There are marked differences in mortality rates in pediatric acute respiratory distress syndrome (ARDS) around the world (1). In Asia, mortality rates range from 44% to 75% (1–3). This contrasts with the lower mortality rates (17–35%) reported in North America, Europe, and Australia-New Zealand (4–6). A recent meta-analysis of pediatric ARDS showed that studies performed in Asia had a higher overall mortality (51.0%; 95% CI, 41.5–62.7%) compared with those performed in the Western hemisphere (27.3%; 95% CI, 22.3–33.5%) (7). Reasons postulated for this difference were differences in resources, case mix, socioeconomic conditions, and management strategies. There remains a paucity of description of the epidemiology of pediatric ARDS across multiple centers/countries in Asia. This current gap in epidemiologic data is a barrier for investigation of plausible reasons for the difference in pediatric ARDS mortality rates in Asia compared with other regions of the world.
In 2015, the Pediatric Acute Lung Injury Consensus Conference (PALICC) developed a pediatric specific definition for acute respiratory distress syndrome (PARDS) (8). The PARDS definition included: (1) hypoxemia with oxygenation index (OI) ≥4 or oxygenation saturation index (OSI) ≥5; (2) new radiological lung infiltrates; (3) occurred within 7 days of a known clinical insult; and (4) not explained by cardiac failure or fluid overload. This definition also allowed children with congenital cyanotic heart disease, chronic lung disease and left ventricular dysfunction to be included as long as they fulfilled the criteria above and the acute deterioration in oxygenation and pulmonary infiltrates cannot be explained by their pre-existing diseases. Perinatal lung disease was excluded. However, to date, a comprehensive validation of this definition is lacking. Given the current limitations in the medical literature, we undertook this multicenter retrospective study to 1) determine the epidemiology of PARDS in Asia, 2) describe the mechanical ventilation (MV) strategies and adjunct therapies in children with PARDS, and 3) evaluate whether the newly proposed PALICC risk stratification categories accurately predict clinical outcomes in PARDS.
MATERIALS AND METHODS
This is a retrospective study of children with PARDS admitted to PICUs in the Pediatric Acute and Critical Care Asian Medicine Network (PACCMAN) centers. PACCMAN is a collaboration network formed in 2016 by dedicated pediatric intensivists from Asia with the aims of sharing experience, developing best practices, and promoting research within the region. This study was approved by all participating hospital’s institutional review boards, and waiver of consent was granted in all sites.
Patients who fulfilled the PALICC criteria for PARDS over the study period 2009–2015 were included. A list of patients was generated using International Classification of Diseases (International Classification of Diseases, 9th Edition, Clinical Modification [before 2012] or International Classification of Diseases, 10th Edition, Australian Modification [from 2012 onward]) diagnostic codes, discharge diagnosis with keywords of “ARDS” or “acute lung injury (ALI),” or by manually screening through the PICU registry at the respective centers. Patients were included if they met the inclusion/exclusion criteria described below. A secure online central database using the Research Electronic Data Capture system was developed by the main coordinating center in Singapore and was accessible to all site investigators (9). This was used for remote multisite data entry and centralized data management.
Management of PARDS among centers was at the discretion of individual sites and physicians. All data were kept secure following local regulations. Data were then merged and analyzed at the coordinating site.
We included patients 1 month to 18 years old who fulfilled the PALICC definition for PARDS and intubated (8). Patients on noninvasive ventilation (NIV) were excluded. Premature neonates with corrected age less than 35 weeks, neonates being cared for in the neonatal ICU and neonates deemed to have perinatal lung disease were excluded. Multiple organ dysfunction was defined as more than one extrapulmonary organ dysfunction according to the International Pediatric Sepsis Consensus Conference criteria for each organ dysfunction (10).
The data collection form was adapted from a previous single-center pediatric ARDS study (2). Data extracted included demographics, severity scores (Pediatric Index of Mortality 2 [PIM 2] score and Pediatric Logistic Organ dysfunction score), presence of organ dysfunction, MV data, adjunct PARDS therapies, and PICU supportive therapies. MV data included ventilation mode, peak inspiratory pressures (PIPs), positive end-expiratory pressure (PEEP), mean airway pressure (MAP), tidal volume (TV), and FIO2 for days 1–7 of PARDS. For standardization, routine morning blood gasses (measured at 6:00 AMto 8:00 AM) and corresponding MV settings were recorded on each day of PARDS. Patients were classified into mild, moderate, and severe groups according to the PALICC oxygenation index (OI) or OSI criteria on day 1 of PARDS. Conventional MV referred to both pressure-controlled and volume-controlled ventilation. Driving pressure was the difference between PIP and PEEP. Use of adjuncts and other PICU support therapies were recorded throughout the course of PARDS up to 28 days or until PICU discharge whichever was earlier.
Primary outcome was 100-day mortality. Patients with 100-day mortality were patients who died within 100 days from PARDS diagnosis. In survivors, the most recent hospital-wide medical records were retrieved and the last hospital visit (inpatient or outpatient) was taken as the last date of contact. If this occurred beyond 100 days, the patient was analyzed as a survivor. Patients who were transferred to another medical facility or lost to follow-up (last date of contact prior to 100 d) were censored. Secondary outcomes included PICU mortality, 28-day ventilator-free days (VFDs), and 28-day ICU-free days (IFDs). VFD is defined as days alive and free from invasive MV up to 28 days. IFD is defined as days alive and discharged from the PICU up to 28 days. This is to eliminate mortality as a competing interest in evaluating MV and PICU duration.
Categorical variables were presented as frequency with corresponding proportion. Continuous variables were presented as mean (SD) or as median (interquartile range [IQR]) as appropriate depending on distribution of data. Differences between categorical and continuous variables between PARDS severity were tested using the chi-square test and Mann-Whitney U test, respectively. Hundred-day survival rate based on severity groups were plotted using Kaplan-Meier model and were compared using log-rank test. Both univariate and multivariate Cox proportional hazard (CPH) regression model were used to quantify association between 100-day mortality and other covariates (site, presence of comorbidities, and severity of illness [PIM 2 score]). Association from CPH was characterized as hazard ratio (HR) with corresponding 95% CI. VFD and IFD were also analyzed with respect to PARDS severity groups. Statistical significance was taken as p value of less than 0.05 for all tests. SAS version 9.3 software (SAS Institute, Cary, NC) was used for the analysis.
Characteristics of Patients
Data were entered for 438 patients with PARDS between 2009 and 2015 from 10 multidisciplinary PICUs (Supplementary Table 1, Supplemental Digital Content 1, http://links.lww.com/CCM/C774). However, after data verification, 427 were confirmed to fulfill the PALICC definition. There were 54 of 427 patients (12.6%) on NIV on day 1 of PARDS and thus excluded. Hence, a total of 373 patients were included in this study (Supplementary Fig. 1, Supplemental Digital Content 2, http://links.lww.com/CCM/C775; legend, Supplemental Digital Content 1, http://links.lww.com/CCM/C774).
There were 89 of 373 (23.9%), 149 of 373 (39.9%), and 135 of 373 (36.2%) patients with mild, moderate, and severe PARDS, respectively (Table 1). The median OI and OSI were 12.8 (8.2–19.6) and 10.3 (6.8–16.1), respectively. Pneumonia/lower respiratory tract infection (309/373 [82.8%]) was the most common risk factor for PARDS.
MV and Adjunct Therapies in PARDS
On day 1 of PARDS, 309 of 373 (82.8%) were supported on conventional MV (Table 2). Over the duration of PARDS, the severe and moderate groups received high-frequency ventilation more frequently compared with the mild group (66/135 [48.9%], 45/149 [30.2%], and 10/89 [11.2%], respectively; [p > 0.001]). Patients with greater severity of PARDS were ventilated with higher PIP throughout the first 7 days (Fig. 1A). These patients also had higher driving pressures throughout the first 7 days (Fig. 1D). However, there were no differences in TV between the severity groups of PARDS (Fig. 1E). Approximately one fifth (68/309 [22.0%]) of patients were ventilated with high PIP (> 28 cm H2O). Compared with patients exposed to PIP less than 28 cm H2O, patients ventilated with high PIP had reduced VFD and IFD (11.0 [0.0–21.0] vs 19.0 [0.0–23.0] d; p = 0.0125 and 9.0 [0.0–20.0] vs 17.0 [0.0–22.0]; p = 0.0235, respectively). There was, however, no difference in PICU or 100-day mortality between these two groups. Majority of patients (51/139 [63.3%]) were ventilated with high TVs (> 8 mL/kg actual body weight). However, high TV itself was not adversely associated with any of the primary or secondary outcomes.
PEEP and MAP were observed to be higher with increasing PARDS severity throughout the first 7 days (Figs. 1B and 1C). Majority of patients were ventilated with PEEP less than 10 cm H2O regardless the severity of PARDS. Overall median FIO2 was high in our cohort (60 [50–95] %), and there was stepwise increase across the severity groups. However, there was no association between the trend of FIO2 and mortality in the three severity groups.
Among those with arterial blood gas measurements (311/373[83.4%]), pH and PaO2 worsened with increasing severity of PARDS but not PaCO2 (Table 2). Distribution of pH, PaO2, and PaCO2 throughout the 7 days showed a normal distribution with mean (SD) of 7.4 (0.11), 83.25 (37.10) mm Hg, and 48.1 (16.9) mm Hg, respectively (Supplementary Fig. 2, Supplemental Digital Content 3, http://links.lww.com/CCM/C776; legend, Supplemental Digital Content 1, http://links.lww.com/CCM/C774). Oxygen saturation distribution, as expected, was skewed with median (IQR) as 96.0 (92.0–99.0) % (Supplement Digital Content - Figure 2).
The commonly used adjunct therapies were diuretics (252/373 [67.6%]), blood product transfusions [245/373 (65.7%)], and systemic steroids (192/373 [51.5%]) (Table 2). A total of 12 of 373 (3.2%) patients required extracorporeal membrane oxygenation (ECMO). Of these, 9 of 12 (75.0%) were on veno-arterial ECMO with a mortality of seven of nine (77.8%). The remainder, three of 12 (25.0%) were on veno-venous ECMO, and all survived.
Overall PICU mortality for PARDS was 113 of 373 (30.3%). There was a stepwise increase in mortality rate across the severity groups (Table 3). After PICU discharge, an additional 14 patients died. Twenty patients were transferred to another hospital, and 36 patients were lost to follow-up after discharge from the PICU (censored). Thus, 100-day mortality was 126 of 317 (39.7%). After controlling for site, presence of comorbidities and PIM 2 score in the multivariate CPH regression model, 100-day mortality stratified according to the PARDS severity groups demonstrated an adjusted HR (95% CI) of 1.88 (1.03–3.45) in the moderate group and 3.18 (1.68–6.02) in the severe group. The Kaplan-Meier curves showed a stepwise decrease in survival in the moderate and severe groups compared with the mild group (Fig. 2). A sensitivity analysis conducted to evaluate the PARDS classification according to OI and OSI classification showed that both oxygenation indices were good at discriminating mortality. OI was, however, more robust with better differentiation of the moderate and severe groups (Figs. 2B and 2C).
The overall VFD and IFD were 16.0 (0.0–23.0) and 11.0 (6.0–22.0) days, respectively. There was a stepwise increase in ventilator duration and PICU duration in increasing severity groups (Table 3). Correspondingly, there was also a reduction in VFD and IFD in increasing severity groups. A sensitivity analysis was conducted excluding patients on volume-controlled mode (10/373 [2.7%]). The PIP, PEEP, MAP, FIO2, and TV trend over 7 days remain unchanged. In this sensitivity analysis, results of the primary and secondary outcomes remain unchanged.
This first multicenter study of PARDS in Asia demonstrated that the PALICC criteria for stratification into mild, moderate, and severe groups was associated with a stepwise decrease in VFD and IFD with increase in short-term and intermediate-term mortality. After adjusting for site, presence of comorbidities and severity of illness, patients with moderate and severe PARDS had approximately two and three times increased risk of dying, respectively, compared to children with mild PARDS. Higher ventilatory pressures were associated with reduced VFD and IFD, but we did not find any association between driving pressures and TV with poor clinical outcomes.
The PICU mortality rate in this study (30.3%) is comparable with the overall PARDS mortality reported by the recent meta-analysis that included patients identified by the American-European Consensus Conference or Berlin definition (7). To date, there are only two retrospective studies using the PALICC definition for mortality reporting and risk stratification (11, 12). The first study (n = 254) showed that mild and moderate PARDS groups had similar mortality (16.7% vs 18.6%, respectively), but the severe PARDS group had significantly higher mortality (37.0%) (11).The second study (n = 211) of PARDS in children with pediatric allogeneic hematopoietic stem cell transplant demonstrated that mortality increased with increasing severity (36.4%, 47.2%, and 75.4%; p < 0.001 in the mild, moderate and severe groups, respectively) (12). Taken collectively with our study, the predictive ability of the PALICC definition with mortality based on severity groupings was fairly robust.
In children, the beneficial effects of low TV have not been shown to be consistent. A retrospective study of 160 children with ALI showed that a higher TV (10.2 ± 1.7 vs. 8.1 ± 1.4 mL/kg; p < 0.001) was associated with mortality (odds ratio, 1.59; 95% CI, 1.20–2.10; p < 0.001) (13). However, a meta-analysis of mechanically ventilated children (n = 1756) including a subgroup analysis of pediatric ALI/ARDS (n = 799) failed to show any association between TV (up to 12 mL/kg) and mortality (14). The findings from our study echo the lack of association between higher TV and mortality. As the predominant mode of ventilation was pressure-limited, one would expect lower TVs as lung injury worsened. TV, however, remained constant across PARDS severity groups except on day 7 when the severe group received lower TV. In this ventilation approach, TV did not vary because the pressure limit was increased as lung compliance (lung injury) worsened.
Alveolar recruitment from the use of high PEEP is an open lung strategy that leads to reduction in sheer stress during lung over-inflation (15). A meta-analysis in adult ALI/ARDS showed that the use of high PEEP was associated with improvements in oxygenation (16). However, a subgroup analysis of ARDS patients alone (n = 205) showed a reduction in mortality (risk reduction [RR] 0.67; 95% CI, 0.48–0.95) with high PEEP. Despite recommendations for the use of moderate levels of PEEP (10–15 cm H2O) in severe PARDS, the levels of PEEP used in pediatric studies remained relatively low (6–10 cm H2O) (17, 18). Our study showed that higher levels of PEEP was used in the moderate and severe PARDS groups, but even in the severe category, it remained relatively low (median PEEP of 9.5 [7–12]). Instead, overall FIO2 in the entire cohort was relatively high (median FIO2 of 60 [50–95] %). This current practice in our cohort suggests a preference for the use of FIO2 to treat hypoxemia rather than PEEP which is not consistent with current recommendations that emphasize minimizing FIO2 and using PEEP effectively (8). However, the clinical impact of using high FIO2 remains controversial (19, 20). We did not find any association between higher FIO2 and mortality. We postulate that one possible reason for this practice is the lack of established PEEP-adjusted FIO2 values for different age groups/sizes in the pediatric population and perhaps “PEEP phobia.”
A post hoc analysis of 3,562 adult ARDS patients demonstrated that the individual beneficial changes in TV and PEEP were due to mediation effects of driving pressure (21). Each one SD increments in driving pressure (7 cm H2O) conferred an increased risk of mortality (RR, 1.4 [95% CI, 1.31–1.51]; p < 0.001). Driving pressure was also demonstrated to be independently associated with mortality in a prospective observational PARDS study (18). Our study showed no differences in driving pressures across severity groups in our cohort (Fig. 1D) and no association between higher driving pressures and mortality. It is possible that this difference is because the prior study had a lower proportion of pulmonary PARDS (54/84 [64.3%]) compared with our study (309/373 [82.8%]). Further larger studies are required to address this issue in a more robust manner.
This study is the first multicenter study on PARDS in Asia, and it included a large cohort of patients. It provided data to validate the recent PALICC definition for PARDS. Detailed MV data over the first 7 days of PARDS allowed us to examine the pattern of use and effect of each ventilation variable. We also determined intermediate-term mortality (100-d mortality) which was higher than short-term mortality (PICU mortality) indicating that PARDS has an impact on survival even beyond the PICU stay. Data were verified by manual review for gross inconsistencies at the coordinating center and corrected prior to analysis. Despite the strengths, the results of this study must be viewed in light of its limitations. First, patients with PARDS were retrospectively identified and stratified. Identifying PARDS especially in mild cases, by diagnostic codes/keywords, may have led to under-recognition as many patients were coded according to their underlying disease (e.g., pneumonia, sepsis) (22). With this limitation in mind, a manual search of the PICU registry was conducted at centers which could not reliably identify patients solely by diagnostic codes. The retrospective design also precluded us from making any cause-and-effect association. We can only surmise that there was an association between ventilation strategy of higher TV, higher FIO2, and relatively lower PEEP and higher mortality in our cohort. Future trials are needed to investigate the impact of these MV strategies on improving mortality in PARDS. Second, electronic medical records that facilitate patient identification and data retrieval were only available at some participating sites. Hence, these sites may have contributed data for a greater number of patients and compared with other sites, leading to selection bias. Sites were not mandated to recruit consecutive patients throughout the 6-year study period. Third, MV strategies and other interventions were not standardized between centers. We attempted to control for this by including site as a covariate in our multivariate analysis. The lack of data on exact cause of death in nonsurvivors and development of new morbidities in survivors is also another limitation of this study. Last, results derived from data from these 10 centers may not be representative of all centers across Asia.
This study demonstrates that PARDS carry a high mortality (30.1%) in Asia and that the severity groups of PARDS as defined by the PALICC criteria are robust in risk stratification with stepwise worsening of clinically significant outcomes from mild to severe. Future therapeutic studies in PARDS should consider using the PALICC risk classification to guide a stepwise increase in management across the severity spectrum of PARDS.
We acknowledge the administrative support from Academic Medicine Research Institute, Singhealth and Duke-NUS Medical School, and the National University of Singapore. We thank the network team from Singapore Clinical Research Institute for setup up and maintenance of the electronic data capture system and website. We thank Ira M. Cheifetz (Duke Children’s Hospital, Durham, NC) for critically appraising this article.
1. Lodha R, Kabra SK, Pandey RM. Acute respiratory distress syndrome: Experience at a tertiary care hospital. Indian Pediatr 2001; 38:1154–1159
2. Wong JJ, Loh TF, Testoni D, et al. Epidemiology of pediatric acute respiratory distress syndrome in singapore: Risk factors and predictive respiratory indices for mortality
. Front Pediatr 2014; 2:78
3. Hu X, Qian S, Xu F, et al; Chinese Collaborative Study Group for Pediatric Respiratory Failure: Incidence, management and mortality
of acute hypoxemic respiratory failure
and acute respiratory distress syndrome from a prospective study of Chinese paediatric intensive care network. Acta Paediatr 2010; 99:715–721
4. Costil J, Cloup M, Leclerc F, et al. Acute respiratory distress syndrome (ARDS) in children
: Multicenter Collaborative Study of the French Group of Pediatric Intensive Care. Pediatr Pulmonol Suppl 1995; 11:106–107
5. Flori HR, Glidden DV, Rutherford GW, et al. Pediatric acute lung injury
: Prospective evaluation of risk factors associated with mortality
. Am J Respir Crit Care Med 2005; 171:995–1001
6. Erickson S, Schibler A, Numa A, et al; Paediatric Study Group; Australian and New Zealand Intensive Care Society: Acute lung injury
in pediatric intensive care in Australia and New Zealand: A prospective, multicenter, observational study. Pediatr Crit Care Med 2007; 8:317–323
7. Fuchs H, Mendler MR, Scharnbeck D, et al. Very low tidal volume ventilation with associated hypercapnia–effects on lung injury in a model for acute respiratory distress syndrome. PLoS One 2011; 6:e23816
8. Pediatric Acute Lung Injury
Consensus Conference Group: Pediatric acute respiratory distress syndrome: Consensus recommendations from the pediatric acute lung injury
consensus conference. Pediatr Crit Care Med 2015; 16:428–439
9. Harris PA, Taylor R, Thielke R, et al. Research electronic data capture (REDCap)–a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform 2009; 42:377–381
10. Goldstein B, Giroir B, Randolph A; International Consensus Conference on Pediatric Sepsis: International pediatric sepsis consensus conference: Definitions for sepsis and organ dysfunction in pediatrics. Pediatr Crit Care Med 2005; 6:2–8
11. Parvathaneni K, Belani S, Leung D, et al. Evaluating the performance of the Pediatric Acute Lung Injury
Consensus Conference Definition of Acute Respiratory Distress Syndrome. Pediatr Crit Care Med 2017; 18:17–25
12. Rowan CM, Smith LS, Loomis A, et al; Investigators of the Pediatric Acute Lung Injury
and Sepsis Network: Pediatric acute respiratory distress syndrome in pediatric allogeneic hematopoietic stem cell transplants: A multicenter study. Pediatr Crit Care Med 2017; 18:304–309
13. Wunsch H, Mapstone J. High-frequency ventilation versus conventional ventilation for the treatment of acute lung injury
and acute respiratory distress syndrome: A systematic review and cochrane analysis. Anesth Analg 2005; 100:1765–1772
14. de Jager P, Burgerhof JG, van Heerde M, et al. Tidal volume and mortality
in mechanically ventilated children
: A systematic review and meta-analysis of observational studies. Crit Care Med 2014; 42:2461–2472
15. Amato MB, Barbas CS, Medeiros DM, et al. Beneficial effects of the “open lung approach” with low distending pressures in acute respiratory distress syndrome. A prospective randomized study on mechanical ventilation. Am J Respir Crit Care Med 1995; 152(6 Pt 1):1835–1846
16. Santa Cruz R, Rojas JI, Nervi R, et al. High versus low positive end-expiratory pressure (PEEP) levels for mechanically ventilated adult patients with acute lung injury
and acute respiratory distress syndrome. Cochrane Database Syst Rev 2013:CD009098
17. Ward SL, Quinn CM, Valentine SL, et al. Poor adherence to lung-protective mechanical ventilation in pediatric acute respiratory distress syndrome. Pediatr Crit Care Med 2016; 17:917–923
18. Panico FF, Troster EJ, Oliveira CS, et al. Risk factors for mortality
and outcomes in pediatric acute lung injury
/acute respiratory distress syndrome. Pediatr Crit Care Med 2015; 16:e194–e200
19. de Jonge E, Peelen L, Keijzers PJ, et al. Association between administered oxygen, arterial partial oxygen pressure and mortality
in mechanically ventilated intensive care unit patients. Crit Care 2008; 12:R156
20. Eastwood G, Bellomo R, Bailey M, et al. Arterial oxygen tension and mortality
in mechanically ventilated patients. Intensive Care Med 2012; 38:91–98
21. Amato MB, Meade MO, Slutsky AS, et al. Driving pressure and survival in the acute respiratory distress syndrome. N Engl J Med 2015; 372:747–755
22. Bellani G, Laffey JG, Pham T, et al; LUNG SAFE Investigators; ESICM Trials Group: Epidemiology, patterns of care, and mortality
for patients with acute respiratory distress syndrome in intensive care units in 50 countries. JAMA 2016; 315:788–800
acute hypoxemic respiratory failure; acute lung injury; children; mortality; pediatric intensive care units
Supplemental Digital Content
Copyright © by 2017 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.