Comparative Safety of Medications for Severe Agitation: A Geriatric Emergency Department Guidelines 2.0 Systematic Review

ABSTRACT

Background

Managing undifferentiated, severe agitation in older adults may require antipsychotic or sedative medications to prevent harm to self or others. Unfortunately, these medications are associated with serious adverse events in older adults, and little is known about their comparative safety.

Methods

We conducted a systematic review to identify comparative effectiveness studies on the safety of medications used in the treatment of severe agitation among older adults in the prehospital or emergency department (ED) setting. We searched eight databases including PubMed, EMBASE, SCOPUS, Cochrane library, CINAHL, Proquest Central, Ageline, and PsycInfo published in or before February 2024. Studies were included if they examined 1st generation antipsychotics, 2nd generation antipsychotics, benzodiazepines, or ketamine. Data were extracted on adverse respiratory events (apnea, hypoxemia, intubation) and other adverse events (arrhythmia, hypotension, worsening delirium, cardiac arrest, and mortality). We report the aggregate occurrence of any adverse events pooled by drug and report odds ratios (ORs) using haloperidol as the reference group.

Results

Among 8600 studies identified, eight observational studies and one randomized clinical trial met eligibility for further qualitative and quantitative analysis. The observational studies included 838 older adults receiving haloperidol (n = 117), droperidol (n = 129), lorazepam (n = 350), midazolam (n = 68), olanzapine (n = 101), quetiapine (n = 56), and ziprasidone (n = 17). Any adverse events were observed in 16.8% of the patients (141/838). Adverse events were most common among patients receiving midazolam (53%; 36/68). Relative to haloperidol, midazolam significantly increased the risk for any adverse events (OR 5.25 [95% CI: 2.64–10.45]). Quetiapine was the only drug observed to have a lower frequency of adverse events (OR 0.27 [95% CI: 0.08, 0.97]).

Conclusions

Adverse drug events are common among older adults receiving antipsychotic or anxiolytic medications for severe agitation. Benzodiazepines, particularly midazolam, pose an excessive risk to older adults requiring pharmacologic treatment for severe agitation.

Summary.

• Key points

  • ○
    Utilize nonpharmacologic approaches in the care of older adults with severe agitation as serious adverse events occurred in 16.8% of older adults who were administered antipsychotic or anxiolytic medications, though rates and types of events varied widely across drugs.
  • ○
    Midazolam posed an elevated risk of serious adverse events without any gains in clinical efficacy.
  • ○
    Oral medications, particularly quetiapine, had the best safety profile among the reported medications.
  • ○
    We observed an increased risk of adverse events with escalating doses of lorazepam. A comparable and clinically relevant signal in dose–response, though lacking statistical difference, was also observed for droperidol and quetiapine.
  • ○
    Intravenous and intramuscular use of antipsychotic or anxiolytic medications increased the risk for adverse drug events relative to oral administration.
• Why does this paper matter?

  • ○
    Emergency clinicians should be aware of excessive risk in using antipsychotic or anxiolytic medications to treat severe agitation in older adults. Where possible, emergency clinicians should avoid benzodiazepines, favor oral medications, and use the lowest effective dose.

1. Introduction

In emergency medicine, antipsychotic or anxiolytic medications are administered to patients with undifferentiated, severe agitation in an effort to prevent harm to patients or others [1]. Although cognitive function and coordination may be impacted by these medications, the goal of the administration of such medications is to achieve a level of anxiolysis, often referred to as minimal sedation, such that patients are able to remain safely engaged in and participatory with their care [2].

Severe, undifferentiated agitation in older adults presents unique challenges in the emergency department (ED). Prehospital and ED providers are often expected to achieve rapid calming of a patient, either to ensure patient and staff safety or facilitate time‐sensitive clinical evaluation, commonly in the absence of information to guide treatment selection. The diagnostic differential—including hyperactive delirium, intoxication, responsive behavioral/psychiatric symptoms of dementia, or psychosis—impacts the preferred pharmacologic approach [3, 4]. Further, nonpharmacologic approaches, such as person‐centered communication or sensory therapy [4, 5], are always preferred as both antipsychotics and anxiolytics pose significant risks. These medications have been linked to short‐term adverse events including respiratory depression [6, 7], dysrhythmia [8, 9], hypotension [3, 10], and exacerbation of delirium [4, 7]. Unfortunately, nonpharmacologic approaches do fail in the management of severe agitation, thus necessitating the use of high‐risk medications. Such use of antipsychotics and anxiolytics has been observed in as many as 1 out of 20 ED visits by an older adult [11].

Despite our understanding of the harms and benefits of antipsychotic or anxiolytic medications, clinical trials on the management of severe, undifferentiated agitation, particularly in the ED, have largely omitted older adults [12]. Hence, we systematically reviewed the comparative safety of using medications to facilitate calming and anxiolysis in older adults with severe agitation. Our objective was to identify which medications were associated with the lowest rates of adverse drug events in the management of severe, undifferentiated agitation among older adults in the prehospital or ED setting.

2. Methods

This study was conducted under the purview of the Geriatric ED Guideline (GEDG) 2.0 Committee [13, 14, 15, 16]. We conducted and report our review in accordance with the Preferred Reporting Items for Systematic Review and Meta‐Analyses (PRISMA) 2020 Guidelines [17].

2.1. Eligibility Criteria and Search Strategy

In defining our eligibility criteria, our team adapted the primary study question (“which antipsychotic or anxiolytic medications are associated with least harm in treating older adults with severe, undifferentiated agitation?”) using the PICO framework. Our PICO framework was:

• Population: Older adults (≥ 65 years old) presenting, in the prehospital or ED setting, with severe, undifferentiated agitation requiring pharmacologic intervention.

• Intervention: Use of medications including 1st generation antipsychotics, 2nd generation antipsychotics, benzodiazepines, or ketamine.

• Comparator: Active pharmacologic comparator such as alternative medication or dosing/formulation strategy.

• Outcome: Probable adverse drug events observed within the hospital encounter of interest include respiratory depression (defined as apnea, airway obstruction, oxygen desaturation < 90%, or intubation), prolonged QTc (> 500), arrhythmia, hypotension, dystonia, in‐hospital fall, delirium, cardiac arrest, and mortality.

Notably, our comparator group did not include placebo‐controlled trials due to ethical concerns in managing a behavioral emergency with a placebo [18]. Feedback for the PICO framework was provided by all participants in the GEDG writing collaborative, including > 50 experts in geriatric emergency medicine.

Consistent with our PICO framework, we identified and modified previously implemented search terms to find studies that discussed older patients [19], agitation [20], prehospital or ED setting [21], and antipsychotic or anxiolytic medications [22]. The search terms were applied to eight databases—including PubMed, Scopus, Embase, AgeLine, CINAHL, PsycINFO, Cochrane Database of Systematic Reviews, and ProQuest Central—in July 2022 and updated in February 2024. Search terms are provided in Figure S1.

To ensure an adequate sample of subjects, we included any randomized clinical trial or observational study in which we could extract relevant outcomes for the subgroup of individuals over the age of 65 years. We excluded studies that compared the use of antihistamines, given strong evidence that these agents should be avoided due to their anticholinergic properties [7, 23]. We further excluded studies from the psychiatric or inpatient setting, as these generally involved the sedation of patients with an established clinical phenotype rather than undifferentiated agitation. Case reports/series and review articles were excluded.

2.2. Study Selection Process and Data Extraction

After removal of duplicate studies, citations were exported and managed in Covidence [24]. Two study authors reviewed the title and abstract of each citation to determine eligibility for full‐text review. In the event of disagreement between the initial reviewers, a third study author provided a tie‐breaking evaluation of the citation in question.

The remaining citations each received a full‐text review from a pair of study authors for final eligibility in the review. Each pair of reviewers worked independently initially, though conflicting recommendations on citation eligibility were discussed and resolved within reviewing pairs. Unresolved conflicts on citation eligibility after full‐text review by pairs were discussed by all study authors and resolved by consensus opinion.

We created a standardized sheet to extract data from citations deemed eligible after full text review. We collected data on author, study year, study country, sample size, and patient demographics. Outcomes data are presented using two categories: (1) adverse respiratory events including apnea, airway obstruction, oxygen desaturation to < 90%, or intubation; and (2) other adverse events including arrhythmia, hypotension, worsening delirium, cardiac arrest, and mortality. Of note, QTc prolongation, dystonia, and in‐hospital falls were included in the preplanned definition of other adverse events, though they were excluded on a post hoc basis due to inconsistent reporting in the identified studies. We primarily extracted outcomes as binary events (e.g., arrhythmia—yes/no).

We also collected data on several secondary outcomes on clinical efficacy. Prior measures of clinical efficacy in adults aged 18–64 years old focused on rapid control of severe agitation (specifically at 15 min). Older adults may require a clinical approach that utilizes lower doses and greater tolerance of time‐to‐sedation to reduce adverse events and over‐sedation [25]. As such, we report efficacy in terms of the need for redosing and median/mean time‐to‐sedation. In instances where studies examined the use of sedatives in a broad age range, we directly contacted the citation authors and requested the release of data for the subgroup older than 65 years. We preferentially requested de‐identified patient level data, though we accepted aggregate subgroup data due to variations in data release regulations across Institutional Review Boards.

2.3. Synthesis Methods and Statistical Analysis

We reported the aggregate occurrence of any adverse event, adverse respiratory events, and other (nonrespiratory) adverse events for each individual study included in the systematic review. Adverse events were pooled by drug across studies of the same study type, with observational studies and clinical trials presented separately. Drugs with inadequate samples (n < 15) were not included in the pooled analysis as confidence intervals (CIs) would be large and noninterpretable. Odds ratios (ORs) are reported for the occurrence of any adverse events, adverse respiratory events, other adverse events, and need for redosing with haloperidol as a reference group. Haloperidol was selected as the reference given its common use in clinical practice [26] and historical precedent as 1st line treatment in calming older adults with agitation [27, 28]. The 95% CI of ORs was calculated using $e^{\ln \left(\mathit{\text{OR}}\right)\pm 1.96\sqrt{\frac{1}{a}+\frac{1}{b}+\frac{1}{c}+\frac{1}{d}}}$, where a, b, c, d are number of cases and noncases within each medication group. We pooled data across the included studies and examined associations of antipsychotic or anxiolytic medications with time to sedation using weighted linear regression, adjusting for study. Weights were assigned as 1 to studies that provided individual‐level data. For studies that provided aggregated data, we assigned weights to the data that were proportional to the number of participants on which the data were aggregated. All statistical analyses were performed in R 4.2.3.

Subgroup dose‐response analyses were also pursued within each drug to assess the impact on any adverse events, adverse respiratory events, and other adverse events by dosing and formulation. Within each drug, a low dosing strategy was used as the reference group. Low dosing strategies were defined, based on study authors' consensus, as 0.5–2 mg for haloperidol, 2.5–5 mg for droperidol, 0.25–2.5 mg for midazolam, 0.25–1 mg for lorazepam, 1–5 mg for olanzapine, and 12.5–25 mg for quetiapine. In analyzing the impact of formulation, we pooled results from drugs available in both oral and parenteral formulations (haloperidol, olanzapine, and lorazepam) to compare safety across oral (PO), intramuscular (IM), and intravenous (IV) administrations.

2.4. Assessment for Bias

The risk of bias for each individual study was assessed using the Cochrane Risk of Bias tools. Accordingly, eligible clinical trials and observational studies were reviewed by two reviewers independently using the Revised Cochrane Risk of Bias Tool for Randomized Trials (RoB 2.0) and the Risk of Bias in Non‐randomized Studies of Interventions (ROBINS‐I) tools, respectively [29, 30].

3. Results

3.1. Study Characteristics

The search strategy identified a total of 8600 studies. There were 8514 studies that were excluded for being a duplicate study (2478) or not meeting our inclusion criteria during abstract screening (6036). Two studies were not reviewed because one was not able to be retrieved and one was not available in English. Eighty‐four studies underwent full‐text review, resulting in nine studies being deemed appropriate for inclusion in this systematic review. Details of the selection process are displayed in Figure 1.

FIGURE 1.

PRISMA flow diagram of the study selection process.

In total, nine studies [31, 32, 33, 34, 35, 36, 37, 38, 39] were identified that examined haloperidol [33, 34, 35, 36, 37, 39], droperidol [31, 34, 36, 38, 39], lorazepam [39], midazolam [31, 34, 36, 39], diazepam [36, 39], alprazolam [39], olanzapine [32, 34, 36, 39], quetiapine [39], ziprasidone [35, 39], risperidone [39], loxapine [35], ketamine [33, 37], or co‐administration of a combination of these medications [36, 37]. All identified studies were prospective cohort studies with the exceptions of Lin et al. [37] (randomized trial), McDowell et al. [35] (retrospective cohort), and Engstrom et al. [39] (retrospective cohort). Similarly, all studies were conducted within a single institution (either paramedic system or hospital) with the exception of Calver et al. [31] (a subgroup analysis of the multicenter DORM II Study [40]) and Engstrom et al. [39] (a multicenter cohort study across 21 EDs in 4 states). The nine studies were conducted in either Australia or the United States. A single study, Engstrom et al. [39], accounted for roughly 78% of the sample. Additional information on study attributes and observed adverse events within each study is reported in Table 1.

TABLE 1.

Baseline characteristics of the included observational studies.

Author (publication year)	Study design	Country	No. of participant	No. of female (%)	Medication a	No. of participant a	Any adverse events b (n)	Respiratory adverse events (n)	Redosing (n)
Calver (2013) [31]	Prospective study	Australia	49	16 (33%)	Droperidol	47	3	1	17
					Midazolam	2	1	0	1
Cole (2016) [33]	Prospective study	US	3	1 (33%)	Haloperidol	2	0	0	0
Cole (2017) [32]	Prospective study	US	26	11 (42%)	Olanzapine	26	0	0	10
Cole (2021) [34]	Prospective study	US	78	16 (21%)	Haloperidol	3	1	1	2
					Droperidol	36	2	1	8
					Olanzapine	38	10	9	11
					Midazolam	1	0	0	0
McDowell (2019) [35]	Retrospective study	US	15	8 (53%)	Haloperidol	5	0	0	2
					Ziprasidone	9	0	0	2
Oliver (2019) [36]	Prospective study	Australia	11	1 (9%)	Droperidol	4	0	0	4
					Haloperidol	2	0	0	2
					Midazolam	1	1	1	1
					Olanzapine	1	1	0	0
Isoardi (2021) [38]	Prospective study	Australia	6	1 (17%)	Droperidol	6	1	1	6
Engstrom (2023) [39]	Retrospective study	US	684	343 (50%)	Haloperidol	105	18	2	25
					Droperidol	36	7	4	6
					Olanzapine	36	4	2	10
					Ziprasidone	8	0	0	1
					Midazolam	64	34	14	27
					Quetiapine	56	3	1	7
					Lorazepam	350	55	26	61

^a Administration of diazepam (n = 11), alprazolam (n = 12), risperidone (n = 6), loxapine (n = 1), ketamine (n = 2), and co‐administration of medications (n = 7) not depicted.

^b Any adverse events include adverse respiratory events (apnea, airway obstruction, oxygen saturation < 90%, and intubation) and other adverse events (arrhythmia, hypotension, worsening delirium, cardiac arrest, and mortality).

As depicted in Table 2, the total sample across observational studies included a total of 872 older adults receiving haloperidol (n = 117), droperidol (n = 129), lorazepam (n = 350), midazolam (n = 68), diazepam (n = 11), alprazolam (n = 12), olanzapine (n = 101), quetiapine (n = 56), ziprasidone (n = 17), risperidone (n = 6), loxapine (n = 1), ketamine (n = 1), and co‐administration of a combination of medications (n = 3). Diazepam, alprazolam, risperidone, loxapine, ketamine, and co‐administration of a combination of medications were excluded from the pooled analysis. The remaining study cohort included 838 patients.

TABLE 2.

Associations of initial medication with adverse events, need for redosing, and time to sedation.

		Haloperidol	Droperidol	Lorazepam	Midazolam	Olanzapine	Quetiapine	Ziprasidone
	No. of participants	117	129	350	68	101	56	17
Any adverse events	No. of cases, (%)	19 (16%)	13 (10%)	55 (16%)	36 (53%)	15 (15%)	3 (5%)	0 (0%)
	OR (95% CI)	1	0.73 (0.30, 1.80)	0.9 (0.51, 1.61)	5.25 (2.64, 10.45)	0.99 (0.41, 2.36)	0.27 (0.08, 0.97)	NA
Adverse respiratory events a	No. of cases, (%)	3 (3%)	7 (5%)	26 (7%)	15 (22%)	10 (10%)	1 (2%)	0 (0%)
	OR (95% CI)	1	2.27 (0.49, 10.52)	3 (0.89, 10.16)	10.69 (2.96, 38.60)	3.39 (0.78, 14.79)	0.68 (0.07, 6.71)	NA
Other adverse events b	No. of cases, (%)	16 (14%)	6 (5%)	29 (8%)	21 (31%)	5 (5%)	2 (4%)	0 (0%)
	OR (95% CI)	1	0.42 (0.13, 1.39)	0.52 (0.27, 1.01)	2.64 (1.26, 5.54)	0.54 (0.17, 1.74)	0.21 (0.05, 0.97)	NA
Redosing	No. in need of redosing, (%)	31 (26%)	41 (32%)	61 (17%)	29 (43%)	31 (31%)	7 (13%)	3 (18%)
	OR (95% CI)	1	0.69 (0.32, 1.48)	0.65 (0.39, 1.08)	2.15 (1.13, 4.11)	1.09 (0.53, 2.26)	0.44 (0.18, 1.08)	0.44 (0.09, 2.15)
Sedation time	Median value, minutes	50	27	—	10	17	—	—
	Difference in sedation time, minutes (95% CI)	Ref.	−11 (−32, 10)	—	−18 (−72, 37)	−10 (−32, 11)	—	—

Note: Logistic regression model adjusted for study. Bolded numbers—95% confidence interval does not include 1 (i.e., p < 0.05).

^a Adverse respiratory events included apnea, airway obstruction, oxygen saturation < 90%, and intubation.

^b Other adverse events include arrhythmia, hypotension, worsening delirium, cardiac arrest, and mortality.

In the randomized clinical trial by Lin et al. [37], there was a total sample of six older adults, with five subjects receiving co‐administration of haloperidol‐lorazepam and one receiving ketamine. No pooled analyses of the randomized clinical trial data were pursued due to the paucity of trial data.

3.2. Synthesis of Results

Any adverse events (respiratory or other) were observed in a total of 16.8% (141/838) patients. Any adverse events by drug were observed in 53% of patients receiving midazolam (36/68), 16% lorazepam (55/350), 16% haloperidol (19/118), 15% olanzapine (15/101), 10% droperidol (13/129), 5% quetiapine (3/56), and 0% ziprasidone (0/17). Adverse respiratory events varied widely across individual drugs, from 0% for ziprasidone (0/17) to 22% for midazolam (15/68). Among adverse respiratory events, there were 14 observed intubations among 1 patient given haloperidol, 1 olanzapine, 5 lorazepam, and 7 midazolam. Similarly, other adverse events varied widely across drugs, from 0% for ziprasidone (0/17) to 31% for midazolam (21/68). Of note, neither tachydysrhythmias, including Torsades des Pointes, nor episodes of cardiac arrest were observed.

Relative to haloperidol, midazolam significantly increased the risk for any adverse events (OR 5.25; 95% CI 2.64–10.45), adverse respiratory events (OR 10.69; 95% CI: 2.96–38.60), and other adverse events (OR 2.46; 95 CI: 1.26–5.54). Quetiapine was the only drug observed to have a lower frequency of any adverse events (OR 0.27; 95% CI: 0.08–0.97), and other adverse events (OR 0.21; 95% CI: 0.05–0.97). However, no difference was appreciated in adverse respiratory events comparing quetiapine to haloperidol. Otherwise, no statistical differences in any adverse events, adverse respiratory events, and other adverse events were observed when comparing haloperidol to droperidol, lorazepam, or olanzapine. Relative comparisons of haloperidol to ziprasidone were not calculable as no adverse events (respiratory or other) were observed among patients receiving ziprasidone.

Treatment failure requiring redosing of medications was variable across drugs, ranging from 13% for quetiapine to 43% for midazolam (Table 2). An increased risk for treatment failure requiring redosing was observed among older adults receiving midazolam (OR 2.15; 95% CI: 1.13–4.11) as compared to haloperidol. No other statistical differences were appreciated in redosing frequency in comparing the remaining medications to haloperidol.

Time to sedation data was only available for haloperidol, droperidol, midazolam, and olanzapine. Observed median sedation time was 50, 27, 10, and 17 min for haloperidol, droperidol, midazolam, and olanzapine, respectively. No significant differences in time‐to‐sedation were observed in the pooled analysis.

The impact of initial dosing of medications on safety is reported in Table S1. Only lorazepam was observed to have an increase in any adverse events (OR 2.38; 95% CI: 1.13–4.98) and other adverse events (OR 2.52; 95% CI: 1.01–6.31) with higher dose (i.e., 1.5–2 mg relative to 0.25–1 mg). No other significant differences were appreciated in the dose–response analyses for any other drug. Though statistical differences were not observed, noticeable signals were observed in any adverse events for droperidol (OR 4.26 [95% CI: 0.56–32.64] for 10 mg relative to 2.5–5 mg) and quetiapine (OR 9.78 [95% CI: 0.80–119.75] for 50–200 mg relative to 12.5–25 mg).

Analyses of formulation of medications on safety are depicted in Table 3. An increase in any adverse events was observed for both intramuscular (OR 7.63 [95% CI: 1.64–35.57]) and intravenous (OR 10.17 [95% CI: 2.36–43.79]) administrations relative to oral administrations of antipsychotic or anxiolytic medications.

TABLE 3.

Associations of formulation of antipsychotic or anxiolytic medications (Haloperidol, Olanzapine, Lorazepam) with adverse events and the need for redosing.

	Formulation	Oral	Intramuscular	Intravenous
	No. of participants	85	127	352
Any adverse events	No. of cases (%)	2 (2%)	19 (15%)	67 (19%)
	OR (95% CI)	1.00	7.63 (1.64, 35.57)	10.17 (2.36, 43.79)
Adverse respiratory events a	No. of cases (%)	1 (1%)	11 (9%)	27 (8%)
	OR(95% CI)	1.00	9.49 (1.05, 85.54)	6.59 (0.87, 50.01)
Other adverse events b	No. of cases (%)	1 (1%)	8 (6%)	40 (11%)
	OR (95% CI)	1.00	5.59 (0.65, 47.81)	13.30 (1.65, 107.49)

Note: Logistic regression model adjusted for study and medication. Bolded numbers—95% confidence interval does not include 1 (i.e., p < 0.05).

^a Adverse respiratory events include respiratory depression, airway obstruction, oxygen saturation < 90%, and intubation.

^b Other adverse events include arrhythmia, hypotension, dystonia, in‐hospital fall, delirium, cardiac arrest, and mortality.

3.3. Risk of Bias

When applying the ROBINS‐I tool, we found that all eight observational studies included in the study had a serious risk of bias (Figure 2). This finding was primarily driven by serious risk of confounding, which occurred in all seven studies, as there was inadequate adjustment for confounding. In applying the RoB 2.0 Tool, we found that Lin et [37] had moderate risk of bias due to concerns with the randomization process, deviations from intended intervention, outcome measurement, and selection of the reported result.

FIGURE 2.

Risk of bias assessment (ROB) for studies included in the meta‐analysis for (a) the eight observational studies using the ROB‐I tool and (b) randomized controlled trial using the ROB 2 tool.

4. Discussion

In the first systematic review on the safety of antipsychotic or anxiolytic medications in the treatment of acute, undifferentiated agitation in older ED adults, we report: (1) a high frequency (16.8%) of adverse events among older ED adults who were administered medications for acute undifferentiated agitation; (2) benzodiazepines, particularly midazolam, had the worst safety profile of any medications examined; and (3) the lowest effective dose, preferably in oral formulation, should be utilized over higher doses and parenteral formulations. Overall, we found that there is a paucity of studies—all of which had moderate or serious risk of bias—on the use of medications to calm older adults with severe agitation.

Our findings support that co‐administration of a parenteral antipsychotic and benzodiazepine (as endorsed by an American College of Emergency Physicians Clinical Policy for adults aged 18–64 [41]) poses increased risk to older adults, thus necessitating a modified clinical approach. The high rates of serious adverse events observed in our study validate the long‐held belief that antipsychotic or anxiolytic medications pose significant risk to older adults with severe agitation, even when only one or two doses are used acutely [3, 4, 27, 28]. Our findings support that early nonpharmacologic approaches (verbal de‐escalation, reorientation, caregiver engagement) should be exhausted prior to the use of a medication [4]. ED providers should consider sensory reduction and person‐centered communication when handling older adults with dementia and severe agitation [5]. Furthermore, when medications are deemed necessary, oral medications, which includes consideration of both tablets and liquid, should be offered prior to pursuing parenteral administration. Relatedly, our data demonstrate that oral quetiapine has a favorable safety profile relative to alternative options and thus a reasonable 1st line agent. These findings are consistent with expert consensus statements which assert that oral medications should be offered over parenteral medications when possible [4, 22] as they are the least coercive method of administering medications [42, 43].

Our study suggests that clinical scenarios requiring the use of parenteral medications—either intramuscular or intravenous—should be minimized and, when absolutely necessary, approached with caution. Specifically, our data do not support the co‐administration of antipsychotic and benzodiazepine to older adults. While we did not observe a significant sample of co‐administered antipsychotic‐anxiolytic for direct comparison, the excessive risk posed by midazolam alone is too high to justify its use in older adults either alone or in conjunction with an antipsychotic. Moreover, there were no additional benefits gained with midazolam administration, as redosing rates were higher. It is unclear why increased redosing was observed among such patients, though it may be potentially attributable to paradoxical worsening of agitation, which has been described with older adults given benzodiazepines [44]. While midazolam demonstrated an unfavorable safety profile, the same pattern was not replicated with the only other benzodiazepine, lorazepam, reported in our review. Unlike midazolam, lorazepam is readily available in both oral and parenteral formulations in most EDs, and the highly prevalent use of oral formulations likely confounded our pooled analyses. Regardless, concerningly high frequencies of adverse events were seen at higher doses of lorazepam and with parenteral administrations. Additionally, there is adequate evidence from alternative clinical settings that the American Geriatrics Society Beers Criteria recommends against the use of benzodiazepines independent of condition/diagnosis (though such recommendations also exist for antipsychotics) [7]. However, emergency clinicians still may be confronted with patients with agitation necessitating the use of benzodiazepines, namely older adults with known alcohol or benzodiazepine withdrawal.

Selecting a parenteral antipsychotic in the treatment of an older adult with severe agitation who declines oral medications remains complicated. While quetiapine performed admirably in our study, there is no short‐acting parenteral formulation available for use in EDs. Ziprasidone also performed favorably in our study with no observed adverse events (0/17) and is available in parenteral formulations. However, there are too few observations from studies with a high risk of bias to accurately comment on ziprasidone's safety profile. The remainder of our drug‐to‐drug comparisons did not show any significant differences. Though our review cannot provide guidance on selecting an appropriate agent, there is useful data on dosing and formulation that can help inform clinicians.

It is not surprising that we found that dose and route of administration impacted adverse events. Older patients are more susceptible to adverse events due to altered physiology impacting pharmacokinetics and drug metabolism [45, 46]. Indeed, we demonstrated that higher doses of lorazepam were associated with a greater frequency of adverse events, primarily respiratory. While these findings were not replicated for other medications, there were clinically relevant signals that such dose–response relationships also exist for droperidol and quetiapine. The lack of statistical significance in our dose–response analyses is likely attributable to loss of power with the small subgroups. Regardless, our study strongly supports the idea that starting with the lowest effective dose represents the safest strategy to achieving therapeutic effect while minimizing the risk for adverse effects [47]. Relatedly, ED providers should utilize more conservative treatment targets consistent with minimal sedation to reduce the required dose [10].

The findings from our review and implications discussed should be interpreted in consideration of the study limitations. Regarding the evidence included, our conclusions are based on data from just nine studies. A potential limitation of our approach is that initial and full‐text screening was conducted by several members of a large team, which may have introduced inconsistencies in determining study eligibility. However, inclusion and exclusion criteria were clearly defined at the outset of the review, and members of the study team were in regular contact to resolve discrepancies or uncertainties that arose throughout the process. Among these, all eight observational studies had serious concerns for bias, and the one randomized study had at least some concerns. Given the heterogeneity across small studies with high risk of bias, we were limited in our ability to draw strong conclusions regarding the safety or effectiveness of individual medications. Another limitation is that most of the data came from a single retrospective cohort study, Enstrom et al. [39] However, this one study did have a broadly generalizable sample as it analyzed 21 EDs across four states in the US. Among the pool of potentially screened studies included for full‐text review, the most common reason for exclusion was due to the lack of data addressing adults aged 65 and older. There is a need for additional research examining the comparative effectiveness and safety of antipsychotic or anxiolytic medications in older adults, who are more vulnerable to the potential risks of these medications.

5. Conclusion

Addressing acute, undifferentiated agitation among older adults requires thoughtful consideration of polypharmacy challenges, the existence of comorbid conditions, the preservation of patient autonomy, and of course, medication characteristics while acknowledging increased risk for adverse effects. ED clinicians should first exhaust nonpharmacologic alternatives. When medications are necessary, clinicians should use oral medications (where possible), avoid benzodiazepines, and use the lowest effective dose required in the treatment of older adults with agitation.

Author Contributions

Conception: All. Designed and developed the protocol: M.F.C., N.M.E., J.N., D.K., E.M., J.L.K., K.T., R.M.S., and S.W.L. Data extraction: M.F.C., N.M.E., A.F., J.N., D.K., Z.C., N.G., E.M., J.L.K., K.T., R.M.S., and S.W.L. Citation review: M.F.C., N.M.E., A.F., J.N., D.K., Z.C., N.G., E.M., J.L.K., K.T., R.M.S., and S.W.L. Risk of bias assessment: M.D. and S.W.L. Data analysis: M.F.C., M.D., and S.W.L. Initial manuscript draft: M.F.C., N.M.E., J.N., J.B.C., K.T., and S.W.L. Revisions and final approval: All.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1. Search terms that were applied to PubMed, Scopus, Embase, AgeLine, CINAHL, PsycINFO, Cochrane Database of Systematic Reviews, and ProQuest Central.

Table S1a. Associations of dose of antipsychotic or anxiolytic medications for haloperidol, droperidol, and olanzapine with adverse events and the need of redosing.

Table S1b. Associations of dose of antipsychotic or anxiolytic medications for midazolam, quetiapine, and lorazepam with adverse events and the need of redosing.

Casey M. F., Elder N. M., Fenn A., et al., “Comparative Safety of Medications for Severe Agitation: A Geriatric Emergency Department Guidelines 2.0 Systematic Review,” Journal of the American Geriatrics Society 73, no. 9 (2025): 2893–2904, 10.1111/jgs.19485.