FormalPara Take home message

During the withholding and withdrawing of life support in critically ill patients at the end-of-life, morphine and midazolam were the most common drugs used for pain management.

Introduction

In the USA, 20% of deaths occur in intensive care units (ICUs), and more than 50% of these deaths involve the withholding or withdrawing of life-sustaining interventions [1, 2]. Investigations of the characteristics of this population and the identification of shifts over time in clinical practice are allowing intensivists to better understand and manage patients at the end-of-life [2]. Several guidelines aiming to improve care for these patients have been developed; however, there is significant variation in the care critically ill patients receive at the end-of-life [2, 3].

In 2003, the Robert Wood Johnson Foundation Critical Care End-of-Life peer workgroup and 15 associated physician and nurse teams in North America established seven key elements for quality improvement at the end-of-life, and one of them was symptom management [4]. Pain is one of the most frequently reported symptoms in critically ill patients at the end-of-life, and perception of adequate pain control at the end-of-life reported by next of kin has been observed to range from 47 to 87% [5,6,7,8]. This suggests that at least half of these patients die with pain and indicates that pain management is still an area for significant improvement in ICUs and that prioritization of pain management is a marker of medical quality and compassionate care [9].

Pharmacotherapy is considered the cornerstone of palliation, and opioids are the most commonly used medications owing to their capacity to decrease pain, anxiety, and shortness of breath [10]. Concerns, however, arise about the possibility of hastening death due to overmedication. In these cases, the principle of double effect is taken into account, and the foreseen but unintended consequence of shortening the end-of-life process may be accepted as part of the physician’s provision of medication intended to relieve suffering [11]. Although ethically validated, the possibility of hastening death deters many clinicians from appropriately increasing doses of analgesics and sedatives at the end-of-life [12]. This concern may be counteracted by implementing multimodal analgesia (the combination of different medications with distinct mechanisms of action), which can minimize the required dosages and associated side effects of opioids [13]. Although using multimodal analgesia at the end-of-life may be beneficial, there is a lack of clarity in how to apply it in patients at the end-of-life, and the literature summarizing pharmacologic interventions for these patients in a systematic way is scarce to nonexistent.

To review and summarize the most frequent medications and dosages used during withholding and withdrawing of life-prolonging measures critically ill patients in the ICU, we conducted a comprehensive systematic review of the literature.

Methods

Protocol and registration

We followed the Cochrane methodology [14]. The protocol was published in the International Prospective Register of Systematic Reviews (PROSPERO), #CRD42018089487. Our report is based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [15].

Eligibility criteria

We included any original study evaluating pain management during the withholding and withdrawal of life support in adult patients at the end-of-life who were admitted to the ICU. We considered conference abstracts, if they provided sufficient information. We excluded narrative reviews, guidelines, and case reports.

Information sources

We searched electronic databases, including PubMed, EMBASE, Cochrane Database of Systematic Reviews, and the World Health Organization’s Biblioteca Virtual en Salud, from database inception through March 2019. No restrictions on language or date were used. Additionally, two reviewers manually searched the bibliography of every relevant citation to identify additional eligible articles otherwise not found.

Search

A professional librarian built a sensitive and specific search strategy based on the terms defined by an anesthesiologist specializing in pain management and two specialists in critical care medicine. The terms used were “pain management” and its synonyms plus specific pharmacologic interventions, combined with the term “end-of-life” and its synonyms, plus “intensive care” and “critical care unit” and their synonyms, with all terms exploded (see Appendix).

Study selection

Two independent investigators (KO and DPD) screened the studies. When disagreements arose on including a specific article, a third investigator (AL or AD) helped to make the final decision.

Data collection process

Two investigators (AL and AD) independently abstracted information from the included studies using a standard table. Then, the results were compared and a third investigator (JC) resolved disagreements.

Data items

Our primary outcomes were the utilization rates of opioids and analgesics used in critically ill patients in the ICU after withholding or withdrawing life support. Secondary outcomes included doses used in patients at the end-of-life (prior to withholding or withdrawal of life support at 24, 12, and 4 h prior to death and just before death). We extracted information regarding the title, author, country, number of centers involved, objectives, tools implemented, outcomes, and conclusions.

Risk of bias in individual studies

Two independent reviewers assessed the risk of bias in the included studies using the Newcastle–Ottawa Scale for observational studies. The scale evaluates three domains of bias: selection of participants (i.e., representativeness of the cohorts, ascertainment of exposure, and outcome of interest not being present at the start of study), comparability (to accounting for one or more factors (confounders) that predict the outcome of interest), and measurement of outcomes (i.e., methods for outcome assessment, appropriateness of the length of time to assess the outcome, and losses of follow-up do not compromise the validity). This scale consists of a grading system with a maximum score of 9. A score of ≥ 7 points indicates that a study is of high quality [16].

Summary measures

For our primary outcomes, we calculated the utilization rates for each type of medication (either opioids or analgesics). Our denominator was considered as the number of critically ill patients undergoing a specific type of pain management during the withholding and withdrawal of life support at the end-of-life.

Synthesis of results

Data were pooled if sufficient data existed for the primary outcomes (at least two studies reporting utilization rates for the same drug). We used a random-effects model to calculate a combined utilization rate and a 95% confidence interval. We used the Freeman–Tukey arcsine transformation to stabilize variances and conducted a meta-analysis using inverse variance weights [17, 18]. This methodology has been reported in multiple meta-analysis of binomial data [19,20,21,22]. We used a Chi-square test to assess statistical heterogeneity, and we also calculated the percent of the variability in effect estimates that is due to heterogeneity rather than sampling error (I2 statistic). A Chi-square test with p < 0.10 or I-squared statistic > 50% indicated statistically significant heterogeneity [14]. All analyses were performed using STATA 15 (StataCorp LP, College Station, TX, USA). The strength of the evidence resulting from our meta-analysis was assessed using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach [23]. Findings were classified into high-, moderate-, low-, and very-low-quality evidence according to our confidence in the estimate of effect. Our confidence in the evidence decreased for several reasons, including study limitations, inconsistency of results, indirectness of evidence, imprecision, and reporting bias.

For the secondary outcomes, we reviewed and summarized the evidence using only a narrative synthesis with tabular analysis.

Additional analyses

We planned performing a meta-regression to explore the influence of study characteristics (i.e., eligibility criteria used study design and type of funding) on the utilization rates of opioid and analgesics. Similarly, we planned to perform a funnel plot and a regression asymmetry test to assess small-study bias for the meta-analysis. However, due to the small amount of studies providing utilization rates per drug (< 10 studies) this was not possible.

Results

Study selection

We found 6981 unique citations; of these, 112 were retrieved as full text and 13 studies were ultimately included in our analysis [8, 24,25,26,27,28,29,30,31,32,33,34,35] (Fig. 1; Table 1). For one of the included studies, results regarding pain management or pain control were reported in a post hoc analysis [27, 36].

Fig. 1
figure 1

PRISMA flowchart

Full size image
Table 1 Characteristics of the reviewed articles (13 studies)a
Full size table

Study characteristics

Five studies were conducted in the USA [24, 27, 28, 31, 33], four in Canada [8, 29, 30, 32], and four in The Netherlands [25, 26, 34, 35]. All were observational studies (Table 1). Eight studies did not report a funding source, although two received financial support from foundations, one from the National Institutes of Health and one from a critical care trauma center; one did not have financial support. The percentage of male patients in the studies ranged from 43 to 60%. The mean age ranged from 59 to 71 years. A range of 59–100% of included patients had mechanical ventilation, and life support was actively withdrawn in 47–100% of patients within the included studies.

Risk of bias within studies

Of the 13 studies, 12 had a total Newcastle–Ottawa Scale score < 7 (Supplementary Table 1). Eleven studies evaluated a single cohort of patients without analyzing or comparing with unexposed groups. In eight studies, the population was representative or somewhat representative of the average population. In all studies, the assessment of outcome was confirmed with medical records or linked through a database and the follow-up of the patients was considered adequate.

Results of the individual studies

Pain management medications used

Eleven studies provided data for our primary outcomes. The utilization rates for each analgesic or adjunctive sedative drug in critically ill patients in the ICU after withholding or withdrawing life support are presented in Figs. 2 and 3, respectively. Eight studies reported the use of opioids without specifying the drug used. The pooled utilization rate of an opioid (unspecified) in critically ill patients in the ICU after withholding or withdrawing life support was 82% (75–88%). For studies reporting the utilization of specific opioids, morphine was used in 60% of the patients receiving pain management medications (95% CI 48–71%). Other opioids used were sufentanil (58%; 95% CI 52–65%), fentanyl (19%; 95% CI 2–48%), and hydromorphone and meperidine (≤ 2% of the patients received it).

Fig. 2
figure 2

Frequency of analgesic drugs used during end-of-life care in the intensive care unit (horizontal bars indicate 95% confidence intervals; black squares [ES; proportion used], effect estimate referred as utilization rate; size of the squares, weight of the effect; red diamonds, pooled percentage used for each type of drug)

Full size image
Fig. 3
figure 3

Frequency of sedative drugs used during end-of-life care in the intensive care unit (horizontal bars indicate 95% confidence intervals; black squares [ES; proportion used], effect estimate referred as utilization rate; size of the squares, weight of the effect; red diamonds, pooled percentage used for each type of drug)

Full size image

Similarly, six studies reported the use of an adjunctive sedative without specifying the drug used. The pooled utilization rate of a sedative (unspecified) in critically ill patients in the ICU after withholding or withdrawing life support was 62% (48–75%). For studies reporting the utilization of specific sedatives, midazolam was used in 28% of the patients receiving pain management medications (95% CI 23–32%). Other adjunctive medications included propofol (24%; 95% CI 12–39%), diazepam (16%; 95% CI 3–36%), and lorazepam (14%, 95% CI 8–21%).

Drug doses for pain management

Five studies compared morphine dosing before the withdrawal of life support with dosing at the moment before death [24, 26, 32, 33, 35]. All five showed a consistent dose increase during the end-of-life process (Table 2). One of these studies reported that in opioid-naïve patients, there was also an increase in mean dose of morphine provided before and after terminal extubation (from 5.2 to 10.6 mg/h, p = 0.001) [33]. The pooled mean increase in dose was 2.6 mg/h (95% confidence interval 1.2–4) [26, 33]. Three studies also compared midazolam doses [26, 32, 35], two compared doses of fentanyl and lorazepam [32, 35], and two compared doses of propofol [26, 35]; one study compared benzodiazepine dose changes in diazepam equivalents [24].

Table 2 Differences in opioid and sedative dosing between the moment prior to withholding or withdrawal of life support (WWLS) and the moment before death
Full size table

One study showed a significant increase in the average hourly doses of opioids and benzodiazepines used from the 24 h preceding death to the hour before ventilator withdrawal [32]. Furthermore, the average doses used in the 2 h before death were higher than the average doses used during the previous 22 h (p < 0.001 for both opioids and benzodiazepines). In another study in patients with withdrawal of life support, morphine doses increased over the 12-h period, particularly in the final 4 h of life [30]. That study also showed a statistically significant difference in the dosages of morphine and lorazepam between patients who died undergoing withdrawal of life support and those who died without limitations of life support (p < 0.007 and p < 0.006, respectively). Doses of other medications did not statistically differ between these two groups [30]. Rocker et al. found that dosages used did not differ among three time periods before death (4 h, 4–8 h, and 12 h), and the cumulative dose in the last 4 h did not differ between modes of withdrawal of life support (e.g., extubation, T-piece or trach hood, weaning mechanical ventilation, full ventilator support) [8].

Discussion

With this systematic review of the literature, we were able to identify published pharmacologic practices of pain management during the withholding or withdrawal of life support in the ICU. We found that opioids were the most common analgesic used (up to 88%), with morphine the most common, followed by fentanyl. Adjunctive benzodiazepines were used in 45–82% of patients; midazolam was the most common sedative used in most of the studies. Regarding medication dosages, we found that the mean dosing of analgesics and sedatives both before and after withholding or withdrawal of life support was consistent with the ranges used for withdrawal of life support in previous studies [24, 29].

Doses reported in many of the reviewed documents were higher than recommended dosages in guidelines for critically ill patients at the end-of-life [3]. This could be explained by the development of opioid tolerance in a considerable proportion of critically ill patients who are at the end-of-life, as evidenced by one study in which more than half of patients were already receiving opioids prior to withholding or withdrawing of life support [33]. We identified a progressive increase in drug dosing preceding the withdrawal of life support, which may reflect adjustments needed to alleviate discomfort during the end-of-life, despite unease about shortening the process [11]. This concern has been questioned by some recent studies showing prolongation of or no effect on time until death with the use of these medications [24, 26, 32, 33, 37], and a Cochrane systematic review found no statistically significant difference in the duration of the end-of-life process between sedated and non-sedated patients in the hospice palliative care setting [38].

Although there is homogeneity among end-of-life critical care guidelines, our systematic review showed a wide variation in practice and medication use, which could be attributed to the individualization of care recommended by guidelines, underuse of standardized protocols, or lack of guideline specificity in the recommendations [3, 39,40,41]. Of note, among the more recent studies included in this review, we observed that the frequency of use was lower for morphine and higher for fentanyl compared to the earlier studies included. This observation may be attributable to fentanyl’s easier titration, faster onset of action, and more favorable side effects profile [42]. Sedatives were employed during end-of-life care in the majority of included studies, suggesting the appropriate association of sedation to analgesia, a key pillar of palliative care medicine [43, 44].

Early cooperation with palliative care services could improve the final care these patients receive; additionally, the implementation of multimodal analgesia may also be beneficial at the end-of-life [2, 3, 45,46,47]. While the latter is favorable for patients in other circumstances, such as perioperative and trauma cases, we found a lack of evidence investigating this therapeutic approach during the withholding and withdrawal of life support [48, 49]. The use of non-opioid medications can reduce opioid requirements and their side effects, such as pruritus, nausea, and constipation, which are undesirable during the end-of-life process. Several case reports have shown positive effects of medications such as ketamine, dexmedetomidine, and regional anesthesia at the end-of-life in the ICU [50,51,52], but the lack of clinical studies and practice parameters limits the applicability of this approach in patients at the end-of-life, and thus, further investigation should be prioritized.

Our review has some limitations. First, citations may have been disregarded by our search strategy, mainly due to the variability of terms used in the literature to describe the end-of-life processes in the ICU. However, to overcome this limitation, all terms were reviewed and complemented by two anesthesiologists and a research librarian who performed reference and hand searches before the final database search was started. A second limitation was the heterogeneity of the reviewed studies and their relatively small sample sizes; all of them were non-controlled, observational studies. The lack of randomized controlled trials may be explained by the difficulty in measuring outcomes of pain management in this population. In addition, due to the small number of studies reporting data on the utilization rates of opioids or analgesics, we were not able to explore the influence of potential confounders (e.g., eligibility criteria used, study design, type of funding, etc.) on the pooled percent of patients using each drug. Therefore, our results should be interpreted with caution. Further research studies are needed and likely may have an important impact on our confidence in the reported pooled utilization rates of the drugs.

In summary, the current systematic review outlines the frequency and dosing of medications used for pain management during the withholding and withdrawal of life support in the ICU. We found that pain control centered on opioid and adjunctive benzodiazepine use with dosages exceeding those recommended by guidelines and that while current guidelines are homogenous, there is significant heterogeneity among practices in this setting.