Article Text

Original research
Continuous palliative sedation in terminally ill patients with cancer: a retrospective observational cohort study from a Chinese palliative care unit
  1. Fang Tan,
  2. Shan Chen,
  3. Lan Huang,
  4. Yang Chen,
  5. Yan Wu
  1. Department of Palliative Medicine, West China School of Public Health and West China Fourth Hospital, Sichuan University, Chengdu, Sichuan, China
  1. Correspondence to Dr Fang Tan; 171633168{at}


Objective This study aimed to describe a 4-year practice of continuous palliative sedation (CPS) in a palliative medicine ward of an academic hospital in China. To compare the survival time of patients with cancer with and without CPS during end-of-life care, we used the propensity score matching method and explored potential patient-related factors.

Design A retrospective observational cohort study.

Setting The palliative ward at a tertiary teaching hospital between January 2018 and 10 May 2022, in Chengdu, Sichuan, China.

Participants The palliative care unit had 1445 deaths. We excluded 283 patients who were sedated on admission due to mechanical ventilation or non-invasive ventilators, 122 patients who were sedated due to epilepsy and sleep disorders, 69 patients without cancer, 26 patients who were younger than 18 years, 435 patients with end-of-life intervention when the patients’ vital signs were unstable and 5 patients with unavailable medical records. Finally, we included 505 patients with cancer who met our requirements.

Main outcome measures The survival time and analysis of sedation potential factors between the two groups were compared.

Results The total prevalence of CPS was 39.7%. Patients who were sedated more commonly experienced delirium, dyspnoea, refractory existential or psychological distress, and pain. After propensity score matching, the median survival was 10 (IQR: 5–17.75) and 9 days (IQR: 4–16) with and without CPS, respectively. After matching, the two survival curves of the sedated and non-sedated groups were no different (HR 0.82; 95% CI 0.64 to 0.84; log-rank p=0.10).

Conclusions Developing countries also practise palliative sedation. Median survival was not different between patients who were and were not sedated.

  • palliative care
  • quality in health care
  • public health
  • cancer pain
  • adult oncology
  • adult palliative care

Data availability statement

Data are available upon reasonable request. Data are available upon reasonable request. Data are available on reasonable request by emailing Fang Tan (email address:

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • We conducted this in a special palliative care unit in China, and we have a relatively large and representative sample size.

  • We used the propensity score matching method to balance selection bias.

  • We used Kaplan-Meier chart plots to compare survival time in sedated and non-sedated groups and survival subgroup analysis with the matched Cox proportional hazards model.

  • Due to the shorter overall survival time, the survival time was calculated from the time of admission in non-sedated and sedated groups.

  • Due to the retrospective study, the specific sedation details and the time point at which sedation began were lacking.


In the last stage of life, patients with serious and fatal diseases often experience physically and psychologically intolerable pain.1–3 Pain cannot be reduced by traditional medical intervention. Thus, palliative sedation is considered in these patients.4 5 Palliative sedation therapy (PST) is defined as an intervention that aims to alleviate intolerable suffering resulting from one or a combination of symptoms.6–9 The European Association for Palliative Care (EAPC) defines PST as ‘palliative sedation in the context of palliative medicine is the monitored use of medications intended to induce a state of decreased or absent unconsciousness to relieve the burden of otherwise intractable suffering in a manner that is ethically acceptable to the patient, family and healthcare providers’.7 Furthermore, palliative sedation can be light or deep, as well as intermittent or continuous.

Currently, PST is considered one of the most important components of end-of-life care,8 9 but its ethics have been a topic of discussion.10 For example, how do we identify patients who need early palliative sedation? After analysing the most frequently used palliative sedation guidelines as reported by experts from eight European countries, Surges et al6 found that the criterion of refractoriness is still debatable. Subsequently, will the patient’s survival period be shortened after sedation?11 12 Only a few studies have investigated this question. Van Deijck et al13 reported patient-related determinants during palliative sedation. Moreover, many empirical studies showed that palliative sedation has no adverse effect and may even prolong survival time in patients with advanced cancer.14–18 However, most studies focused on areas with mature palliative care development, whereas only a few relevant studies focused on developing areas of palliative care. It is well-known that many factors, such as country, culture, subjects and local nursing environment, affect palliative sedation.19 Therefore, the practice of palliative sedation should be equally understood in different regions, especially in countries where palliative care is developing.

This study described the practice of continuous palliative sedation (CPS) in a large academic hospital in a Chinese city over a 4-year period, using the propensity score matching (PSM) method to compare the survival time of patients with cancer with and without palliative sedation during end-of-life care and exploring the potential patient-related factors.


Design, setting and participants

We conducted this retrospective cohort study in the palliative ward at a tertiary teaching hospital in Chengdu, Sichuan, China. West China Fourth Hospital of Sichuan University is a 600-bed capacity hospital. In 1995, the Department of Palliative Medicine was established, which was a pioneer institution of palliative medicine in China. It initially started with one doctor and two beds and developed into three-level linkage service model with a palliative care ward, outpatient service and home service (a charity project) after 27 years of continuous development. Palliative care services were provided to >1500 terminally ill patients annually. Currently, the unit has 40 beds, 15 doctors, 26 nurses, a social worker, a full-time driver and a number of mobile volunteers. The team includes aromatherapists, lymphoedema masseurs, wound therapists and psychological counsellors, as well as a multidisciplinary team. They cater for patients with cancer and those without advanced cancer, mainly from the intensive care unit, oncology department and general wards of local hospitals.

We retrospectively analysed the electronic medical records of those who died in the palliative ward between 1 January 2018 and 10 May 2022. We excluded patients without tumours and younger than 18 years, patients who were given normal sedation doses to relieve insomnia or anxiety, patients who were sedated by mechanical ventilation on admission (the accurate reason for the use of sedatives was unknown), patients who used end-of-life intervention when their vital signs were unstable (because there were some differences in survival time between different end-of-life interventions) and patients who used sedation intermittently for seizures.

Procedures for CPS

To relieve symptoms of intractable distress in terminally ill patients, we chose CPS as the deliberate and continuous use of sedative drugs.20 These include benzodiazepines (midazolam and diazepam), barbiturate phenobarbital and dopamine receptor blocker chlorpromazine. Furthermore, haloperidol ≥5 mg/day was used for sedation (lower doses were less likely to cause sedation).21

The West China Fourth Hospital of Sichuan University has been practising CPS in the palliative ward for a long time. With strict procedures and systems, the department has set up a palliative sedation medical team composed of the department director, head nurse, doctor in charge, nurse in charge and medical team leader. In the course of treatment, relevant regulations of China and the EAPC palliative sedation guidelines were strictly followed. When conventional medical intervention cannot alleviate intractable pain symptoms, the physician in charge suggests palliative sedation after a discussion with the medical team. The advantages and disadvantages of treatment must be actively communicated with patients and their families (the chief communicator), and informed consent of palliative sedation should be signed and implemented. Subsequently, the patients’ vital signs are closely monitored during treatment and recorded. To adjust the sedative medication, the patient’s level of consciousness is measured using the Ramsay Scale.22

In mainland China, doctors are still unable to accurately obtain the medical decision of most patients before their death due to the insufficient promotion of advance care planning in terminally ill patients. Family members play an important role in a patient’s medical decisions. Due to cultural differences, although patients or family members choose to receive CPS, some family members choose end-of-life intervention (also known as end-of-life medical decisions) when the patients’ vital signs are unstable in the palliative ward, which is generally limited to vasoactive medications or non-invasive ventilators (maintaining oxygen saturation to improve dyspnoea), but they select do not resuscitate (DNR) once the patients’ vital signs are unstable. This is somewhat different from the international consensus on palliative sedation, where care procedures and medications are generally discontinued if deemed incompatible or irrelevant to the patient’s comfort goals. The disease and prognosis are usually communicated to the patient or their family during admission to the palliative unit, including end-of-life DNR and advanced directive plans, but ultimately the patient or their family make a choice. Some families are hoping to keep the patient’s life as long as possible with pain relief. We consider that end-of-life interventions have large confounding factors, because different end-of-life interventions may lead to different survival times, and individuals may choose different drugs and rescue methods. So we excluded this subset of patients and only analysed them in the sensitivity analysis.

In our study, CPS was classified as light or deep sedation, with continuous light and deep sedation provided with a Ramsay Scale rating of 2–3 and ≥4, respectively. Sedation programmes can be either proportional or rapidly induced sedation. Proportional sedation means administration of a relatively low dose of midazolam (0.5–2 mg/hour) or chlorpromazine (1–5 mg/hour) in a continuous micropump, and the dose was titrated until the patient’s symptoms were relieved if the patient’s agitation persists or pain not alleviated. Moreover, rapid sedation induction refers to administering relatively high doses of sedative drugs. Midazolam was used as a sedative inducer, with 2–4 mg of rapid intravenous injection to bring patients into short-term sedation, whereas a certain dose of sedative drugs was administered to maintain the sedation.

Sample size

Generally, when studying multiple risk factors, the size of the sample size is approximately equal to the number of risk factors multiplied by 5–10.22 We studied 10 risk factors. Taking the maximum multiple as an example, the total sample size of the control and sedation groups was 100–200. After computing the loss of follow-up rate at 10%–20%, a maximum of 240 cases were required. Eventually, we included 505 subjects, which is larger than the number calculated. Therefore, there will be no error due to small sample size.

Data collection

Demographic data (sex and age at admission), date of admission, date of death and marital status were collected. The primary tumour site was determined using the International Classification of Diseases 10th Revision. The patient’s daily living ability was assessed using the Barthel index of activities of daily living (ADLs).23 Barthel index score was 0–100 points, with >60, 60~41, and ≤40 considered as good, medium, and poor, respectively. We divide ADLs into two levels, namely 0–40 and ≥41. In addition, we had collected the patients’ pain and opioid use, artificial nutrition, artificial hydration, as well as the total amount of liquid intake daily 3 days before death.

Specific sedation details of sedated patients, such as sedation types, sedation programme, the main cause of terminal sedation (dyspnoea, uncontrollable pain, delirium, etc), sedation agent used, with or without end-of-life interventions, and sedation dosage from beginning to end, and the Ramsay Scale score were also collected.

PSM analysis

In observational studies, the propensity score is defined as the conditional probability of receiving treatment and is usually used to adjust for selection bias.24 25 Since patients were not randomly assigned to non-sedated or sedated groups, we used the PSM method to control for selection bias. In total, 505 patients with cancer were derived using 1:1 greedy nearest neighbour matching in a propensity score range of 0.02. The groups (sedation and non-sedation) were used as dependent variables, and the predictors were sex, age, marriage, primary tumour site, pain and opioid use, ADLs, artificial nutrition, artificial hydration and total amount of liquid intake. This strategy resulted in 148 matched pairs in each group.


The main result was CPS. The patients who died were divided into those who received CPS and those who did not. Categorical variables were used to represent absolute numbers and percentages. Χ2 tests or Fisher’s tests were used to compare if they were non-hierarchical data. Continuous variables were represented as mean and SD and compared using an independent t-test (Mann-Whitney U test). Potential confounding factors were controlled using a 1:1 PSM method. Eventually, 148 patients each were matched to sedated and non-sedated patients.

Using Kaplan-Meier plots and log-rank tests, hospital survival times were compared for patients with and without palliative sedation and compared before and after PSM. The patient’s survival time in days was calculated as the date of death minus the start date of admission, whichever was less than 1 day. Because our patients with advanced cancers were often sicker, had a shorter survival time after admission, and had a shorter interval between admission and initiation of sedation, we used survival time from admission to the hospital until death.

In addition, we tested effect modifications based on patient characteristics (age, sex, ADLs, the primary tumour site and total amount of liquid) by the insertion of a first-order interaction term into the matched Cox proportional hazards model. Propensity score-matched HRs for each subgroup were visualised using plots with error bars showing 95% CIs. Finally, sensitivity analysis was performed to compare the statistical difference between the type of sedation, the reasons of sedation, sedative drugs, tumour primary site and end-of-life intervention in the sedation group in survival time from onset of sedation to death (online supplemental material 1).

Our study showed few patients with attributable missing data for independent variables, so we did not have attributed missing data, and missing data for independent variables were assumed to be random. SPSS V.26.0 and MedCalc V.20.0 (Ostend, Belgium) were used for all analyses. All statistical tests were bilateral, and p<0.05 was statistically significant. To ensure clear and comprehensive reporting of the best quality, the full text adopts the Enhanced Observational Epidemiological Research Reporting Initiative.26

Patient and public involvement

Patients or the public were not involved in the design, conduct, reporting or dissemination plans of our research.


From 1 January 2018 to 10 May 2022, 1445 patients died in the palliative care unit (figure 1). We excluded 283 patients who were sedated on admission due to mechanical ventilation or non-invasive ventilators, 122 patients sedated due to epilepsy and sleep disorders, 69 patients without cancer, 26 patients who were younger than <18 years, 435 patients with end-of-life intervention when the patients’ vital signs were unstable and 5 patients with unavailable medical records. The remaining 505 patients met our requirements figure 1.

Figure 1

Patient selection flow chart.

Table 1 shows the baseline characteristics of patients in CPS and non-CPS groups. Overall, we included 505 terminally ill patients with cancer, of which 51.5% were men and 63.8% were ≤70 years old. Before PSM, patients who received CPS were significantly younger, with different primary tumour sites, had pain and used opioid, had poor ADLs, had artificial nutrition and more total amount of liquid intake. After PSM, these factors were well balanced between the two groups, except for ADLs (p<0.001).

Table 1

Characteristics of terminally ill patients based on use or non-use of sedation

Table 2 shows the CPS group characteristics. The overall prevalence of CPS was 39.7% (n=573). The median duration of sedation was 5 days (IQR: 2.0–9.5). The most frequent indications for starting CPS were agitated delirium, dyspnoea, pain, and refractory existential or psychological distress. Continuous light sedation and proportional sedation were selected in 92.8% and 93.0% of patients, respectively. Two sedatives were administered to 47.1% of patients. Sedative doses of chlorpromazine and midazolam ranged from 37.5 to 300 and from 12.5 to 204 mg/day, respectively, except for one ultra-high dose (online supplemental material 2).

Table 2

Characteristics of continuous palliative sedation

In terms of survival time, the unmatched median survival was 12 (IQR: 6–20) and 9 days (IQR: 3–16.25) in the CPS and no sedation groups, respectively. After PSM, these values slightly changed, and the matched median survival was 10 (IQR: 5–17.75) and 9 days (IQR: 4–16) in the CPS and non-CPS groups, respectively. The two survival curves of the sedated and non-sedated groups were different before matching (HR 0.73; 95% CI 0.59 to 0.90; log-rank p=0.003)(figure 2A) and were not different after matching (HR 0.82; 95% CI 0.64 to 1.04; log-rank p=0.10) (figure 2B). We performed a matched subgroup analysis comparing the survival in the non-CPS group versus that in the CPS group according to the baseline covariates. We found no significant heterogeneity in any of the subgroups, while CPS was associated with a significant survival risk in patients with primary lung cancer (figure 3).

Figure 2

Kaplan-Meier plots of survival in patients with and without CPS. CPS, continuous palliative sedation.

Figure 3

Matched subgroup analyses of the association between CPS and survival time in terminal patients with cancer. ADL, activity of daily living; CPS, continuous palliative sedation.

In our sensitivity analysis, survival time after sedation between different sedatives (p=0.57), different types of sedation (p=0.15), with or without end-of-life intervention (p=0.18) and different primary tumour sites (p=0.29) were not significantly different. However, a significant difference was observed between different reasons for sedation (p=0.01) (online supplemental material 1).


This study described the practice of CPS in a specialised palliative care unit during a period of 4 years. After adjusting for patient characteristics using PSM, our study showed that median survival was not different between patients with advanced cancer who were and were not sedated. After PSM, age, sex, disease diagnosis, opioid use, artificial hydration and total hydration were not significantly different between the sedated and non-sedated groups.

Although international guidelines and frameworks were implemented for palliative sedation, the development of palliative care is different in different countries, and the cultural traditions, religious beliefs and values of each society make palliative sedation different. Particularly, many ethical issues are discussed repeatedly,6 such as the identification of refractory symptoms and hydration decisions. Moreover, previous studies14–16 27 28 found shorter survival after admission to the hospital or participation in a palliative care programme was not associated with CPS or more extensive palliative sedation.

Our study analysed the survival time of the sedated and non-sedated groups using the actual length of stay from admission to death. Before PSM in our study, the survival time of the sedated group was longer than that of the non-sedated group. To balance out selection bias, we used PSM; the matched median survival was 10 (IQR: 5–17.75) in the CPS and 9 days (IQR: 4–16) in non-sedation groups, and this was not significantly different after matching. We found no significant heterogeneity in any of the matched subgroups. It is worth noting that CPS was associated with a significant survival risk in patients with primary lung cancer. The potential interpretation is that patients with primary lung cancer might have a poorer organ function than other patients. In addition, patients with lung cancer are prone to breathing failure. However, these results were exploratory and need further confirmation. In our sensitivity analysis, the survival time after sedation was statistically different for different reasons for sedation. We divided patients with refractory physical suffering (agitated delirium, pain, dyspnoea and others) into one group and patients with refractory existential distress into one group, and we found that there was a longer survival time in patients with refractory existential distress. The potential explanation is that patients with refractory physical suffering have poorer organ function than those with refractory existential distress.

Other studies29 showed that sedated patients more commonly experience agitated delirium, pain, dyspnoea, and refractory existential or psychological distress, which was consistent with our study. More studies have assessed refractory existential distress although difficult.30 Our study showed that the sedation rate due to refractory existential distress was very high (21.8%). In our study, the rate of sedation use was 39.7%, consistent with previous reports (2.5%–54.2%).15 31 Moreover, the median survival time of our sedation was 2.0–9.5 days, which was longer than the previously reported median time of 0.9–2 days of palliative sedation use.16 32 This may be due to the limited measures we can take for patients with end-stage refractory physical or psychological suffering in China at present, such as lack of more psychological support. Most patients were treated with proportional sedation (93%) and continuous light sedation (92.8%). Continuous light sedation was selected because of the cautious working attitude of the doctors in the palliative ward due to the lack of death education in China and the misunderstanding of the family members, as well as the anxiety and restlessness after sedation. To the best of our knowledge, midazolam was the drug of choice in accordance with the guidelines,1 but we preferred using chlorpromazine for CPS,33 34 especially in patients with psychological or existential suffering. Delirium and dyspnoea were more commonly controlled using midazolam.

To date, an aggressive care plan can be implemented by understanding the determinants associated with the management of CPS, which can help physicians identify patients at risk of developing refractory symptoms. Seale35 showed significantly higher rates of reported continuous deep sedation among deaths from cancer, young patients and physicians working in ‘other hospital’ specialties. In a 2013 systematic review, van Deijck et al36 reported factors associated with CPS management. Furthermore, a prospective study in 2016 reported that using opioid alone at admission was significantly associated with CPS use,13 but these studies did not control confounding factors well. We also collected potential factors related to palliative sedation. Before PSM, univariate analysis showed that the patients were likely to use CPS if they were younger, with different primary tumour sites, had pain and used opioid, had poor ADLs, had artificial nutrition and more total amount of liquid intake. The difference in these factors may be due to selection bias, and no significant difference was found after PSM, except for lower ADL index. The patients admitted with lower ADL levels were more likely to be sedated. This may due to the patient with lower ADLs usually showing severe and more refractory symptoms, but further research is needed to explore this analysis.

In addition, the need for artificial nutrition and hydration after palliative sedation is controversial.37 38 Furthermore, more studies confirmed that the existence or lack of artificial hydration before death has no clear positive effect on patients’ health.15 Our study showed that total fluid intake daily 3 days before death had no significant difference in survival time between the sedated and non-sedated groups. The selection of administration of artificial nutrition and artificial hydration, as well as reducing or increasing total fluid intake, are more ethical considerations for doctors and family members.

To balance out selection bias, we have the advantage of using PSM. However, PSM does not allow duplicate matching, which causes us to lose part of the sample size; this might lead to bias. Beyond that, our study had some other limitations. First, in the survival analysis, we used the day of admission as the survival time, which led to bias. Because there is a risk of large differences in the general condition and organ function between the two groups, the actual duration of sedation often begins some time after admission. Therefore, we should find a suitable time node to evaluate the survival time of patients, whether sedated patients or non-sedated patients. It is more scientific and reasonable to evaluate the survival time at the same level. Yokomichie et al18 used the patient’s Palliative Performance Scale down to 20 to measure the impact of continued sedation on patient survival. This is a method worth learning for us; in the future, we can explore this in some prospective studies. Second, although we used PSM, the patient’s Barthel index of ADLs might be another important confounding factor. The evaluation of ADLs is subject to many factors. Because this was a retrospective study, there may be some bias among different assessors. Another reason is that due to the incompleteness of a few ADL data (n=8), we did not fill in the missing data. In addition, ADL mainly reflects patients’ daily living ability and is mostly used in elderly patients. It may be more accurate to use internationally recognised Kamofsky and Eastern Cooperative Oncology Group for the physical function changes of patients with tumour. Besides, our study is a single-centre study, so our findings may not be applied to other situations. Because of the retrospective study design, detailed information about the specific decision-making process of palliative sedation patients and their families, reasons for the interruption of sedation and complications after sedation was not obtained. Lastly, the generalisation of our findings was limited by the lack of standardisation of sedation indications, sedative drug selection and dosages.


In conclusion, our study suggests that palliative sedation is equally well practised in countries with developing palliative care. Similar to other studies, patients receiving palliative sedation more commonly experience delirium, dyspnoea, refractory pain, and psychological and existential distress. Additionally, median survival was not different between patients who were and were not sedated. Therefore, CPS can be used to relieve the patients’ end-stage intractable symptoms in specialised palliative care units.

Data availability statement

Data are available upon reasonable request. Data are available upon reasonable request. Data are available on reasonable request by emailing Fang Tan (email address:

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and was approved by the Ethics Committee of the West China Fourth Hospital of Sichuan University (approval number: HXSY-EC-2022036). We waived the requirement for patient informed consent because our study presented minimal risk to patients, and the data are anonymous.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors FT—study concept and design, data acquisition, quality control of data, statistical analysis, interpretation, and writing and editing of the manuscript. SC—quality control of data, statistical analysis, interpretation, and writing and editing of the manuscript. LH—project administration and organisation of the research team, study concept and design, data analysis and interpretation, and reviewing and editing of the manuscript. YW—study concept and design, data analysis and interpretation, and reviewing and editing of the manuscript. YC—data acquisition, data interpretation, and reviewing and editing of the manuscript. All authors read and approved the final manuscript. FT acts a guarantor.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.