Impact of bilateral infiltrates on inflammatory biomarker levels and clinical outcomes of children with oxygenation defect

Abstract

Objectives:

The 2015 definition for Pediatric Acute Respiratory Distress Syndrome (PARDS) did not require the presence of bilateral infiltrates. We tested the hypothesis that pediatric patients meeting oxygenation criteria for PARDS but without bilateral infiltrates would have different inflammatory biomarker levels and clinical outcomes than those with bilateral infiltrates.

Design:

Secondary analysis of a prospective cohort study.

Setting:

Twenty-two pediatric intensive care units.

Patients:

446 patients age 2wks-17yrs intubated for respiratory failure with oxygenation index ≥4 or oxygenation saturation index ≥5 on the day of intubation or the day after.

Interventions:

None.

Measurements and Main Results:

Patients with bilateral infiltrates, either on the day of intubation or within the following two days, were compared to children who never developed bilateral infiltrates. Two analyses were performed to test (1) whether bilateral infiltrates are associated with elevated interleukin-1 receptor antagonist (IL-1ra) or IL-8 and (2) whether bilateral infiltrates are associated with worse clinical outcomes. Patients with bilateral infiltrates more often had a primary diagnosis of pneumonia (41% vs 28%, p=0.02) and less often asthma (8% vs 23%, p<0.01). After controlling for age, gender, and primary diagnosis, IL-1ra was higher on study days 1 and 2 in patients with bilateral infiltrates. There was no difference in IL-8 levels. After adjusting for age, gender, PRISM score, and severity of oxygenation defect, presence of bilateral infiltrates was associated with longer duration of mechanical ventilation in survivors (hazard ratio 0.64, 95% CI: 0.49–0.82, p<0.01); this association was independent of primary diagnosis. Overall mortality was 9%; mortality was higher in those without bilateral infiltrates (14% vs 8%, p=0.04, respectively).

Conclusions:

Children meeting PARDS oxygenation criteria with bilateral infiltrates on chest radiograph experience a more intense early inflammatory response. Bilateral infiltrates are associated with longer time on the ventilator in survivors independent of oxygenation defect severity.

Introduction

Radiographic findings have been part of the definition of Acute Respiratory Distress Syndrome (ARDS) since first described by Ashbaugh et al in 1967 (1). For years, pediatric intensivists identified children with ARDS using consensus definitions based on ARDS in adults, initially the American-European Consensus Conference definition and later the Berlin definition (2, 3). Both definitions include the requirement for bilateral infiltrates. In 2015, the Pediatric Acute Lung Injury Consensus Conference (PALICC) created a new definition of pediatric Acute Respiratory Distress Syndrome (PARDS) (4). The PALICC authors diverged from the Berlin criteria by using oxygenation index (OI) or oxygenation saturation index (OSI) (rather than PaO2/FiO2 or SaO2/FiO2) to measure the degree of the oxygenation defect and required “chest imaging findings of new infiltrate(s) consistent with pulmonary parenchymal disease” rather than the presence of bilateral infiltrates. The PALICC authors argued that the inclusion of bilateral infiltrates in the PARDS definition was unnecessary as severe unilateral disease also purports acute lung injury, there is no evidence that bilateral infiltrates impart additional risk, and chest radiographs are insensitive (4). There is significant evidence that interpretation of frontal radiographs is unreliable, so a definition less reliant on chest radiograph findings is desirable (5–8).

There is minimal data on the impact of chest radiograph findings on pathophysiology and clinical outcomes, especially in pediatric patients. Biomarker studies have provided important information about the biology of adult and pediatric ARDS (9). Biomarkers related to inflammation, including interleukin 1 receptor antagonist (IL-1ra), an inhibitor of IL-1β elaborated early in the inflammatory cascade, and the chemo-attractant interleukin 8 (IL-8) elaborated further downstream in the inflammatory process, are associated with ARDS outcomes in children and adults (10–17). IL-8 has also been used to help define a “hyper-inflammatory” ARDS subphenotype in adults (18, 19). To date, no studies have utilized biomarker data to examine the relationship between biomarker level and chest radiograph findings as related to the PARDS diagnosis. Using data from the Genetic Variation and Biomarkers in Children with Acute Lung Injury (BALI) study, which prospectively collected data and longitudinal biological specimens from children with acute hypoxemic respiratory failure, we sought to explore the relationship between the presence of bilateral infiltrates and inflammatory biomarkers or clinical outcomes in children who met the oxygenation criteria for PARDS. We hypothesized that patients with bilateral infiltrates would have higher levels of inflammation and worse clinical outcomes than those without bilateral infiltrates.

Materials and Methods

Study Design

This is a secondary analysis of the Genetic Variation and Biomarkers in Children with Acute Lung Injury (BALI) study (R01HL095410) (10), an ancillary study to the Randomized Evaluation for Sedation Titration for Respiratory Failure (RESTORE) trial (U01 HL086622) (20). This study was approved by the University of Michigan Institutional Review Board and all subjects signed written informed consent. RESTORE enrolled patients from age 2 weeks to 17 years from 2009 to 2013 requiring invasive mechanical ventilation (MV) for acute airways and/or pulmonary parenchymal lung disease. As the RESTORE intervention was a sedation protocol, patients were excluded if the length of MV was unlikely to be altered by sedation (eg. expected intubation <24 hours or chronic assisted ventilation). Twenty-two of the 31 centers participating in RESTORE participated in BALI. The study was approved by the Institutional Review Boards at all participating sites. Patients had blood samples collected within 24 hours of consent and again 24 and 48 hours later, with the first blood sample drawn within three days of intubation (study day 0) in most patients (70%). Plasma samples were assayed for IL-8 (15) and IL-1ra (21) by ELISA. 350 patients had at least one sample assayed for IL-1ra with 56% of patients having the first sample drawn on day 0 or 1. 352 patients had at least one sample assayed for IL-8 with 58% having the first sample drawn on day 0 or 1.

We excluded patients who did not have an oxygenation index (OI) ≥ 4, or if no OI was available, an oxygenation saturation index (OSI) ≥ 5 (oxygenation criteria for PARDS by PALICC criteria) on the day of intubation or the day after. Patients were also excluded if they had left atrial hypertension identified during study days 0, 1 or 2. For each patient, the degree of oxygenation defect (mild: 4≤OI<8 or 5≤OSI<7.5, moderate:8≤OI<16 or 7.5≤OSI<12.3, severe: OI≥16 or OSI≥12.3) was determined using the highest OI, or OSI if no OI was available, on days 0 or 1. Severity of illness was determined by PRISM-III (22).

RESTORE enrolled patients prior to the development of the PALICC criteria during the time frame that the AECC definition was used. Consequently, data was not collected about the presence of unilateral lung disease or number of quadrants involved, rather only the presence of new bilateral infiltrates was documented. Study site investigators were asked whether the patient had new onset bilateral infiltrates on the day of intubation and every subsequent day. Patients with chest radiographs demonstrating bilateral infiltrates at the time of intubation or during study day 0, 1 or 2 were coded as having bilateral infiltrates. Patients were considered to have no bilateral infiltrates if they did not have new onset bilateral infiltrates at any point during the 28-day study period. Patients who developed bilateral infiltrates on study day 3 or later were excluded from the analysis.

Statistical Analyses

All analyses were completed using either STATA 15.0 or SAS 9.4. Basic descriptive analyses of all demographic and key dependent variables (including frequency distributions or medians and interquartile ranges) were conducted on both groups. Bivariate analyses were done using Wilcoxon rank-sum, chi-squared, or Fisher’s exact tests.

Two separate analyses were performed. The first analysis was to determine if there was an association between the presence of bilateral infiltrates and an elevated inflammatory response. A separate mixed effects regression model using repeated measures was run for each biomarker. All biomarker levels in the regression models were log-transformed. Study days 0–5 were assessed by repeated measures analysis in order to capture three days of biomarker levels for most patients, with the mixed models accounting for any missing data. Given the equally spaced time points, an auto-regressive covariance structure was used. The models were run with study day as a categorical variable to investigate differences in biomarker levels in patients with and without bilateral infiltrates using differences of least square means. Biomarker analyses were adjusted for relevant covariates (age, gender, primary diagnosis [pneumonia, sepsis, bronchiolitis, asthma, aspiration, or other]), which were chosen a priori.

The second analysis evaluated whether the presence of bilateral infiltrates was associated with worse clinical outcomes. The effect of the presence of bilateral infiltrates on clinical outcomes was evaluated using logistic regression models (mortality) or Cox proportional hazard regression (PICU length of stay and duration of MV). Mortality was defined as all-cause in-hospital 90-day mortality. Patients who were transferred without extubation for >24 hours were assigned a 28 day length of MV. Survivors present in the ICU on day 28 were assigned a 28 day LOS. Analyses for clinical outcomes were adjusted for relevant covariates chosen a priori: age, gender, PRISM, and severity of oxygenation defect.

Sensitivity analyses eliminating patients with a primary diagnosis of asthma were also done for the two analyses described above as these patients were believed to be the least likely to meet the x-ray criteria in a diagnosis of PARDS.

Results

Patient Characteristics

There were 549 patients in the BALI cohort (Supplemental Figure 1). Eighty-five percent (455/538) of patients without left atrial hypertension met oxygenation criteria for PARDS with OI ≥4 or OSI ≥5 on the day of intubation or the day after. Of the group meeting PARDS oxygenation criteria in the first two study days, 78% (356/455) developed bilateral infiltrates in the first 2 study days; 20% (90/455) never developed bilateral infiltrates. Nine patients (2%) developed bilateral infiltrates on study day 3 or later and were excluded. Consequently, 446 patients were considered in the final analyses. Among the 446 patients, there were no significant differences in age, gender, ethnicity, and PRISM III between those with and without bilateral infiltrates (Table 1). More patients with bilateral infiltrates had a primary diagnosis of pneumonia (41% vs 28%, p=0.02) and fewer had a primary diagnosis of asthma (8% vs 23%, p<0.01). More patients with bilateral infiltrates had a severe oxygenation defect (p<0.01). Baseline characteristics of patients who did not meet oxygenation criteria are presented in Supplemental Table 1.

Table 1.

Patient Characteristics of Those Meeting Oxygenation Criteria With and Without Bilateral Infiltrates

Characteristics	Bilateral Infiltrates (n=356)	Unilateral or No Infiltrates (n=90)	p-valuea
Female, n (%)	152 (43)	46 (51)	0.15
Non-Hispanic White, n (%)	191 (54)	43 (48)	0.28
Age(y), median (IQR)	4.15 (0.8–11.4)	7.54 (1.6–12.1)	0.06
Intervention Siteb, n (%)	206 (58)	50 (56)	0.69
Primary Diagnosis, n (%)
Acute Respiratory Failure related to sepsis	72 (20)	17 (19)	0.78
Aspiration Pneumonia	19 (5)	7 (8)	0.38
Asthma or RAD	27 (8)	21 (23)	<0.01
Bronchiolitis	62 (17)	12 (13)	0.35
Pneumonia	147 (41)	25 (28)	0.02
Otherc	29 (8)	8 (9)	0.82
Medical History, n (%)
Prematurity	48 (13)	10 (11)	0.55
Asthma	51 (14)	25 (28)	<0.01
Seizure disorder	36 (10)	8 (9)	0.73
Immunodeficiency	8 (2)	4 (4)	0.27
Cancer	25 (7)	9 (10)	0.34
Chromosomal abnormality	29 (8)	5 (6)	0.41
Oxygen Severity Category, n (%)
Mild	75 (21)	24 (27)	0.25
Moderate	113 (32)	39 (43)	0.04
Severe	168 (47)	27 (30)	<0.01
PRISM III, median (IQR)	9 (4–14)	9 (4–13)	0.75
Died, n (%)	27 (8)	13 (14)	0.04

^aStatistical significance determined using chi-square, Wilcoxon rank-sum or Fisher’s exact test as appropriate

^bRefers to whether the patient was in a hospital randomized to receive the sedation protocol in the original RESTORE trial

^cOther includes acute chest syndrome, post-BMT, croup, pertussis, pulmonary edema, pulmonary hemorrhage, and trauma

IQR- interquartile range; PRISM- Pediatric Risk of Mortality

Association between Bilateral Infiltrates and Inflammatory Biomarkers

For the IL-1ra analysis, there were 277 patients meeting oxygenation criteria with bilateral infiltrates and IL-1ra measured between days 0–5 and 73 patients without bilateral infiltrates who had IL-1ra measured between days 0–5. Three samples were available for 79% of patients. Mixed effects modeling adjusting for age, gender, and primary diagnosis, showed that log-transformed levels of IL-1ra decreased by day in a quadratic pattern with larger decreases seen during the first study days (Figure 1). IL-1ra was higher in patients with bilateral infiltrates on study day 1 (p=0.03) and study day 2 (p=0.01). A sensitivity analysis excluding patients with a primary diagnosis of asthma showed significantly elevated levels of IL-1ra on days 0, 1, and 2 in patients with bilateral infiltrates (Supplemental Figure 2.)

Figure 1:

Adjusted least square mean of IL1-ra level by presence or absence of bilateral infiltrates. Adjusted for age, gender, and primary diagnosis. Biomarker levels are log-transformed. Error bars represent the standard error of the mean. Day represents study day with day 0 representing the day of intubation. * indicates p<0.05

For the IL-8 analysis, there were 278 patients meeting oxygenation criteria with bilateral infiltrates and IL-8 measurements on days 0–5 and 74 patients without bilateral infiltrates who had a sample on days 0–5. Three samples were available for 82% of patients. Mixed effects modeling adjusting for age, gender, and primary diagnosis, showed that log-transformed levels of IL-8 also decreased by day in a quadratic fashion similar to that seen with IL-1ra. Unlike IL-1ra, there was no difference in IL-8 between patients with bilateral infiltrates and those with unilateral or no infiltrates on any study day (Figure 2). Similarly, a sensitivity analysis excluding patients with a primary diagnosis of asthma showed no difference in IL-8 on any day (Supplemental Figure 3.)

Figure 2:

Adjusted least square mean of IL-8 level by presence or absence of bilateral infiltrates. Adjusted for age, gender, and primary diagnosis. Biomarker levels are log-transformed. Error bars represent the standard error of the mean. Day represents study day with day 0 representing the day of intubation. * indicates p<0.05

Association Between Bilateral Infiltrates and Clinical Outcomes

We also examined whether the presence of bilateral infiltrates impacted relevant clinical outcomes in children meeting the PARDS oxygenation defect requirement. After adjustment for age, gender, PRISM III, and severity of oxygenation defect, survivors with bilateral infiltrates had a longer duration of MV than patients without bilateral infiltrates (median 7.5d, IQR 4.4–12.1 vs median 4.9d, IQR 3.2–8.5) with a hazard ratio of 0.64 (95% CI: 0.49–0.82, p=0.001) (Figure 3). If the six-category primary diagnoses were added to the proportional hazards regression model for duration of ventilation described above, the hazard ratio for presence of bilateral infiltrates was similar (0.68 95% CI: 0.53–0.89, p<0.01); Supplemental Table 2). There was no statistical difference in PICU length of stay in survivors with or without bilateral infiltrates (median 11.9d, IQR 7.7–19.9 vs median 8.2d, IQR 5.3–17.8, respectively, HR: 0.81, 95% CI: 0.63–1.04, p=0.10) (Supplemental Figure 4).

Figure 3:

Kaplan-Meier plot of survivors remaining mechanically ventilated with bilateral (solid) or without bilateral infiltrates (dashed).

Mortality was low in our cohort (9% overall). Surprisingly, bivariate analysis indicated that mortality was lower in patients with bilateral infiltrates (Table 1). The presence of bilateral infiltrates was associated with decreased odds of mortality even after adjustment for age, gender, PRISM, and the severity of oxygenation defect (OR 0.3, 95% CI 0.13–0.67, p<0.01, Table 2). Given the overall low mortality in the cohort, we could not include the six-category primary diagnosis used in the other models in the mortality model. Instead, the six-category primary diagnosis groups were reduced into a four-category model: direct lung injury, asthma, sepsis, and other. The association of bilateral infiltrates with decreased odds of mortality was independent of primary diagnosis (using the four-group categorization) (Supplemental Table 3). Non-survivors without bilateral infiltrates most often died from multisystem organ failure (6 out of 13 deaths) followed by respiratory failure (5 out of 13 deaths). Non-survivors with bilateral infiltrates most often died from respiratory failure (13 out of 27 deaths), followed by cancer (4 out of 27 deaths), and multisystem organ failure (3 out of 27 deaths).

Table 2.

Multivariable Analysis of Association of Bilateral Infiltrates with 90-day All-Cause Mortality

Variable	OR	95% CI	p
Bilateral infiltrates on CXR	0.30	0.13–0.67	<0.01
Age (years)	1.09	1.03–1.15	<0.01
Gender	0.91	0.46–1.84	0.80
PRISM III	1.04	0.99–1.08	0.10
Oxygenation Defect
Mild	Reference
Moderate	4.56	0.54–38.4	0.16
Severe	20.1	2.61–155.20	<0.01

OR= Odds Ratio, PRISM= Pediatric Risk of Mortality, Mild oxygenation defect= 4≤OI<8 or 5≤OSI<7.5, moderate oxygenation defect= 8≤OI<16 or 7.5≤OSI<12.3, severe oxygenation defect= OI≥16 or OSI≥12.3

Sensitivity analyses excluding patients with a diagnosis of asthma showed similar effect of bilateral infiltrates on duration of ventilation and mortality (Supplemental Tables 4 and 5).

Discussion

The question addressed by the present study is whether patients who otherwise meet criteria for PARDS but do not have bilateral infiltrates on frontal radiograph exhibit a differing inflammatory profile and whether the presence of bilateral infiltrates is associated with worse clinical outcomes. Collection of biomarkers across different study days and multiple specimens from the same patient allowed us to both assess marker levels on individual days, but also to describe the trajectory across days to more richly assess the inflammatory profiles for each group. We analyzed two inflammatory biomarkers, IL-1ra, a marker of early inflammation, and IL-8, elaborated later in the inflammatory cascade.

By having multiple samples from patients, we were able to use a novel analytic technique of mixed effects modeling to examine the trajectory of the biomarker and level by day. Both IL-1ra and IL-8 show a decreasing early trajectory and then stabilized regardless of the chest radiograph findings. This finding suggests that the escalation period occurs before intubation with inflammation peaking sometime before, or at the time of intubation. After adjusting for relevant covariates, IL-1ra levels were higher on days 1 and 2 in patients meeting PARDS oxygenation criteria with bilateral infiltrates compared to those without bilateral infiltrates; by day 3 there was no significant difference between levels in children with or without bilateral infiltrates. The two groups did not have significantly different levels of IL-8 at any time point. These data suggest that bilateral infiltrates may be associated with a more profound early inflammatory response as indicated by higher IL-1ra levels, which may not be observed with later markers of inflammation such as IL-8. Future studies with additional inflammatory biomarkers will be required to further clarify this finding. Interestingly, our previous analysis using the entire cohort of children irrespective of OI levels or presence or absence of bilateral infiltrates, indicated that IL-1ra is associated with PARDS development, higher OI, and longer duration of MV and ICU length of stay in pediatric patients with acute respiratory failure, highlighting its potential importance in PARDS risk stratification (21).

The question posed by the PALICC authors was: “it is not clear whether the presence of bilateral pulmonary infiltrates helps with risk stratification for children with hypoxemic respiratory failure, that is, does it capture additional risk to the degree of oxygenation impairment?”(4). Even after adjusting for age, PRISM III, and the severity of oxygenation defect, our data suggest that patients with bilateral infiltrates experience a differing clinical course (evidenced by longer duration of ventilation in survivors) than those with acute hypoxemic respiratory failure without bilateral infiltrates. This finding is independent of the primary diagnosis that lead to the patient’s respiratory failure. This information is important to consider in future discussions of the definition of pediatric ARDS and risk stratification of patients with acute hypoxemic respiratory failure.

The 2015 PALICC definition of PARDS allowed for classification of more patients as having ARDS by removing the requirement for bilateral infiltrates. Although this “lumping” approach likely introduces additional heterogeneity, it is attractive because it can lead to patients who would have otherwise been excluded from a PARDS diagnosis being considered for management with strategies such as prone positioning and low-tidal volume ventilation (23). In addition, eliminating the bilateral infiltrate requirement could also lead to the earlier diagnosis of PARDS if the chest imaging findings lag behind the clinical changes. However, there is evidence that even within the narrower AECC/Berlin definitions of ARDS, there exists heterogeneity and hidden phenotypes that may impact response to treatment (18, 19, 24). The use of biomarkers in these studies is providing insight into features of the disease that may not be clinically apparent, so it is of interest to understand how the biomarker profile relates to the chest radiograph. The differences observed in the IL-1ra levels and clinical outcomes in children with acute hypoxemic respiratory failure with or without bilateral infiltrates suggest that eliminating the requirement for bilateral infiltrates may actually introduce more heterogeneity into PARDS. Adding heterogeneity may be important when considering patients for “higher risk” therapies. Future studies combining the clinically available data such as chest radiograph and “hidden” features of the disease made apparent by biomarker analysis may be able to uncover this heterogeneity.

Although only 9% of patients in our cohort died, our observation that mortality is higher in children without bilateral infiltrates is of interest. This finding is similar to data presented in the PALICC definition of PARDS, which suggested higher mortality in patients with oxygenation deficits without bilateral infiltrates compared to those with bilateral infiltrates (4), but differs from the recently published Pediatric Acute Respiratory Distress Syndrome Incidence and Epidemiology Study (PARDIE) where patients with bilateral infiltrates experienced higher mortality than those with unilateral infiltrates (25). These conflicting results may be partly explained by the fact that patients enrolled in PARDIE met all PARDS criteria and had unilateral or bilateral lung infiltrates, whereas RESTORE enrolled patients with acute respiratory failure with or without bilateral disease. Although patients in the group with bilateral infiltrates more frequently had a primary cause of death of respiratory failure (48% vs 38%), it is well established that many ARDS patients die from causes other than refractory hypoxemia (26, 27). While the deaths in both groups were infrequent, the similar risk of mortality as determined by PRISM-III between the two groups of patients and consistency with previous findings makes this seeming paradox worthy of further investigation. Other studies have shown that duration of ventilation (or ventilator-free days) do not always correspond with mortality data. In adults, using the Radiographic Assessment of Lung Oedema (RALE) scoring system, Warren et al demonstrated that a higher score (worse edema) was associated with higher mortality but that it had no association with ventilator-free days (28).

This study has several limitations. Because the chest radiographs were not collected as part of the initial study, the determination of presence or absence of bilateral pulmonary infiltrates was made by the site investigators, without multiple interpretations or separate radiologist review. In addition, there is no data available regarding whether the chest radiographs that did not have bilateral infiltrates contained pulmonary parenchymal disease that would now qualify for the PALICC definition of PARDS. Future studies would certainly benefit from a more detailed description of imaging findings in patients with acute hypoxemic respiratory failure. PALICC allowed for inclusion of patients with chronic lung disease and a worsening from baseline to be diagnosed with PARDS, however RESTORE excluded patients with chronic assisted ventilation, so these patients are not represented in this study. In addition, this study only enrolled children who were intubated for airways or pulmonary parenchymal disease, so other patients at risk for ARDS such as trauma patients, post-operative patients, and patients with neurologic injury may not be included in this cohort.

Conclusions

Patients with hypoxemia consistent with PALICC definitions for PARDS and bilateral infiltrates on chest radiograph appear to have exaggerated early inflammation as measured by elevated IL-1ra in the days immediately after intubation. These same patients have improved survival but at the expense of prolonged mechanical ventilation compared to those without bilateral infiltrates. These findings suggest inclusion of patients without bilateral infiltrates into the definition of PARDS may introduce heterogeneity into the disease and warrants further research.