Background
Patient-ventilator asynchronies are frequent during invasive mechanical ventilation (MV) [
1,
2]. Poor patient-ventilator interaction could be associated with prolongation of MV, longer intensive care unit (ICU) and hospital stays [
2], and increased mortality [
1,
3]. Thus, optimizing patient-ventilator interaction may improve outcomes [
4].
Many aspects of clinical management affect patient-ventilator interaction. Adjusting ventilator settings can decrease asynchronies and associated anxiety and dyspnea [
5,
6]. Sedatives can cause ventilatory depression affecting respiratory drive and timing, worsening patient-ventilator interaction in proportion to decreasing level of consciousness [
7,
8]; these effects appear to differ with different drugs [
9,
10]. Sedation is associated with deleterious side effects. Deep sedation is associated with worse short- and long-term outcomes [
11‐
14]. Forgoing or minimizing sedatives during MV is increasingly recommended [
15‐
17].
The relationship between asynchronies, level of consciousness (sedation level), and sedatives and opioids is poorly understood. Increasing sedatives in ICU patients with double cycling (DC) not only failed to correct this asynchrony [
6], but also prolonged MV and ICU and hospital stays. The rate of ineffective inspiratory efforts increases proportionately with the depth of sedation [
8]. Different studies have reported disparate effects of sedative dosage on the overall rate of asynchronies. Whereas one study found that light sedation with propofol did not affect the rate of asynchronies but deep sedation with propofol increased it [
10], another found that deep sedation reduced but did not eliminate asynchronies [
18]. A recent study showed that deep sedation, benzodiazepines, and cumulative doses of benzodiazepines were associated with higher mortality [
19]. In another trial, patients on dexmedetomidine had slightly fewer asynchronies than those on propofol [
20]. However, these studies did not take opioid administration into account.
Opioids are commonly used in ICU patients. In a study comparing midazolam vs. fentanyl plus midazolam, patients receiving fentanyl had fewer asynchronies than those receiving only midazolam [
21]. Thus, the effects of opioids alone or together with sedatives on asynchronies warrant investigation. Adequate opioid treatment with minimal doses of sedatives might enable more spontaneous breathing and improve patient-ventilator interaction; however, the relationships between opioids, level of consciousness, and asynchronies remain to be elucidated. We hypothesized that opioids alone would improve trigger and cycling asynchronies and result in more wakeful patients than sedatives alone or sedatives-plus-opioids.
Methods
Study population and design
We obtained data from an ongoing database started in 2011 in four centers in Spain. The database was constructed prospectively with funding for a project to develop a connectivity platform to interoperate signals from different ventilators and monitors and subsequently compute algorithms to diagnose patient-ventilator asynchronies (ClinicalTrial.gov, NCT03451461); each institution’s review board approved the database.
This prospective observational study included adult patients admitted to four ICUs between October 2011 and January 2013. The institutional review boards approved the protocol, waiving informed consent because the study was non-interventional, posed no added risk to patients, and did not interfere with usual care.
Patients were prospectively included when the following criteria were met: admission to a bed equipped with BetterCare™ software and intubated for MV expected to last > 24 h. To avoid selection bias, members of the research team were not involved in assigning patients to a bed equipped with BetterCare™ software. Exclusion criteria were < 48 h of data, age < 18 years, pregnancy, do-not-resuscitate orders, admission for organ donation, and chest tubes with suspected bronchopleural fistula.
Patient management and data collection
Demographic and clinical data were obtained from medical records. Level of consciousness was assessed every 4 h with the Riker Sedation-Agitation Scale (SAS), and the mean value of these assessments was computed to obtain a daily average. Illness severity was assessed daily with the Sequential Organ Failure Assessment (SOFA). ICU teams were aware of the recording system, but not of the study hypothesis. All patients were managed with similar processes of care and lung-protective ventilation strategies (tidal volume 6 mL/kg ideal body weight and plateau pressure under 30 cmH
2O) following the quality indicators of the Spanish Society of Intensive Care Medicine (
https://semicyuc.org/wp-content/uploads/2018/10/quality_indicators_update_2011.pdf) throughout the study. Patients were ventilated with Evita 4 (Dräger, Lübeck, Germany), Puritan Bennet 840 (Covidien, Plymouth, MN, USA), or Servo I (Maquet, Fairfield, NJ, Sweden) ventilators, receiving volume assist/control, pressure assist control, or pressure support based on clinicians’ assessment of clinical status. Ventilatory modes were analyzed as previously described [
22]. The predominant mode for each day analyzed was classified as assist-control or pressure-support if the patient remained in that mode for ≥ 70% of the time. Other ventilator parameters were also adjusted at the discretion of the attending physician following national recommendations. Clinicians adjusted ventilator settings when asynchronies were observed at the bedside, but adjustments were not protocolized.
Total doses of opioids (morphine and fentanyl) and sedatives (midazolam, propofol, and lorazepam) administered each day were recorded and converted to morphine and midazolam equivalents [
6]. We classified each day of MV for a given patient as (1) no-drugs, (2) sedatives-only, (3) opioids-only, or (4) sedatives-plus-opioids. To avoid misleading classifications due to residual treatment effects, days were classified according to all medications administered during the day; thus, a day in which a patient received sedatives-plus-opioids for > 2 h and opioids-only thereafter would be classified as “sedatives-plus-opioids.” Days in which patients were treated with neuromuscular blockers were excluded from the analysis. Sedatives and analgesia were managed following each ICU’s protocols based on the Spanish Society of Intensive Care Medicine recommendations [
23], SAS level, and pain and discomfort assessments.
For the analyses, data was structured as averaged measures per day. Therefore, every treatment group could include a different number of patients depending on the day.
Analysis of asynchronies
Asynchronies were detected by BetterCare™ software (Barcelona, Spain), which continuously records airflow, airway pressure, and tidal volume from admission to extubation or death. BetterCare™ identifies the beginnings of inspiration and expiration to analyze and store data breath by breath. It analyzes each breath to detect four types of asynchronies (ineffective inspiratory efforts during expiration (IEE), DC, short cycling/aborted inspiration, and prolonged cycling [
1,
2]) and computes the asynchrony index (AI) (Additional file
1) [
24]. Periods in which recording was interrupted due to clinical interventions, out-of-ICU transfers, technical problems, or other issues were excluded from the analysis, which was done on the remaining valid periods.
All asynchronies were averaged per day. The rates of IEE, DC, and the overall AI were computed considering the total number of breaths (ventilator-delivered cycles plus IEE), enabling us to compare days, despite varying respiratory rates.
Statistical analysis
Patients’ characteristics are summarized as medians (25th–75th percentiles) or percentages. Sample size calculation was considered unnecessary for this exploratory study.
To analyze the level of consciousness and illness severity by treatment groups, we used linear mixed-effects (LME) models with random intercepts for the patients. This approach takes inter- and intra-subject variability in longitudinal data into account; each patient differs from the overall mean response by an individual-specific constant that applies equally over time [
25]. To analyze asynchronies by treatment groups, we used generalized LME (GLME) models assuming a negative binomial distribution for the response variable (number of asynchrony events) because the variable response was discrete, limited to non-negative values, and positively skewed with most observations having values near zero. Negative binomial distributions are often used in regression models with count data. Furthermore, to analyze the number of asynchrony events as a rate, we incorporated an exposure term (total number of respiratory cycles per day) that indicates the number of times a particular event occurred.
To assess the effects of level of consciousness and severity of illness on each type of asynchrony, we also used GLME models with random intercepts for the patients and allowing variation in SAS and SOFA slopes by treatment groups for the population mean.
To explore the effects of dosage on level of consciousness and asynchronies, we used a LME model and GLME models, respectively. This analysis included the opioids-only, sedatives-only, and sedatives-plus-opioids groups. These models used random intercepts only for the patients and allowed variations in slopes for dose equivalents by treatment groups for the population mean. Additionally, we investigated the effect of severity (SOFA) as a potential confounding variable that could influence both the asynchronies and the treatment group.
We used R 3.3.1 (R Core Team, Vienna, Austria, URL
http://www.R-project.org) for all analyses, building the mixed-effects models with the lme4 package [
25] and summarizing the mean (95% CI) effects by treatment groups with the lsmeans package. When using LME models for the continuous response variables, we checked the normality assumptions for the estimated random effects and for the within-subject residuals by graphical methods (normal Q-Q plots). When the response variable was discrete, we assessed overdispersion by graphical comparison of the standardized residuals versus the fitted values. Significance was set at
p < 0.01. Pairwise comparisons among treatment groups were two-sided and adjusted by the Bonferroni method to maintain the significance level.
Discussion
This is the first study to present data relating asynchronies and treatment with sedatives and opioids throughout the complete MV period. The overall rate of asynchronies did not differ between days classified as opioids-only, sedatives-only, and sedatives-plus-opioids. Patients receiving sedatives had a lower level of consciousness than those receiving opioids-only; sedatives-plus-opioids decreased the level of consciousness, but did not result in fewer asynchronies than the other treatments. Interestingly, in sedatives-plus-opioids days, the sedative dose was directly associated with the rate of asynchronies and with a lower level of consciousness, whereas higher opioid doses were associated with a lower AI without worsening level of consciousness. Thus, opioid administration seems a clinically sound approach to improve patient-ventilator synchrony while preserving consciousness.
Patients with shock or severe respiratory failure often require sedatives. Moreover, sedatives are sometimes administered in attempts to improve patient-ventilator interaction. However, deep sedation has been associated with worse outcomes [
13]. Lighter or no-sedation is favored in partial ventilatory support modes, where patient-ventilator synchrony is crucial. Physiological studies show that sedatives have varying effects on asynchronies. Increasing sedatives/analgesia is relatively ineffective in abolishing severe breath-stacking [
6]. Deep sedation with propofol increases asynchronies during pressure support ventilation, whereas light sedation does not [
10]; deeper sedation is associated with increased ineffective triggering events [
8] and with increased mortality [
26]. Thus, sedative management in MV is a modifiable variable that could improve outcomes [
26].
Associations between sedatives and asynchronies are probably confounded by many factors, especially by clinicians’ ventilator adjustments. In our study, the incidence of asynchronies was associated with the drugs used (sedatives and opioids), irrespective of the ventilatory mode. However, sedatives lowered patients’ level of consciousness without decreasing asynchronies beyond opioids alone. Inadequate pain control worsens patient-ventilator synchrony [
21] and is associated with agitation, which negatively affects outcomes [
27]. Richman et al. [
21] found that patients receiving midazolam-plus-opioids had fewer asynchronies/day over a 3-day period than those receiving midazolam alone. Our results support these findings, showing that opioids could help improve asynchronies beyond sedatives, although prospective trials are necessary to determine whether appropriate opioids favor better synchrony by ensuring adequate analgesia without depressing consciousness and without affecting respiratory drive or minute ventilation [
28‐
30].
Increasing sedative doses prolongs MV and ICU and hospital stays [
12,
14]. Paired with pain management protocols [
17], sedation protocols including light sedation or daily interruptions of sedation [
13,
15,
16] improve ICU patients’ outcomes. Our results suggest that, compared with treatments including sedatives, treatment with opioids-only enables patients to be more awake without increasing asynchronies. Opioids-only treatment resulted in a higher level of consciousness than treatment with sedatives-plus-opioids. Our findings on the effects of sedatives and opioids throughout MV are in line with those of a randomized clinical trial where MV patients receiving morphine had more ventilator-free days and shorter ICU stays than those receiving sedatives-plus-morphine, without increases in accidental extubation or ventilator-associated pneumonia [
16]. Moreover, we found that in assist-control modes, compared with no-drugs, treatment with sedatives-plus-opioids and opioids-only favored better patient-ventilation interaction (lower AI), suggesting that opioids might improve patient comfort. Thus, it might be beneficial to maintain opioid treatment until liberation from MV.
We also found significantly lower AI, IEE, and DC in sedatives-plus-opioids than in no-drugs days. However, asynchronies in MV patients who do not require sedatives or opioids probably are intrinsically different from those that occur in patients who require these treatments, and therefore, they probably require a different clinical approach. Rue et al. [
31,
32] recently used a Bayesian joint model incorporating longitudinal ICU stay markers to evaluate outcome determinants. Finding that overall AI was not associated with severity of illness or vital status, they postulated that some asynchronies could be a marker of life. In our study, we considered the effect of severity of illness to better understand this possible confounding factor. We found no associations between severity and AI or IEE (Table
2), so patients with more asynchronies were not necessarily more severely ill. The lower severity and higher rates of asynchronies in no-drugs days compared with sedatives-only, opioids-only, and sedatives-plus-opioids days are likely due to differences in the origin and behavior of the asynchronies that occur in the different groups. Therefore, from a clinical perspective, it makes no sense to compare asynchronies and their management in no-drugs versus in treatment days [
3,
33].
Finally, our analysis of the relationship between asynchronies and drug dosage found that, unlike sedatives, increasing doses of opioids were associated with decreasing rates of asynchronies, without significantly affecting the level of consciousness, independently of the level of severity. This finding highlights the importance of titrating opioids for comfort, and possibly asynchronies, in addition to pain control.
Unlike some physiological studies, we found that sedatives, alone or together with opioids, did not decrease asynchronies more than opioids-only. On the other hand, our findings of increased rates of asynchronies, especially IEE, with an increased daily dose of sedatives are in line with physiological studies [
8,
10]. Whereas physiological studies analyze only short time periods, our study considered the entire period of MV, making it closer to clinical practice. One strength of our study is that it is based on prospectively accrued physiological data with enough breadth and depth to characterize a patient’s condition throughout MV. Observational trials like this are increasingly being used because conclusions drawn from data collected in real-world situations can be more generalizable than the more restricted, if more vigorous, conclusions of randomized clinical trials [
34‐
36]. Our findings add to the growing body of evidence pointing to the inability of sedatives to prevent and/or correct asynchronies in daily practice and the association of opioids with improved asynchrony rates, thus supporting the strategy of managing pain while maintaining the lightest sedation possible [
17,
37].
Our study has several limitations. First, patients were not randomized to each drug regimen. Furthermore, patients received opioids, sedatives, or no-drugs in any sequence or combination as deemed clinically necessary. Granular data for sedative or opioid doses over smaller time intervals were unavailable, thus precluding analyses of temporal associations between sedatives/opioids and asynchronies that might have enabled causal inferences and greater insight. Moreover, we did not consider factors that influenced clinicians to modify sedative or opioid dosing. Patients often receive more drugs early in MV and less when approaching weaning; however, to counterbalance this bias, we explored the relationship between severity and drugs and asynchronies, but found no relevant associations. Additionally, we did not analyze other painkillers such as acetaminophen and nonsteroidal anti-inflammatory drugs, which may influence patient-ventilator interaction differently, so our findings cannot be extrapolated to other non-opioid drugs. We used no objective measures of pain levels, precluding the analysis of associations between pain and asynchronies. Likewise, we did not measure surrogates of respiratory center activity, so we cannot evaluate associations between respiratory drive and different asynchronies. As individual patients could be considered in more than one group because management strategies evolved from day-to-day, our analysis of drug doses could not take into account the prolonged half-life and accumulation of some sedatives [
38]. Nevertheless, we found a good relationship between SAS score and the sedative treatment group. We analyzed only IEE and DC because they are the most relevant asynchronies; we did not analyze flow asynchronies because they cannot be established from ventilator airway pressure and flow scalars alone. Thus, our findings cannot be extrapolated to these asynchronies. Additionally, because the measure unit was days rather than patients, it was difficult to analyze the effect of the underlying disease on the results. In an attempt to overcome this difficulty, we adjusted the results by including SOFA score as a marker of severity, but we found no effect of severity in the incidence of asynchronies per group. Finally, clinically detected asynchronies were treated according to each ICU’s protocols; thus, differences between centers might have affected the results. We performed an analysis to evaluate the influence of each center and we did not find significant differences between centers.
Acknowledgements
The authors thank Mercè Ruiz and John Giba for their invaluable support in editing the manuscript.
ASYNICU group: Candelaria de Haro, Josefina López-Aguilar, Rudys Magrans, Sol Fernández-Gonzalo, Gemma Gomà, Encarna Chacón, Anna Estruga, Ana Ochagavia, Lluís Blanch (Critical Care Center, Hospital Universitari Parc Taulí, Institut d’Investigació i Innovació Parc Taulí I3PT, Universitat Autònoma de Barcelona. Sabadell, Spain); Jaume Montanya, Bernat Sales (Better Care. Barcelona, Spain); Enrico Lena, Umberto Lucangelo (Department of Perioperative Medicine, Intensive Care and Emergency, Cattinara Hospital, Trieste University. Trieste, Italy); Rafael Fernández, Carles Subirà (ICU, Fundació Althaia, Universitat Internacional de Catalunya. Manresa, Spain); Guillermo M Albaiceta (Unidad de Cuidados Intensivos Cardiológicos, Hospital Universitario Central de Asturias, Oviedo, and Departamento de Biología Funcional, Instituto Universitario de Oncología del Principado de Asturias, Universidad de Oviedo. Oviedo, and CIBERES, Instituto de Salud Carlos III, Madrid, Spain); Enrique Fernández-Mondejar (Department of Emergency and Critical Care Medicine, Hospital Virgen de las Nieves, Universidad de Granada, Granada, Spain); Gastón Murias (Critical Care Department, Hospital Británico. Buenos Aires, Argentina); Robert M Kacmarek (Department of Respiratory Care, Department of Anesthesiology, Massachusetts General Hospital, Harvard Medical School. Boston, MA, USA).
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.