 |. CONCLUSION

In this retrospective cohort study, alveolar dead space fraction at PARDS onset was not associated with RV systolic dysfunction or pulmonary hypertension identified by echocardiography within 24 h of PARDS onset and may be more closely associated with pulmonary causes of ventilation-perfusion mismatch. Future investigations should focus on clarifying the clinical utility of AVDSf in relation to existing metrics throughout the course of PARDS and for high-risk subpopulations.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.