Article Text

Anticholinergic drugs for death rattle in dying patients with cancer: multicentre prospective cohort study
  1. Takashi Yamaguchi1,
  2. Naosuke Yokomichi2,
  3. Takuhiro Yamaguchi3,
  4. Isseki Maeda4,
  5. Ryo Matsunuma5,
  6. Yukako Tanaka-Yagi5,
  7. Asami Akatani1,
  8. Kozue Suzuki6,
  9. Hiroyuki Kohara7,
  10. Tomohiko Taniyama8,
  11. Yosuke Matsuda9,
  12. Nobuhisa Nakajima10,
  13. Tatsuya Morita2,
  14. Satoru Tsuneto11 and
  15. Masanori Mori2
  1. 1 Department of Palliative Medicine, Kobe University Graduate School of Medicine School of Medicine, Kobe, Hyogo, Japan
  2. 2 Department of Palliative and Supportive Care, Seirei Mikatahara General Hospital, Hamamatsu, Shizuoka, Japan
  3. 3 Division of Biostatistics, Tohoku University Graduate School of Medicine, Sendai, Miyagi, Japan
  4. 4 Department of Palliative Care, Senri Chuo Hospital, Toyonaka, Osaka, Japan
  5. 5 Department of Palliative Care, Konan Medical Center, Kobe, Hyogo, Japan
  6. 6 Department of Palliative Care, Tokyo Metropolitan Cancer and Infectious Diseases Center Komagome Hospital, Bunkyo-ku, Tokyo, Japan
  7. 7 Department of Palliative Care, Hatsukaichi Memorial Hospital, Hatsukaichi, Hiroshima, Japan
  8. 8 Department of Oncology and Palliative Care, Mitsubishi Kyoto Hospital, Kyoto, Japan
  9. 9 Department of Palliative Care, St Luke's International University, Chuo-ku, Tokyo, Japan
  10. 10 Division of Community Medicine and Internal Medicine, University of the Ryukyus Hospital, Nishihara, Okinawa, Japan
  11. 11 Department of Palliative Medicine, Kyoto University, Kyoto, Japan
  1. Correspondence to Professor Takashi Yamaguchi, Department of Palliative Medicine, Kobe University Graduate School of Medicine School of Medicine, Kobe 650-0017, Hyogo, Japan; ikagoro{at}


Background This study aimed to investigate the effectiveness of anticholinergics (AC) for death rattle in dying patients with cancer.

Methods This is a prospective cohort study enrolled Terminally ill adult (20 years or older) patients with cancer who developed substantial death rattle (Back score ≥2) from 23 palliative care units in Japan. AC treatment for death rattle was prescribed according to primary physician’s decision. The primary outcome was the proportion of patients whose death rattle improved, which was defined as a Back score of ≤1. We compared the proportion of improved cases in patients treated with (AC group) and without (non-AC group) AC, controlling potential confounders by employing propensity score weighting.

Results Of the 1896 patients enrolled, we included 196 who developed a substantial death rattle. Of these, 81 received AC. 56.8% in the AC group and 35.4% in the non-AC group had an improved death rattle at 8 hours after baseline. In the weighted analysis, AC group showed significant improvements in death rattle, with an adjusted OR of 4.47 (95% CI 2.04 to 9.78; p=0.0024). All sensitivity analyses achieved essentially the same results. In the subgroup analysis, ACs were strongly associated with death rattle improvement in men, patients with lung cancer, and type 1 death rattle (adjusted OR 5.81, 8.38 and 9.32, respectively).

Conclusions In this propensity score-weighted analysis, ACs were associated with death rattle improvement in terminally ill patients with cancer who developed substantial death rattle.

Trial registration number UMIN-CTR (UMIN00002545).

  • End of life care
  • Hospice care
  • Terminal care

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.


  • Death rattle is often distressing for patients’ families and for healthcare providers caring for these patients.

  • Anticholinergics are often prescribed for death rattle in daily practice, despite insufficient evidence for its efficacy and effectiveness for death rattle.


  • The improvement of death rattle was seen significantly more in patients prescribed anticholinergics than those who did not.

  • This improvement was strongly associated with the subgroups of men, lung cancer and type 1 death rattle.


  • Clinical research on specific populations should be conducted to confirm the efficacy of anticholinergics for death rattle in terminally ill patients with cancer.


Death rattle is noisy ventilation due to accumulation of secretions in the pharynx and/or airways. death rattle typically occurs in the last few days of life,1 2 with a reported prevalence of 13%–92% in dying patients.3 Previous studies have reported that death rattle was often distressing for patients’ families4–7 and for healthcare providers caring for these patients.8 Thus, management of death rattle is an important issue in end-of-life care.

Although several randomised controlled trials (RCTs) have failed to show efficacy,9–11 anticholinergics (ACs) are often prescribed for death rattle in daily practice.12 There are several possible reasons. First, death rattle has been proposed to be classified into types 1 and 2.13 Type 1 predominantly occurs due to the accumulation of salivary secretions in the pharynx in the absence of effective swallowing reflexes due to decreased consciousness; this typically develops in the last days of life.14 Type 2 is predominantly the accumulation of bronchial secretions due to deterioration or weakness of cough, and patients can sometimes still be conscious with this type. ACs are generally considered to be more effective for type 1.15 However, previous studies have not clearly distinguished these two subtypes. Second, considering the pharmacological properties, ACs might decrease the production of saliva and not affect existing salivary accumulation.16 Therefore, ACs were thought to be ineffective for eliminating the existing accumulation of secretions in the pharynx and proposed to be used preemptively or after suctioning.17 However, previous studies did not review the influence of death rattle intensity or preceding suctioning on the effectiveness of ACs for death rattle. Third, the natural course of death rattle and the effectiveness of ACs in real-world practice have not been sufficiently investigated.

We aimed to investigate the effectiveness of AC for death rattle in real-world practice after controlling for potential confounders with propensity score (PS) weighting and investigate factors influencing the effectiveness of ACs.


This study was conducted as a part of the East Asian Collaborative Cross-Cultural Study to Elucidate the Dying Process (EASED), an international, multicentre, prospective cohort study on patients with advanced cancer at palliative care units (PCUs) in Japan, South Korea and Taiwan.18 Briefly, the EASED study consecutively enrolled adult patients with cancers admitted to 38 PCUs (23 in Japan, 11 in South Korea and 4 in Taiwan). We used the data from 23 Japanese PCUs in this analysis.

Setting and participants

We consecutively enrolled patients with cancer ≥18 years of age who were admitted to participating PCUs for the first time and had locally advanced or metastatic cancer (histological, cytological or clinical diagnosis). The exclusion criteria were as follows: (1) scheduled discharge within 1 week and (2) refusal of patients or their families to participate. The participants were enrolled from January 2017 to December 2017. For this analysis, we included patients who developed death rattle with a Back score ≥2 during their PCU stay.


We defined death rattle as audible sounds at the bedside produced by movement of secretions in the hypopharynx or the bronchial tree in association with respiration. The primary physicians typically visited patients at least twice daily and evaluated whether they had death rattle. Physicians directly ordered AC according to the clinical guidelines.19 Although these guidelines do not recommend routine use of AC for death rattle, it allows AC use as an option when death rattle is refractory to other measures. When physicians prescribed AC, the choice of the type and dose of AC were at the primary physician’s discretion. Suctioning for death rattle was performed at the discretion of the physician or nursing staff.


All measurements were evaluated by primary physicians within daily practice. The intensity of death rattle and treatments were recorded every 4 hours after substantial death rattle development (T0) until 24 hours after (T6) or the patient’s death, whichever came first.

Death rattle intensity

Death rattle intensity was evaluated with the Back score.13 The Back score consists of four categories: ‘inaudible’ (0), ‘audible only very close to the patient’ (1), ‘clearly audible at the end of the bed in a quiet room’ (2) and ‘clearly audible at about 6 m or at the door of the room’ (3). We defined substantial death rattle as a Back score of 2 or higher in this study.20 21

Death rattle treatment

We recorded whether ACs were prescribed, as well as the type of AC at each time point. We also recorded whether suctioning was performed at 4 hours ahead of each time point.

Patient characteristics

We collected patients’ baseline characteristics at admission, including age, sex, primary tumour site, metastatic lesions (ie, brain, liver and lung), and a history of heart, lung and neuromuscular disease. We also obtained the following data at T0: death rattle subtype, character of secretion (ie, serous or purulent), presence of crackles on lung auscultation, presence of fluid retention signs (eg, pleural effusion, ascites or peripheral oedema), hydration volume and consciousness level. The subtype of death rattle was classified as one of three categories (type 1, type 2 or mixed) based on clinical judgement by the primary physicians.14 Consciousness level was assessed using the modified Richmond Agitation and Sedation Scale (RASS), which measured the severity of agitation and sedation on a 10-point scale (+4: combative; +3: very agitated; +2: agitated; +1: restless; 0: alert and calm; −1: drowsy; −2: light sedation; −3: moderate sedation; −4: deep sedation and −5: unarousable).22 23 The date of death was recorded at the time of the patient’s death.

Statistical analysis

As the primary endpoint, we compared the percentages of improved patients (defined as a Back score ≤1) at 8 hours after baseline between patients treated with (AC group) and without (non-AC group) ACs. We defined patients in the AC group as those who started ACs between T0 and T4. The baseline time point of the non-AC group was T0, whereas that of the AC group was the time of starting ACs.

First, we constructed two models for PS (ie, the conditional probability of receiving AC) by selecting a set of confounders between treatment assignment (receiving AC) and outcome (death rattle improvement) based on previous studies’ results4 14 15 21 24 25 and clinical knowledge. Models 1 and 2 included 18 and 7 variables, respectively (online supplemental table 1). Model 2 was used when the regression model failed to converge with model 1.

Supplemental material

Next, under the missing at random assumption, we performed multiple imputation by chained equations to impute missing covariates.26 The variables included in the imputation models were the same variables as in the PS model. We generated 10 complete datasets for subsequent analyses. Missing outcome values were imputed with the last observation.

To account for confounding biases, the observed differences in baseline covariates between the two groups were adjusted by the inverse probability of treatment weighting (IPTW) method.27 28 With this method, we estimated the PS for each patient using a multivariate logistic regression with the set of confounders after imputation. The PSs from 10 imputed datasets were then pooled according to Rubin’s rule.29 Patients in the AC group were weighted by the average treatment effect weight (1/PS), whereas those in the non-AC group were weighted by 1/(1-PS).

Then, a univariate inverse probability weighted logistic regression model was used to estimate the IPTW-adjusted OR for death rattle improvement of the AC group vs the non-AC group.

We further performed exploratory subgroup analyses to investigate the IPTW-adjusted OR of the AC vs non-AC group according to the baseline covariates.

In addition, we explored the effect of suctioning on death rattle improvement before starting AC using an AC group cohort.30 Following multiple imputations of the missing values, the PS for receiving suctioning was estimated. Then, patients treated with and without suctioning were weighted and IPTW-adjusted OR for death rattle improvement of suctioning group vs non-suctioning group was calculated.

Lastly, we conducted six sensitivity analyses to assess the robustness of the results: (1) analysing patients with a baseline Back score of only 2 or more, (2) defining the AC group as those who started AC at T0 and T1 only, (3) analysing with listwise deletion of missing values, (4) fitting logistic regression with model 2 in calculating the PS, (5) fitting a traditional multivariate logistic regression model to estimate the OR of AC versus non-AC by adjusting the same covariates as in the primary analysis and (6) calculating the E-value, which represents the minimum strength of association that an unmeasured confounder would need to have with both the treatment and the outcome to fully explain the estimated treatment–outcome association.31

All statistical analyses were performed with R V.3.5.3 (R Core Team 2019, Vienna, Austria). All p values were two sided. A p<0.05 was considered significant.

Patient and public involvement

Patients and the public were not involved in setting the research question or outcome measures or in the writing of the results.


Patient characteristics

A total of 1896 patients were enrolled in the main study (figure 1). Of these, we analysed 196 (10.3%) who developed substantial death rattle (115 in the non-AC group and 81 in the AC group).

Figure 1

Patient selection flow chart per STROBE. AC, anticholinergic; STROBE, Strengthening the Reporting of Observational Studies in Epidemiology.

The missing covariate values imputed by multiple imputations were baseline Back score (1.0%), presence of suctioning (1.0%), secretion character (1.0%), presence of crackles (3.1%) and hydration volume (1.0%). 12.8% (25/196) of the patients did not have a Back score at 8 hours after baseline because they had died before then; these were imputed by the last observation values.

Patient characteristics after imputation are summarised in table 1. The mean age was 71.3 years; 38.8% were female. The most common primary tumour site was the gastrointestinal tract (40.8%). The modified RASS was −3 or less in 62.2%, and 29.1% had type 1 death rattle. The baseline Back score was 2 in 57.2% and 3 in 25.2%. 27% received 500 mL/day or more hydration. The median time from T0 to death was 1 day (IQR 1–3): 1 day (1–4) in the non-AC group and 1 day (1–3) in the AC group.

Table 1

Characteristics of patients who developed death rattle and covariate balance between those treated with or without ACs

In the AC group, ACs were started at T0 in 31 patients, T1 in 34, T2 in 8, T3 in 5 and T4 in 3. Scopolamine butylbromide was administered to 59 patients and scopolamine hydrobromide to 22.

Balance of covariates between the non-AC and AC groups

Compared with patients in the non-AC group, the AC group had significantly less history of heart or lung disease, asymptomatic ascites, and type 2 death rattle and higher symptomatic pleural effusion, prevalence of baseline Back score of 3, crackles and receiving ≥500 mL hydration. After PS weighting, standardised differences for all covariates were <0.1, except for liver metastasis (0.11), which indicated that the weighted population in the two groups was comparable (table 1).

Comparison of death rattle improvement

In both the AC and non-AC group, the mean Back score decreased over time (figure 2). In the unweighted analysis, the proportion of improved patients at 8 hours after baseline was 35.4% (40/113) in the non-AC group and 56.8% (46/81) in the AC group (unadjusted OR 2.40; 95% CI 1.34 to 4.30; p=0.034). In the weighted analyses, the adjusted OR was 4.47 (95% CI 2.04 to 9.78; p=0.00024; table 2).

Figure 2

Change of Back’s score in AC and non-AC groups. AC, anticholinergic.

Table 2

Association of anticholinergics on the severity of death rattle

Subgroup analysis

We performed a weighted subgroup analysis comparing the ORs of improved patients in the non-AC group versus the AC group according to the baseline covariates. No significant heterogeneity was found in any subgroup, whereas AC were strongly associated with death rattle improvement, especially in subgroups of men, lung cancer and type 1 death rattle (ORs 5.81, 8.38 and 9.32, respectively; figure 3).

Figure 3

Subgroup analysis. AC, anticholinergic drugs; NA, not available; RASS, modified Richmond Agitation and Sedation Scale.

Effect of suctioning on death rattle intensity

Of 81 patients in the AC group, 34 did not receive suction before starting AC (non-suctioning group), 46 received suction (suctioning group) and 1 had a missing value. The patient characteristics after imputation and balance between the weighted groups are shown in table 3. The percentage of improved patients at 8 hours after baseline was 67.6% in the non-suctioning group and 48.9% in the suctioning group (OR 0.48; 95% CI 0.19 to 1.22; p=0.13). In the weighted analysis, the adjusted OR was 0.53 (95% CI 0.19 to 1.51; p=0.24).

Table 3

Characteristics of patients who received anticholinergics and covariate balance between those treated with or without suctioning before starting anticholinergics

Sensitivity analyses

The percentage of improved patients at 8 hours after baseline in the AC group was significantly higher than the non-AC group in the following sensitivity analyses: (1) the cohort with a baseline Back score of ≥2 only (OR 3.60; 95% CI 1.28 to 10.11; p=0.016), (2) the cohort of those who started AC at T0 and T1 only (OR 3.10; 95% CI 1.64 to 5.87; p=0.00063), (3) the analysis with deletion of missing outcome value (OR 4.62; 95% CI 1.70 to 12.57; p=0.0031), (4) the analysis with PS model 2 (OR 3.39; 95% CI 1.79 to 6.41; p=0.00024) and (5) the multivariate logistic regression (OR 3.48; 95% CI 1.77 to 6.86; p=0.00041). We applied the E-value method that produced E=3.65 for the estimate (table 2).


To the best of our knowledge, this is the largest study investigating the effectiveness of AC for death rattle in real-world terminally ill patients with cancer. This study has several major findings. First, AC reduced death rattle more than the natural course in terminally ill patients with cancer receiving care in PCUs. The previous two placebo-controlled RCTs did not find efficacy of AC for death rattle.10 11 However, one of the studies, including only unconscious terminally ill patients with cancer, showed a tendency for AC superiority, despite it not reaching statistical significance.11 The other study was prematurely terminated due to futility in the interim analysis. However, most of the included patients in that study were terminally ill patients with non-cancer.24 Heart and lung disease tend to develop type 2 death rattle which is considered to be less responsive to ACs.15 Indeed, death rattle improvement after starting AC was observed more frequently in type 1 than type 2 or mixed cases in this study. Moreover, the previous study also included mild death rattle (Back score of 1), whereas this study included the patients only substantial death rattle (Back score of 2 or more), which might have influenced the result. Thus, ACs could have significant role in managing death rattle in terminally ill patients with cancer, selecting cases with type 1 death rattle and substantial intensity, after appropriate non-pharmacological care. Second, suctioning before starting AC and the severity of death rattle did not influence the effectiveness of ACs in this study. Recently, two RCTs showed the efficacy of prophylactic use of AC for the prevention of death rattle.32 33 However, approximately 40%–70% of the control group (placebo or observed) did not develop death rattle in these studies. Moreover, in present large-scale real-world study, the incidence of substantial death rattle was only 10.3% in PCUs. Thus, we are not sure whether it is appropriate to use AC prophylactically for all terminally ill patients with cancer. Furthermore, suctioning appears to be invasive or distressing for these patients,4 34 which could also distress patients’ families.5 According to the results of this study, AC might not be necessarily used prophylactically or started after suctioning in the management of death rattle in terminally ill patients with cancer. Instead, minimal and proper use of AC based on careful evaluation and selection of the patient in need might be more appropriate.

This study has several strengths. First, we included the largest scale of real-world patients to date, and the results were adjusted with IPTW to minimise the influence of potential confounders. Thus, the results of this study are reliable and broadly applicable to terminally patients with cancer in daily clinical practice. Second, although few previous studies had evaluated the subtype of death rattle, this study distinguished the subtypes and showed that ACs were more effective in type 1.

Despite these strengths, this study had limitations. First, due to its observational nature, causality between ACs and the intensity of death rattle could not be confirmed. Second, although the results of the E-value method produced moderately robust results, we cannot rule out unmeasured confounders affecting these results. Third, given that this was an observational study, the indications and dosages of ACs were not completely standardised despite following AC treatment according to clinical guidelines.19 Fourth, the Back score was a physician-reported outcome measure, which might be biased in this unblinded study. Thus, we should conduct a blinded RCT focusing on terminally ill patients with cancer with type 1 death rattle of substantial intensity to confirm the efficacy of ACs. Fifth, although we set the inception point as a Back score of ≥2, the baseline Back score was ≤1 in some patients, which might reflect the fact that the intensity of death rattle could quickly change. To minimise influence of this phenomenon, we conducted a sensitivity analysis excluding patients with a baseline Back score of 0–1, which demonstrated essentially the same results. Sixth, we identified missing values in the outcomes and covariates, mainly due to the patients’ death. Given that death rattle develops in the dying phase, missing data due to death are inevitable. We processed missing outcomes with the last observation carried forward in the primary analysis and deleted cases with missing values in a sensitivity analysis, which confirmed the consistency of the results. Seventh, patients in the AC group included those who started AC between T0 and T4, which could have led to a time bias. However, we do not believe that this seriously affected the results because the results of the sensitivity analysis including patients started ACs at T0 and T1 only were consistent with the main analysis. Eighth, misspecification of the PS model was possible. We attempted to address this by conducting sensitivity analyses with another PS model and multivariate logistic regression, which showed the consistency of the results. Lastly, our results might not be generalised to patients who are not admitted to PCUs.


ACs were associated with the improvement of death rattle in terminally ill patients with cancer in PCUs. We need to conduct RCTs on specific populations to confirm the efficacy of ACs and perform a larger real-world observational study to find the appropriate population for prescribing ACs in the future.

Ethics statements

Patient consent for publication

Ethics approval

This study was approved by the institutional review board of Seirei Mikatahara General Hospital (Research No. 16-22) and all participating institutions.In accordance with the ethical guidelines for human research of the Ministry of Health, Labor, and Welfare in Japan, informed consent from the patients was waived due to the observational nature of the study.


We are grateful to the cooperation of members of the East-Asian collaborative cross-cultural Study to Elucidate the Dying process (EASED) study Group.


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 TY, IM, TM, ST and MM were responsible for conception and design. RM, AA, YT-Y, KS, HK, TT, YM and NN were responsible for collection and assembly of data. TY, NY, TY and MM were responsible for data analysis and interpretation. All authors were responsible for manuscript writing and final approval of manuscript.

  • Funding This study was supported by JPPS KAKENHI (Grant number 21H02829) and a Grant-in-Aid from the Japan Hospice Palliative Care Foundation.

  • Disclaimer The sponsor did not involve the study design, the collection and management of the data, and manuscript writing.

  • Competing interests None declared.

  • 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.