This prospective, randomised controlled trial (RCT) was conducted in the 12-bed respiratory intensive care unit (ICU) of Beijing Chao-Yang Hospital Western Branch in China. The protocol was approved by the ethics committee at Beijing Chao-Yang Hospital (reference no. 2019-KE-263), and written informed consent was obtained from all patients, their next of kin, or other surrogate decision-makers as appropriate. The trial was registered with clinicaltrials.gov (identifier: NCT04044625).
We screened all COPD patients admitted to the respiratory ICU. Patients were considered eligible for the trial if they had been diagnosed with AECOPD as defined by the 2019 criteria of the Global Initiative for Chronic Obstructive Lung Disease , had arterial pH < 7.35 and PaCO2 > 45 mmHg at ICU admission, and still had PaCO2 > 45 mmHg after a 6 h screening period while receiving low-intensity NPPV. Exclusion criteria are provided in Additional file 1: Supplementary methods.
Randomisation and blinding
Randomisation was accomplished by a computer-generated random number sequence. Each allocation sequence was concealed through the use of numbered, opaque, sealed envelopes until the intervention assignment was finished and was managed by an independent employee who was not involved in the trial. Eligible patients were assigned at a 1:1 ratio to undergo either high-intensity NPPV or low-intensity NPPV. At least two investigators per patient conducted the study: One performed the intervention defined in the protocol, and the other performed the outcome measurements. All data analyses were performed by the trial statistician, who was not involved in the trial.
In the high-intensity NPPV group, IPAP was initially adjusted in increments/decrements of 1–2 cmH2O, typically ranging from 20 to 30 cmH2O (or a tolerated maximum), to obtain a tidal volume (VT) 10–15 mL/kg of predicted body weight (PBW) and a respiratory rate (RR) < 25 breaths/min. Subsequent adjustments to IPAP were based on the results of arterial blood gases (ABGs; up to 30 cmH2O) to achieve either normocapnia (if possible) or a maximal reduction in PaCO2. If PaCO2 decreased to less than 35 mmHg, IPAP was decreased to achieve normocapnia. Patients were encouraged to use NPPV as continuously as possible, but brief disconnection from the ventilator was allowed to clear secretions, drink water, or eat.
In the low-intensity NPPV group, as well as during the 6-h screening period, IPAP was initially adjusted in increments/decrements of 1–2 cmH2O (up to 20 cmH2O), according to patients’ tolerance, to obtain a VT 6–10 mL/kg of PBW and an RR < 25 breaths/min. Subsequent adjustments to IPAP were based on the results of ABGs (up to 20 cmH2O) to achieve a pH of ≥ 7.35 and to reduce PaCO2 to an extent accepted by the attending physician. Patients were encouraged to use NPPV as much as possible during the first 6 h after randomisation and at least 10 h per day. Brief disconnection from the ventilator was allowed to clear secretions, drink water, or eat but was not scheduled.
In both groups, expiratory positive airway pressure (EPAP) was set at 5–8 cmH2O, the backup RR was set at 12 breaths/min, the inspiratory time was set at 0.8–1.2 s, the rise slope was set at level 1 or 2, and fraction of inspired oxygen (FiO2) was adjusted to obtain an oxygen saturation measured by pulse oximetry of 90–95%. All patients used the same noninvasive ventilator (Respironics V60, Philips Respironics, Carlsbad, CA, USA) in the bilevel positive airway pressure (spontaneous/timed) mode. An oronasal mask (Philips Respironics) was used as a first choice, but a nasal mask (Philips Respironics) was optional if patients did not tolerate the oronasal mask.
Medical treatments other than NPPV were based on the 2019 guidelines of the Global Initiative for Chronic Obstructive Lung Disease  and routine clinical practice in the respiratory ICU. If severe alkalosis occurred (pH > 7.55), arginine was provided.
The primary outcome was PaCO2 24 h after randomisation. Secondary outcomes included gas exchange other than PaCO2 24 h after randomisation, inspiratory effort, dyspnoea, consciousness, NPPV tolerance, patient–ventilator asynchrony, cardiac function, ventilator-induced lung injury (VILI), and NPPV-related adverse events.
For gas exchange, we recorded pH, arterial oxygen tension [PaO2], PaCO2, and bicarbonates at baseline and 2, 6, 24, 48, and 72 h after randomisation, and calculated the differences in PaCO2 between baseline and these other time points, respectively.
For inspiratory effort, we measured inspiratory oesophageal pressure swing (ΔPes), oesophageal pressure–time product (PTPes)/breath, PTPes/min, and PTPes/L over the last 3–5 min of oesophageal pressure (Pes) recording within 24 h after randomisation (see Additional file 1: Supplementary methods for more details).
For dyspnoea, consciousness, and NPPV tolerance, we recorded accessory muscle use, dyspnoea score, Glasgow Coma Scale (GCS) score, Kelly–Matthay score, and NPPV tolerance score at baseline and 2, 6, 24, 48, and 72 h after randomisation (see Additional file 1: Supplementary methods for more details).
All asynchrony events (including ineffective efforts, auto-triggering, double-triggering, premature cycling, and delayed cycling) were determined by visual inspection of the tracings of Pes, airway pressure, and flow over the last 10 min of these recordings within 24 h after randomisation, and an asynchrony index was computed (see Additional file 1: Supplementary methods and Fig. S1 for more details).
Regarding cardiac function, we recorded heart rate and blood pressure; measured serum N-terminal pro-B-type natriuretic peptide, troponin I, and creatine kinase isoenzyme; and performed bedside echocardiographic examination at baseline and 24, 28, and 72 h after randomisation.
To assess VILI, we measured plasma levels of VILI-related inflammatory mediators, including tumour necrosis factor (TNF)-α, interleukin (IL)-1β, IL-6, IL-8, IL-10, and macrophage inflammatory protein (MIP)-2 at baseline and 24, 28, and 72 h after randomisation (see Additional file 1: Supplementary methods for more details).
We expected that the mean (± standard deviation [SD]) PaCO2 24 h after randomisation would be 65 ± 15 mmHg in the low-intensity NPPV group, based on our clinical experience and previous studies [1, 6,7,8, 19]. Based on the assumption that mean PaCO2 24 h after randomisation would be 45 ± 15 mmHg in the high-intensity NPPV group, a sample of 12 patients in each group was required to detect an absolute between-groups difference of 20 mmHg in PaCO2 24 h after randomisation. We used a superiority test to compare the means of the two groups, with a superiority margin of 3 mmHg, 85% power, and a one-tailed alpha of 0.05.
Continuous variables are presented as means ± SD with normal distributions or as medians (25th–75th percentiles) with non-normal distributions unless otherwise specified. Qualitative or categorical variables are presented as absolute frequencies with percentages. The test of normality was performed with the Kolmogorov–Smirnov test, and the test of homogeneity of variances was performed with Levene’s test. Continuous variables were compared between the two groups with Student’s t test for normally distributed variables and the Mann–Whitney U test for non-normally distributed variables. Qualitative or categorical variables were compared with the Fisher’s exact test. All tests were two sided. Differences with p < 0.05 were considered statistically significant. Statistical analyses were performed with SPSS (version 25.0; IBM, Chicago, IL, USA).