Driving pressure is associated with outcome in pediatric acute respiratory failure

Abstract

Objective:

Driving pressure (DP) [ratio of tidal volume over respiratory system compliance] is associated with mortality in acute respiratory distress syndrome (ARDS). We sought to evaluate if such association could be identified in critically ill children.

Design:

We studied the association between DP on day 1 of mechanical ventilation and ventilator-free days at day 28 (VFD-28) through secondary analyses of prospectively collected physiology data.

Setting:

Medical-surgical university hospital pediatric intensive care unit.

Patients:

Children younger than 18 years (stratified by Pediatric Mechanical Ventilation Consensus Conference clinical phenotype definitions) without evidence of spontaneous respiration.

Interventions:

Inspiratory hold maneuvers

Measurements and Main Results:

Data of 222 patients with median age 11 (2 – 51) months was analyzed. Sixty-five (29.3%) patients met PEMVECC criteria for restrictive and 78 (35.1%) for mixed lung disease, and 10.4% of all patients had ARDS. DP calculated by the ratio of tidal volume (Vt) over respiratory system compliance (Crs) [Vt/Cr] for the whole cohort was 16 (12 – 21) cmH₂O and correlated with the static airway pressure gradient (plateau pressure minus positive end-expiratory pressure [PEEP]) (r_s .797, p < .001). Bland-Altman analysis showed that the dynamic pressure gradient (peak inspiratory pressure minus PEEP) overestimated driving pressure (levels of agreement −2.295 to 7.268). Rematching the cohort through a double stratification procedure (obtaining subgroups of patients with matched mean levels for one variable but different mean levels for another ranking variable) showed a reduction in VFD-28 with increasing driving pressure in patients ventilated for a direct pulmonary indication. Competing risk regression analysis showed that increasing driving pressure remained independently associated with increased time to extubation (p < 0.001) after adjusting for PRISM III-24hr score, presence of direct pulmonary indication jury and oxygenation index.

Conclusions:

Higher driving pressure was independently associated with increased time to extubation in mechanically ventilated children. Dynamic assessments of driving pressure should be cautiously interpreted.

Introduction

Mechanical ventilation is one of the most practiced interventions in the pediatric intensive care unit (PICU). Although unmistakable lifesaving, it can also injure the lung (i.e., ventilator induced lung injury [VILI]) when used inappropriately. Key mechanisms involved in VILI include volutrauma (i.e. the delivery of too large tidal volume (Vt)) and atelectrauma (i.e. the repetitive opening and closure of alveoli) (1). The importance of volutrauma became apparent after the publication of the National Heart, Lung and Blood Institute ARDS Network trial, reporting lower mortality rates in critically ill adult ARDS patients randomized to low tidal volumes (Vt) of 6 mL/kg ideal bodyweight (IBW) and plateau pressures (Pplat) less than 30 cmH₂O compared to 12 mL/kg IBW (2) and Pplat < 50 cmH₂O (2). Positive end-expiratory pressure (PEEP) may minimize atelectrauma, but it is unclear how much PEEP is required (3). DP is the ratio of Vt over respiratory system compliance (Crs) – or simplified plateau pressure [Pplat] minus PEEP – and may better reflect lung stress (4, 5). The concept was introduced by Amato and colleagues after reanalyzes of data from nine randomized trials comparing ventilation strategies in adult ARDS patients showing that DP best predicted outcome (6). Subsequent investigations reported DP thresholds ranging from 14 to 18 cmH₂O being associated with adverse patient outcome (7–11).

The clinical implications of DP in children have little been studied. Yehya and colleagues reported that DP was not retained as independent predictor for mortality in children with ARDS identified using the Berlin Definition, but no specific threshold was tested (12, 13). In contrast, one group of investigators recently showed that DP 15 ≥ cmH₂O was associated with significant morbidity (i.e., duration of ventilation >7 days and duration of ICU stay >10 days, or occurrence of death) in 101 children with acute hypoxemic respiratory failure (AHRF) (14). Importantly, DP examined in these studies was defined as peak inspiratory pressure (PIP) instead of Pplat minus PEEP which is not actual driving pressure because PIP is significantly influenced by the resistive properties of the respiratory system leading to overestimating the actual Pplat as we have shown in bench testing (15). We therefore sought to test the agreement between DP calculated by Vt/Crs versus Pplat – PEEP (i.e., static pressure gradient) or PIP – PEEP (i.e., dynamic pressure gradient), and to study if there was an association between DP and ventilator-free days at day 28 (VFD-28) in a heterogeneous group of mechanically ventilated children after stratification for disease category proposed by the Pediatric Mechanical Ventilation Consensus Conference (16).

Methods

This study was designed as a secondary analysis of routinely prospectively collected demographic, physiological and laboratory data from sedated and/or paralyzed children < 18 years with and without lung injury on weekdays at 8am. Data from patients with congenital or acquired heart disease, obstructive airway disease, documented chronic lung disease, neuromuscular disorders, severe traumatic brain injury (i.e., Glasgow Coma Scale < 8), chronic lung disease (i.e., tracheostomy ventilation), severe pulmonary hypertension, managed with high frequency oscillation ventilation or with an ETT leakage > 18% were excluded. We also excluded patients if the chronological age reduced by the number of weeks born before 40 weeks of gestation was < 0. Severity of disease was assessed by the 24 – hr Pediatric Risk of Mortality III (PRISM III-24hr) (17). The Pediatric Mechanical Ventilation Consensus Conference (PEMVECC) definition was used to stratify patients based on their admission diagnosis and PARDS was identified using the pediatric acute lung injury consensus conference (PALICC) definition (see electronic supplemental material for definition of the PEMVECC criteria) (16, 18). The Institutional Review Board (IRB UMCG; number 2017/599) approved the study and waived the need for informed consent.

We exclusively ventilate patients in a PC ventilation mode (AVEA^™, Vyaire, Mettawa, Ill, USA), limiting inspiratory pressures ≤ 28 cmH₂O ( ≤ 32 cmH₂O when there was clinically suspected decreased chest wall compliance) and expiratory Vt (Vt-exp) 5 – 7 mL/kg actual bodyweight (as there was no obesity in the patient cohort [i.e., no patient with SDS < −2 SD or > +1 SD]). Vt-exp was measured near the Y-piece in children < 10 kg using a proximal flow sensor (VarFlex^™, Vyaire, Mettawa, Ill, USA). We only use cuffed endotracheal tubes (ETT). Setting the ventilator breath rate is dictated by the underlying pathology and age; we routinely carefully monitor the flow-time scalar to identify if the inspiratory time setting was appropriate and there was no development of intrinsic PEEP. This means that in our unit the I:E ratio is not fixed. Initial PEEP is 4–6 cm H₂O and further titrated at the discretion of the attending physician, targeting SpO₂ 88 – 92% for patients with lung injury. Unless dictated otherwise, target pH is > 7.20. Patients who breathed spontaneously were not studied as the AVEA does not allow for Pplat to rise above PIP during an inspiratory hold (19). High-frequency oscillatory ventilation (HFOV) is used per unit-specific algorithm as described elsewhere in patients who cannot be supported within the aforementioned conventional ventilation targets (20). Use of neuromuscular blocking agents was at the discretion of the bedside team.

Demographic, physiological and laboratory data were manually extracted from the patient’s electronic medical record. Ventilator settings and parameters were read from the ventilator by the study team not involved in the daily care for the patient. The ETT cuff was inflated prior to the measurements to prevent air leak. Plateau pressure (Pplat) and quasi-static compliance (Crs) were measured at end-inspiration by a manual inspiratory hold maneuver of 3 seconds. Absence of flow was confirmed by visual inspection of the flow-time scalar.

Data analysis

To assess oxygenation (with SpO₂ < 98%), we calculated the OI ([mean airway pressure * FiO₂ * 100] / PaO₂) or the oxygen saturation index [OSI] [mean airway pressure * FiO₂ * 100] / SpO₂) when no indwelling arterial line was available. We calculated ventilator-free days (VFD) through day 28, defined as the number of days within 28 days that a subject is alive and free of MV (21). Patients were assigned 0 VFD if they remained intubated or died prior to day 28 without remaining extubated for more than 24 hours.

DP was calculated by Vt / Crs. The dynamic pressure gradient was calculated by PIP minus PEEP and the static pressure gradient by Pplat – PEEP. For analytical purposes, we grouped patients as “ventilated for a direct pulmonary indication” if they met PEMVECC clinical phenotype criteria for restrictive and mixed lung disease, and patients with normal lung mechanics as “ventilated for indirect pulmonary indication (i.e., post-operative patients, patients with alterations in respiratory drive or airway protective reflexes).

Statistical analysis

Normality of data was assessed using the Kolmogorov-Smirnov test. Continuous data are presented as median and interquartile range (IQR) and analyzed using the Mann Whitney-U (for comparing two groups) or Kruskal Wallis test; Spearman correlation coefficient (r_s) was calculated to analyze correlations. The χ² test with Yates continuity correction was used to analyze categorical data. The primary outcome measure was VFD at day 28 (VFD-28). Mirroring the study by Amato et al, we wanted to study if DP would provide prognostic information independent of Pplat or PEEP (6). For that purpose, we performed a similar double-stratification resampling, producing subgroups of patients with matched mean levels of one variable but distinct mean values for another ranking variable. Based on our total sample size, for meaningful comparisons we chose to create three clusters. Competing risk regression analysis (Fine and Gray model) was used to analyze time to extubation with failed extubation at day 28 as failure interest and death as competing risk (22). All statistical analyses were performed using SPSS v25.0 (IBM Corp., Chicago, III, USA) or STATA v16 (StataCorp LLC, College Station Texas, USA). P <.05 was considered statistically significant.

Results

Data of 222 patients with median age 11 (2 – 51) months was analyzed (Table 1). Of these, 119 patients (53.6%) were younger than 12 months of age. Primary admission diagnosis was respiratory failure in 143 (64.4%) patients. Sixty-five (29.7%) patients met PEMVECC clinical phenotype criteria for restrictive and 78 (35.1%) for mixed lung disease, and 10.4% of all patients met PALICC criteria for PARDS. Duration of MV was 95 (49 – 162) hours. Nine (4.1%) patients died. Patients were ventilated for a median of 15 hrs (11 – 21) before the first measurement on day 1 was made. There was an overall use of low PEEP and discordance with the low PEEP/FiO₂ grid, especially when the FiO₂ > .45.

Table 1.

Characteristics of the study population of 222 mechanically ventilated children with and without lung injury. Data are presented as median (25–75 interquartile range) or percentage (%) of total. Ventilator-free days (VFD) through day 28 were defined as the number of days within 28 days that a patient was alive and free of mechanical ventilation. Patients were assigned 0 VFD if they remained intubated or died prior to day 28 without remaining extubated for more than 24 h.

	Patient stratification per Pediatric Mechanical Ventilation Consensus Conference clinical phenotype based on admission diagnosis
PEMVECC stratum	Normal respiratory system mechanics	Restrictive respiratory system mechanics	Mixed respiratory system mechanics
Ventilation indication	Indirect pulmonary indication	Direct pulmonary indication
	N = 79	N = 65	N = 78
Age (months) *	49 (9 – 122)	23 (8 – 58.0)	2.0 (1 – 9)
≤12 months (%)	23 (29.1)	26 (40.0)	66 (84.6)
Male gender (%)	49 (62.0)	36 (55.4)	49 (62.8)
Weight (kg) *	16.0 (8.0 – 32.0)	12.4 (8.4 – 17.5)	5.0 (4.0 – 6.7)
PRISM III 24 hr score	9 (5. – 15)	10 (5 – 15)	11 (8 – 15)
PARDS (%)	0 (0.0)	4 (6.1)	19 (24.4)
Cstat (cmH2O/L/kg) *	0.60 (0.45 – 0.79)	0.58 (0.43 – 0.73)	0.34 (0.28 – 0.44)
PIP (cmH2O) *	19 (17 – 22)	20 (17 – 24)	27 (24 – 29)
Pplat (cmH2O) *	16 (14 – 19)	17 (15 – 22)	24 (21 – 26)
PEEP (cmH2O) *	5 (5 – 6)	5 (5 – 6)	6 (5 – 7)
Vt-exp (mL/kg) *	7.5 (6.5 – 8.6)	7.6 (6.7 – 8.6)	6.6 (6.1 – 7.7)
Inspiratory time (sec) *	0.75 (0.64 – 0.90)	0.70 (0.60 – 0.75)	0.55 (0.50 – 0.60)
Mandatory breath rate (/min) *	22 (17 – 30)	23 (20 – 30)	40 (32 – 40)
Oxygenation index	2.7 (2.0 – 4.1)	2.7 (2.2 – 4.0)	6.5 (5.1 – 9.8)
Ventilation time (hr) *	89 (46 – 163)	74 (25 – 188)	108 (78 – 152)
VFD day 28	24 (20 – 26)	24 (19 – 26)	23 (21 – 24)
PICU mortality (%)	4 (5.1)	3 (4.5)	2 (2.6)

^*Denotes p < 0.005 (Kruskal Wallis test). PRISM Pediatric Risk of Mortality; OI oxygenation index

DP analysis whole cohort

DP for the whole cohort was 16 (12 – 21) cmH₂O, comparable between male and female patients and not statistically significantly correlated with age. DP correlated significantly with Plat (r_s .745, p < .001) and PEEP (r_s .271 p < .001) but not statistically significantly with Vt (mL/kg). Sixty-one patients received neuromuscular blocking agents on day 1. These patients had a significantly higher DP (21 [17 – 23] vs 14 [12 – 19] cmH₂O, p < .001). DP significantly correlated with the static pressure gradient (r_s .797, p < .001) and dynamic pressure gradient (r_s .853, p < .001). Average bias between DP and dynamic pressure gradient was −0.09 ± 3.21 [levels of agreement −6.370 to 6.192] and 2.49 ± 2.44 [levels of agreement −2.295 – 7.268] for the static pressure gradient. (Figure 1). There was a significant correlation between DP and OI (r_s .644, p < .001), PaO₂/FiO₂ (r_s −.527, p < .001) and the SpO₂/FiO₂ ratio (r_s −.404, p < .001). Patients who met PARDS criteria had a significantly higher DP (20 [17 – 22] vs 16 [12 – 20], p = .002). DP was significantly higher among N = 27 patients subsequently managed with HFOV during PICU admission (23 [20 – 24] vs 16 [12 – 19] cmH₂O, p < .001). In patients with restrictive lung mechanics clinical phenotype, the bias (SD) between Vt/Crs and the dynamic pressure gradient was 0.39 (3.74), for patients with mixed lung mechanics clinical phenotype this was −0.04 (2.91).

Figure 1.

Level of agreement between driving pressure calculated as the ratio of tidal volume over respiratory system compliance versus calculated by plateau pressure minus positive end-expiratory pressure (panel A) and between driving pressure calculated by plateau pressure minus positive end-expiratory pressure versus peak inspiratory pressure minus positive end-expiratory pressure (panel B). The dashed lines represent the upper and lower level of agreement, the shaded areas depict the 95% confidence interval of these levels.

DP PEMVECC clinical phenotype analysis

The distribution of DP on day 1 was significantly different (p < 0.001) across the cohort after stratification by PEMVECC criteria. Patients meeting criteria for the clinical phenotype “normal lung mechanics” had the lowest DP (13 [10 – 17] cmH₂O), whereas patients with “mixed lung mechanics” showed the highest DP (21 [18 – 23] cmH₂O). We found that patients who were ventilated for an indirect pulmonary indication had a significantly lower DP (13 [11 – 17] vs 18 [14 – 22] cmH₂O, p < .001).

Figure 2 graphically depicts VFD-28 in patients mechanically ventilated for an indirect and direct pulmonary indication matched by PEEP, DP and Pplat. Pplat increased significantly (p < 0.001) in all three resampling groups, whereas DP significantly increased (p < 0.001) when matched for PEEP and decreased when matched for Pplat. We observed a significant decrease in VDF-28 when there was increasing DP only in patients ventilated for a direct pulmonary indication. There was no relationship in the groups matched for DP or Pplat.

Figure 2.

Ventilator free days at day 28 after grouping the cohort (N = 222) into terciles. A double stratification procedure (obtaining subgroups of patients with matched mean levels for one variable but very different mean levels for another ranking variable) was used. The upper left panel represents the median (interquartile range) values for matched positive end-expiratory pressure (PEEP) [black bars] and driving pressure (ΔP) [grey bars], the upper middle panel for matched DP and the upper right panel for matched plateau pressure (Pplat). The lower panels show ventilator free days at day 28 (median [interquartile range]) for each matching. Ventilator free days at day 28 was significantly reduced in patients with lung injury when driving pressure increased. The numbers on the X-axis represents the size of tercile. S1 Sampling tercile 1, S2 Sampling tercile 2, S3 Sampling tercile 3

In an additional analysis including only in patients ventilated for a direct pulmonary indication (N = 143), we observed that VFD-28 was significantly different when DP was dichotomized at 15 cmH₂O (25 [IQR 22 – 27] vs 23 [IQR 20 – 24], p < .001). We observed a statistically significant difference in VFD-28 between the four DP strata (p < .001) in these patients (Figure 3). Competing risk regression analysis of the whole cohort showed that DP (modelled per 1 cm H₂O) remained independently associated with ventilation time (hazard ratio 0.58 [95% CI 0.44 – 0.79], p < 0.001) after adjusting for PRISM III-24hr score, presence of lung injury and OI on day 1 (Supplemental Table). Sensitivity analysis excluding 27 patients managed with HFOV during PICU admission yielded similar results (data not shown).

Figure 3.

Relationship between stratified driving pressure (DP) and ventilator free days at day 28 (VFD-28) in patients ventilated for a direct pulmonary indication (N = 143). Data are depicted as median [interquartile range]. A Kruskal-Wallis H test showed that there was a statistically significant difference in VFD-28 between the four DP strata.

Discussion

This is the first pediatric study examining the clinical relevance of DP in mechanically ventilated children. The main finding was that higher DP was significantly associated with an increased time to extubation after adjusting for disease severity and MV indication. Lack of higher PEEP levels did not allow to study if lowering DP by increasing PEEP would be associated with improved outcome. The dynamic airway pressure gradient overestimated DP.

Following the pioneer work by Amato et al, targeting DP < 15 cmH₂O in ARDS has been identified as evolving standard of care (6, 23). The LUNG-SAFE study reported increased mortality in adults with moderate or severe PARDS with DP 14 cmH₂O or above, and others reported that patients who went on to develop ARDS during ICU admission had higher baseline DP values (24, 25). However, it remains to be studied if DP should be interpreted as determinant of outcome or as marker of disease severity (26, 27). On top of that, randomized trials demonstrating the superiority of a ventilatory strategy focused on limiting DP have not been performed so far except for one small trial examining feasibility of a DP guided strategy (28). For children, whether a strategy limiting DP would be beneficial remains to be studied. Unlike adults, in pediatric patients the ratio of transpulmonary pressure to lung strain is different between patients with ARDS and healthy controls, making DP thresholds safe in one lung condition but harmful in another (29, 30). We found an increased time to extubation associated with increasing DP after adjusting for disease severity and presence of a direct pulmonary ventilation indication, but obviously our study could not detect causality. In fact, several arguments may be proposed questioning a rapid implementation in pediatric ARDS. First, the concept of DP was derived from adult ARDS patients who were ventilated with a volume controlled (VC) mode of ventilation. With VC, the operator selects the Vt in the range of 5 – 7 mL/kg predicted bodyweight, although the allowable Vt decreases with decreasing lung compliance (31). Simplifying this pathophysiological mechanism into DP makes it easier for the bedside team because DP is a function of both Vt and compliance. In pediatrics, pressure controlled (PC) is the predominant mode of ventilation. The operator limits the inspiratory pressure and the delivered Vt is dictated by respiratory system compliance (32). It may therefore be surmised that in pediatrics the concept of DP unknowingly already has been applied to a certain degree and provide some explanation as to why in a meta-analysis that included pediatrics patients not a single Vt threshold was associated with increased mortality (33). Second, improving outcome by limiting DP can be achieved by increasing PEEP (6). However, pediatric critical care practitioners tend to use low levels of PEEP and inherently accept higher FiO2₂ (34, 35). Worrisome were the findings by Khemani et al, showing an increase in mortality when there was poor adherence to the ARDSNet PEEP/FiO₂ grid in a large sample of PARDS patients (36). It may therefore be surmised that better PEEP strategies are required to make limiting DP a concept that might be adopted into pediatric critical care. In our cohort, we also found low PEEP being used. On the other hand, there are at present no pediatric randomized controlled trials showing that higher PEEP improved patient outcome and a recent systematic review of adult data reported that the routine use of higher PEEP did not reduce mortality in unselected patients with ARDS (3).

Our unit has a liberal use of HFOV driven by a unit-specific algorithm, which may also explain why we could not ascertain an association between lowering DP and improved patient outcome in patients ventilated for a direct pulmonary indication (20). A sensitivity analysis excluding the 27 patients who were managed with HFOV during PICU admission was not different from the results of primary analyses that included these patients.

To mirror the pivotal work by Amato et al, we chose the ratio of Vt over Crs as our primary approach to measuring DP, although we acknowledge that volume measurements can be less reliable than pressure measurements in children. Reassuringly, we observed a good correlation with DP calculated by Pplat – PEEP. Our data also confirmed that using PIP instead of Pplat to calculate DP in the passive patient overestimates true DP. Furthermore, younger pediatric patients are often ventilated for lung diseases that are characterized by amongst others increased airway resistance, further adding to an overestimation of DP if PIP is used (15). This signifies the need for (manually applied) inspiratory holds even in the youngest pediatric patient.

Our study has several limitations. First, it was designed a single-center study that may limit the generalizability of our findings, although our unit is comparable to most large units globally. In addition, the clinical relevance of DP has only been studied in ARDS, yet our heterogeneous cohort included only a small group of PARDS patients. Furthermore, most patients in our cohort were infants, calling for studies examining the clinical implications of DP in selected groups of patients including also older children. Second, although we showed that DP on day 1 was independently associated with VFD-28 after adjusting for disease severity and PEMVECC clinical phenotype lung pathology type, this does not mean that there is a causative relationship between DP and patient outcome since other confounders may not have been picked up. For example, the liberal use of HFOV as alternative mode of ventilation may have impacted total ventilation time and thus our primary outcome measure (37, 38). Also, there was no use of extubation readiness testing and no clinical algorithm guiding NMBA use, thereby affecting total ventilation time (39). Aside from, VFDs offer inferential benefits but also has several limitations, with the main criticism being that it does not adequately distinguish between component risks (21). Many events unaccounted for during the PICU admission may impact our primary outcome. Third, we grouped our patients by PEMVECC criteria using the admission diagnosis (i.e., the clinical phenotype) and not by baseline Crs, potentially leading to incorrect labelling. This may explain why Crs was not different between patients with the clinical phenotype “normal lung mechanics” and “restrictive lung mechanics”. Reassuringly, our analyses were done with patients grouped as being ventilated for a direct versus indirect pulmonary indication. Fourth, at the time of the study routine esophageal pressure measurements were not implemented so we could not ascertain the effect of altered chest wall mechanics in our cohort. Lastly, we did not measure auto-PEEP, although we carefully reviewed the flow-time scalars when the patients were studied so that we could exclude measurements where auto-PEEP occurred (30).

Conclusions

Higher driving pressure was independently associated with an increased time to extubation. The dynamic airway pressure gradient overestimated driving pressure. Future studies are required to understand if driving pressure is a determinant or marker of severity in pediatric acute respiratory failure. Injurious thresholds for driving pressure in patients with or at risk for PARDS need to be identified.

Research in Context.

• Driving Pressure, the ratio of tidal volume over compliance of the respiratory system, reflects lung stress and is associated with outcome in adults with acute respiratory distress syndrome. It is unclear if these observations also hold true for children

• It has been proposed that the dynamic driving pressure may be used as an alternative when patients are ventilated using modes that do not have a zero-flow state, as often is the case in pediatrics

• We found In patients with respiratory failure, higher driving pressure was independently associated with an increased time to extubation but injurious thresholds need to be identified. The dynamic pressure gradient overestimated the actual driving pressure and its use should therefore be discouraged