Introduction

Worldwide, an increasing number of deaths occur in the intensive care unit (ICU) after a decision to withdraw life support [1,2,3,4,5]. A major goal is to avoid unnecessary suffering due to prolongation of the dying process. Relatives of patients who die in the ICU have been reported to experience psychological distress manifesting as posttraumatic stress syndrome (PTSD)-related symptoms, complicated grief, anxiety, and/or depression, to degrees that vary according to the treatments provided and the quality of dying of the patient [6,7,8].

Withdrawal of mechanical ventilation holds a special place in the process of discontinuing life-sustaining treatments in ICU patients. Mechanical ventilation is withdrawn either by immediate extubation or by terminal weaning (gradual decrease in ventilatory support). The choice between these two methods is controversial, and whether it influences the experience of the patients and relatives is unclear [9,10,11]. The main challenge to mechanical ventilation withdrawal consists in preventing discomfort to the patient and, therefore, additional distress in the relatives [12,13,14]. In addition, mechanical ventilation has a deep psychological meaning, as it maintains breathing, which symbolizes life in many cultures. Compared to terminal weaning, immediate extubation is generally viewed as providing a more natural dying process, with less ambiguity, but with a higher risk of patient discomfort related to airway obstruction [14]. Previous studies suggested better satisfaction and lower rates of complicated grief among relatives of patients who died without the endotracheal tube [15, 16]. Other data, however, suggest greater family satisfaction after terminal weaning and worse patient distress after immediate extubation [15, 17]. No study has compared the psychological variables of relatives after terminal weaning versus immediate extubation [2, 10]. ICU staff members also express concern about choosing the best procedure for mechanical ventilation withdrawal [13, 18, 19].

We designed the prospective observational multicentre ARREVE study to compare immediate extubation versus terminal weaning regarding the long-term presence in relatives of PTSD-related symptoms, complicated grief, and symptoms of anxiety and depression. We also compared comfort of patients during the dying process and well-being of ICU staff members between the two methods.

Methods

Study design

ARREVE was a prospective, observational, multicentre study conducted in 43 French ICUs (in 20 university and 23 non-university hospitals) from February 2013 through April 2014. Critically ill adults (older than 18 years) with a decision to withdraw invasive mechanical ventilation, and the main adult relative of each, were enrolled in the study.

Participating relatives were informed that the clinical data of the patients would be collected and were asked for consent to a phone interview by a psychologist 3, 6, and 12 months after the death. Exclusion criteria for patients were non-invasive ventilation, brain death, and death within 48 h after ICU admission or before the initiation of mechanical ventilation withdrawal; relatives with insufficient knowledge of French for a phone interview and those who declined participation were excluded. Consent to the study was obtained from the relatives after the decision to withdraw mechanical ventilation but before its implementation. For each patient, we included the closest relative among those actively involved in exchanges with the ICU team, as identified by the ICU physicians.

For each patient, we included one nurse, one nursing assistant, one senior ICU physician, and one resident, all of whom had provided care to the patient during the last day before death or ICU discharge. The ethics committee of the French Intensive Care Society (FICS-SRLF) approved the study (see supplementary Appendix).

Terminal weaning and immediate extubation

Immediate extubation consisted in extubation with no previous decrease in ventilatory assistance. Terminal weaning was defined as a decrease in the amount of ventilatory assistance (oxygen supply and/or tidal volume and/or positive expiratory pressure and/or respiratory rate) and/or as initiation of spontaneous ventilation with a T-piece. The terminal weaning procedure could include the discontinuation of treatments used as adjuncts to mechanical ventilation (i.e. prone position and/or nitric oxide and/or almitrine), and secondary extubation performed in the event of prolonged dying causing distress to the patient and/or the family.

The choice between immediate extubation and terminal weaning was made by the ICU physician and other staff members when withdrawal of mechanical ventilation was decided, according to local practices and preferences of both relatives and ICU staff. Concomitant decisions to withdraw or withhold other treatments were at the discretion of the ICU staff. In this observational study, no specific recommendations were made about end-of-life care, including the use of sedative agents.

Study outcomes

Patient outcomes

Comfort of patients during the dying process was assessed on the basis of both the proportions of patients with airway obstruction and/or gasping and the Behavioural Pain Scale (BPS) score [20].

Psychological assessments of relatives: primary outcome

Relatives were interviewed over the phone by a psychologist 3, 6, and 12 months after the death, with three validated instruments widely used in relatives of ICU patients. The Impact of Events Scale-Revised (IES-R) was completed to assess PTSD-related symptoms 3 and 12 months after the death; scores can range from 0 (no symptoms) to 88 (severe symptoms), and a score greater than 32 indicates PTSD-related symptoms. The IES-R score 3 months after the death was the primary study outcome [8, 21,22,23].

The Hospital Anxiety and Depression Scale (HADS) includes two subscales for symptoms of anxiety and of depression, respectively. The total score can range from 0 (no anxiety or depression) to 42 (severe anxiety and depression). An anxiety or depression subscore greater than 8 indicates clinically meaningful symptoms [7, 24]. Relatives completed the HADS 3, 6, and 12 months after the death. Finally, the Inventory of Complicated Grief (ICG) was completed by the relatives 6 and 12 months after the death [16, 25]. ICG scores can range from 0 (no complicated grief) to 76, and scores greater than 25 indicate complicated grief.

An additional questionnaire of three general items evaluating the relative’s satisfaction with the end of life of the patient was completed during the interview 3 months after the death.

A relative was considered unreachable after 10 unanswered telephone calls. Relatives unreachable 3 months after the death were called 6 months after the death. Relatives who were unreachable 3 and 6 months after the death were not called for the 12-month interview.

Psychological assessments of ICU staff

The ICU staff members completed the Job Strain Score (JSS). This 12-item tool explores three domains (job demand, control, and social support) and has been validated in ICU staff [26, 27]. For each patient, the nurse, nursing assistant, senior physician, and resident included in the study completed the JSS either shortly after the death or on the day of ICU discharge if the patient did not die in the ICU. Total JSS values can range from −3 to 9, with higher scores indicating lower job strain. In addition, each staff member completed a questionnaire of three general items evaluating their satisfaction with the patient’s end of life.

Sample size

No reliable data were available for anticipating the IES-R difference between groups. Given the observational design, we planned to perform adjusted analyses and therefore required a number of observations appropriate for the large number of covariates taken into account, i.e. about 40 continuous and 10 qualitative covariates. We consequently planned to recruit 400 relatives in all [28].

Statistical analysis

Continuous variables were reported as mean ± standard deviation (SD) or median and interquartile range (IQR) and categorical variables as number and proportions. Missing data were handled as follows: when a single quantitative item was missing, we imputed the median of observed values for this item then calculated the score; otherwise, we eliminated the relative from the analysis of the relevant outcome measure. The Student test and Chi square test were applied to compare the two groups. Between-group mean differences in scores were estimated, as well as their 95% confidence intervals. Adjusted analyses were performed using linear and logistic regressions. The covariates are listed in the supplementary Appendix. We also performed three sensitivity analyses of the primary outcome, by using a propensity score instead of conventional multivariate analysis, by adding a random centre effect then using a mixed model approach, and by combining these two approaches. Statistical analyses were performed with SAS version 9.2 (SAS Institute, Cary, NC) and R 3.0.2 (http://www.r-project.org).

Results

Characteristics of the patients and relatives

Among 1674 patients treated with mechanical ventilation, 458 patients with a decision to withdraw invasive mechanical ventilation were included (Fig. 1). The distribution of terminal weaning and immediate extubation among the 43 participating centres is shown in eTable 1. We were able to include 190 (90.5%) relatives of the 210 patients with immediate extubation and 212 (85.5%) relatives of the 248 patients with terminal weaning. Demographic characteristics of enrolled relatives did not differ between groups (Table 1).

Fig. 1
figure 1

Study flowchart. ICU intensive care unit, IES-R Impact of Events Scale-Revised, HADS Hospital Anxiety and Depression Scale, ICG Inventory of Complicated Grief. * For each patient, a single relative completed the questionnaires. Relatives who did not respond to phone calls 3 months after the patient’s death were called again 3 months later. If they did not respond then, they were not called 12 months after the death. Those relatives who completed questionnaires 3 months after the death but did not answer the phone calls 6 months after the death were called 12 months after the death and invited to complete questionnaires

Full size image
Table 1 Characteristics of the patients and enrolled relatives
Full size table

Patient characteristics are reported in Table 1 and eTable 2. Compared to patients with immediate extubation, more patients with terminal weaning had previous activity limitation, previous ICU admission, surgical diagnosis, and acute respiratory failure at admission. At the time of the withdrawal decision, compared to patients with immediate extubation, those with terminal weaning had a longer ICU stay, worse Sequential Organ Failure Assessment (SOFA) score, greater vasoactive drug use, higher inspired oxygen fraction, and higher end-expiratory pressure.

Decisions to withdraw mechanical ventilation and comfort of patients during the dying process

Compared to relatives in the terminal weaning group, those in the immediate extubation group were more often involved in the withdrawal decision and in choosing between the two methods (eTable 4). Concomitant decisions to withdraw life-sustaining treatments were more common in the terminal weaning group than in the immediate extubation group. Other circumstances surrounding decisions to withdraw mechanical ventilation did not differ between groups.

The dying process is described in Table 2 and eTables 5 and 6. Presence of a relative in the room was twice as common during immediate extubation than during the first step of terminal weaning. Immediate extubation was associated with gasping or symptomatic airway obstruction and with a higher mean BPS score, compared to terminal weaning. Use of opioids, hypnotic drugs, and neuromuscular blocking agents was more common in the terminal weaning group. Secondary extubation was performed eventually in 26 (10.8%) patients receiving terminal weaning, usually because of a prolonged dying process. All patients with terminal weaning died in the ICU, whereas 11 (5.2%) patients were transferred to another ward after immediate extubation. Time to death in the ICU did not differ after terminal weaning initiation and after immediate extubation. Proportions of patients with relatives at their bedside at death in the ICU were similar in both groups.

Table 2 Assessment of the dying process after the decision to withdraw mechanical ventilation
Full size table

Psychological variables of relatives in the two groups

Primary outcome

Three months after the death, the mean IES-R score in the relatives was not significantly different between groups. The proportion of relatives with PTSD-related symptoms was also similar (Table 3 and eTables 7–8). These results were unchanged both after adjustment and in the sensitivity analyses.

Table 3 Assessment of posttraumatic stress syndrome, complicated grief, anxiety, and depression in relatives after the decision to withdraw mechanical ventilation
Full size table

Secondary outcomes

One year after the death, neither the mean IES-R score nor the frequency of PTSD-related symptoms differed significantly between groups (Table 3 and eTable 7). Similarly, ICG scores showed no significant differences between groups after 6 months or 1 year. Finally, neither the HADS scores nor the frequency of anxiety or depression symptoms was significantly different between groups 3 months, 6 months, or 1 year after the death. Satisfaction of relatives with end-of-life care in the ICU, participation in decisions, and respect of the patient’s wishes were high in both groups, with no significant difference (eTable 9).

Psychological variables of ICU staff in the two groups

Total JSS values in the assistant nurses were better with immediate extubation compared to terminal weaning (Table 4 and eTable 10). Total scores for other ICU staff categories were not significantly different between groups. Subscores indicated higher demand in nurses of the terminal weaning group compared to the immediate extubation group. Conversely, higher control and stronger social support were reported in assistant nurses and physicians, respectively, with immediate extubation compared to terminal weaning. Satisfaction of residents with end-of-life care was lower in the immediate extubation group, whereas in the other three ICU staff categories no significant differences were found (eTable 14).

Table 4 Job Strain Scores of nurses, nursing assistants, senior physicians, and residents
Full size table

Discussion

This large pragmatic multicentre observational study is the first comparison of immediate extubation versus terminal weaning in terms of quality of death of critically ill patients, as assessed by the comfort of dying in patients and the psychological well-being of relatives and ICU staff. Immediate extubation was associated with greater airway obstruction, a higher frequency of gasping, and higher BPS scores. In the relatives, PTSD-related symptoms, complicated grief, and symptoms of anxiety and depression up to 1 year after the death were not significantly different between the two groups. In the staff, job strain of assistant nurses was lower with immediate extubation compared to terminal weaning.

Considerable variation exists in practices for mechanical ventilation withdrawal in the ICU [11, 12]. Few data are available for determining whether one method is superior over the other. Consequently, the choice between the two methods is mainly a matter of opinion. In the current large study under the conditions of everyday practice, immediate extubation and terminal weaning were preferred by nearly identical proportions of ICU staff; however, the between-group differences in admission diagnoses suggest a preference for immediate extubation in comatose patients and for terminal weaning in patients with respiratory failure. In contrast to previously stated opinions, in our observational study, immediate extubation was not associated with a greater burden on the relatives compared to terminal weaning [18]. Importantly, satisfaction was very high among relatives and staff, with no difference between methods, except in the residents. Thus, our study suggests that, for the relatives, the two methods may result in similar experiences, provided the staff members are well trained in, and comfortable with, the method they apply.

Symptomatic airway obstruction and gasps were more common and the mean BPS score was higher in the immediate extubation group. This finding can be ascribed to airway compromise directly related to removal of the endotracheal tube with subsequent obstruction by the tongue and/or inability to remove secretions. Another factor may be underuse of analgesics and sedatives in patients undergoing immediate extubation. Higher doses of opioids and sedatives were used in the patients undergoing terminal weaning, in whom previous respiratory and/or multi-organ failure was more severe than in the immediate extubation group. Interestingly, time to death from extubation or first change in ventilator settings for terminal weaning did not differ between groups, in keeping with previous findings [29]. Conceivably, physicians may be concerned about active shortening of the dying process related to pre-emptive deep sedation after extubation [30]. On the other hand, in the terminal weaning group, the greater use of opioids and sedatives, and the administration to some patients of neuromuscular blocking agents, may reflect a willingness to shorten the dying process, despite the double-effect principle and guidelines discouraging the use of neuromuscular blocking agents at the end of life [31, 32]. Both hypotheses suggest room for improving end-of-life care. However, the use of neuromuscular blockade may also reflect an attempt by ICU staff to avoid additional suffering in relatives of patients who are deeply comatose with severe myoclonic status epilepticus or distressing agonal gasps despite high doses of sedative drugs [32].

Finally, we found that the psychological welfare of the ICU staff was better with immediate extubation than with terminal weaning. The emotional responses of staff members to death may affect their beliefs about whether withdrawing life support is legitimate and whether comfort care should be offered as an option [33]. Differences in perceptions of the two methods have been reported, but their associations with markers for psychological burden in staff members had not been evaluated previously [13, 14, 18]. In ICU nurses, both caring for dying patients and the number of decisions to forego life-sustaining treatments are associated with increased burnout, which may also influence nurses’ beliefs and behaviours [34]. In a recent study of ICU staff perceptions, a preference for terminal weaning was related to an unfavourable perception of immediate extubation, whereas both staff members who preferred extubation and those with no preference perceived extubation as providing a less medicalised death and minimising ambiguity [18]. Thus, identifying personal beliefs that might constitute barriers to mechanical ventilation withdrawal is crucial when seeking to implement protocols for patient care [35,36,37,38].

The main limitation of this study is the absence of randomisation. However, a major concern when performing end-of-life studies with psychological assessments of relatives and staff is the risk of adverse effects induced by modifying the usual practice of the ICUs in a direction contrary to the convictions of the participating ICU teams. Indeed, ICU staff members differ in their perceptions of immediate extubation and terminal weaning, suggesting a risk of poor compliance with random allocation of the method of mechanical ventilation withdrawal [19]. Another limitation is that end-of-life care was not standardised, as shown by the differences in this regard between the two groups. However, dictating end-of-life practices for the study might have generated bias due to reluctance of ICU staff to apply methods with which they felt uncomfortable. Moreover, we used a clear definition of terminal weaning and immediate extubation. Immediate extubation was not preceded by interventions on ventilator settings and very few patients in the terminal weaning group underwent secondary extubation. These conditions avoided any mismatch between the two practices. Last, neither the adjusted nor the sensitivity analyses suggested any associations linking the patients’ baseline characteristics or end-of-life practices to the study results. A third limitation is that communication between ICU staff and relatives was not evaluated in detail, and no specific recommendations were provided. Studies have emphasised the importance of good communication with relatives of dying patients [16, 39]. However, as previously stated, our aim was to interfere as little as possible with the everyday practice of the ICU teams. The absence of difference in psychological welfare of relatives between groups suggests that communication between ICU staff and relatives in both groups was in line with patient health state. The fourth limitation is that all participating ICUs were in France. Whether our findings apply to other countries is unclear. However, the ICUs were distributed throughout France and located in both university and community hospitals. Moreover, frequencies and levels of PTSD-related symptoms, complicated grief, and symptoms of anxiety and depression in the relatives were consistent with previous reports [6]. Last, the large size and varied case-mix of our patient population support the general applicability of our findings to other countries and settings.

In conclusion, immediate extubation for mechanical ventilation withdrawal was not associated with differences in psychological welfare of relatives compared to terminal weaning, when each method constituted standard practice in the ICU where it was applied. Compared to terminal weaning, immediate extubation was associated with less job strain in ICU staff. Patients had more airway obstruction and gasps with immediate extubation, indicating a need for better palliative care.