High-frequency oscillatory ventilation and short-term outcome in neonates and infants undergoing cardiac surgery: a propensity score analysis

High-frequency oscillatory ventilation (HFOV) is an established treatment for acute respiratory distress in preterm neonates. However, there is no evidence that it improves outcome in term or near-term neonates with pulmonary disease [1]. HFOV is considered as a rescue therapy in children with severe acute respiratory distress syndrome (ARDS), but to date there is lack of evidence to support it [2, 3]. HFOV is also used to achieve lung recruitment and improve oxygenation when recruitment maneuvers have failed, as part of the "open lung" and lung protective ventilation strategies in adults with severe ARDS; early initiation of HFOV has been associated with improved outcome [4‐6]. Mild acute lung injury occurs in 12% of adults following cardiopulmonary bypass (CPB), and more severe lung injury, indistinguishable from ARDS, in 0.4% [7, 8], as a result of accumulation of excessive extrapulmonary lung water, decreased lung compliance, atelectasis and increased shunting.

Experience with HFOV following cardiac surgery is limited, due to concerns about hemodynamic impairment in animal and human studies [6, 9‐13]. However, HFOV has been associated with a significant reduction in pulmonary vascular resistance (PVR) after the Fontan procedure in children [14]. Thought to be beneficial for gas exchange and PVR, the present authors have used HFOV in neonates and infants with respiratory distress following cardiac surgery since January 2007. The aim of the present study was to assess associations between commencement of HFOV on the day of surgery (Day 0) and the length of mechanical ventilation and Intensive Care Unit (ICU) stay, and mortality in this population.

This retrospective cohort study was conducted at the Necker University Hospital in Paris, France. It was reviewed and approved by the Ethics Committee of the French Society of Thoracic and Cardiovascular Surgery, which waived the requirement for consent to use anonymized records. All parents had provided informed consent to surgery.

Records of all neonates and infants who underwent cardiac surgery between 1 January 2001 and 30 June 2010 were reviewed; patients switched to HFOV on Day 0 were identified as the HFOV group. Those switched to HFOV after Day 0, as a rescue therapy, were not analyzed. The remaining patients were included in the control group. Data for each patient were extracted retrospectively from a prospective database, which is updated daily by the clinical staff. These concerned: demographics, surgical and CPB techniques, short-term outcome variables accounting for the severity of the postoperative illness, such as re-operation, delayed sternal closure, extracorporeal membrane oxygenation (ECMO), acute kidney injury requiring renal replacement therapy (RRT), and hospital-acquired pneumonia [15], length of mechanical ventilation, length of ICU stay and in-hospital mortality. Normothermic CPB with intermittent warm blood cardioplegia was performed in every patient during the study period, except in cases where deep hypothermic circulatory arrest (DHCA) was indicated [16]. Pulmonary arterial pressure was measured in every patient, either continuously by a catheter inserted into the pulmonary artery by the end of surgery, or by serial echocardiography. Persistent pulmonary hypertension was noted whenever it occurred during the postoperative course, and inhaled nitric oxide was administered [17].

All patients were initially commenced on pressure controlled conventional mechanical ventilation (CMV) using a SERVO-300 (Siemens-Elema AB, Sweden) before 2002, then a SERVO-i ventilator (Maquet GmbH&Co.KG, Rastatt, Germany). This was set to provide a positive end expiratory pressure of 2 cmH₂O, a tidal volume of 6 to 8 ml Kg^-1, and a fraction of inspired oxygen, which was dependent upon the underlying cardiac disease. In the event of severe respiratory failure, recruitment maneuvers by stepwise increase in the mean airway pressure were applied. The patients were switched to HFOV when hypoxemia and acidosis occurred despite increasing alveolar ventilation on CMV, when the tidal volume exceeded 10 ml kg^-1, or when there was evidence of pulmonary hypertension and right ventricular failure. The decision to switch was made by the attending intensivist. A SLE 2000 or a SLE 5000 HFO ventilator (SLE Ltd, South Croydon, UK) was used. This was set to a mean airway pressure (Paw) of 12 cmH₂O, an inspiratory to expiratory ratio of 33%, and an oscillation frequency of 8 Hz. Amplitude was adapted to achieve adequate chest wall vibrations. All parameters were adjusted to achieve optimal inflation, a PaCO2 of 35 to 45 mmHg and a pH > 7.35. The adequacy of the PaO2 level was judged according to the underlying cardiac disease. Patients were switched back to CMV when these conditions had been achieved with an oscillation frequency ≥10 Hz and a mean Paw ≤10 cmH₂O. Sedation was achieved through a continuous infusion of midazolam and morphine. Whenever possible, muscular relaxants were avoided and spontaneous breathing was maintained. Catecholamine support (milrinone and epinephrine), fluid support and diuretics were administered as appropriate to achieve hemodynamic stability and a negative fluid balance. All patients were weaned from mechanical ventilation when the underlying indication had resolved and following a successful one-hour trial of spontaneous breathing with a continuous positive pressure of 2 cmH₂O and a pressure support of 10 cmH₂O.

Overall, 3,549 neonates and infants underwent cardiac surgery during the study period. Life support was withdrawn from four patients with obstructed total anomalous pulmonary venous connection and one patient with severe pulmonary hypoplasia, with hopeless prognosis secondary to pulmonary lymphangienctasia. Another two patients died periprocedural, leaving 3,542 cases to be analyzed. The number of patients, their length of mechanical ventilation and ICU stay across the study period are shown in Figures 1 and 2.

Patient characteristics are shown in Table 1. The 120 neonates and infants switched to HFOV on Day 0 were younger and smaller, had undergone more complex surgery and had experienced more severe postoperative illness. Patients switched to HFOV had longer durations of both mechanical ventilation, median 7 days, inter-quartile ranges (IQR) 5 to 11 vs. 1 day, IQR 0.3 to 4 in controls (P < 0.001), and ICU stay, median 11 days, IQR 7 to 15.7 vs. 4 days, IQR 3 to 7 (P < 0.001). The median duration of HFOV was 4 days, IQR 2 to 7. The in-hospital mortality rates for the two groups were similar, 8.3% in patients switched to HFOV vs. 4.8% in controls (P = 0.08).

Table 2 shows the most prevalent procedures and their "HFOV indexes", range 0 to 0.82, median 0.04, IQR 0.03 to 0.08. Table 3 shows the variables included in the propensity score. When data were missing, the median of the respective variable was used (< 5% of all information concerning CPB technique was missing). The propensity score model was well calibrated (Hosmer Lemeshow test, P = 0.14) and discriminated well between patients on HFOV and the others (c index = 0.82). Patients were more likely to be switched to HFOV on Day 0 if they were small, had undergone a procedure with high "HFOV index", required CPB and DHCA, had hemodynamic impairment precluding closure of the sternum or required RRT on Day 0. Patients commenced on postoperative ECMO were not matched due to their high mortality rate (39.1%).

Matching resulted in two well-balanced groups of 120 patients respectively: the HFOV and the CMV groups (Table 1). The HFOV group had shorter durations of mechanical ventilation, 7 days, IQR 5 to 11 vs 9 days, IQR 5 to 17 in the CMV group (P = 0.03), shorter durations of ICU stay, 11 days, IQR 5 to 17 vs 14 days, IQR 9 to 22 (P = 0.009), a higher prevalence of pulmonary hypertension, 36.7%, compared to 23.3% (P = 0.03) and a similar prevalence of hospital-acquired pneumonia, 49.2% in the HFOV group compared to 47.5% in the CMV group (P = 0.80). Ten patients in the HFOV group (8.3%) died during ICU stay, compared to 18 in the CMV group (15.8%) (P = 0.08). Median follow-up was 411 days, IQR 32 to 3,060, 18 patients (15.8%) died during follow-up in the HFOV group, compared to 22 in the CMV group (18.3%) (P = 0.66).

Kaplan-Meier plots of the probability of successful weaning from mechanical ventilation over time are shown in Figure 3. The median length of mechanical ventilation was 7 days in the HFOV group, IQR 5 to 11, and 9 days in the CMV group, IQR 5 to 17 (P = 0.01). Four patients in the HFOV group underwent mechanical ventilation for ≥ 30 days. Of these, one developed tracheal stenosis and underwent slide tracheoplasty, another required tracheostomy. Ten patients in the CMV group underwent mechanical ventilation for ≥ 30 days. Of these, two developed tracheal stenosis, one of whom died, one developed oeso-tracheal fistula and died, and seven developed chronic lung disease, of whom four required tracheostomy and two died. Kaplan-Meier plots of the probability of ICU delivery over time are shown in Figure 4. The median length of ICU stay was 11 days in the HFOV group, IQR 7.2 to 15.7 and 14 days in the CMV group, IQR 9 to 22 (P = 0.002). Differences between length of ventilation and ICU stay were found significant with a statistical power of 0.77 and 0.89, respectively.

Cox proportional-hazards regression analysis, adjusted for the delay to sternal closure, duration of RRT, occurrence of pulmonary hypertension and year of surgery, showed that patients in the HFOV group had a higher probability of successful weaning over time, adjusted HR 1.63; 95% confidence interval (CI) 1.17 to 2.23 (P = 0.004) (Table 4). The probability of ICU delivery over time was also higher in the HFOV group, adjusted HR 1.65, 95% CI 1.20 to 2.28 (P = 0.002) (Table 4). Longer delay to sternal closure was independently associated with longer length of mechanical ventilation and ICU stay.

The present study reports experience with HFOV in a population of neonates and infants with respiratory distress following several cardiac surgery procedures. Previous findings reported from randomized trials of HFOV in term or near-term neonates with pulmonary disease showed no benefit in terms of 28-day mortality [1], and our findings were similar. But, unlike previous research on elective use of HFOV, length of mechanical ventilation and length of stay were reduced among patients with a similar severity of illness when they were switched to HFOV on the day of surgery.

When commenced on the day of surgery, HFOV was associated with a shorter duration of mechanical ventilation and ICU stay in this population of neonates and infants with respiratory distress following congenital cardiac surgery. No association was observed between the use of HFOV and mortality.