We conducted a prospective cohort study in the 14-bed PICU of a tertiary care hospital between January 2005 and December 2009. We used consecutive sampling to recruit patients aged 0–18 years with a diagnosis of ARF admitted to the PICU that received NIV as initial treatment. Non-invasive ventilation was initiated in patients that met the clinical criteria for ARF and at high probability of intubation in the next 6h. This decision was made by the clinician in charge of the patient if the patient presented any of the following signs: respiratory rate (RR) more than 2 standard deviations (SDs) above normal, need of a fraction of inspired oxygen (FiO2) greater than 0.5 to maintain the percentage of oxygenated haemoglobin (peripheral capillary oxygen saturation, SpO2) above 94% or progressive dyspnoea at rest. We categorised respiratory failure as type I (changes in ventilation/perfusion without impairment of gas exchange at the alveolar level) or type II (changes accompanied by impaired gas exchange in alveoli) based on the classification proposed by Newth.14 The exclusion criteria were the presence of any contraindication for the use of NIV15: cardiac arrest, haemodynamic instability requiring inotropic therapy, severe arrhythmia, Glasgow coma scale score of less than 10, facial trauma, surgery, vocal cord paralysis, undrained pneumothorax and active upper gastrointestinal bleeding. We also excluded patients that were intubated or ventilated through a tracheostomy tube at the time of admission, as well as patients that received NIV after extubation or palliative NIV (such as patients with diseases where intubation is considered a futile intervention). The study was approved by the Ethics and Research Committee of the Hospital Sant Joan de Déu and considered exempt of the need for informed consent, as NIV is a routine treatment in the PICU and consent for the use of this type of data was not necessary at the time of the study.

The physician in charge of the patient was responsible for the decision to choose between continuous positive airway pressure (CPAP) and bilevel positive airway pressure (BiPAP). Continuous positive airway pressure was used in patients with apnoea or mild dyspnoea, and BiPAP in patients with moderate to severe dyspnoea, significantly increased work of breathing (based on lung sounds and diaphragmatic excursion) despite the use of CPAP, low inspiratory capacity or progressive hypercapnia.

We used the following interfaces: nasal prongs for the Bubble CPAP system (Fisher & Paykel Healthcare Ltd., New Zealand) and the Giulia™ ventilator (GINEVRI, Italy); Profile Lite™ and Comfortclassic™ Philips Respironics nasal masks (Philips Respironics, Netherlands) used as oronasal masks; the Mirage™ (ResMed, California, USA), PerformaTrak® (Philips Respironics, Netherlands) and 7500® (Hans Rudolph Inc., Kansas, USA) oronasal masks; and endotracheal tubes used as nasopharyngeal tubes. Table 1 details the criteria used to select the interface and ventilator. Comfeel™ hydrocolloid dressings (Coloplast, Denmark) and Mepentol™ hyperoxygenated fatty acid solution (Bama-Geve, Spain) were used to prevent pressure ulcers. Active humidification (Fisher & Paykel) was used in every patient starting in 2007.

The ventilation strategy was selected according to the protocol established by the Working Group on Respiratory Care of the Sociedad Española de Cuidados Intensivos Pediátricos (Spanish Society of Paediatric Intensive Care).15 The initial pressure set for CPAP was 4–5cmH2O. In patients that did not experience improvement in work of breathing or apnoea, the pressure was increased to up to 8cmH2O. The initial settings for BiPAP were an inspiratory positive airway pressure (IPAP) or maximum inspiratory pressure of 6–8cmH2O and an expiratory positive airway pressure (EPAP) or positive pressure at the end of expiration of 4cmH2O. The IPAP was increased in 2cmH2O increments to up to 22cmH2O in case of an inadequate tidal volume, insufficient improvement of work of breathing or hypercapnia. The EPAP was increased to up to 8cmH2O to improve alveolar recruitment and oxygenation. There was no set maximum FiO2.

If any sign of treatment failure was detected, NIV was discontinued and the patient underwent intubation and started conventional mechanical ventilation. The following were considered signs of failure: SpO2<85% and a partial pressure of carbon dioxide (pCO2) greater than 65mmHg despite application of maximum settings in NIV, increase in work of breathing, apnoea or development of exclusion criteria. Patients were weaned of NIV gradually, with stepwise decreases in IPAP by 2cmH2O and in EPAP by 1cmH2O when the patient experienced a decrease in RR and work of breathing and the FiO2 was less than 0.4. The patients underwent placement of a nasogastric tube to avoid abdominal distension and for administration of enteral feedings.

All patients underwent continuous monitoring of heart rate, RR and SpO2. Peripheral venous blood gas analysis was only performed when the physician in charge considered it necessary.

When nonpharmacological measures were insufficient to prevent patient-ventilator asynchrony, the patient received sedatives (in monotherapy or as combination therapy), such as oral levomepromazine (1mg/kg/dose), continuous intravenous infusion of midazolam (0.05–0.1mg/kg/h) or propofol (1–2mg/kg/h) for less than 24h.

For each episode, we recorded the age and sex of the patient, the underlying disease and the severity score (Paediatric Risk of Mortality Score II [PRISM II]). We also collected data on variables related to NIV: aetiology of ARF, type of respiratory failure, mode of ventilation, ventilatory parameters (FiO2, IPAP, EPAP), physiological variables (HR, RR and SpO2) prior to initiation of NIV and at 2, 8, 12 and 24h of treatment, changes (numerical difference in values) in HR and RR relative to baseline, peripheral venous blood gas values (when obtained), changes in the mode of ventilation, ventilator or interfaces used, administration of sedatives, development of complications or contraindications, mortality, duration of NIV, outcome of NIV (success/failure) and length of stay in the PICU.

We defined NIV failure as the need for intubation or rescue with BiPAP in the group initially treated with CPAP. The very strong correlation between the SaO2/FiO2 (S/F) and PaO2/FiO2 (P/F) ratios was confirmed in the paediatric population in 2009.16 Therefore, we calculated the S/F ratio retrospectively and assessed its potential usefulness as a predictor of NIV failure. We calculated the S/F ratio for the baseline before initiation of NIV and at 2, 4, 8, 12 and 24 of NIV, excluding any values of SpO2 greater than 97%.

We performed a descriptive analysis, expressing qualitative variables as percentages and 95% confidence intervals (CIs) and quantitative variables as mean or median as a measure of central tendency, and standard deviation or interquartile range (IQR) as a measure of dispersion, depending on whether or not the data distribution was normal. We assessed the correlation of outcome measures with different variables under study by means of the chi square test, the Student t test or the Mann–Whitney U test, depending on the shape of the distribution. We performed a multivariate analysis fitting forward stepwise binary logistic regression and Cox regression models to identify potential predictors of NIV success or failure, calculating the odds ratios and hazard ratios, respectively, with their corresponding 95% CIs. The variables selected for the model were those corresponding to p-values of less than 0.1 in the univariate analysis or which had been found to be significant in the previous literature. In case of potential colinearity, we selected the variable for which there was data earlier in time. We provide a measure of the predictive power of the model, the −2 log-likelihood statistic (−2LL), which is better the lower the value of the statistic, to better inform the reader.

In the period under study, 5101 patients were admitted to the PICU, of who 2238 (43.9%) required ventilatory support. Within this subset, 331 (14.8%) received NIV as the initial mode of ventilation. We excluded 48 patients that received NIV for palliative care and 1 case where a test ventilator was used. Thus, the final sample comprised 282 patients (12.6%), of who 71 were managed with CPAP and the rest with BiPAP (Fig. 1). In the CPAP group, 14% of patients had type I respiratory failure, compared to 40% in the BiPAP group.

Figure 1.

Flow chart of patients included in the study.

(0.11MB).

The most frequent relevant medical conditions in the sample were congenital heart disease (12.4%), chronic neuromuscular disease (9.1%), cardiovascular postoperative period (8.7%), history of preterm birth (3.2%) and Down syndrome (2%); 47% of the patients were previously healthy. The diagnoses associated with ARF most frequently were bronchiolitis (31%), pneumonia (15%), viral respiratory infection (9%) and asthma (7%).

Sedation was used in 47% of cases to promote patient-ventilator synchronisation, and oral levomepromazine was the drug used most frequently for the purpose. Fifteen patients required a change of interface due to leaks. Complications were mild and infrequent: pressure ulcers (n=11), pneumomediastinum (n=1), gastric distention (n=1), hypercapnia (n=1) and conjunctivitis (n=1). Four patients died, and none of these deaths could be attributed to the previous use of NIV. The proportion of success in our study was 71%, although there were differences based on the initial treatment, with the lowest proportion (55%) in the group treated exclusively with CPAP (Fig. 1). As Fig. 1 shows, 18 patients required rescue with BiPAP, which succeeded in preventing intubation in 11 (61%). Table 2 presents the baseline data for the total sample comparing the NIV success and NIV failure subsets. We found significant differences between the success and failure subsets at baseline for age (21 vs. 3.4 months), blood pH, type of respiratory failure and frequency of bronchiolitis diagnosis, and in the PRISM score at 24h (5.5 vs. 8).

Table 3 compares the patients treated with CPAP versus BiPAP, highlighting that patients that received CPAP were significantly younger and had lower oxygen requirements compared to patients treated with BiPAP.

Tables 4 and 5 present the differences between the NIV success and NIV failure subsets of patients in the CPAP group and the BiPAP group, respectively. Although we ought to highlight the difference in efficacy between NIV devices (Benveniste, 55%; BiPAP Vision® 71%) it was not statistically significant. In the BiPAP group, patients in whom NIV failed were significantly younger (10 vs. 28.5 months, P=.001), had significantly higher RRs (51 vs. 41bpm, P=.010) and had significantly lower blood pH values (pH 7.30 vs. 7.35, P=.014) prior to initiation of NIV compared to patients that did not require intubation. As can be seen in Fig. 2, approximately 70% of patients were intubated in the first 24h. The most frequent causes of intubation were the presence of clinical signs of respiratory muscle fatigue (absence of improvement in HR and RR, thoracoabdominal asynchrony) and persistent hypoxaemia.

Figure 2.

Duration of NIV in success and failure groups until NIV discontinuation or endotracheal intubation. PARCHEOS: Cambiar las comas a puntos en los decimales del eje vertical.

(0.1MB).

Lastly, the length of stay in the PICU was significantly longer in patients that experienced BiPAP failure.

We performed a multivariate analysis of the data for the total sample by means of binary logistic regression (BLR). We analysed the following variables: mode of ventilation, type of respiratory failure, age 6 months (greater than/less than), RR, HR and S/F ration at 2h of treatment and bronchiolitis (yes/no) (predictive power, −2LL=232). We also performed a multivariate analysis of the subset of patients that received BiPAP at some point during the course of disease by means of Cox regression. We analysed the following variables: mode of ventilation age 6 months (greater than/less than) and HR, S/F ad IPAP at 2h. Good predictive power (−2LL=22).

Another multivariate analysis involved the group treated exclusively with BiPAP.

• -

  BLR: the variables analysed were age 6 months (greater than/less than), values of IPAP, S/F ratio and HR and the decrease in the RR at 2h. Good predictive power (−2LL=131).

• -

  Cox regression analysis: the variables analysed were age 6 months (greater than/less than), values of IPAP, HR and S/F ratio at 2h. Lesser predictive power (−2LL=240).

Although we also performed a multivariate analysis in the CPAP group, we did not find statistically significant differences due to the small size of this subset of patients.