Study Design
In accordance with our pre-published protocol(12), we conducted this investigator-initiated single-centre randomized controlled trial using concealed allocation, assessor-blinding and intention-to-treat analysis to compare inspiratory muscle training with usual care in ICU patients who were ventilator-dependent for at least 7 days. The study was approved by the Australian Capital Territory Health Human Research Ethics Committee (ETH.10.10.370) and the University of Queensland Medical Research Ethics Committee (2010001498). The published protocol(12) (trial registration ACTRN12610001089022) complied with the CONSORT guidelines for clinical trials(13).
Patients were eligible for inclusion if they had been invasively mechanically ventilated (via endotracheal tube or tracheostomy) for at least 7 days, were aged ≥ 16 years, and were sufficiently alert to provide informed consent and participate actively in training (Riker Sedation-Agitation Scale(14) score of 4). Exclusion criteria included refusal to participate, pregnancy, significant pain or distress affecting breathing, medical instability or anticipated death within weeks. All participants provided written consent to participate in the study.
The study was conducted in a 31-bed Australian mixed medical/surgical/trauma ICU where minimal sedation and early rehabilitation(15) are well-established. The medical officers making ventilator liberation decisions were blinded to group allocation. Training was conducted by physiotherapists in line with our previously-published protocol, which is safe and feasible in ICU patients(6). Due to the nature of the supervised training, therapists could not be blinded to group allocation.
Intervention
Using a computer-generated random number sequence (with concealed allocation), participants were randomized to usual care (control group) or inspiratory muscle training in addition to usual care (IMT group). Usual care included secretion clearance techniques (e.g. percussion, hyperinflation and suction) but did not include inspiratory resisted breathing of any kind.
The IMT device used was the Threshold inspiratory muscle trainer (Threshold IMT device HS730, Respironics NJ, USA). This spring-loaded one-way valve provides titratable inspiratory resistance in a range of 9–41 cmH2O and can readily be connected to an endotracheal tube or tracheostomy (Fig. 1).
For training, a high-intensity low-repetition method was used as previously described(1, 5, 16). Intensity was prescribed at a minimum of 50% of maximal inspiratory pressure (MIP) at the highest tolerable intensity where the participant could just complete the 6th breath in a set of 6 breaths. One treatment session consisted of 5 sets of 6 breaths, where resistance was increased between sets as appropriate. Participants were returned to the ventilator between sets, where they typically required only a few minutes’ rest.
Training commenced following randomisation and continued once daily (weekdays only) until 1 week following successful liberation from mechanical ventilation (defined as 24 hours without positive pressure). We did not use a sham device for comparison due to the risk of a sham device providing a training stimulus in participants with very low inspiratory muscle strength(17).
Measures
Primary outcomes were measured by specifically-trained research nurses blinded to group allocation. Initial measurements were conducted following enrolment and prior to randomisation; interim measurements were obtained following successful liberation from the ventilator (24 hours spontaneously breathing without positive pressure); and final measurements were recorded 1 week following liberation. Inspiratory muscle strength (MIP) was measured from residual volume using a portable MicroRPM Respiratory Pressure meter (CareFusion, San Diego, USA) in accordance with the protocol described by the American Thoracic Society and European Respiratory Society(18). This device has excellent reliability (ICC 0.83–0.90)(19).
Following successful ventilator liberation, inspiratory muscle fatigue was measured using the fatigue-resistance index (FRI) previously described in ICU survivors(20). This technique, based on the Maximum Incremental Threshold Loading test(21), requires participants to breathe against 30% resistance for 2 minutes, and MIP measures before and after the loading test are compared. FRI was also measured 1 week following successful ventilator liberation.
Participants’ quality of life was measured on enrolment and completion (1 week following ventilator liberation) by research nurses blinded to group allocation. Quality of Life was measured using both the SF-36v2 tool (acute 1 week time frame) (under license QualityMetric USA) and the EQ-5D-3L tool (under license EuroQol International). The SF-36 is reliable, responsive, and has both construct and criterion validity in intensive care patients(22). The EQ-5D-3L tool has been used extensively in ICU patient follow-up(23) and gives a more general measure of health-related quality of life than the SF-36.
Dyspnea was measured using a Modified Borg Dyspnea scale, where dyspnea is a patient-reported categorical score out of 10. This scale has acceptable reliability and validity in patients undergoing mechanical ventilation(24). Dyspnea was recorded both at rest (sitting comfortably in the chair or bed) and during exercise (the peak exercise activity experienced in the previous 24 hours) by research nurses blinded to group allocation, at both enrolment and study completion.
Physical function was assessed using the Acute Care Index of Function (ACIF)(25). This tool captures mental status, bed mobility, transfers and mobility, and has excellent inter-rater reliability in ICU patients (ICC = 0.94)(26). On enrolment, ACIF scores were completed by ICU physiotherapists prior to randomisation (thus blinded to group allocation), however follow-up ACIF scores were recorded by the ward physiotherapist who was not blinded.
Other outcomes extracted from the hospital databases included the number of training sessions (intended and completed), any requirement for reintubation, duration of mechanical ventilation, duration of pressure support ventilation, ICU length of stay, post-ICU hospital length of stay and in-hospital mortality.
Data analysis
The sample size was calculated a priori for the primary outcome measures (MIP). To detect a 10% change in MIP with a power of 0.80, 70 participants were required (inflating group size by 15% due to anticipated mortality of 12.8%(12)). In the absence of an established minimal clinically important difference in MIP in ICU patients, the 10% level was selected to facilitate comparison with previous studies of ICU survivors(20, 27). Raw MIP scores were normalised(28) to account for variations of MIP with age and gender.
Paired t-tests were used to compare within group differences. Mixed linear models were used to assess the between-group difference of the changes between enrolment and follow-up measures, including age, gender, APACHE II scores and ‘ventilation time prior to randomisation’ as covariates. Diagnostic plots (predicted means versus Pearson's residuals) were generated to assess model assumptions. Mortality and reintubation data were analysed using Fisher’s exact test. Post-ICU length of stay was analysed using a Wilcoxon rank-sum test, with exclusion of patients who died in-hospital.
Statistical significance was set as p < 0.05. All analyses were done using R 3.6.1.