Severe respiratory distress is a common and serious complication of premature birth, neonatal pneumonia, and neonatal sepsis, which together account for over one-half of all neonatal deaths globally [1]. More than 50% of babies born at ≤31 weeks of gestation will develop respiratory distress syndrome (RDS) [2]. Respiratory distress is associated with over 80% of cases of neonatal pneumonia [3] and most cases of neonatal sepsis [4].

In the developed world, respiratory support is provided to neonates using either mechanical ventilation or Continuous Positive Airway Pressure (CPAP). Unfortunately, ventilators and CPAP machines are too expensive and technically complex for many resource-limited settings [5]. As a result, respiratory illness remains one of the most common causes of neonatal death in the developing world.

CPAP is a gentle and effective tool to treat even preterm and low birth weight infants in respiratory distress [6]–[11]. Well-resourced hospitals use ventilators, stand-alone CPAP devices, or tubing, wall air and oxygen to set up CPAP at the bedside [8]. In bubble CPAP (bCPAP), pressure is safely regulated by submerging the end of the tubing in a bottle of water. The depth of water determines the pressure in the system. This pressure helps recruit alveoli and increase functional residual lung capacity [12], thus lowering the baby's work of breathing. The result is better compliance, reduced airway resistance, conservation of surfactant, and stabilized chest and diaphragm [13]. Bubble CPAP has been used in developed countries for decades [8]. It reduces morbidity [6] and mortality [5], [8], as well as the need for mechanical ventilation [6], [7], [9]–[11], [14], [15]. It can be administered by trained nurses [14], [16], and is safer than mechanical ventilation [6], [14]. It also reduces hospital stay [6] and up-referrals [17]. As a result CPAP is increasingly used as a first choice for ventilatory support in tertiary centres [9], [14], [18].

A number of observational studies have shown that CPAP can be safely implemented in low-resource settings using commercially available devices designed for high-resource settings [14], [19]–[22]. Unfortunately, the cost and complexity of currently available CPAP devices is prohibitive for many low-resource settings.

We recently developed a novel, low-cost bCPAP system for low-resource settings that can be assembled for approximately $350, a fifteen-fold cost reduction compared to the average stand-alone CPAP. The device delivers the same therapeutic flow and pressure as bCPAP systems used in high-resource settings [23]. Here, we report a study to evaluate the efficacy of the new low-cost bCPAP system in improving survival in newborns with severe respiratory distress at Queen Elizabeth Central Hospital in Malawi. Malawi has a population of 15.4 million [24], with a preterm birth rate of 18% [25], neonatal mortality rate of 27 and infant mortality rate of 53 per 1000 live births [26]. The GDP is $805 per capita and 73.9% of the population live on less than $1.25 per day [27]. Queen Elizabeth Central Hospital is the main referral hospital in the southern region of the country; each year, about 10,000 babies are delivered at the hospital and over 3,000 are admitted to the neonatal ward [28].

Eighty-nine neonates were enrolled from January to October 2012 (Table 1). Chart review identified two neonates who did not meet eligibility criteria because they suffered severe birth asphyxia. Of the 87 eligible participants, 62 were treated with bCPAP therapy and 25 received oxygen therapy. The 62 neonates receiving bCPAP included nine who were initially assigned to the control group, but ultimately received bCPAP therapy when a device became available. Data were analysed for all 87 eligible neonates.

Table 1 summarizes the demographic and clinical covariates for the eligible neonates. Differences in the mean values of covariates for the group receiving bCPAP were not statistically significant compared to those for the control group with one exception: the fraction of babies who required bag and mask ventilation before recruitment was higher in the bCPAP group than in the control group (p = 0.015). Most participants had very low birth weight (VLBW) and were born at QECH. Over 25% of neonates in each group were from multiple births, and approximately 20% of neonates in each group were exposed to HIV.

Figure 2 compares survival to discharge in the two groups. The survival rate in the control group was 44.0% (95% CI: 26–63%); survival in the bCPAP group was 71.0% (95% CI: 59–81%) (p = 0.018). Without adjustment, bCPAP is associated with a 3.1 fold increase in odds of survival (95% CI: 1.2–8.1, p = 0.02). Figure 2 also shows Kaplan-Meier survival curves of cumulative survival vs. time following initiation of treatment. These are different by the log rank test at p = 0.013.

Univariate logistic regression analysis indicated that five of the covariates in Table 1 were related to survival: a primary diagnosis of RDS (p = 0.002); co-morbidity of sepsis (p = 0.039); birth weight (p = 0.003); a birth weight in the VLBW range (p = 0.016); and gestational age (p = 0.023). Further subgroup analyses were performed by birth weight, presence of RDS and of sepsis. As gestational age was not available for approximately 10% of subjects, no subgroup analysis was conducted for this variable.

Figure 3 compares survival for eligible neonates with a primary diagnosis of RDS and those with sepsis. For neonates with RDS, survival was 23.5% (95% CI: 9–49%) in the control group, compared to 64.6% (95% CI: 50–77%) in the bCPAP group (p = 0.006). None of the seven neonates with sepsis in the control group survived, while survival was 61.5% (95% CI: 42–78%) in the bCPAP group (p = 0.005).

Figure 4 shows survival rates for eligible neonates in the two groups, stratified by birth weight. Improvements in survival are greatest for VLBW infants (≥1.0–<1.5 kg); in whom survival was 15.4% (95% CI: 4–45%) in the control group compared to 65.5% (95% CI: 47–80%) in the bCPAP group (p≤0.001). For babies with low birth weight (≥1.5 kg–<2.5 kg) and babies with birth weight ≥2.5 kg, differences in survival were not statistically different for the two groups.

Multiple logistic regression analysis was performed to adjust for baseline differences in RDS, sepsis, and VLBW; the adjusted survival rate for babies receiving bCPAP is 85% (95% CI: 69–94%), while that for standard care is 47% (95% CI: 23–73%) and bCPAP is associated with an adjusted 6.5 fold increase in odds of survival (95% CI: 1.9–22.7, p = 0.003).

Table 2 compares the secondary outcomes of treatment time and duration of hospital stay for all eligible subjects and for those who survived to discharge. On average, neonates receiving bCPAP spent five more days in the hospital and nearly three more days on treatment than those in the control group. On average, neonates who were transitioned from nasal oxygen to bCPAP spent twice as long in the hospital and twice as long on treatment as those who received bCPAP initially; the average delay in receiving bCPAP was 3.1 days. Similar trends in increased treatment and hospitalization time were seen for neonates who survived to discharge. The increase in treatment and hospitalization time in bCPAP could be explained by the differences in primary diagnosis between the two groups; children receiving bCPAP were more likely to suffer from RDS than those receiving nasal oxygen.

Only mild complications were associated with bCPAP (Table S1), including nasal irritation, facial irritation, and epistaxis. Similar complication rates were observed in the control group. No clinical diagnoses of pneumothorax were made, but as routine x-rays and cranial ultrasound were unavailable, silent complications cannot be ruled out.

This study demonstrates the potential impact of introducing a low-cost, appropriate bCPAP system in a low-resource setting where standard therapy is nasal oxygen. Using bCPAP to treat newborns with severe respiratory distress resulted in a 27% absolute improvement in survival to discharge. The benefit was more pronounced in neonates with VLBW, RDS, or sepsis. Only 24% of neonates with RDS treated with nasal oxygen survived to discharge, compared to 65% receiving bCPAP, reflecting similar rates of survival as observed in the US during the transition from nasal oxygen to bCPAP for treatment of RDS [5].

The value of CPAP therapy for neonatal RDS is well-accepted in high-resource settings but very few studies have examined the value of CPAP therapy in resource-limited settings lacking advanced respiratory support [16], [20], [29]. These studies have either included older children [16], examined immediate improvements in clinical endpoints rather than survival to hospital discharge [16], or been small, observational studies [20], [29].

Our study was carried out in a typical low-resource setting using a bCPAP that was conceived, designed and evaluated with input from local physicians and nurses, ensuring that the device addresses competencies and capacity of this under-resourced environment [23], The technical performance of the device was monitored throughout the study. Therapeutic flow and pressure met performance standards reliably without preventive maintenance, and no device failures occurred. In contrast, 40% of study oxygen concentrators failed during this same period when circuit boards were damaged by line voltage spikes. To meet cost and infrastructure constraints, the bCPAP device did not heat or humidify the mix of pressurized air and oxygen delivered via the nasal prongs. Instead, humidification was provided through routine use of nasal saline drops. This did not appear to result in significant complications (Table S1).

Our study has a number of limitations. As bCPAP is known to be an effective therapy [5], [6], [8], we did not perform a randomised trial. Instead, the decision to treat with bCPAP was based on availability of a bCPAP device. This design allowed potential bias. We planned to recruit equal numbers of patients in each arm to detect a difference between the two groups. However, since allocation of CPAP to eligible subjects was based on device availability actual enrollment was dependent on available equipment and staffing resources. Our original estimate was that CPAP devices would be available approximately half the time needed. To allow for potential loss due to data collection errors, we planned to enroll up to 110 patients. In practice, CPAP devices were available more frequently than not and thus more study participants received CPAP than nasal oxygen. The study was stopped when over 60 participants received CPAP; at this point, the projected number of CPAP patients had been enrolled plus the entire additional allowed margin to account for data loss.

In addition, this design could result in treating infants with more severe illness with bCPAP and those with less severe illness with nasal oxygen. Indeed a significantly higher proportion of infants treated with bCPAP required resuscitation prior to therapy (24.5%) than those who received nasal oxygen (4.0%). We monitored the status of children receiving nasal oxygen. Nine of them were transitioned to bCPAP when they failed to improve and a device became available. The outcomes of these nine children were analysed with the bCPAP group. All of the babies in the CPAP group received CPAP within 36 hours of being identified as needing CPAP, with a majority receiving it immediately. Five of the nine babies in the group that transitioned to CPAP also received CPAP treatment within 36 hours of being identified as needing CPAP. Because, in general, preterm infants who survive the first few postnatal days have an increased chance of long-term survival, transitioning babies into the CPAP group after a few days in the oxygen group might have created bias in favour of CPAP.

To account for this potential bias, we performed a per-protocol analysis, comparing outcomes for infants who received bCPAP to those who received nasal oxygen; babies who initially received oxygen but were later transitioned to CPAP are not included in the per-protocol analysis. The survival rate in the control group was 44.0%; survival in the bCPAP group was 69.8% (p = 0.014). For neonates with RDS, survival was 23.5% in the control group, compared to 61.5% in the bCPAP group (p = 0.004). Again, improvements in survival were found to be greatest for VLBW infants (≥1.0–<1.5 kg); in whom survival was 15.4% in the control group compared to 57.1% in the bCPAP group (p = 0.006).

Analysing these nine babies as part of the standard care group, analogous to intent-to-treat, we find that 69.8% (37/53) of children receiving bCPAP survived to discharge, compared to 52.9% (18/34) for standard care; the improvement in survival approaches significance (p = 0.087). Finally, we did not determine the optimal time to initiate or terminate treatment with bCPAP. Additional studies are needed to determine if earlier commencement of bCPAP could further improve outcomes in this setting.

Neonatal deaths account for 41% of global child mortality; the neonatal mortality rate has changed little in the last decade [30]. If our results are generalizable, we estimate that on the African continent, where nearly one million babies die each year within a week of birth [31], providing low-cost bCPAP in central and district hospitals could prevent 178,000 neonatal deaths. While the cost of the bCPAP device has been reduced, the cost and availability of consumables, staff support and support equipment remain a barrier to scale-up. Nonetheless, implementing such a system has the potential to improve neonatal care and health outcomes in low-resource settings.