Prospective Randomized Crossover Study of a New Closed-loop Control System versus Pressure Support during Weaning from Mechanical Ventilation

PRESSURE support ventilation (PSV) is the most widely used assisted mode of ventilation during the weaning process, in both medical and surgical critically ill patients.¹  However, PSV provides a fixed inspiratory pressure (P_INSP), regardless of the patient’s ventilatory demand or gas exchange, which limits breathing pattern variability.^2–5  Given the high variability in disease processes and states, the application of predefined, uniform values for ventilator parameters, such as a fixed P_INSP or tidal volume (VT), is unlikely to provide optimal assist at all times.⁶  In contrast, variability in the breathing pattern may be useful in improving gas exchange as suggested by previous reports in animals^7,8  or recently in humans.² 

New ventilatory modes can offer ventilation automatically adjusted to the patient’s ventilatory demand.^2,5,9,10  Studies that evaluated these modes have shown benefits on the optimization of ventilation,^2,11  burden of care,^12,13  and duration of weaning.^12,14  No automatic management of oxygenation (i.e., both fraction of inspired oxygen [Fio₂] and of the positive end-expiratory pressure [PEEP]) was available to date.^15,16 

Intellivent is a new full closed-loop solution for passive and active breathing patients receiving invasive mechanical ventilation and includes automatic adjustment of minute ventilation (MV), Fio₂, and PEEP. Intellivent has been studied in sedated, passively ventilated critically ill adult patients with acute respiratory failure, but only for short duration (2–4 h).¹⁷  To our knowledge, no physiological study has been performed to evaluate Intellivent for a longer ventilation period in nonsedated actively breathing critically ill patients and during the weaning period.

The aim of this prospective, randomized, crossover study was to compare ventilatory parameters and gas exchange between Intellivent and PSV given more than 24 h, in critically ill patients, during the weaning phase.

Compared with PSV, which provides a fixed level of assistance (P_INSP) regardless of the patient’s ventilatory demand, we hypothesized that Intellivent would improve oxygenation by offering more variable ventilation.^18,19 

This single-site study was carried out in the 16-bed medical–surgical intensive care unit of the St Eloi Hospital, a 660-bed teaching and referral facility of the University of Montpellier in France. The experimental protocol was approved by the Ethics Committee of the Saint-Eloi Teaching Hospital (Comité de Protection des Personnes Sud Méditerranée IV, Montpellier, France), and written informed consent was provided by patient or next of kin. This study followed the CONSORT recommendations concerning randomized trial reporting.²⁰ 

From March 2011 to May 2011 (2.5 months), 50 consecutive patients were screened and 16 enrolled. The patients were included if they were in spontaneous mode (active patient, i.e., able to trigger a breath) with an expected duration of invasive mechanical ventilation longer than 48 h. Patients were not included in case of clinical instability, whatever the reason, and when a decision to withhold life-sustaining treatment was made. Pregnant women and children younger than 18 yr were also not included.

The two ventilation modes (PSV and Intellivent) were given by the same ventilator (Hamilton S1; Hamilton Medical, Rhäzuns, Switzerland). The rise time (50 ms), inspiratory flow trigger (2 l/min), and expiratory trigger sensitivity (30% of the peak flow) were identical in the two modes.

In PSV, the Fio₂ was set to achieve a pulse oxygen saturation (Spo₂) greater than 92% and the PEEP level was set between 5 and 10 cm H₂O. The level of P_INSP was adjusted to obtain a VT between 6 and 8 ml/kg of predicted body weight (PBW)²¹  (as calculated with the following formula for men: PBW (kg) = 50 + 2.3 ([height (cm)/2.54] − 60) and for women: PBW (kg) = 45.5 + 2.3 ([height (cm)/2.54] − 60) and a respiratory rate (RR) between 20 and 30 breaths/min.

In Intellivent, initial MV²²  is automatically determined by the ventilator according to the PBW set by the clinician. The MV is automatically adjusted to maintain end-tidal partial pressure of carbon dioxide (Petco₂) within expert-based acceptable ranges²³  when the patient is not triggering the breath or to maintain the patient’s RR within acceptable ranges, as defined by the Otis least work of breathing concept,²⁴  when the patient is triggering the breath. To adjust MV in order to maintain an acceptable range of Petco₂ or RR, the ventilator adjusts both the VT and the RR as it is in adaptive support ventilation.^4,25  In brief, based on the breath-by-breath expiratory time constant (RC_EXP) estimation, optimal VT and RR are derived. When MV needs to be adjusted to keep the patient’s Petco₂ or RR within the defined range, the ventilator adjusts P_INSP and the mandatory breath to target optimal VT and RR to the patient (appendix 1).

To avoid extreme and potentially dangerous values of VT and RR, Intellivent uses, on a breath-by-breath basis, a safety window for the given VT and RR values. The minimal target VT is defined as twice the anatomical dead space estimated from the PBW. The maximal target VT is defined as the maximal pressure (set by the clinician) times the dynamic compliance of the total respiratory system. The minimal value for the target RR is 5 breaths/min. The maximal value for the target RR is defined as the ratio 20/RC_EXP.²⁶ 

PEEP and Fio₂ are automatically adjusted based on the ARDSnetwork PEEP–Fio₂ tables²⁷  to maintain an Spo₂ within expert-based acceptable ranges (appendices 1 and 2). The tables are user adjustable by selecting the maximal PEEP delivered. The PEEP/Fio₂ tables from the ARDSnetwork are used only as starting values in the algorithm; but adjustable by the user according to the local policies by setting the maximal PEEP value (appendix 2). Setting low maximal PEEP makes the algorithm adjusting the Fio₂ more than the PEEP. In case of moderate decrease in Spo₂, Fio₂ increases by 10% of actual value every 30 s and PEEP increases by 1 cm H₂O every 6 min. If Spo₂ is above the target range, Fio₂ decreases by 5% of the actual value every minute and PEEP decreases by 1 cm H₂O every 6 min. If Petco₂ and Spo₂ informations are of poor quality or lost, the controllers automatically pause and an alarm is generated. Automatic control is resolved when signal of good quality is measured again. In addition, Fio₂ is automatically increased to 100% if Spo₂ is below 85%, and 100% Fio₂ manual bypass is still available. In addition to the adaptive support ventilation safety windows, minimal and maximal MV, Fio₂, and PEEP settings are set by the users before starting Intellivent. By adjusting, on a breath-by-breath basis, the level of P_INSP, RR, PEEP, and Fio₂, Intellivent may generate more variability than conventional ventilation such as PSV.

We applied a prospective, randomized, single-blind crossover study design very similar to that previously reported.^2,4,28  Determination of the first used ventilatory mode (PSV or Intellivent) was randomized. Randomization was carried out using a random-number table. Each patient was consecutively ventilated for 24 h with the PSV mode and with the Intellivent mode in a random order. Throughout the protocol, suctioning via the endotracheal tube was performed on a per need basis and routine care, such as physiotherapy and nursing was performed as usual in the unit.

Standard three-lead monitoring electrodes continuously monitored heart rate and rhythm. Spo₂ was continuously monitored using pulse oximetry. Systolic and diastolic arterial blood pressures were continuously monitored through a 20-gauge catheter inserted in a radial or femoral artery. Blood samples were obtained at baseline (in the first hour after mechanical ventilation for each mode), after 8, 16, and 24 h of mechanical ventilation for arterial blood gas analysis (GEM Premier 3000 analyzer; Instrumentation Laboratory, Lexington, MA) through the arterial catheter.

The following variables such as airway pressures (P_INSP as inspiratory pressure level above PEEP delivered by the ventilator, P_MEAN as mean airway pressure, and P_MAX as maximal airway pressure), VT, RR, MV, RC_EXP, PEEP, Fio₂, and Petco₂ were collected continuously breath-by-breath, by a dedicated software (Study recorder software; Hamilton Medical) via a RS32 cable, exported through a Universal Serial Bus support and analyzed using a customized software based on Microsoft Excel®(Redmond, WA).

Every 8 h, according to the unit protocol, the nurse in charge of the patient evaluated the pain and comfort using the Behavioral Pain Scale and the sedation and agitation level using the Richmond Agitation Sedation Scale.^29,30  Setting changes made by the attending physician were also recorded.

The primary endpoint was oxygenation, estimated by the calculation of the difference between the Pao₂/Fio₂ ratio obtained after 24 h of ventilation and the Pao₂/Fio₂ ratio obtained at baseline in each mode. To calculate the number of patients needed, we used data previously published^2,31  showing in postoperative patients a mean Pao₂/Fio₂ ratio of 202 ± 48 mmHg in PSV. Assuming an α risk of 0.05 and a β risk of 0.20, we calculated that at least 14 patients would be required to identify, after 24 h of mechanical ventilation, a difference of 20% between the variation of the Pao₂/Fio₂ ratio obtained in Intellivent in comparison to PSV. Therefore, we decided to include 16 patients. The secondary endpoints were the variability in the ventilation variables and the time spent with acceptable ventilation. Ventilation variables were collected continuously breath-by-breath during 24 h (see above). The variability in the ventilation parameters was evaluated by the coefficients of variation for P_INSP, RR, VT, MV, PEEP, Fio₂, RC_EXP, and Petco₂ calculated as the ratio of the SD to the mean multiplied by 100 as previously described.^2,32  Acceptable ventilation was defined with very permissive ranges, and calculated as the number of breath with RR between 12 and 35 breaths/min, a VT between 5 and 12 ml/kg of PBW, and Petco₂ less than 55 mmHg, over the total number of breath collected.^2,27,28  Values are expressed as median [interquartile range, IQR] or mean ± SD according to the type of variable distribution, from data collected breath-by-breath for 24 h. Normality of the distribution was assessed with Kolmogorov–Smirnov test. Comparisons were performed using Wilcoxon and Mann–Whitney tests according, and two-tailed P values less than 0.05 were considered significant. Statistical analysis was performed using SAS/STAT software version 8.1 (SAS Institute, Cary, NC) by an independent statistician.

Among the 16 enrolled patients, 2 did not complete the study because of early extubation, and 14 patients were finally analyzed (fig. 1). There were no safety issues requiring premature interruption of Intellivent for the studied patients. Diagnosis at the time of admission in intensive care unit and clinical characteristics of the patients are shown in table 1. Ventilation settings and main monitored parameters obtained during the first hour after inclusion were shown in table 2.

Arterial blood gases were shown in the table 3. The Pao₂/Fio₂ ratio improved significantly from 245 ± 75 at baseline to 294 ± 123 mmHg (P = 0.035) after 24 h of Intellivent, whereas no significant change was observed with PSV (fig. 2). The Pao₂/Fio₂ ratio variation after 24 h of mechanical ventilation was significantly higher in Intellivent than in PSV mode (+18 ± 32% vs. −3 ± 20%; P = 0.026).

The 24-h average values of the ventilation parameters are reported in table 4. VT was 7.6 ml/kg [IQR, 6.6–9.0] during PSV period compared with 8.4 ml/kg [IQR, 7.9–8.6] PBW during Intellivent period (P = 0.04). The RR was 22 breaths/min [IQR, 19–27] during PSV period compared with 19 breaths/min [IQR, 15–22] during Intellivent period (P = 0.01). Typical tracings obtained during 24 h of both PSV and Intellivent are shown in figure 3. There is obviously more variability in P_INSP, PEEP, and Fio₂ with Intellivent as compared with PSV. The coefficient of variation of P_INSP and PEEP was significantly higher with Intellivent as compared with PSV (table 4).

Times spent in different ranges of VT, RR, and Petco₂ in both modalities are shown in figure 4. The number of changes in P_INSP, PEEP, and Fio₂ was significantly higher with Intellivent compared with PSV (table 5). No significant differences were observed between the two modes for the Behavioral Pain Scale and Richmond Agitation Sedation Scale scores over the study period.

The current study demonstrates that over a 24-h study period, ventilation with Intellivent was associated with a higher Pao₂/Fio₂ ratio and a greater variability in P_INSP and PEEP compared with PSV. In addition, the study suggests that the use of Intellivent for 24 h of mechanical ventilation is feasible and safe for critically ill patients during the weaning period.

The current study is the first to report long-term (i.e., 24-h period) safe use of the Intellivent mode and to associate its use with improvement in oxygenation. The improvement in oxygenation observed with Intellivent is probably not related to a single mechanism, but we could speculate that it is because of more complex association of different features of Intellivent. Some features related to Intellivent can be proposed such as a slight, but significantly, higher airway pressures (P_INSP and P_MAX, not P_MEAN) leading a higher VT (table 4), an increase in variability of inspiratory pressures and PEEP, which may assimilated to more physiological sigh, and be considered as repeated alveolar auto-recruitment. Although, median VT was significantly higher in Intellivent than in PSV (8.4 [IQR, 7.9–8.6] vs. 7.6 [IQR, 6.6–9.0] ml/kg PBW; P = 0.04; table 4), VT remained lower than 10 ml/kg PBW more than 90% of the time spent in each modes with no significant difference between both Intellivent and PSV (fig. 4). Indeed, risk factors to develop acute lung injury or acute respiratory distress syndrom in passive ventilated patients are VT above 10 ml/kg PBW and an end-inspiratory pressure above 30 cm H₂O. In the current study, no patient was passively ventilated, and the association of a VT above 10 ml/kg PBW and an end-inspiratory pressure above 30 cm H₂O occurred together exceptionally either in Intellivent or PSV. Low VT with subsequently high transpulmonary pressure may basically be more dangerous than higher VT with lower transpulmonary pressure. In combination with more variability in airway pressures (P_INSP, P_MAX, and P_MEAN) and in PEEP with Intellivent, it suggests causality between Intellivent-induced variability and improvement in oxygenation. Several publications have already reported better oxygenation when pressure or volume applied to the respiratory system is variable, whatever the mode of ventilation being used.^2,7,8,33  The current study can only speculate on the mechanism responsible for such improvement. The Jensen inequality, i.e., the local nonlinear pressure–volume relationship, has been suggested as a mechanistic explanation,^34,35  heterogeneity in local time constant,^17,36  improved surfactant production,³⁷  and improved ventilation–perfusion matching³⁸  may also play a role in better oxygenation with more variable ventilation.

More ventilator adjustments were made during Intellivent ventilation compared with PSV. Although expected with closed-loop systems, which are by design able to continuously adjust the ventilatory parameters according to the changes in patient’s condition, it is worthwhile discussing the possible impact of adjusting the ventilator more often. First, as already discussed earlier, more adjustment makes ventilation more variable, and this may have clinical impacts (such as oxygenation in the current study). Second, it may help to maintain the patient within predefined acceptable ventilation ranges. In the current study, time spent by the patient with acceptable ventilation during Intellivent was not very different compared with PSV (fig. 4). However, during Intellivent, 11 of 14 patients (76%) spent more than 90% of ventilation time in acceptable ventilation (as defined in the methodology section) as compared with only 5 patients with PSV (P = 0.04). In other studies,^39,40  Intellivent was able to keep the patient more often with acceptable ventilation. In these studies,^39,40  most of the patients were not triggering the breath and were therefore fully controlled by the ventilator. In the current study, all patients were able to trigger the breath and to control their breathing pattern at least partially. The definition of “acceptable ventilation”, which was relatively permissive in the current study (because of the selected population), is indeed also important in analyzing the results and can be extensively discussed. Finally, although not analyzed in the current study, the number of ventilator adjustments with Intellivent may reflect how often the ventilator should be adjusted, whereas the number of changes in PSV may reflect basically how often the ventilator can be adjusted manually considering human resources and knowledge available at the bedside day and night. The current study is definitely not able to draw any conclusions on possible clinical impact of continuous and more frequent adjustments of the ventilator but give enough confidence to design large randomized controlled trials to address such a question. In addition, the current study is the first to report the use of Intellivent for more than couple of hours, in adult patients with variable conditions and during the weaning phase.

All patients were able to complete the study, and no patients were removed from Intellivent for major safety issues, suggesting that 24-h use of Intellivent may be safe in critically ill patients during the weaning period.

This study has some limitations. The study was not designed to evaluate the safeness and effectiveness of Intellivent as a routine mode of ventilation in patients in intensive care unit and therefore is underpowered for that. Interestingly, in three patients, Fio₂ had to be adjusted manually because of a discrepancy between Spo₂ and Sao₂ obtained from arterial blood sampling. However, because of the population selection (few hypoxemic patients), the Fio₂ and PEEP algorithms and the robustness of Spo₂ information (filtering, artifact, and motion rejections, and so on) for running the loops were not really challenged in the current study. Moreover, although we evaluated the agitation and sedation–analgesia levels every 4 h (using Richmond Agitation Sedation Scale and Behavioral Pain Scale scores), there was no specific auto-evaluation of the ventilatory comfort.

The current prospective study is the first to report 24-h use of Intellivent in spontaneously breathing patients during the weaning process. As compared with PSV, the Pao₂/Fio₂ ratio at 24 h was improved in Intellivent with more variability in airway pressures, PEEP, and Fio₂. Adjustment of the ventilator was much more frequent with Intellivent as compared with PSV, which may explain the variability and ultimately the better oxygenation observed with Intellivent. The current study definitely warrants further prospective controlled studies to estimate the potential clinical impact of Intellivent as compared with conventional modes of ventilation.