Prolonged mechanical ventilation secondary to weaning failure demands a significant amount of intensive care unit (ICU) resources, thus increasing the economic burden of public healthcare costs. One of the proposed mechanisms accounting for weaning failure is the concept that excessive work of breathing for weak respiratory muscles during the liberation from mechanical ventilation compromises cerebral blood flow, thereby predisposing the brain to dysfunction. Restriction in brain perfusion could have an adverse impact on the function of the respiratory muscles by impairing the output of the respiratory centre thus promoting respiratory muscle fatigue, leading to weaning failure. Inspiratory muscle training (IMT) has been shown to improve the functional capacity of the inspiratory muscles in patients with respiratory muscle weakness whilst has been recently proposed as a possible additional component of weaning strategies. Therefore, this project aims to identify both a mechanism that might be linked to prolong ICU length of stay and that at the same time might be amenable to treatment.

1. Introduction.

Weaning covers the entire process of liberating the patient from mechanical ventilation (MV) support and the endotracheal tube or tracheostomy. For the majority of mechanically-ventilated patients spontaneous breathing (weaning) can be successfully accomplished quickly and easily, however, this is not the case for 15-30% of ventilated patients. The most common factors leading to weaning failure are i) respiratory muscle fatigue ii) hypercapnia iii) dyspnoea and iv) anxiety. Along these lines, a proposed a mechanism that may contribute to weaning failure namely the ''stealing effect'' theory. More specifically, the ''stealing effect'' theory hypothesizes that during the weaning process the energy demand of the respiratory muscles can increase to such an extent that respiratory muscles may deprive oxygen and blood of other tissues such as the brain. In fact, at least for the brain, the decrease in blood and oxygen delivery could theoretically predispose this organ to dysfunction. Restriction in oxygen delivery to the brain could have an adverse impact on the function of the respiratory muscles per se by impairing the output of the respiratory centre and of neuromuscular competence of the ventilatory pump, rendering the ventilatory pump incapable of inflating the lungs thus exaggerating respiratory muscle fatigue and leading to weaning failure. Inspiratory muscle training (ΙΜΤ) is a non-pharmacological and cost-effective treatment which has been shown to improve the functional capacity of the inspiratory muscles in patients with respiratory muscle weakness whilst has been recently proposed as a possible additional component of weaning strategies. Implementation of an IMT program may decrease the energy demands of inspiratory muscles by improving inspiratory muscles breathing efficiency on the one hand and by optimizing the respiratory muscle oxygen utilization on the other hand. Enhancement in muscle strength and contractile function, as well as oxidative capacity in response to IMT that have been documented in patients with chronic lung diseases in response to IMT, may increase the oxygen utilization by the inspiratory muscles thus decreasing the demand of blood flow and oxygen in favour of the cerebral cortex.

The aims of the project are:

In an observational study design:

1. To investigate whether the inspiratory muscles steal blood and oxygen from the cerebral cortex (''stealing effect'').

2. To investigate the effect of the ''stealing effect'' theory on weaning outcomes.

   In a randomized controlled trial design:

3. To evaluate the effects of IMT using a recently developed TFRL device on weaning outcomes and on the ''stealing effect'' in difficult to wean patients in the ICU compared to a sham low-intensity endurance training group.

2. Primary methodologies

Patients with difficult weaning will be included according to inclusion criteria. In addition, a group of "simple weaning" defined as patients who succeed in the first weaning trial will be used as a control group for addressing the 1st and 2nd objective of the project, in an observational study. Simple weaning patients, difficult to wean, and prolonged weaning patients will be classified according to criteria that have been proposed in the recent WIND study. Specifically, patients will be classified in group 1 (i.e., simple weaning patients-weaning successfully achieved at first weaning attempt resulting in successful extubation within 1 day. Patients will be included if classified in group 2 (i.e., difficult to wean - weaning successfully achieved after more than 1 day but in less than 1 week after the first separation attempt (successful separation or death) or group 3 (i.e., prolonged weaning - weaning still not terminated 7 days after the first separation attempt by success or death). After being informed about the study written informed consent will be obtained from all patients if awake and adequate or from a family member if unconscious. In order to address the 3rd objective of the project, patients with weaning difficulties will then be randomized in one of the two groups: IMT group or endurance training (SHAM) group. Block randomization will be performed stratifying on two factors: score on the APACHE II and the presence of COPD. Envelopes will be prepared and sealed (80 for the treatment group and 80 for endurance training group). Piles of envelopes will be separated in APACHE<18, APACHE>18, non-COPD or COPD and its combinations (20 for each condition) considering 10 IMT envelopes and 10 endurance training envelopes for each pile.

2.1. Regional blood flow, oxygen delivery and oxygen availability measurements by NIRS-ICG derived BFI.

Inspiratory and cerebral cortex blood flow index (BFI) will simultaneously be measured by near-infrared spectroscopy (NIRS) in combination with injections of the tracer indocyanine green dye (ICG). ICG that is widely used as hemodynamic tracer elicits an absorption peak at 805 nm and, following intravenous infusion, is restricted to the intravascular compartment by >95% binding to plasma proteins. NIRS-ICG derived BFI as relative measurement of local perfusion in respiratory and peripheral muscles and cerebral cortex will be calculated by dividing the muscle ICG peak concentration (assessed by NIRS-ICG curve) by the rise time from 10 to 90% of peak. To measure inspiratory muscles and cerebral cortex ICG concentrations, four sets of NIRS optodes will be transcutaneously positioned as follows: for the brain over the prefrontal cortex area (at an adequate distance to avoid interference with the midline sinus), on the scalene muscles and on the upper rectus abdominis. An additional probe (4th) will be placed on a non-working muscle group (i.e., thenar eminence) which will be used as a control measurement spot. Skeletal muscle and cerebral tissue oxygen delivery will be calculated by multiplying BFI to arterial oxygen content; the latter will be calculated by arterial and venous blood gas samples. Inspiratory and thebar muscle and cerebral cortex oxygen saturation (Stio2,%) -a noninvasive index of local tissue oxygen availability which reflects the balance between oxygen supply and utilization- will be recorded continuously by NIRS.

2.2 Hemodynamic status

Cardiac output will be assessed continuously by pulse contour analyses using a sensor (Pulsioflex Monitor, Pulsion Medical Systems SE) connected to an existing arterial catheter32 (see online supplement). The calculation of cardiac output is performed beat-by-beat by simply multiplying the stroke volume that is calculated by arterial pressure waveform analysis with the recorded heart rate. Pulse contour analyses method has been validated against cardiac output calculations using gold-standard methods. The results show that this method can provide a clinically acceptable cardiac output trend assessment in hemodynamically stable ICU patients. Cardiac output and heart rate data will be averaged over 60 seconds during each BFI determination.

2.3. Inspiratory muscle training program

Both groups (i.e., IMT group or endurance training (SHAM) group) will participate in usual care training interventions aimed at improving respiratory muscle endurance such as spontaneous breathing trials (SBT) performed with pressure support ventilation (PSV), continuous positive airway pressure (CPAP), T-tube as well as early mobilization. After inclusion and randomization patients in both groups will start the training with the TFRL device (POWERBREATHE KH2, HaB International Ltd, UK). Before and after every session, respiratory rate, heart rate, oxygen hemoglobin saturation and blood pressure will be recorded. Sessions will consist of 4 sets of minimum 6 and maximum 10 breathes per set with resting periods of at least 2 minutes where the patient will be placed in a sitting position (45 degrees), aspirated if necessary, and the cuff will be inflated to prevent air leakage. Patients will be instructed to achieve full inspiration and expiration at every breath and to perform a fast and forceful inspiration. Instructions and encouragements during the sessions are standardized. A feedback screen on the computer will be available for patients on the IMT group to show patients' performance during the training (BreatheLink Software, HaB International Ltd, UK). The training session will be interrupted if patient reports symptoms of dyspnea, anxiety or coughing, discomfort or when the transcutaneous oxygen desaturation falls below 85%. The training load in the strength training group will be adjusted to maintain the highest tolerable intensity throughout the course of the study until weaning of MV. Patients in the endurance training group will perform the same regimen of training without adjustments of the training resistance (<10%MIP). After the sessions Borg-scale scores on perceived breathing effort and dyspnea will be recorded. Experienced physiotherapists will perform the sessions in all groups.Training will last for 28 days or until the patient is successfully weaned from MV. Patients still in MV after 28 days of IMT will be considered failure to wean. MIP and VC will be measured weekly to follow lung function progression and adequate resistance of training.

3. Statistical Analysis plan

3.1. Statistical analysis plan for the 1st and 2nd objective of the study

For the sample size calculation of this study, we used the change in cerebral cortex BFI from MV to the SBT between SBT success and SBT failure patients. An expected effect size [Cohens d] of 0.467 was calculated from the mean difference of cerebral cortex BFI (i.e., 6.70 nM/s) and the corresponding pooled standard deviation (i.e., 14.0 nM/s), from a previous study that investigated interhemispheric differences in cerebral cortex BFI in critically ill patients. Accordingly, using this effect size, the critical sample size is calculated to be 20 SBT failure patients on the basis of using an ANOVA as the statistical analysis method. Anticipating that approximately only 1 out of 5 patients (20% rate)1 is expected to fail the SBT, an estimated number of (i.e., 20*5=100) patients in total is considered to be included in order to identify the 20 weaning failure patients. Continuous variables will be presented as mean values with standard deviations if normally distributed or as a median with an interquartile range if not. Categorical values will be presented as numbers and proportion. For comparisons between SBT success or failure group, continuous variables will be compared using Student t-test or Mann Whitney U test based on the distribution of the variables. Categorical values will be compared using Chi-square or Fisher's exact test, as appropriate. Two-way analysis of variance (ANOVA) will be applied to examine the interaction amongst respiratory, hemodynamic, blood gases and peripheral circulation and oxygenation responses and different time points (i.e., T1-T4, see Table 1) between SBT success and SBT failure group. One-way ANOVA with repeated measures will be used for the comparison of the different time measurements (i.e., T1-T4, see Table 1) for each group. Independent association between changes in cerebral cortex BFI from MV to different time measurements during SBT and SBT outcomes (failure, success) will be explored by logistic regression analysis. Further exploration of independent associations between SBT outcomes (failure, success) and all respiratory, hemodynamic, blood gases and peripheral circulation and oxygenation variables will be also explored by logistic regression analysis. Finally, multiple logistic regression analysis including all significant independent predictors (after checking them for collinearity) will be performed to identify determinants of SBT outcomes. Data will be analyzed using the SPSS Software (Chicago, IL, USA). Statistical significance will be defined as p<0.05. Additionally, analogous analyses will be performed for weaning outcomes (failure, success).

3.2. Statistical analysis plan for the 3rd objective of the study

The sample size calculation for the effectiveness of IMT (IMT group versus endurance training group) on weaning outcome is based on a previous literature review that evaluated the effects of IMT in difficult to wean patients. According to the proportions of weaning success observed in this meta-analysis, 78% in the IMT group and 50% in the control group, and assuming [a]- and [β]-risks of 5% and 20%, a sample of 45 patients should be included in the IMT group and 45 in the endurance training group. A true intention-to-treat analysis compared weaning success encompassing all randomized patients irrespectively of whether they performed IMT or not. Modified intention-to-treat analyses were performed for secondary outcomes that were not lost to follow-up for the primary outcome. This entailed excluding patients randomized but not receiving any intervention session (Hi-IMT or Lo-IMT), while all other patients, even those only performing one IMT session, were included. Data distribution was tested with Shapiro-Wilk tests. Between-group differences were assessed with unpaired t-tests. Multiple analysis of variance (MANOVA) was conducted to analyse the effects of group allocation (Hi-IMT and Lo-IMT) and time (pre- and post-intervention) on breathing pattern and respiratory parameters. Post-hoc within-group comparisons were conducted using Holm-Šídák tests for within-group multiple comparisons. Multiple analysis of variance (MANOVA) was opted instead of analysis of covariance (ANCOVA),which was specified in the protocol ( Hoffman M, Van Hollebeke M, Clerckx B, Muller J, Louvaris Z, Gosselink R, et al. Can inspiratory muscle training improve weaning outcomes in difficult to wean patients? A protocol for a randomised controlled trial (IMweanT study). BMJ Open 2018; 8: e021091), as the focus of the analysis was on the difference between pre- and post-intervention values. Statistical significance will be set at p<0.05

Inclusion Criteria:

• Difficult and prolonged weaning patients
• Simple weaning patients
• Adequate oxygenation
• Febrile temperature < 38ºC
• Hemodynamic stability
• Stable blood pressure
• No or minimal pressors
• No myocardial ischemia
• Adequate hemoglobin and mentation
• Resolution of disease acute phase
• Able to follow simple verbal commands related to IMT
• Mechanically ventilated via a tracheostomy or endotracheal tube

Exclusion Criteria:

• Pre-existing neuromuscular disease
• Agitation
• Hemodynamically instable (arrhythmia, decompensated heart failure, coronary insufficiency)
• Hemoptysis
• Diaphoresis
• Spinal cord injury above T8
• Use of any type of home MV support prior to hospitalization
• Skeletal pathology that impairs chest wall movements
• Poor general prognosis or fatal outcome