Infants create positive end-expiratory pressure by grunting to try and prevent atelectasis [23]. Gregory et al. [21] used that principle to develop continuous positive airway pressure (CPAP) and described its first successful use in neonates. Since then, the use of CPAP has become widespread. CPAP can increase functional residual capacity thereby improving oxygenation [47], reduce airway resistance and the work of breathing, stabilise the chest wall enhancing respiratory thoraco-abdominal synchrony [35], improve diaphragmatic function [4] and reduce upper airway resistance decreasing obstructive apnoea [37, 38]. CPAP, however, does have side effects, such as ‘CPAP belly syndrome’ [27], nasal irritation and skin damage [63], and nasal deformities if CPAP use is prolonged [48]. There are different methods of CPAP generation and delivery and CPAP is now used to support prematurely born infants at different stages in their respiratory illness. The aim of this review is to critically analyse the literature to determine if there is sufficient evidence to determine the optimum method of CPAP generation or delivery and when prematurely born infants would most benefit from CPAP.

Methods of CPAP generation

Devices which generate CPAP can broadly be divided into two categories, continuous flow or variable flow devices. Continuous flow devices include conventional ventilators, jet ventilation systems and bubble CPAP. Conventional ventilators provide a constant flow of gas and the pressure is controlled by the exhalation valve. With a jet system, a small jet is produced either at the nostrils or in a prechamber in front of the nasal prongs. During bubble CPAP, the pressure is set by immersing the expiratory limb of the CPAP device in an underwater chamber to a depth equal to the desired CPAP level. The gas flows through the system causing bubbling in the chamber; this causes variability in the mean CPAP pressure. Using a variable flow device, the CPAP level is dependent on the flow of gas.

High-flow nasal cannulae (HFNC) are small, thin, tapered cannulae which deliver flows in excess of 1 L/min [61]. HFNC may deliver positive end-expiratory pressure [19]. There have been four randomised or quasi-randomised trials which were included in a Cochrane review [61]. In one trial [42], HFNC (Vapotherm ®) was compared to bubble CPAP in infants requiring CPAP in the first 6 h after birth, no significant difference was found in the requirement for intubation (RR 1.03; 95% CI 0.28, 3.78). In a second study [9], HFNC compared to nasal CPAP post-extubation was associated with a significantly higher rate of re-intubation (RR 4.00; 95% CI 1.33, 12.05). In a further study [62], Vapotherm ® performed better than standard unhumidified high flow nasal cannula with regard to maintaining a normal appearing nasal mucosa, a lower respiratory effort score and averting reintubation. In the fourth trial [39], two different HFNC delivery systems (Vapotherm ® and Fisher and Paykel) post-extubation were compared in 40 infants, no significant difference was found in the rate of extubation failure. The Cochrane review [61] concluded that the trials differed in their design and interventions, such that meta-analysis was not possible and, at present, there is insufficient evidence to establish the safety or efficacy of HFNC for prematurely born infants.

There have been few randomised trials comparing methods of CPAP generation, the majority are of crossover design (Table 1). The results of those studies have suggested that variable-flow CPAP has advantages over ventilator-delivered CPAP being associated with lower duration of supplementary oxygen, length of stay and work of breathing and greater stability of flow and tidal volume [25, 44, 54]. In subgroup analysis of infants ventilated for less than 2 weeks, bubble CPAP compared to variable-flow CPAP was associated with a significantly lower rate of extubation failure, as well as a significantly reduced duration of CPAP, but overall, there were no significant differences in the rates of successful extubation [22]. In a smaller randomised crossover study, bubble compared to variable-flow CPAP was associated with significantly greater work of breathing, respiratory rate and thoraco-abdominal asynchrony [34].

Table 1 Studies comparing CPAP systems

CPAP delivery

Short binasal prongs and nasal masks are two of the most commonly used interfaces for delivering CPAP [30]. Other delivery methods include via a nasopharyngeal prong, an endotracheal tube and a helmet. Short binasal prongs impose a lower work of breathing compared to nasopharyngeal prongs [31, 40]. The results of a meta-analysis highlighted that binasal prongs were more effective in preventing reintubation in prematurely born infants when compared to either single nasal or nasopharyngeal prongs (RR 0.59, 95% CI 0.41–0.85) [16]. Nasal injury following CPAP delivered by short nasal prongs has been documented [48]. In a randomised study of 89 very low-birthweight infants, in which nasal prongs were compared to nasal mask, no significant difference in the incidence of nasal injury was demonstrated, but the incidence of nasal injury was significantly associated with the duration of CPAP [63].

In a short-term randomised crossover study, ‘helmet’ CPAP was found to be as effective as nasal CPAP with respect to oxygen requirement, oxygen saturation, heart rate, respiratory rate, mean arterial blood pressure and transcutaneous carbon dioxide. In addition, helmet CPAP use was associated with a significant increase in infant comfort assessed by the Neonatal Infant Pain Scale [59]. There are, however, concerns that it may adversely affect cerebral blood flow [64] and it is associated with significantly higher noise levels than nasal CPAP [58].

CPAP for initial respiratory support

Avery et al.’s survey of eight tertiary neonatal centres in the USA [6] demonstrated that the centre which used CPAP as initial respiratory support had the lowest rate of bronchopulmonary dysplasia (BPD) compared to centres in which the initial mode was intubation and ventilation. A subsequent, retrospective study showed that Columbia’s primary CPAP approach continued to be associated with a lower rate of BPD compared to that seen in centres which favoured mechanical ventilation and surfactant administration [60]. Other observational studies have highlighted that use of early CPAP was associated with significantly less need for intubation and ventilation [2, 15, 20, 26, 36, 43] and surfactant therapy [1, 15] and lower incidences of intraventricular haemorrhage (IVH) [3, 26, 33] and BPD [2, 3, 7, 15, 33, 36].

There have been a number of randomised trials comparing early CPAP to other respiratory mode strategies. In a small randomised study [12] of 27 infants of less than 30 weeks of gestation, prophylactic surfactant followed by immediate extubation to CPAP compared to prophylactic surfactant followed by mechanical ventilation was associated with less need for mechanical ventilation at 7 days (p = 0.026). The CPAP group also had significantly lower durations of supplementary oxygen, CPAP and mechanical ventilation, less need for a second dose of surfactant and a shorter duration of intensive care, but there were no significant differences in death or BPD at 28 days or 36 weeks postmenstrual age (PMA). Sandri et al. [49] compared prophylactic to rescue CPAP in a randomised study of 230 infants born between 28 and 31 weeks of gestation, who did not require intubation in the delivery room. No significant differences were found in the need for surfactant therapy or mechanical ventilation or the incidence of air leaks. There was a trend towards an increased risk of IVH grades 3–4 (RR 3.0, 95% CI 0.96, 28.42) in the early CPAP group. Three recently published randomised trials have assessed early CPAP [41, 50, 55] and only one reported a reduced risk of death or BPD but only at 28 days and not 36 weeks PMA (Table 2). In the COIN trial [41], the combined outcome of BPD or death was lower in the CPAP group at 28 days (OR 0.63; 95% CI, 0.46–0.88; p = 0.006), but not at 36 weeks PMA (OR 0.76; 95% CI, 0.54–1.09). A significantly lower proportion of the CPAP group required surfactant therapy (38% vs. 77%, p < 0.001) and the infants in the CPAP group required fewer days on mechanical ventilation (3 vs. 4 days, p < 0.001). The CPAP group, however, had a significantly increased incidence of pneumothorax (9.1% vs. 3%, p = 0.001), which might relate to the CPAP pressure of 8 cm H2O or the lower use of prophylactic surfactant. In the SUPPORT trial [55], there was no significant difference in the primary outcome of death or BPD, although there was higher postnatal steroid use in the intubation–surfactant group (13.2% vs. 7.2%, p < 0.001). The CPAP group required significantly less days of mechanical ventilation (p = 0.03) and had a significantly higher rate of survival free of mechanical ventilation at 7 days (RR 1.14, 95% CI, 1.03–1.25). In the CURPAP trial [50], no significant difference was found in the primary outcome, the need for mechanical ventilation in the first 5 days or in the secondary outcomes including death, BPD, air leaks, grades 3–4 IVH, sepsis, retinopathy of prematurity and pulmonary haemorrhage.

Table 2 Recent trials assessing early CPAP

CPAP following extubation

Meta-analysis of the results of nine randomised trials (Table 3) highlighted that extubation to nasal CPAP rather than into headbox oxygen was associated with a decreased incidence of respiratory failure defined as respiratory acidosis, apnoea or an increased oxygen requirement needing additional ventilatory support (RR 0.62, 95% CI 0.51–0.76) [14]. In a subgroup analysis, in the trials in which CPAP levels of 5 cm H2O or greater were used, CPAP significantly reduced extubation failure (RR 0.49 (95% CI 0.37–0.66)) [5, 13, 18, 24, 52].

Table 3 Randomised trials of CPAP following extubation

Weaning from CPAP

A survey of 58 neonatal units in the North of England [8] revealed that only 3 units had a weaning protocol. Some respondents practised abrupt discontinuation of CPAP, but the majority of units (66%) weaned by ‘time off’, 4% by weaning pressure and 30% indicated no set method. In a survey of 124 Australian tertiary neonatal units [28], 70% of respondents reported using graded time off CPAP as part of weaning; 74% of all respondents (whether using graded time off or not) replied that they gradually reduced the airway pressure before coming off CPAP. Three randomised controlled trials have examined different methods of weaning CPAP [51, 53, 57], two have been reported only as abstracts (Table 4). Weaning by reduction in pressure rather than by time cycling was associated with a shorter duration of weaning, less days on CPAP and a lower incidence of chronic lung disease [51, 53].

Table 4 Trials of CPAP weaning