Study design
This single-center, open-label, randomized controlled study was conducted in the post-anesthesia care unit, surgical intensive care unit, and thoracic surgical ward of a tertiary university hospital in Italy, between September 2015 and April 2018. The protocol was approved by the local ethics committee and was registered on ClinicalTrials.gov (NCT02544477) before trial initiation. The study was conducted in accordance with the declaration of Helsinki, and written informed consent was obtained from all enrolled subjects according to committee recommendations.
Participants
All adult patients scheduled for elective thoracotomic pulmonary lobar resection for malignant disease were eligible for study inclusion. Exclusion criteria were refusal of informed consent, pregnancy, body mass index ≥ 35 kg/m2, history of obstructive sleep apnea syndrome, long-term oxygen therapy due to chronic pulmonary disease, presence of tracheostomy, and any nasal/facial defect that could impede HFNC or Venturi mask use.
Patients enrolled in the study were randomized to receive oxygen therapy by a Venturi face mask or a treatment with HFNC continuously over the course of 48 h after surgery. A computer-generated random allocation list was used to allocate enrolled patients to study arms.
Patient management during surgery
All enrolled patients received general anesthesia according to the following standard protocols: induction with propofol 2–3 mg/kg, fentanyl 1.5–2.5 mcg/kg, and rocuronium bromide 0.9 mg/kg; maintenance provided by sevoflurane titrated to keep bi-spectral index values between 40 and 60%, continuous infusion remifentanil 0.05–0.4 mcg/kg/min, repeated boluses of rocuronium bromide 0.02 mg kg−1 to maintain a train of four of 1–3 by neuromuscular monitoring, and 3–5 ml/kg/h of intravenous crystalloids and antibiotic prophylaxis; and postoperative analgesia was obtained by intercostal nerve block at the end of the procedure and paracetamol at a standard dose for the first three postoperative days. During surgery, all patients were ventilated with a tidal volume of 6–8 ml/kg of predicted body weight [21] for two-lung ventilation and of 5 ml/kg for one-lung ventilation; PEEP was set at 5 cmH2O throughout the whole surgical procedure. Recruitment maneuvers were performed once (i.e., after lobectomy) in all patients.
At the end of surgery, patients were extubated as the following criteria were met: spontaneous respiratory activity with exhaled tidal volume between 5 and 8 ml/kg; respiratory frequency ranging between 12 and 30 breaths/min; absence of residual neuromuscular blockade, as assessed by train-of-four monitoring; peripheral oxygen saturation (SpO2) ≥ 92%; hemodynamic stability (heart rate < 120/min; systolic blood pressure between 90 and 160 mmHg; no signs of cardiac ischemia, no hemodynamically significant arrhythmias and absence of catecholamines); body temperature ≥ 36 °C; adequate cough reflex; and absence of copious secretions.
After extubation, enrolled patients were transferred to either the post-anesthesia care unit or the intensive care unit, according to the decision of the attending anesthesiologist, who was not aware of the randomization arm. According to department guidelines, early intensive care unit admission was reserved to American Society of Anesthesiologists’ three patients who were deemed at high-risk of postoperative complications. Patients treated in the post-anesthesia care unit were transferred to the surgical ward within 4–8 h after surgery, unless deemed clinically inappropriate. Patients treated in the intensive care unit were transferred to the surgical ward on postoperative day 1, unless clinically contraindicated.
Study treatments
All patients had to undergo the assigned treatment within 30 min after extubation.
Patients in the control group received oxygen therapy via a Venturi mask (OS/60 K, FIAB, Florence, Italy); pure O2 flow was set depending on the needed FiO2 according to manufacturer recommendations. Patients in the intervention group received HFNC by AIRVO™ (Fisher & Paykel Healthcare Ltd., Auckland, New Zealand). The initial flow rate was 50 l/min and was eventually diminished in case of intolerance. Humidification chamber temperature was set at 37 °C and eventually diminished in case of intolerance.
In both groups, SpO2 was monitored continuously and FiO2 was titrated on an hour basis to maintain SpO2 between 92% and 98%. The assigned treatment was administered continuously until day 2 after surgery, 9.00 a.m., when patients were assessed for treatment interruption; study treatments were discontinued and patients were deemed weaned from oxygen therapy as the following criteria were met: respiratory rate ≤ 35 breaths/min; no recruitment of accessory muscles during calm breathing; hemodynamic stability (heart rate < 120/min; systolic blood pressure between 90 and 160 mmHg; no signs of cardiac ischemia, no hemodynamically significant arrhythmias, and absence of catecholamines); and core body temperature < 38.5 °C. After day 2, in case of failure to be weaned from oxygen therapy, all enrolled patients received Venturi Mask oxygen therapy, as long as deemed appropriate by the attending physician.
In the surgical ward, patients from both groups underwent a standard physiotherapy protocol: over the initial 24 h, this consisted of upright positioning, sitting on the edge of the bed or on the chair, non-resistance leg exercises, and lung expansion maneuvers twice a day (i.e., deep diaphragmatic breathing, thoracic expansion exercises, and incentive spirometry). As soon as the patient was weaned from oxygen therapy, a walking program was also adopted.
Measurements
Baseline blood gas analysis was obtained in the preoperative period. Postoperative blood gas analyses and dyspnea assessment were performed 1, 3, and 24 h after extubation and then on a daily basis up to day 4. Chest X-ray was obtained 2 h after surgery and then on a daily basis up to postoperative day 3.
Self-assessment of dyspnea (i.e., respiratory discomfort-shortness of breath) was performed by a visual analog scale (VAS) ranging from 0 (no dyspnea) to 10 (maximum dyspnea) (in Additional file 1: Figure S1).
Endpoints
The primary endpoint of the study was the overall incidence of patients developing clinically relevant hypoxemia (i.e., ratio of the partial pressure of arterial oxygen (PaO2) to FiO2 (PaO2/FiO2) < 300 mmHg) during the first four postoperative days.
Secondary outcomes were (i) the need for supplemental oxygen after study treatment discontinuation and within 7 days from randomization (i.e., a peripheral arterial oxygen saturation (SpO2) < 93% while breathing on room air); (ii) the occurrence of postoperative severe acute respiratory failure requiring ventilatory support; (iii) the degree of dyspnea over the course of the first four postoperative days; and (iv) the rate of pulmonary complications within 7 days after surgery.
In non-prespecified post hoc analyses, we also assessed (i) the overall incidence of patients developing moderate-to-severe hypoxemia (PaO2/FiO2 < 200 mmHg) over the first 96 h after surgery; (ii) the cumulative incidence of clinically relevant hypercapnia (i.e., PaCO2 > 45 mmHg) over the course of the first four postoperative days; (iii) PaO2/FiO2 and PaCO2 over the course of the first four postoperative days; (iv) the length of hospital stay; and (v) all-cause 30-day mortality.
Postoperative severe acute respiratory failure requiring ventilatory support was defined as the presence of at least two of the followings: respiratory acidosis (arterial pH ≤ 7.35 with PaCO2 > 45 mmHg); SpO2 < 90% or PaO2 < 60 mmHg at an FiO2 ≥ 0.5; respiratory frequency > 35/min; altered state of consciousness; and clinical signs of respiratory muscle fatigue [22]. Respiratory failure was initially treated with NIV, except when endotracheal intubation was required (i.e., cardiac arrest, loss of consciousness, psychomotor agitation, massive aspiration, persistent inability to remove respiratory secretions, heart rate < 50/min with loss of alertness, and severe hemodynamic instability without response to fluids and vasoactive drugs [22]). Patients with worsening blood gases and/or persistent tachypnea (respiratory rate > 35 breaths/min) despite NIV received endotracheal intubation.
Postoperative pulmonary complications were defined as sub-lobar or lobar atelectasis, detected by the chest X-rays and scored using the radiological atelectasis score equal or greater than two [23]; nosocomial pneumonia (new-onset or progressive pulmonary infiltrates with at least two of the following: purulent respiratory secretions, temperature > 38 °C or < 36 °C, and white blood cell count > 12,000/mm3 or < 4000/mm3) [24]. Non-pulmonary complications included new-onset cardiac arrhythmias, cardiac ischemia, hemodynamic instability requiring fluid or vasoactive resuscitation, hyperlactatemia, and metabolic acidosis.
The post hoc analyses on PaCO2 and hypercapnia development were conducted under the light of the most recent evidence suggesting a relevant effect of HFNC on CO2 washout in the upper airways [10,11,12, 25,26,27]. Results on these endpoints should be considered merely exploratory in nature.
Statistical analysis
Data on the rate of patients experiencing postoperative hypoxemia (defined as a PaO2/FiO2 ratio < 300 mmHg within 96 h after surgery) during Venturi mask oxygen therapy after lung resection were lacking at the time of study design, but it was known that 50% of them show a PaO2/FiO2 < 320 mmHg 24 h after surgery while on low-flow oxygen [28]. Using a conservative approach, we hypothesized a 45%-incidence of postoperative hypoxemia in the Venturi mask group, and we estimated that 45 patients per group were needed to detect a 60% relative reduction in the rate of the primary endpoint in the intervention group (estimated absolute risk in the intervention group, 18%), with a type I error set at 5% and statistical power of 80%. Given an attrition rate lower than 5%, mostly due to protocol violations and crossover between treatments, we planned to enroll 94 patients.
The analysis was conducted on a “modified intention-to-treat” population that included all patients who underwent the allocated treatment for at least 6 h.
Distribution normality was assessed with the Kolmogorov-Smirnov test. Continuous variables with normal distribution are reported as means (± standard deviation), whilst those with non-normal distributions were expressed as medians (interquartile ranges).
Analysis on the primary efficacy criterion and for other categorical outcomes was performed with the χ2 test, or Fisher’s exact test, as appropriate: Cochran–Mantel–Haenszel statistics are reported for all these results. For other relevant outcomes whose distribution was statistically different in the two groups at the univariate analysis, a logistic regression model was conducted: all variables with p ≤ 0.20 at the univariate analysis were included. Kaplan–Meier curves were plotted to assess the time from enrollment to the primary endpoint or relevant secondary outcomes by means of the log-rank test; for secondary outcomes, Cox regression analysis was also conducted to confirm the independent effect of the treatment on the time from enrollment to occurrence of the endpoint: all variables with a log-rank p ≤ 0.20 were included in the model. Two-way analysis of variance (ANOVA) for repeated measures with Bonferroni correction was used to determine the differences in PaO2/FiO2 ratio, PaCO2, and dyspnea in the two groups. Comparisons between groups regarding these variables at each study timepoint were performed with the Student’s t test or Mann-Whitney test, as appropriate. Mean difference and 95% confidence interval (95% CI) are reported for most significant results.
Two-tail p values ≤ 0.05 were considered significant. Statistical analysis was performed with SPSS software package (SPSS Inc. Released 2009. PASW Statistics for Windows, Version 18.0. Chicago: SPSS Inc.).