Palliative pharmacological sedation for terminally ill adults
Abstract
Background
Terminally ill people experience a variety of symptoms in the last hours and days of life, including delirium, agitation, anxiety, terminal restlessness, dyspnoea, pain, vomiting, and psychological and physical distress. In the terminal phase of life, these symptoms may become refractory, and unable to be controlled by supportive and palliative therapies specifically targeted to these symptoms. Palliative sedation therapy is one potential solution to providing relief from these refractory symptoms. Sedation in terminally ill people is intended to provide relief from refractory symptoms that are not controlled by other methods. Sedative drugs such as benzodiazepines are titrated to achieve the desired level of sedation; the level of sedation can be easily maintained and the effect is reversible.
Objectives
To assess the evidence for the benefit of palliative pharmacological sedation on quality of life, survival, and specific refractory symptoms in terminally ill adults during their last few days of life.
Search methods
We searched the Cochrane Central Register of Controlled Trials (CENTRAL; 2014, Issue 11), MEDLINE (1946 to November 2014), and EMBASE (1974 to December 2014), using search terms representing the sedative drug names and classes, disease stage, and study designs.
Selection criteria
We included randomised controlled trials (RCTs), quasi‐RCTs, non‐RCTs, and observational studies (e.g. before‐and‐after, interrupted‐time‐series) with quantitative outcomes. We excluded studies with only qualitative outcomes or that had no comparison (i.e. no control group or no within‐group comparison) (e.g. single arm case series).
Data collection and analysis
Two review authors independently screened titles and abstracts of citations, and full text of potentially eligible studies. Two review authors independently carried out data extraction using standard data extraction forms. A third review author acted as arbiter for both stages. We carried out no meta‐analyses due to insufficient data for pooling on any outcome; therefore, we reported outcomes narratively.
Main results
The searches resulted in 14 included studies, involving 4167 adults, of whom 1137 received palliative sedation. More than 95% of people had cancer. No studies were randomised or quasi‐randomised. All were consecutive case series, with only three having prospective data collection. Risk of bias was high, due to lack of randomisation. No studies measured quality of life or participant well‐being, which was the primary outcome of the review. Five studies measured symptom control, using four different methods, so pooling was not possible. The results demonstrated that despite sedation, delirium and dyspnoea were still troublesome symptoms in these people in the last few days of life. Control of other symptoms appeared to be similar in sedated and non‐sedated people. Only one study measured unintended adverse effects of sedative drugs and found no major events; however, four of 70 participants appeared to have drug‐induced delirium. The study noticed no respiratory suppression. Thirteen of the 14 studies measured survival time from admission or referral to death, and all demonstrated no statistically significant difference between sedated and non‐sedated groups.
Authors' conclusions
There was insufficient evidence about the efficacy of palliative sedation in terms of a person's quality of life or symptom control. There was evidence that palliative sedation did not hasten death, which has been a concern of physicians and families in prescribing this treatment. However, this evidence comes from low quality studies, so should be interpreted with caution. Further studies that specifically measure the efficacy and quality of life in sedated people, compared with non‐sedated people, and quantify adverse effects are required.
PICOs
Plain language summary
Sedation medication for relieving symptoms at the end of life
Background
People with diseases that are not curable may have a variety of symptoms at the end of life. These symptoms can include confusion (delirium), anxiety, restlessness, breathlessness (dyspnoea), pain, vomiting, and distress. Medicines that reduce consciousness (sedatives) may help relieve these symptoms when people are close to death.
Treatment with sedatives can vary in terms of the level of sedation (mild, intermediate, and deep), and duration (intermittent or continuous).
Study chara cteristics
We searched international databases in October 2012 and again in December 2014 for studies of terminally ill adults who required sedation in order to control symptoms. We found 14 studies of around 4000 people. The studies compared sedation versus non‐sedation. Most people in the studies had cancer (95%). The studies took place in hospices, palliative care units, hospitals, and the home.
Key results
Five studies showed that sedatives did not fully relieve delirium or breathlessness. There was no difference between the groups in terms of the other symptoms. There was no difference in time from admission or referral to death
Only one study reported side effects, and did not report any major problems.
Future studies should focus on how sedatives affect a person's quality of life, or peacefulness and comfort during the dying phase, and how well sedation controls the distressing symptoms. Side effects should be better reported.
Quality of evidence
The studies were not randomised controlled trials (where people are randomly allocated to one of two or more treatment groups), and so we judged the quality of the evidence as poor.
Authors' conclusions
Background
Description of the condition
Terminally ill people experience a variety of symptoms in the last hours and days of life, including delirium, agitation, anxiety, terminal restlessness, dyspnoea, pain, vomiting, and psychological and physical distress. Terminal restlessness is an agitated delirium that occurs in some people during the last few days of life (Doyle 2008). The Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments (SUPPORT) indicated that during their last three days of life, 80% of dying hospitalised people had severe fatigue, 50% severe dyspnoea, and 40% severe pain (Lynn 1997). In another study, the most commonly reported symptoms were fatigue, dyspnoea, and dry mouth, with the most distressing being fatigue, dyspnoea, and pain (Hickman 2001). Other distressing symptoms reported in this and other similar studies were noisy breathing, excess respiratory secretions, agitation, anxiety, constipation, nausea and vomiting, anorexia, incontinence, pressure sores, and insomnia (e.g. Cowan 2006; Morita 2005).
In the terminal phase of life (i.e. when the disease is progressive, far advanced, incurable, and death is imminent), these symptoms may become refractory, unable to be controlled by supportive and palliative therapies specifically targeted to these symptoms. Palliative sedation therapy is one potential solution to providing relief from these refractory symptoms.
Description of the intervention
Palliative sedation therapy has been described as "the use of sedative medications to relieve intolerable suffering from refractory symptoms by a reduction in patient consciousness" (De Graeff 2007). The therapy can vary in terms of level of sedation (mild, intermediate, and deep), and duration (intermittent or continuous). Sedation can be achieved by drugs that are primarily sedatives, and are not designed to treat the underlying condition or symptom, or by drugs that have some effect on the underlying symptom and have a secondary effect of causing somnolence.
Drug classes used for palliative sedation include benzodiazepines (particularly midazolam and clonazepam), antipsychotics, opioids, and hypnotics. They may be administered intravenously or subcutaneously.
How the intervention might work
Sedation in terminally ill people in the last hours or days of life is intended to provide relief from refractory symptoms that are not controlled by other methods. Sedative drugs such as benzodiazepines are titrated to achieve the desired level of sedation, and potentially the desired level of symptom control; the level of sedation can be easily maintained and the effect is reversible. Therefore, sedation may be useful in terminally ill people where symptom control cannot be achieved by drugs targeted at the specific symptom. There is also some concern as to whether this form of treatment may shorten life, and could be used to hasten death intentionally, similarly to euthanasia, so assessing the effects on survival is important.
Why it is important to do this review
There are existing Cochrane reviews on interventions for particular symptoms (e.g. interventions for noisy breathing in people near to death (Wee 2008), opioids for palliation of breathlessness (Jennings 2001), drug therapy for delirium in terminally ill people (Jackson 2004a), benzodiazepines and related drugs for insomnia in palliative care (Hirst 2009), and anxiety in palliative care (Jackson 2004b)), and one systematic review of one drug (propofol) for terminal sedation (McWilliams 2010). One systematic review was published after the commencement of this review (Maltoni 2012a), which reports on most of the studies included in this review. The focus of the review was survival. Our review aimed to bring together in one place the limited information on all drugs used to sedate terminally ill people, for all symptoms, specifically in the terminal phase of life, as distinct from the broader palliative care setting.
Objectives
To assess the evidence for the benefit of palliative pharmacological sedation on quality of life, survival and specific refractory symptoms in terminally ill adults during their last few days of life.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs), quasi‐RCTs, non‐RCTs, and observational studies (e.g. before‐and‐after, interrupted‐time‐series) with quantitative outcomes. We made this choice because we knew prior to starting the review that there would be few, if any, RCTs in this area. We excluded studies with only qualitative outcomes or that had no comparison (i.e. no control group or no within‐group comparison) (e.g. single arm case series).
Types of participants
We included studies of terminally ill adults (aged 15 years or greater) who required sedation in order to control symptom(s) (e.g. agitation, anxiety, insomnia, terminal restlessness, dyspnoea, and pain). We considered all terminal conditions (malignant and non‐malignant), in all settings (e.g. home, hospital, and palliative care institution).
Types of interventions
Any medication with a sedative effect (e.g. benzodiazepines, barbiturates, anaesthesia, opioids, antipsychotics, antihistamines, or other hypnotics) where the intention was sedation for symptom relief. Sedation may have been given continuously or intermittently, with the intention of reducing the level of consciousness to relieve symptoms. Sedation may have been deep (unconscious) or the person may have had periods when they were drowsy, but not unconscious. The comparator was no sedation. Sedative medications may have been given in very low doses (e.g. for sleep at night), but the intention was not to sedate to relieve intractable symptoms.
Types of outcome measures
Primary outcomes
-
Quality of life or a person's well‐being. This would usually be measured by a proxy (e.g. doctor, nurse, carer), but in certain circumstances may have been measured by the person during periods of adequate consciousness. We used the term 'quality of life' to represent any domain related to the quality of the person's experience during the dying phase. This may have included peacefulness or comfort, carer's satisfaction with the person's experience, or a multi‐dimensional assessment of symptom control affecting quality of life, for example.
Secondary outcomes
-
Control of specific symptom(s) (e.g. agitation, anxiety, insomnia, terminal restlessness, dyspnoea, and pain).
-
Duration of symptom control.
-
Time to control of symptoms.
-
Adverse effects of treatment. For example, for antipsychotics these may include: worse drowsiness than intended, extrapyramidal effects, akathisia (restlessness), antipsychotic malignant syndrome, urinary retention, and constipation; and for benzodiazepines: drowsiness, ataxia, confusion, falls, increased restlessness, respiratory depression, and hypotension.
-
Duration of institutional care.
-
Time to death.
-
Carer satisfaction.
Search methods for identification of studies
Electronic searches
We searched the following databases:
-
the Cochrane Central Register of Controlled Trials (CENTRAL; 2014, Issue 11);
-
MEDLINE (Ovid) (1946 to November 2014);
-
EMBASE (Ovid) (1974 to December 2014).
Appendix 1 shows the search strategies. We applied no language or date restrictions.
We also searched clinical trials registries (ClinicalTrials.gov; the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) (apps.who.int/trialsearch/)) in October 2012 and again in December 2014 to find any ongoing trials or to locate other publications that might not have been found in the database searches.
Searching other resources
-
Reference lists of relevant textbooks, review articles, and relevant studies.
-
We wrote to investigators known to be involved in previous studies, seeking information about unpublished or incomplete trials.
Data collection and analysis
Selection of studies
We used the search strategy described to obtain titles and abstracts of studies that may be relevant to the review. Two review authors (EB, MvD) independently screened the titles and abstracts, and discarded studies that were not applicable; however, we initially retained studies and reviews that might have included relevant data or information on studies.
Two review authors (EB, MvD) independently assessed the retrieved full‐text of these studies to determine which studies satisfied the inclusion criteria. A third review author was to act as arbiter if needed (GM).
Data extraction and management
Two review authors (LM, ST) independently carried out data extraction using standard data extraction forms. We translated studies reported in non‐English language journals before assessment. Where more than one publication of one study existed, we grouped reports together and we used the publication with the most complete data in the analyses. Where relevant outcomes were only published in earlier versions, we used these data as well. We highlighted any discrepancy between published versions. A third review author (EB) acted as arbiter. One review author (EB) used Review Manager 5 software to enter data, which we would have used to perform meta‐analyses (RevMan 2014).
Assessment of risk of bias in included studies
For RCTs, two review authors (LM, ST) independently assessed the following items using the Cochrane 'Risk of bias' assessment tool (Higgins 2011). A third review author (EB) acted as arbiter.
-
Adequate sequence generation (selection bias).
-
Allocation concealment (selection bias).
-
Blinding of participants and personnel.
-
Blinding of outcome assessment (detection bias).
-
Incomplete outcome data (attrition bias).
-
Selective reporting (reporting bias).
-
Other bias.
For non‐randomised, comparative trials, the first and second criteria were set to 'high risk of bias'. We included two additional criteria to assess selection bias in non‐randomised studies:
-
were baseline characteristics similar?
-
were baseline outcome measurements similar?
This follows the recommendations of the Cochrane Effective Practice and Organisation of Care (EPOC) review group for assessing risk of bias in non‐randomised studies. For future updates of this review, we plan to use the Cochrane ACROBAT‐NRSi tool (Sterne 2014).
We assessed each of these criteria as low risk of bias, unclear risk of bias, or high risk of bias.
Measures of treatment effect
Primary outcomes
Quality of life or a person's well‐being
We anticipated that studies would measure quality of life or a person's well‐being on a recognised quality of life continuous scale. Since there are many quality of life scales, we planned to use the standardised mean difference (SMD) (with standard deviation) between the intervention and control groups to include all comparative studies with a quality of life scale or well‐being scale as outcome. However, no included studies measured the primary outcome.
Secondary outcomes
Symptom control
Where studies reported symptom control on a continuous scale, we planned to use SMD (with 95% confidence interval (CI)) to combine them in a similar method to the primary outcome.
Where studies reported symptom control on a dichotomous scale (e.g. relief versus no relief), we planned to combine these studies using risk ratio (RR) in the intervention group compared with control (with 95% CI).
If data were reported using a short ordinal scale (e.g. no relief, some relief, moderate relief, good relief), we planned to combine the moderate and good relief categories, and the none and mild categories and include these with the dichotomous outcome studies.
Adverse effects of treatment
We planned to use the proportion of participants experiencing any adverse effect of treatment, and combine studies using RR (and 95% CI).
Duration of symptom control, time to control of symptoms, duration of institutional care, and time to death
If there had been sufficient studies to meta‐analyse duration of symptom control, time to control of symptoms, duration of institutional care, and time to death, we intended to use the generic inverse variance method to pool hazard ratios (Higgins 2011).
Carer satisfaction
We planned to pool studies where carer satisfaction was reported on a continuous scale by calculating the standardised mean difference (SMD) with standard deviation (SD).
Unit of analysis issues
We thought it unlikely that any studies would utilise a cluster or cross‐over design, so the unit of randomisation or allocation would have been the individual participant.
Dealing with missing data
If necessary we requested any further information from the study authors by written correspondence (e.g. emailing or writing (or both) to corresponding author(s)), however no additional relevant information was included in the review. We planned to perform a careful evaluation of important numerical data such as numbers of participants screened; numbers of randomised participants; and number of participants in the intention‐to‐treat (ITT), as‐treated, and per‐protocol (PP) populations; however, there were no randomised or quasi‐randomised studies. We investigated attrition rates, for example drop‐outs, losses to follow‐up, and withdrawals. We planned to appraise issues of missing data and imputation methods (e.g. last‐observation‐carried‐forward (LOCF)) critically (Higgins 2011); however, the level of reporting within studies was generally insufficient for us to assess this.
Assessment of reporting biases
We assessed the potential for publication bias by checking clinical trials registers for unpublished studies. If we had found more than 20 studies, we intended to use the funnel plot statistic and funnel plots to assess for the potential existence of small study bias (Higgins 2011).
Data synthesis
We planned to pool data using the random‐effects model as we expected heterogeneity of treatments and participants, but we also planned to use the fixed‐effect model to evaluate robustness of the model chosen and susceptibility to outliers.
Subgroup analysis and investigation of heterogeneity
We planned to undertake subgroup analyses according to:
-
the condition causing the need for palliative care (i.e. malignant versus non‐malignant);
-
drug class; and
-
main symptom being treated.
There were insufficient studies to permit any of these subgroup analyses.
We planned to assess heterogeneity using the I2 statistic, and make a decision to either:
-
combine studies using a random‐effects model (where the I2 statistic was low to moderate, and the studies appeared to be reasonably similar); or
-
not combine all studies (where either the I2 statistic was moderate and reasons for heterogeneity were plausible and, therefore, indicated that combining studies was not appropriate, or the I2 statistic was high).
When combining studies, we intended to undertake post hoc subgroup analyses if sufficient studies existed in each subgroup of the plausible heterogeneity factor.
In investigating heterogeneity of the included studies, we planned to consider treatment factors such as the continuation or weaning of sedation, degree of sedation achieved, timing and dosage of medication, and study design factors such as the length of follow‐up, type of proxy used for quality of life measurement, and time of outcome measurement.
Sensitivity analysis
Since we planned to include included quasi‐randomised and non‐randomised comparative studies, we planned to investigate the effect of omitting such studies on the results using sensitivity analysis.
Results
Description of studies
Results of the search
The searches resulted in 6685 citations to screen. After title and abstract screening, we reviewed 70 full‐text articles, resulting in 14 included studies and three awaiting classification (see Figure 1). We found no unpublished studies from clinical trials registers.
Study flow diagram.
Included studies
See: Characteristics of included studies table.
We included 14 comparative studies (Alonso‐Babarro 2010; Bulli 2007; Caraceni 2012; Chiu 2001; Fainsinger 1998; Kohara 2005; Maltoni 2009; Maltoni 2012b; Muller‐Busch 2003; Radha Krishna 2012; Rietjens 2008; Stone 1997; Sykes 2003; Vitetta 2005). All studies compared a group of people who received palliative sedation with a concurrent control group who did not receive sedation. All were consecutive case series, however only four had prospective data collection (Bulli 2007; Chiu 2001; Maltoni 2009; Maltoni 2012b). One of these three used matching to select the controls (Maltoni 2009). The other 10 studies were retrospective chart reviews. None was randomised or quasi‐randomised.
The 14 studies included 4167 adults, of whom 1137 received palliative sedation. The proportion of people in each study receiving palliative sedation ranged from 12% to 67%. In all studies, the proportion of people with a cancer diagnosis was greater than 95%. The setting of the studies was hospices (seven studies), palliative care units (five studies), hospital oncology wards (three studies), and home‐based palliative care (two studies). Three studies involved more than one setting; Bulli 2007 was set in both the home and hospice, Chiu 2001 in hospice and palliative care units, and Stone 1997 in a hospital ward and a hospice.
The most commonly used drug to achieve palliative sedation was midazolam, which all 14 studies used. Other drugs were haloperidol (eight studies) and chlorpromazine (five studies). A small proportion of people received only opioids (morphine, fentanyl, and methadone), or propofol, other benzodiazepines (lorazepam, diazepam, clonazepam, flunitrazepam, and levomepromazine/methotrimeprazine), antihistamines (promethazine and chlorphenamine), phenobarbital, scopolamine hydrobromide, or ketamine hydrochloride.
The mean duration of sedation from initiation to death ranged from 19 hours (Rietjens 2008) to 3.4 days (Kohara 2005) in the nine studies that reported duration of sedation, although the Sykes 2003 study had a small group of people who received palliative sedation for seven days prior to death in addition to their larger group who received sedation in the last 48 hours of life.
The 14 comparative studies had control groups of concurrent participants in the same care setting who did not receive palliative sedation. Only the study by Maltoni matched control participants for age group, gender, reason for hospice admission, and Karnofsky performance status (Maltoni 2009). Four studies stated a funding source being their institution or a government granting body (Alonso‐Babarro 2010; Caraceni 2012; Chiu 2001; Maltoni 2009), and three studies stated that there were no competing interests to declare (Maltoni 2012b; Muller‐Busch 2003; Rietjens 2008).
The following table describes the included studies.
| Study | Setting | Study design | Number (%) in sedation group | Number in non‐sedated group | Two most common indications for sedation | Most common sedative(s) used | Type of sedation at commencement | Mean duration of sedation |
| Home‐based care team | Retrospective cohort of consecutive cases | 29 (12%) | 236 | Delirium (62%), dyspnoea (14%) | Midazolam, levomepromazine | Dose titration to effective control of symptoms | 2.6 days (range 1‐10 days) | |
| 4 hospice and home‐based teams | Prospective cohort of consecutive cases | 136 (13%) | 939 | Not reported | Benzodiazepines, opioids, antipsychotics | Continuous, deep | 68% ≤ 1 day, 25% 2‐4 days, 6% 5‐10 days | |
| Palliative care team in tertiary care cancer hospital | Retrospective cohort of consecutive cases | 83 (64%) | 46 | Dyspnoea (37%), delirium (31%) | Benzodiazepine (48%), antipsychotic (45%), antipsychotic plus benzodiazepine (26%) | Not reported | Median 18 hours | |
| Hospice and Palliative care unit | Prospective cohort of consecutive cases | 70 (25%) | 206 | Delirium (57%), dyspnoea (23%) | Haloperidol (50%), midazolam (24%) | Intermittent (63%), continuous (37%) | Median 5 days | |
| Hospice | Retrospective cohort of consecutive cases | 23 (29%) | 53 | Delirium (87%), dyspnoea (4%) | Midazolam (91%), chlorpromazine and lorazepam (9%) | Continuous (61%), intermittent (30%) | 2.5 days (Median 1 day) | |
| Palliative care unit | Retrospective cohort of consecutive cases | 63 (51%) | 61 | Dyspnoea (63%), malaise/restlessness (40%) | Midazolam (98%), haloperidol (84%) | Continuous (69%), intermittent (30%) | 3.4 days | |
| 4 hospices | Prospective matched cohort | 267 | 251 | Uncontrolled symptoms (53%), terminal phase of life (41%) | Lorazepam (38%), chlorpromazine (38%) | Continuous (44%), intermittent (56%), deep (38%), mild (62%) | 4 days (SD 6) (Median 2 days) | |
| 2 palliative care units | Prospective cohort of consecutive cases | 72 (22%) | 255 | Delirium (61%), existential distress (38%) | Benzodiazepines (76%), antipsychotics (38%) | Continuous (92%), intermittent (6%) | 32.2 hours (range 25‐253) | |
| Palliative care unit | Retrospective cohort of consecutive cases | 80 (15%) | 468 | Pain (38%), dyspnoea (23%) | Midazolam | Titrated to symptom control, then intermittent if possible to control symptoms | Approx. 60 hours | |
| Oncology ward in tertiary care hospital | Retrospective cohort of consecutive cases | 68 (29%) | 170 | Anxiety (24%), dyspnoea (21%) | Midazolam, haloperidol | Titrated to symptom control | Not reported | |
| Acute palliative care unit | Retrospective cohort of consecutive cases | 68 (43%) | 89 | Terminal restlessness (62%), dyspnoea (47%) | Midazolam (75%), propofol (15%) | Not reported | Median 19 hours | |
| Hospital support team and hospice | Retrospective cohort of consecutive cases | 30 (26%) | 85 | Delirium (60%), mental anguish (27%) | Midazolam (80%), haloperidol (37%) | Not reported | 1.3 days | |
| Hospice | Retrospective cohort of consecutive cases | 80 (34%) | 157 | Not reported | Midazolam, methotrimeprazine | Not reported | Not reported | |
| Hospice | Retrospective cohort of consecutive cases | 68 (67%) | 34 | Not reported | Benzodiazepines, haloperidol | Titrated to symptom control, then intermittent | Not reported | |
| SD: standard deviation. | ||||||||
Excluded studies
See: Characteristics of excluded studies table.
We found 11 studies that described the use of palliative sedation in case series of people, but they made no comparison with a control group.
Risk of bias in included studies
See Figure 2 and Figure 3 for 'Risk of bias' summary graphs.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Allocation
All 14 studies were at high risk of selection bias, as none was randomised or quasi‐randomised.
Blinding
No studies allocated people to treatment group; blinding of treatment was not possible; and blinded assessment of outcomes was not carried out; however, the main outcome assessed in the studies was survival, which is an objective measure. Therefore, they are all at high risk of performance bias, but low risk of detection bias. One study reported satisfaction with treatment, with people responding to the questionnaire being aware of treatment group, therefore, this study was at high risk of bias for satisfaction with treatment (Chiu 2001). Five studies assessed symptom control, and were at high risk of bias for the outcome symptom control .
Incomplete outcome data
All studies were consecutive case series. Studies did not generally report on the proportion of missing data for outcomes. One study reported that 9% of survival time data were missing (Chiu 2001).
Selective reporting
It is unlikely that any of the outcomes of this review were measured but not reported.
Other potential sources of bias
Since these were all non‐randomised studies, we assessed two further areas for risk of bias. Six studies reported on the difference between the sedated and non‐sedated groups on baseline characteristics (Alonso‐Babarro 2010; Bulli 2007; Fainsinger 1998; Maltoni 2009; Muller‐Busch 2003; Rietjens 2008). The five unmatched studies reported significant differences between the groups, so are at high risk of bias in comparing the outcomes between the groups. Whilst Maltoni 2009 matched participants on several factors, there was a significant difference between the groups in symptoms at admission, with the sedated group having more uncontrolled symptoms, as expected. Therefore, this study was also at high risk of bias. The other eight studies did not report a comparison of groups at baseline, so were at unclear risk of bias.
We also assessed whether the groups were alike at baseline on the outcomes to be measured in the study. Since survival is not an outcome that can be measured at baseline, all studies were at unclear risk of bias for survival. Only five studies measured an outcome other than survival.
Effects of interventions
Primary outcomes
No studies measured quality of life or well‐being.
Secondary outcomes
Five studies reported on symptom control (Caraceni 2012; Chiu 2001; Fainsinger 1998; Muller‐Busch 2003; Rietjens 2008). One study reported symptom control as odds ratios for prevalence of each symptom in the last seven days of life (Caraceni 2012); one as mean scores (Chiu 2001); one as adequacy of control rated good, fair, or poor (Fainsinger 1998); and two as symptom prevalence (Muller‐Busch 2003; Rietjens 2008). Therefore, we were unable to pool results for this outcome.
Caraceni 2012 reported only the odds ratio for comparison of sedated and non‐sedated groups for the prevalence of symptoms during the seven days before death, but did not report the counts or percentages that these were based on. CIs around the odds ratios were wide. There was no statistically significant difference between the sedated and non‐sedated groups in the prevalence of confusion, gastrointestinal symptoms, pain, or psychological distress. The odds ratio for recurrent agitation was 3.5 (95% CI 1.4 to 8.8), for recurrent drowsiness was 0.3 (95% CI 0.2 to 0.7), and for recurrent dyspnoea was 4.2 (95% CI 1.9 to 9.2), indicating that the sedated group was more likely to experience recurrent agitation and dyspnoea, but less likely to experience recurrent drowsiness than the non‐sedated group.
Chiu 2001 measured pain, dyspnoea, and delirium in 70 people in the sedated group and 206 people in the non‐sedated group, and found that pain scores and dyspnoea scores measured two days before death were similar between groups. The mean score for pain (10‐point scale) was 2.5 in the sedated group and 2.1 in the non‐sedated group (P value = 0.27, t‐test). The mean score for dyspnoea (10‐point scale) was 3.0 in the sedated group and 2.9 in the non‐sedated group (P value = 0.78, t‐test). However, mean delirium score two days before death was significantly worse in the sedated group (1.8 in the sedated group compared with 1.1 in the non‐sedated group, 0 to 3 scale, P value < 0.001, t‐test).
Fainsinger 1998 measured adequacy of overall symptom control as good, fair, or poor, daily for the day of death and six days prior in 23 people in the sedated group and 53 people in the non‐sedated group. Symptom control was significantly worse in the sedated group on the day of death and the two days prior (P value < 0.001). The percentage of people in the sedated group with good control was 61% on the day of death, 35% on the day prior, and 38% two days prior, compared with 96% on the day of death, 88% on the day prior, and 87% two days prior in the non‐sedated group.
Muller‐Busch 2003 measured prevalence of pain, dyspnoea, delirium, and anxiety over the course of admission, and during the last 48 hours of life. They did not report between‐group results, but rather the change over time in each group for each symptom. Pain improved significantly in both groups between admission and the last 48 hours of life; however, all other symptoms worsened significantly.
Rietjens 2008 measured symptom prevalence at 0 to 24 hours prior to death, and 25 to 48 hours prior to death. Only participants commencing sedation during this period were reported for the sedated group (45 for the 0 to 24 hours period, and 13 for the 25 to 48 hours period). That is, it was a measure of symptom control shortly after commencing sedation. Pain, constipation, nausea/vomiting, and anxiety were not significantly different between the sedated and non‐sedated groups. However, in both periods, the percentage of people with dyspnoea was significantly higher in the sedated group at 50% of people compared with 31% in the non‐sedated group at 0 to 24 hours prior to death, and 69% in the sedated group compared with 38% in the non‐sedated group 25 to 48 hours prior death. The percentage of people with delirium at 0 to 24 hours prior to death was also significantly worse in the sedated group (29% in the sedated group compared with 13% in the non‐sedated group), but not at 25 to 48 hours (31% in the sedated group compared with 23% in the non‐sedated group).
No studies measured duration of symptom control or time to control of symptoms in using comparative methods.
Only one study reported on unintended adverse effects of sedation (Chiu 2001). It stated that there were no significant adverse events in the sedated group; however, four of 70 (6%) participants appeared to have drug‐induced delirium. It was also reported that no respiratory suppression was noted.
The least biased time comparison possible between intervention and control groups was from admission/referral to death in this set of observational studies. Whilst all studies except one measured time from admission or referral to death, some used mean time, and some used median time. Measures of variance were frequently missing or indicated skewed data distributions; therefore, we were unable to pool results. We have reported these results in tabular form below.
Table: survival time comparison between sedated and non‐sedated groups, from time of admission or referral
| Study | Measurement unit | Survival time in the sedated group | Survival time in the non‐sedated group | Comparison |
| Mean | 64 days (SD 60) | 63 days (SD 88) | P value = 0.963, t‐test | |
| Bulli 2007 | Median | 23 days | 23 days | NS, test not reported |
| Bulli 2007 | Median | 24 days | 17 days | ‐ |
| Mean | 28.5 days | 24.7 days | P value = 0.43, t‐test | |
| Mean | 9 days (SD 5) | 6 days (SD 7) | P value = 0.09, t‐test | |
| Mean | 28.9 days (SD 25.8) | 39.5 days (SD 43.7) | P value = 0.10, t‐test | |
| Median | 12 days | 9 days | P value = 0.95, log‐rank test | |
| Mean | 11 days (95% CI 9 to 11) | 9 days (95% CI 7 to 11) | P value = 0.51, log‐rank test | |
| Mean | 21.5 days (SD 20.3) | 21.1 days (SD 23.6) | NS, t‐test | |
| Median | 8 days (approx.)* | 8 days (approx.)* | P value = 0.78, log‐rank test | |
| Median | 8 days | 7 days | P value = 0.12, test not reported | |
| Mean | 18.6 days | 19.1 days | P value > 0.2, test not reported | |
| Mean | 14.3 days (95% CI 11.2 to 17.4)1 36.6 days (95% CI 31.5 to 41.7)2 | 14.2 days (95% CI 12.7 to 15.7) | P value = 0.23, t‐test P value < 0.001, t‐test | |
| Mean | 36.5 days (SD 66) | 17.0 days (SD 43) | P value = 0.12, t‐test | |
| CI: confidence interval; HR: hazard ratio; NS: not statistically significantly different; SD: standard deviation. * Times not reported; have been interpolated from survival curves. 1 people receiving palliative sedation for last 48 hours of life. 2 people receiving palliative sedation for last 7 days of life. 1 study gave results for the outcome of satisfaction with treatment, as reported by the medical staff, family, and participant (Chiu 2001). For the medical staff, satisfaction was rated as 71% yes, 20% fair, 9% no, 0% unavailable. For the family, satisfaction was rated as 67% yes, 20% fair, 4% no, 9% unavailable. For the participant, satisfaction was rated as 53% yes, 10% fair, 4% no, 33% unavailable. | ||||
Discussion
Summary of main results
No studies reported on the primary outcome of this review (i.e. quality of life or well‐being).
Four studies compared the sedated and non‐sedated groups for control of symptoms, and showed that despite sedation with the intent to control symptoms, delirium and dyspnoea were still troublesome symptoms in these people in the last few days of life, and were significantly worse in the sedated group. Control of other symptoms appeared to be similar in sedated and non‐sedated groups.
All studies except one compared survival time in the sedated and non‐sedated groups, and concluded that there was no statistically significant difference between the groups. This is important, as there has been extensive discussion in the literature about whether palliative sedation might shorten life, therefore leading to uncertainty by some physicians about whether to use this treatment for fear of the perception that they were performing a form of euthanasia (Billings 1996; De Graeff 2007; Rietjens 2006). The use of time from admission to death in comparative groups may be a weak measure of any potential effect of palliative sedation on shortening life. However, it is difficult to determine what the comparison would be in the group who did not receive sedation, and time from admission to death may be the only feasible comparative measure. Although we were unable to meta‐analyse this outcome, and CIs around the point estimates in individual studies were wide, there was consistency in this result over all studies, with 12 of the 13 studies having a longer survival time in the sedated group.
Overall completeness and applicability of evidence
No studies measured the primary outcome of this review, quality of life or well‐being, in a formal way. Many of the study reports discussed the 'settling' of symptoms in an anecdotal way; however, there were no quantitative reports.
Quality of the evidence
There were no randomised or quasi‐randomised trials, and it is unlikely that these will be done. Therefore, the best quality evidence will come from well‐designed observational studies. Only one study in this review attempted to reduce selection bias between the groups by matching groups on baseline characteristics (Maltoni 2009). However, it is likely that even if this is done, the groups will differ significantly in their level of symptom control, with people with more severe symptoms more likely to receive palliative sedation. Hence, even matching of controls cannot adjust for this confounder. It would be possible to adjust for this confounder (and others) statistically, but this was not done in any of the reported studies.
Potential biases in the review process
The search strategy for this review was wide; however, this was a difficult topic to search. It was possible that the search missed some studies, despite screening more than 5000 citations and screening the bibliographies of narrative reviews.
Agreements and disagreements with other studies or reviews
This is the first review, to our knowledge, that attempts to summarise only studies that have compared outcomes for sedated and non‐sedated people.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
