Introduction

Respiratory syncytial virus (RSV) is one of the most frequent cause of respiratory infection in infants and is responsible for 460,000 cases of bronchiolitis each year in France [18]. Fortunately, RSV bronchiolitis is generally self-limited; however, infants with predisposing conditions appear more likely than others to develop severe forms of bronchiolitis leading to hospitalisation or to a need for mechanical ventilation (MV). Prematurity, bronchopulmonary dysplasia (BPD), congenital heart disease (CHD) and younger age have been previously described as risk factors for hospital admission and for controlled MV support in infants with RSV bronchiolitis [1, 13,20]. Although clinical characteristics increasing the risk of hospitalisation or need for MV have been well documented, little is known about the prognostic factors associated with a more severe RSV bronchiolitis when the infants are ventilated. Prolonged duration of MV was associated with CHD and chronic lung disease in a cohort of 72 children [3]. Despite maximal MV management, a small subgroup of children with RSV bronchiolitis go on to develop profound hypoxaemia and a need for extracorporeal membrane oxygenation (ECMO) support. ECMO was first successfully applied to infants with RSV respiratory failure from 1983 to 1988 [17]; since then, it has been recognised as a possible alternative to MV and can provide respiratory support for children with life-threatening RSV bronchiolitis. To our knowledge, no previous studies have analysed risk factors of developing refractory hypoxaemia requiring ECMO support in ventilated children with severe RSV bronchiolitis.

We undertook, over a 8-year period (1996–2003), a retrospective study of 151 children with RSV bronchiolitis for whom MV was required. Our purpose was first to better characterise infants requiring MV and/or ECMO support for severe RSV bronchiolitis. This study was also conducted to look for risk factors for a more severe bronchiolitis assessed by hospital mortality, need for ECMO support and prolonged time course of MV and their independent contributions.

Subjects and methods

Subjects

All children under 1 year of age admitted to the paediatric intensive care unit (PICU) of the Trousseau Hospital in Paris from 1 January 1996 to 31 December 2003 for management of bronchiolitis requiring MV and caused by proven RSV infection were retrospectively reviewed. Patients requiring chronic ventilation by tracheotomy were excluded. Bronchiolitis was defined as acute respiratory failure associated with RSV detection in respiratory secretions. RSV was detected via nasopharyngeal suction by direct immunofluorescence tests using monoclonal antibodies (Argene Biosoft, France). No children received palivizumab before PICU admission.

Clinical data collection and laboratory investigations

The medical records were retrospectively reviewed by a single investigator. Age was corrected for gestation. Premature birth was defined as having a gestational age (GA) of less than 37 weeks. BPD was defined as a requirement for supplementary oxygen for longer than 28 days. Laboratory investigations include aspirations from the endotracheal site for bacterial detection and culture. Only tracheal culture results obtained during the first hours of MV were gathered. Tracheal aspirations were considered positive if >10000 colony-forming units per ml of a single organism were found.

Management

All patients were ventilated with a volumetric ventilator (Drager Evita4). Regarding antimicrobial treatment, antibiotics were given for any child having underlying risk factors for severe RSV disease (BPD, CHD, prematurity) and/or presenting an infiltrate on the chest X-ray film. It was our practice to give firstly amoxicillin associated with clavulanic acid, subsequently adapted according to antimicrobial sensitivity. No patients were given theophylline or ribavirin.

Extracorporeal membrane oxygenation

In France, infants are candidates for ECMO as a rescue treatment when they had persistent hypoxaemia despite high ventilator pressures and 100% inspired oxygen concentration before cannulation that which presumed fatal respiratory failure. Children had to have one of the following signs of severe respiratory failure: (1) oxygenation index >40 on three blood gas values obtained at least 30 min apart; (2) alveolar-arterial oxygen difference >610 mmHg for 8 h; and (3) sudden decompensation with PaO2 (arterial oxygen pressure) <40 mmHg despite maximal management for 2 h. The ECMO-venovenous circuit is detailed in Fig. 1 according to Chevalier et al. [4]. ECMO-venovenous is the treatment of choice for children requiring ECMO support. Nevertheless, in children with persistent respiratory failure, ECMO-venovenous support was switched by a veno-arterial technique which used the same venous catheter and an arterial catheter inserted into the right common carotid artery.

Fig. 1
figure 1

ECMO venovenous circuit. A single lumen venovenous cannula inserted into the right internal jugular vein (with its tip in the right atrium) is connected to an alternative clamp (for drainage and reinfusion) and to a roller pump that delivers venous blood to a membrane lung and returns it through the cannula to the right atrium

Full size image

Statistical analysis

Statistical analysis was performed using Stat View 5.0 (SAS Institute Inc. 1992–1998). Differences in categorical variables were assessed using the chi-squared test or Fisher’s exact test. Continuous variables were compared using the Mann-Whitney U test or Kruskal-Wallis test. Multivariate analyses using a stepwise logistic regression model were used to determine which patient characteristics could predict bronchiolitis severity assessed by prolonged durations of MV, need for ECMO support and mortality. Prolonged duration of MV was defined using a cutoff point of 6 days, corresponding to the median ventilator days value for the study sample. Crude and adjusted odds ratio (OR) and their 95% confidence intervals (95%CI) were calculated. A P value <0.05 was considered statistically significant.

Results

Study population

A total of 152 patients required assisted ventilation for severe RSV bronchiolitis during the 8-year period reviewed (1996–2003). One child requiring chronic ventilation by tracheotomy for a Pierre-Robin sequence was excluded. Therefore, 151 patients were retrospectively included in this study. Only seven patients (4.6%) had CHD, all required oxygen supplementation at birth, and among those, one had DiGeorge syndrome. Types of CHD in patients of our study were ventricular septal defect ( n =4), transposition of the great arteries, Fallot’s tetralogy, and coarctation of the aorta (one each). No patient was known to be immunocompromised.

Entire cohort characteristics

The baseline characteristics of the patients are presented in Table 1. Fifty-eight children were prematurely born (38.4%). Mean peripheral leucocyte count at admission was 10.3×109/l (SD 7.4). Blood gases just before intubation were available in 76 patients: mean pH value was 7.26 (SD 0.09) and mean pCO2 value was 63.4 mmHg (SD 14.1). The initial chest X-ray films showed atelectasia in 43 of 151 patients (28.5%). Antibiotics were given to 140 of 151 of the patients (94%). Baseline characteristics did not vary significantly over the eight RSV seasons that were studied (data not shown) and characteristics of patients in our cohort were similar to those of infants admitted to other French PICUs [11].

Table 1 Characteristics of ventilated infants with RSV infection in our population ( n =151). Statistical comparisons between covariates and need for ECMO support
Full size table

Extracorporeal membrane oxygenation support

Of the study population, 14 infants were treated with ECMO (9.3%). Comparisons between infants requiring ECMO support and those treated only by MV support showed that gender and GA were similar between the two groups (Table 1). Univariate analysis of the neonatal period data and characteristics at admission revealed that BPD was the only predictive factor for ECMO support in ventilated children with bronchiolitis. The frequency of BPD was significantly higher amongst children who required ECMO support as compared with those from the group without ECMO support ( P =0.001, OR =8.9; 95%CI =2.4–33.1). This association was assessed by the multivariate approach after adjustment for potential confounding factors (adjusted P =0.004, OR =11.8; 95%CI =2.2–63.1). Interestingly, age at admission between children requiring ECMO and no ECMO approached significance ( P =0.05) but this association was not confirmed after adjustment (adjusted P =0.13).

Table 2 summarises the ECMO characteristics of the 14 patients with severe RSV bronchiolitis needing ECMO support. No predictive factor of survival was found amongst the cohort of infants supported by ECMO. Gender, GA and neonatal history were not significantly different between the groups of survivors and nonsurvivors ( P =0.59; P =0.52; P =0.73 respectively). In particular, diagnosis of BPD did not predict a worse outcome with ECMO ( P =0.59). The mean ± SD duration of ECMO for survivors (11.8 days ±2.3 days) was not significantly different as compared with the group of nonsurvivors ( P =0.36).

Table 2 Data of RSV infected patients requiring ECMO between 1996 and 2003 ( n =14)
Full size table

Risk factors for prolonged time course of mechanical ventilation

An association between bronchiolitis severity and patient characteristics was first evaluated by comparison of mean duration of MV. No difference between mean ventilator days was observed according to gender ( P =0.32). Low GA at birth was correlated significantly with prolonged duration of MV ( P =0.001). Mean duration of MV was also significantly higher in the group of infants with a neonatal history of BPD ( P <0.0001) and CHD ( P =0.005) and among those infants from whom a bacterial tracheal colonisation was isolated (9.0 days versus 6.6 days; P =0.0006). Time course of MV was not significantly associated with either CRP level at admission ( P =0.35) or WBC at admission ( P =0.87). To further assess the association between risk factors and time course of MV, multivariate regression analysis was carried out using two groups of infants with respect to a threshold value of 6 days (Table 3). This revealed that prolonged time course of MV was significantly associated with low GA, presence of oxygen supplementation, MV support and BPD in the neonatal period (Table 3).

Table 3 Multivariate analysis. Risk factors for prolonged duration of MV in children with RSV bronchiolitis ( n =151)
Full size table

Influence of bacterial colonisation

Aspirations from the endotracheal site within the first hours of MV detected bacterial colonisation in 78 of 151 patients (51.7%). The most common bacteria involved in the mixed RSV-bacterial colonisation was Haemophilus influenzae (31/151=20.5%). Other organisms identified in the tracheal aspiration cultures included Branhamella catarrhalis, Staphylococcus (for each 12/151=7.9%) and Streptococcus pneumoniae (8/151=5.3%). There was no statistically significant difference in the mean WBC and mean CRP level at admission between those infants with a positive versus a negative tracheal aspiration culture ( P =0.48 and P =0.86 respectively). A positive tracheal aspiration culture at admission was associated with prolonged duration of MV support ( P =0.003, OR =2.7; 95%CI =1.3–5.2). This association remained significant after adjustment (Table 3). The influence of bacterial colonisation on bronchiolitis severity was further illustrated using a Kaplan Meier survival plot of the proportion of children requiring MV by day of admission according to tracheal bacterial status (Fig. 2). As compared with the group of children with a negative bacterial tracheal culture, the group of infants colonised by H. influenzae present a great severity of bronchiolitis, assessed by a significant difference in mean ventilator days ( P =0.003). Thus, H. influenzae colonisation was associated with a 3-fold increase in the risk of prolonged MV compared with absence of a pathogen isolated in tracheal culture (adjusted P =0.03, adjusted OR =3.0; 95%CI =1.2–7.5).

Fig. 2
figure 2

Kaplan-Meier survival plot of the proportion of patients remaining on intensive care by day of mechanical ventilation according to tracheal bacterial colonisation

Full size image

Discussion

In this retrospective study of 151 ventilated children with RSV bronchiolitis, we have shown that the risk for prolonged MV was associated with prematurity, neonatal respiratory distress requiring oxygen including BPD and tracheal colonisation. Of these factors, only BPD was associated with a need for ECMO support.

Many studies have shown that preterm neonates or infants with neonatal respiratory distress requiring oxygen, including those with BPD, are among those at risk for higher RSV disease severity [1,20]. In particular, it has been well established that infants with RSV bronchiolitis who were born prematurely have a significantly higher admission rate [13], a greater risk for RSV-related rehospitalisation [12] and for the need of MV than normal full term infants [19]. In our study, we found that prematurity was associated with prolonged duration of MV, with a higher correlation in infants with a very low GA. The need for oxygen supplementation during the neonatal period also seems an important predictive factor for greater bronchiolitis severity, particularly regarding the existence of BPD. Taken together, these data provide evidence that preterm neonates can be considered as having a higher risk for prolonged ventilatory support when MV is required for RSV bronchiolitis, especially if their history includes MV in the neonatal period and/or BPD. In contrast, regarding the influence of CHD, our results differ from those of a previous report in a cohort of 72 children [3]. Nevertheless, it is important to consider that infants with CHD constituted only 4.6% of our population which perhaps explains the lack of significance.

ECMO support was used in more than 9% of infants in our cohort. This is explained by the fact that some children had been referred from other hospitals to our PICU for ECMO. We assessed the impact of BPD as a predictive factor for the need for ECMO support. To the best of our knowledge, only two studies have previously reported on the use of ECMO in a cohort of 12 infants [17] and 24 infants [9], all treated for bronchiolitis. Steinhorn and Green [17] reported that a need for ECMO was observed mainly in patients born prematurely with a survival rate of 58%, whereas children supported by ECMO in the study of Khan et al. [9] were previously healthy infants which most likely contributed in part to their survival rate of 96%. In the present study, infants supported by ECMO accounted for much of the overall mortality rate from RSV infection. Nevertheless, more than 70% of the patients supported by ECMO survived, although each patient would have had little chance of survival if conventional medical therapy were continued. We conclude that ECMO may provide lifesaving support and can be used successfully in cases refractory to conventional management.

The present study provides the first evidence suggesting that tracheal colonisation is frequent and contributes to bronchiolitis severity in children requiring MV. Previous studies reported that bacterial infection was observed in 39% of the children with RSV infection [10], whereas prevalence of serious bacterial infections has been shown to be very low [2,15]. None of these studies had investigated the tracheal colonisation among a cohort of ventilated infants. Importantly, in our population (1) all infants were intubated before admission or in the first hours after admission and (2) tracheal aspirations were performed in the first hours of ventilatory support, which limits the risk of secondary bacterial infections that might occur during hospitalisation [7]. Although in many cases these positive tracheal cultures represented colonisation rather than true infection, the bacterial status was one of the strongest independent predictors of prolonged ventilation, especially for H. influenzae. Interestingly, many studies have shown the strong association between RSV and H. influenzae [5, 10,16] and those in experimental animals have also demonstrated that RSV infection enhances the colonisation with H. influenzae [8,14]. We found that children in our cohort were often treated with antibiotics. This is consistent with previous studies which show high rates of antibiotic use in children ventilated for RSV infection [6] and are concordant with recommendations regarding antibiotic treatment for these children [2]. These recommendations are emphasised by the importance of bacterial colonisation as a risk factor for severe RSV bronchiolitis, although the results of our study should be verified and extended.