Changing Practices in the Use of Continuous Sedation at the End of Life: A Systematic Review of the Literature

Context. The use of continuous sedation until death (CSD) has been highly debated for many years. It is unknown how the use of CSD evolves over time. Reports suggest that there is an international increase in the use of CSD for terminally ill


Introduction
In the last phase of life, patients may suffer from severe symptoms. 1,2 Continuous sedation until death (CSD) is a last option for these patients when intolerable suffering cannot be relieved by regular symptom treatment. The use of CSD has been highly debated for many years. 3e5 The inability of patients during CSD to communicate in the last phase of their lives and the potential of CSD to hasten death are important issues in this debate. 6e8 In addition, the appropriateness of CSD for symptoms of nonphysical origin such as fear, anxiety, and psycho-existential distress is controversial, as determining these symptoms as refractory may be subjective and complex. 8e10 It is unknown how frequencies and reasons to start CSD evolved over time in clinical practice. Reports suggest that there is an increase in the use of CSD. 4,11,12 The aim of this review is two-fold. Our first aim is to explore if there is an increase in the use of CSD between 2000 and 2020. Our second aim is to provide insight into the indications to use CSD during this period. This insight is important as it will contribute to a better understanding of current practices in end-of-life care and inform further discussion on the use of CSD.

Definitions of Sedation
A variety of terms, concepts, and definitions are used in the literature to describe the use of sedation for the relief of intolerable suffering at the end of patients' lives. 7,13,14 Continuous sedation, terminal sedation, palliative sedation, deep sedation, end-of-life sedation, and sedation until death are among these terms. The type of sedation varies from intermittent to continuous until the end of life. The depth of sedation varies from superficial to deep. Despite efforts to achieve consensus in terms and definitions of sedation, there are still many inconsistencies in the literature. 15,16 The same holds for guidelines on the use of CSD. 17,18 These inconsistencies complicate the debate on the use of sedation. In this literature review, we focused on CSD.

Search Strategy
On the 15th of April 2020, we performed a literature search in PubMed, Embase, CINAHL, PsycInfo, and the Cochrane Library, using the preferred reporting items for systematic review and meta-analysis protocols criteria for this report. 19 The search included the following terms: continuous sedation, terminal sedation, palliative sedation, deep sedation, end-oflife sedation, sedation practice, and sedation until death. The complete search, listed in Appendix Table 1, was verified by our information specialist to ensure that the search was correct and complete.
The search was limited to articles in Dutch or English published between January 2000 until April 2020.

Study Selection
After defining the selection criteria with all authors, study selection was performed by M.T.H. and G.J.M.W.v.T. We used the online program Rayyan for the title and abstract screening, a Web application for systematic reviews. 20 We selected studies that reported frequencies of the use of continuous sedation, in English or Dutch language. Studies that described sedation as continuous, and until the end of life, or where the results of the article indicated that the sedation was given continuously, and until the end of life, were included. Articles describing other forms of sedation, articles without frequencies of continuous sedation, studies with less than 100 patients, and comments on articles were excluded. Conflicting judgments in article selection were resolved in discussions between M.T.H. and G.J.M.W.v.T.

Data Extraction
The following data were extracted: title, first author, year of publication, period of data collection, type of study, country, number of patients, number of deaths in the study, place of death, definition of sedation, number and percentage of use of CSD, specialty of the attending physician, whether a palliative care team was involved, patients' symptoms, details on the decision-making process, and characteristics of the sedation.

Synthesis
In our description of changes in the use of CSD over time, we distinguish nationwide studies from studies in subpopulations. The changes in characteristics of sedation and in patients' symptoms requiring sedation are described for all included studies.

Assessment of Methodological Quality
To assess the methodological quality of the reviewed studies, we used an adapted version of the Revised Cochrane risk-of-bias tool for nonrandomized trials (Robins I-tool), see Appendix Table 2. The quality of the reviewed studies was assessed independently by M.T.H. and G.J.M.W.v.T., and inconsistencies in total score of bias were discussed. The tool consists of 6 elements of the study in which bias could have occurred: 1. Bias in selection of participants of the study: The risk of bias was considered as low when a clear description of the selection of participants was given and when patients who received continuous sedation were selected via the same procedure as patients who did not receive continuous sedation.

Bias in classification of interventions:
The risk of bias was considered as low when a clear description of continuous sedation was provided, when sedation was described as continuous and until death, and when continuous sedation was clearly distinguished from intermittent sedation. 3. Bias due to missing data: The risk of bias was considered as low if there was a complete follow-up or a loss to follow-up unlikely to introduce bias. 4. Bias in measurement of outcomes: The risk of bias was considered as low when data were collected prospectively by trained staff (physicians, nurses, researchers). The risk of bias was considered as higher when data were collected retrospectively, obtained from a database, or by self-report. 5. Bias in selection of the reported results: The risk of bias was considered as low when reported results of the study were in line with the research question and when the methods section of the study was well described. 6. Bias due to confounding: The risk of bias was considered as low when confounders were taken into account, and when these confounders were described in the article.
For each element, the risk of bias was considered as low (1 point) or higher (2 points). A total score of #8 was considered as a low risk of bias. A total score of 9 or more was considered as a higher risk of bias. Figure 1 presents an overview of the selected articles. Initially, we found 8128 articles, and after removing duplicates, 4078 articles remained in our search. These articles were screened for eligibility based on title and abstract, which resulted in 160 articles being assessed based on the full text. Sixty-one articles were finally included in our review, 23 articles on 16 nationwide studies, and 38 articles on 37 studies in subpopulations. 21e30,31e45,46e55,56e65,66e75,76e81 Table 1 shows the country, study period, study type, the total of patients investigated, how many patients received sedation, how sedation was defined, and the study population per study. Appendix Table 3 shows the risk-of-bias assessment of the included studies. We considered 22 out of 23 articles on nationwide studies to have a low risk of bias. Most studies had a retrospective design. The questionnaire studies reported a high response rate, included a description of loss to follow-up, and accounted for confounders. Only 11 out of 37 articles on subpopulation studies were considered to have a low risk of bias. In the other studies, definitions of CSD were lacking, missing data were not always described, and when comparing between subgroups, confounders were not taken into account.

Frequencies of Continuous Sedation
We found 23 articles on 17 different nationwide studies that were performed in 7 countries: Belgium, Denmark, Italy, The Netherlands, Sweden, Switzerland, and the United Kingdom (Table 1). Table 2 shows characteristics of patients who received CSD in nationwide studies compared to all patients who died during the observed study period. CSD was more often applied in men than in women, in age groups below 80 years, and in patients with cancer and hospitalized; in four of the studies, these differences were statistically significant. 21,29,38,43 Frequencies of CSD were calculated in the articles by dividing the number of patients that received sedation by all deaths in the study. The frequency of CSD ranged between 3% in 2001 in Denmark and 18% in The Netherlands in 2015. 21,38 Figure 2 displays CSD frequencies by year in each country.
Apart from The Netherlands, where the use of CSD increased from 8% of all deaths in 2005 to 12% in 2010 to 18% in 2015, an increase was also observed in Switzerland, from 5% of all deaths in 2001 to 18% in 2013. 37e39 After an initial increase in Belgium from 8% of all deaths in 2001 to 14% in 2007, the percentage decreased in 2013 to 12%. 29 For Denmark, Sweden, the United Kingdom, and Italy, it was not possible to assess country-specific trends over time. The use of CSD increased in Switzerland, The Netherlands, and less clearly in Belgium between 2000 and 2020.
We found 38 studies that reported frequencies of CSD in subpopulations from 18 different countries (Table 1). Subpopulations were children, patients older than 80 years, cancer patients, patients with dementia, and patients with amyotrophic lateral sclerosis. CSD was delivered at home, in hospices, nursing homes, inpatient palliative care units, and hospitals. In most subpopulation studies, the percentage of CSD was calculated by dividing the number of patients who received CSD by all patients who died during the observed period. In three studies, the frequency of CSD was calculated by dividing the number of patients that received sedation by the number of all admitted patients. 51,65,66 In one study, the percentage of sedation was calculated by dividing the number of patients who received CSD by the consultations by a palliative care team. 54 Frequencies of CSD varied in these subpopulation studies from 1% in Japan between 2005 and 2011 in patients with cancer in a palliative care  Figure 3 shows the reported symptoms requiring sedation over time. Over the years, there was an increase in studies that reported patients' symptoms requiring sedation. The most frequently reported symptoms requiring sedation were dyspnea, agitation or delirium, and pain. Fatigue was mentioned only in four studies (all after 2010). Psycho-existential distress as an indication for sedation was mentioned only once in studies before 2008, and from 2008 and onwards, it was mentioned in 9 studies with percentages ranging from 0 to 32%. Fear as an indication for sedation was mentioned in six studies between 2001 and 2015, with percentages ranging from 0 to 27%. Thus, there is a clear trend for an increased use of CSD for nonphysical symptoms including fear, anxiety, and psycho-existential distress.    The presented nation-wide frequencies from Belgium were based on data collected in Flanders, the Dutch speaking area of the country, and the presented nation-wide frequencies from Switzerland were based on data collected on the German speaking part of the country. b These percentages are not compared to all deaths, but compared to all patients that had received CSD. c Place of death instead of attending physician.

Development of CSD in Clinical Practice
in Germany. 55 In Belgium, this number increased from 10% in 2007 to 15% in 2013. 29 During the same period, the percentage of CSD on requests of the family slightly increased in Belgium from 12% in 2007 to 14% of all deaths in 2013. 29 From 2010 to 2014, there was an increase of the documentation of discussion of continuous sedation with patients, their relatives, and the medical team in a UK hospice. 80 From 2010 to 2014, there was an increase in the number of patients that was aware of their death in an Italian hospice, from 17% to more than 30% in 2014. 63 In all countries, benzodiazepines were used for CSD in most cases, with or without other medication. In the repeated studies, the use of benzodiazepines for CSD increased over time. In Belgium, the use of benzodiazepines alone or in combination with opioids was 54% in 2007 and 57% in 2013. 29 The use of opioids as the only drug for CSD decreased from 31% to 17% of all cases during this period. 29 In The Netherlands, the use of benzodiazepines for CSD increased from 60% of all cases in 2000-2001 to 93% in 2015. 37 The use of morphine without a benzodiazepine for sedation decreased in The Netherlands from 15% in 2005 to 3% in 2015. Over the years, CSD was more frequently provided in the absence of artificial nutrition or hydration. The percentage of cases of CSD in which no artificial nutrition or hydration was provided varied from 33% in 2000 in Italy to 91% in The Netherlands in 2015. 30 29,37,43 In all studies, more than 85% of patients died within a week after starting sedation. In some cases, CSD had been performed with the intention or cointention to hasten a patient's death. In Belgium, the proportion of cases in which there had been a cointention of hastening death increased from 13% in 2007 to 15% in 2013, but this rise was not statistically significant. 29 In Italy in 2007 and in the United Kingdom in 2007-2008, the proportion of cases of CSD was higher when a palliative care team was involved or when the attending physician had followed palliative care training. 30,43 Discussion Our systematic literature review shows that CSD is used in many countries in different settings to relieve the suffering of dying patients and suggests an increase in the use of CSD in at least some countries. Nationwide frequencies of CSD ranged between 3% and 10% in the period between 2000 and 2006 and between 12% and 18% from 2006 until June 2019. 21,29,31,40 Country-specific trends in time could only be assessed for The Netherlands, Belgium, and Switzerland. In The Netherlands and Switzerland, frequencies rose over the period 2001-2015, but in Belgium, the frequency of CSD decreased between 2007 and 2013 after an earlier increase. 29,37,40 Frequencies of CSD in the different subpopulations varied too widely to observe patterns and to observe associations between subpopulations and the use of CSD. Where reported reasons to start CSD used to be mainly of physical origin, over the years, more studies reported nonphysical symptoms as indication for CSD such as fear, anxiety, or psycho-existential distress. Several studies showed an increased frequency of CSD on requests of patients and their families for CSD, which was notable from the beginning of 2000 and onwards. 55 Studies also showed that the use  of CSD was increasingly discussed with patients, their families, and in the medical team. Several hypotheses could explain why the use of CSD seems to increase over the years. First, the broader range of symptoms requiring sedation from only physical to also nonphysical symptoms may explain the increase. Our results showed that over the years, more studies reported nonphysical symptoms such as fear, anxiety, and psycho-existential distress as indication to start CSD. 31,69,70,80 Second, it could be possible that improved palliative care has increased awareness among health-care providers of the refractory symptoms and suffering of terminally ill patients. It could be possible that health-care providers have become more acquainted with the guidelines and that they are increasingly aware of CSD as an option to relieve suffering, resulting in a higher frequency of CSD. 82,83 Third, it could be possible that patients and their relatives are more aware of CSD as a relevant option at the end of life. Our review shows an increase of CSD at the request of the patient or the family. Over the years, several campaigns have been established to make people more aware of their needs and preferences for the last phase of their lives. 84,85 A consequence of these campaigns could be that people are more aware of CSD as an option to relieve suffering in the dying phase and that they are more likely to request for CSD when they suffer of intractable symptoms. 29,55

Strengths and Limitations
To our knowledge, this is the first review comparing frequencies and characteristics of CSD on an international level and in subpopulations over time. This review shows that patients' symptoms requiring CSD evolved over time from only physical symptoms to both physical and psycho-existential symptoms. A limitation of our study is that most subpopulation studies were considered to have a higher risk of bias: Oftentimes, definitions of CSD were lacking, missing data were not always described, and when comparing between subgroups, confounders were not taken into account. Consequently, the comparability of these included studies is limited. A second limitation is that we excluded articles written in other languages than Dutch or English in our review.

Conclusion
The frequency of CSD seems to increase over time, possibly because of the extension of indications for sedation, from only physical symptoms to also nonphysical symptoms. The use of CSD appears to have become an integrated part of end-of-life care in many different countries, and it might have lost its status of ''last resort.'' In-depth studies are needed to explore what the views, expectations, and experiences of health-care professionals, patients, and families are to better understand the changing practices and increase in the use of CSD to maintain CSD as a proportional answer to the relief of unbearable suffering of terminally ill patients.