Long-Term Outcomes after Treatment of Delirium during Critical Illness with Antipsychotics versus Placebo: a Longitudinal Follow-up Study to the MIND-USA Randomized Controlled Trial

Summary

Background

Delirium is common during critical illness and associated with long-term cognitive impairment and disability. Antipsychotics are frequently used to treat delirium, but their effects on long-term outcomes are unknown.

Methods

In a multicentre, randomised, double-blind, placebo-controlled trial, delirious critically ill adults with respiratory failure or shock were treated with intravenous haloperidol, ziprasidone, or placebo for up to 14 days. Three and twelve months after randomisation, we assessed survivors’ cognitive, functional, psychological, quality-of-life, and employment outcomes using validated telephone-administered tests and questionnaires.

Findings

Of 566 patients randomised, 358 survived >3 months and 306 survived >12 months; 316 completed follow-up at least once at 3 and/or 12 months. One-year survival and follow-up rates were similar between groups. Cognitive impairment was common in all three treatment groups with one-third of survivors impaired at follow-up. More than half of surveyed survivors in each group had cognitive and/or physical limitations precluding employment. At 3 months, neither haloperidol (aOR: 1.22; 95% CI: 0.73, 2.04) nor ziprasidone (aOR: 1.07; 95% CI: 0.59, 1.96) significantly altered cognitive outcomes compared to placebo as measured by the TICS score. Similarly, at 12 months, neither haloperidol (aOR: 1.12; 95% CI: 0.60, 2.11) nor ziprasidone (aOR: 0.94; 95% CI: 0.62, 1.44) altered the TICS score versus placebo. We also found no evidence that functional, psychological, quality-of-life, or employment outcomes improved with haloperidol or ziprasidone versus placebo.

Interpretation

In delirious critically ill patients, neither haloperidol nor ziprasidone had a significant effect on cognitive, functional, psychological, or quality-of-life outcomes among survivors.

Funding

National Institutes of Health (AG035117) and the Department of Veterans Affairs (Geriatric Research, Education and Clinical Center).

Introduction

Delirium—a neuropsychiatric syndrome characterized by acute and fluctuating disturbances in attention, awareness, and cognition—affects more than half of mechanically ventilated intensive care unit (ICU) patients and is associated with numerous adverse outcomes.¹ Not only does delirium frequently involve frightening delusions and other distressing symptoms, it interferes with important medical care and delays recovery from critical illness.² Even after the resolution of acute symptoms, the impact of delirium can persist. A lengthy period of delirium during critical illness, in fact, predicts long-term cognitive impairment and disability, which can last months to years after the episode of delirium.^3–5 An ideal treatment for delirium, therefore, would not only alleviate the acute symptoms of delirium but would also improve long-term cognitive, psychological, and functional outcomes for survivors of critical illness.

Based on case series and anecdotal experience, antipsychotic medications have been used to treat delirium in the ICU for several decades. Placebo-controlled trials, however, called into question the efficacy of haloperidol,^6,7 the antipsychotic most commonly used to treat delirium, but these early trials were small and focused only on in-hospital outcomes. We therefore conducted the multicenter, randomised, placebo-controlled Modifying the Impact of ICU-Associated Neurological Dysfunction-USA (MIND-USA) Study⁸ to examine both the short- and long-term effects of haloperidol and ziprasidone when used to treat delirium during critical illness. Long-term cognitive, functional, psychological, and health-related quality of life outcomes were still being assessed when we previously reported that treatment with haloperidol and ziprasidone did not significantly affect duration of delirium or other in-hospital outcomes.⁸ Herein, we report data from a pre-specified long-term follow-up study of the effects of treatment of delirium during critical illness with these antipsychotic medications and the effect on long-term cognitive, functional, psychological, and quality-of-life outcomes.

Methods

Study Design and Participants

As previously described in detail,⁸ we conducted this randomised, double-blind, placebo-controlled, phase 3 trial in 16 medical centers throughout the United States. The trial protocol was approved by the Food and Drug Administration (as part of an Investigational New Drug application), an independent data and safety monitoring board, and the institutional review boards at each participating center. We registered the trial at ClinicalTrials.gov (NCT01211522) before the first participant was recruited.

Adults (≥18 years of age) admitted to a medical or surgical ICU with respiratory failure and/or septic or cardiogenic shock were eligible for inclusion in the long-term follow-up study if they had delirium (and thus were randomized to receive either haloperidol, ziprasidone, or placebo) and none of the following exclusion criteria: ongoing treatment with an antipsychotic medication, allergy to haloperidol or ziprasidone, history of neuroleptic malignant syndrome, QT prolongation (QTc > 550 msec), history of torsades de pointes, pregnancy, breastfeeding, life expectancy <24 hours, rapidly resolving organ failure, enrollment in another trial that prohibited co-enrollment, or severe cognitive impairment at baseline (i.e., prior to their current critical illness). We identified baseline cognitive impairment using both medical record review and the short form of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE),⁹ excluding those with a score of ≥4.5, indicating severe cognitive decline. We also excluded patients who were incarcerated, blind, deaf, or were unable to understand English since these characteristics could impede assessments for delirium and/or cognitive function at follow-up. Finally, we excluded non-comatose patients if informed consent was not obtained within 72 hours after inclusion criteria had been met, and we excluded comatose patients if informed consent was not obtained within 120 hours after inclusion criteria had been met.

Randomisation and blinding

Using a computer-generated, permuted-block randomisation scheme with stratification by trial site, we randomly assigned participants in a 1:1:1 ratio to treatment with placebo, haloperidol, or ziprasidone if patients developed delirium. We administered all treatments intravenously as a colorless preparation prepared by investigational pharmacists and delivered in identical syringes in clear plastic bags so that research personnel, ICU clinicians, participants, and their families were blinded to treatment group assignment.

Procedures

As previously described,⁸ we obtained informed consent from each participant or an authorized representative before the onset of delirium, when possible, and monitored potential participants twice daily for delirium using the validated Confusion Assessment Method for the ICU (CAM-ICU).^10,11 Patients who were not delirious at the time of informed consent but later became delirious were randomised immediately after delirium was diagnosed with the CAM-ICU as long as they had developed no exclusion criteria. Those who were delirious at the time of informed consent were randomised immediately.

After randomisation, participants < 70 years of age received 2.5 mg of haloperidol per 0.5 ml or 5 mg of ziprasidone per 0.5 ml or 0.5 ml of placebo (0.9% saline). Alternatively, those ≥ 70 years received initial doses that were half that received by younger participants. We administered all study drugs intravenously, with subsequent doses given every 12 hours. We doubled the volume (and mg) of each dose if a participant remained delirious, was not yet receiving the maximum dose, and had no criteria requiring study drug be held. The maximum individual (and daily) dose received in the haloperidol group was 10 mg (20 mg per day), and the maximum dose in the ziprasidone group was 20 mg (40 mg per day).

In addition to assessing participants for delirium twice daily, trained study personnel assessed participants at least once daily for side effects, including oversedation, extrapyramidal symptoms, akathisia, dystonia, QT prolongation > 550 msec, torsades de pointes, neuroleptic malignant syndrome, and drug reaction with eosinophilia and systemic symptoms (DRESS) syndrome. When a potential side effect was noted, we temporarily or permanently discontinued study drug according to a previously reported protocol.⁸ Additionally, we halved the volume (and dose) of study drug if a participant was not delirious for two consecutive assessments and was not yet receiving the minimum dose. We stopped study drug if a participant was not delirious for four consecutive assessments or was discharged from the ICU.

Outcomes

At three months and twelve months after randomisation, we assessed participants’ cognitive, functional, psychological, and quality-of-life outcomes using validated tests and questionnaires, which were administered via telephone by trained psychology professionals who were blinded to treatment assignment.

Though all post-discharge outcomes were considered secondary endpoints, we prespecified that participants’ T scores on the Telephone Interview for Cognitive Status (TICS),¹² a validated test of global cognition that correlates highly with the Mini-Mental State Examination (MMSE), would be the long-term outcome of primary interest given the association between delirium during critical illness and long-term global cognition.^3,4 We considered an age-adjusted TICS T-score ≤35 to represent cognitive impairment.

We assessed participants for disability in activities of daily living using the Katz Index of Independence in Activities of Daily Living (ADL)¹⁴ and Functional Activities Questionnaire (FAQ),¹⁵ posttraumatic stress disorder (PTSD) using the PTSD CheckList Civilian Version (PCL), and reduced quality of life using the EQ-5D.⁸ We identified hospital readmissions using a healthcare utilization questionnaire and determined employment status via a questionnaire. The employment questionnaire was added to the follow-up assessments approximately 2 years after the start of recruitment, so fewer participants were eligible to be assessed for this outcome.

Statistical Analysis

In keeping with the intention-to-treat principle, we analyzed all participants according to the group to which they were randomised. The analysis of 3-month outcomes included all participants for whom at least one partial follow-up assessment was completed at 3 months. The analysis of 12-month outcomes included all participants for whom at least one partial follow-up assessment was completed at 12 months. Since our previous analysis of 90-day survival found no evidence that haloperidol or ziprasidone affected mortality in the trial population¹⁶ and follow-up was similar through 12 months between treatment groups, we conducted survivors-only analyses.¹⁷ We used the same sample size for the analyses of all outcomes at a time point, imputing missing data with model-based multiple imputation that used model covariates and the outcome of interest in the imputation model. We also performed post-hoc sensitivity analyses using only complete cases.

We analyzed all outcomes using multivariable regression. Because adjustment for baseline (prerandomisation) characteristics can improve statistical power by increasing precision,¹⁸ we prespecified adjusted analyses as our primary analyses of long-term outcomes and selected the following covariates based on previous research and clinical judgment: age at enrollment, sex, race, baseline frailty, level of education, and baseline short IQCODE, Katz ADL, or FAQ (depending on the outcome being analyzed).

We used proportional odds logistic regression to analyze the effects of haloperidol and ziprasidone versus placebo on the primary long-term outcome, TICS T-score, and other continuous and ordinal secondary and exploratory outcomes. We evaluated the proportional odds assumptions graphically and found that the proportional odds assumption was generally met. We report adjusted medians and odds ratios (OR) with 95% confidence intervals (CI) for continuous outcomes. We used multivariable logistic regression to analyze the effects of haloperidol and ziprasidone versus placebo on the exploratory outcome, post-discharge inpatient care assessed by health utilization survey, and report adjusted OR with 95% CI.

To evaluate multicollinearity, which can inflate standard errors, we used redundancy analysis with an adjusted R² cutoff of 0.7 prior to regression modelling.¹⁸ We modeled continuous covariates using restricted cubic splines to allow for nonlinear associations and accounted for correlation among participants within a given site by adjusting standard errors using the Huber-White robust sandwich estimator.

We set the level of significance for all analyses at 5% and made no adjustments for multiple comparisons. Lastly, we used interaction terms between treatment group and age, education, and baseline IQCODE to evaluate for heterogeneity of treatment effect. We used statistical software R (version 4.3.2) for all analyses.

Role of the Funding Source

The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. This material is based upon work supported by the National Institutes of Health, the U.S. Department of Veterans Affairs (VA) Office of Academic Affiliations, the U.S. Department of Veterans Affairs (VA) Office of Research and Development, and with resources and use of facilities at VA Tennessee Valley Healthcare System in Nashville, Tennessee. The contents of this paper are solely the responsibility of the authors and do not necessarily represent those of the National Institutes of Health, the Department of Veterans Affairs, or any enrolling clinical site. The senior and corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Results

From December 2011 to August 2017, we recruited and randomised 566 patients. The participant flow during follow-up is shown in figure 1. Long-term survival after critical illness was similar in the three treatment groups; 66 (37%), 78 (41%), and 64 (34%) of participants in the placebo, haloperidol, and ziprasidone groups, respectively, had died before 3-month follow-up, and 82 (46%), 97 (51%), and 81 (44%) had died before 12-month follow-up. Among those who survived to 3 months, did not withdraw, and spoke English and thus were eligible to complete 3-month follow-up, 103 (95%) participants in the placebo group, 94 (89%) in the haloperidol group, and 107 (93%) in the ziprasidone group were assessed at 3 months. Similarly, among those who survived to 12 months, did not withdraw, and spoke English and thus were eligible to complete 12-month follow-up, 87 (95%), 73 (85%), and 91 (95%) participants in the placebo, haloperidol, and ziprasidone groups, respectively, were assessed at 12 months. Mortality did not differ between groups at 3-month or 12-month follow-up in post-hoc analyses (see Appendix, Figures S1 and S2). Baseline characteristics of participants who were assessed at either 3- or 12-month follow-up were similar in the three treatment groups (table 1).

Figure 1:

Participant flow during follow-up

Table 1:

Baseline characteristics of participants assessed at either 3 or 12 months by treatment group

	Placebo n=107	Haloperidol n=97	Ziprasidone n=112
Age (years)	59 [50–65]	57 [48–64]	57 [48–66]
Female sex	47 (44)	40 (41)	46 (41)
Race
White	86 (80)	82 (85)	89 (80)
Black or other race(s)	21 (20)	15 (16)	23 (21)
Short-form IQCODE	3.0 [3.0–3.2]	3.0 [3.0–3.2]	3.0 [3.0–3.3]
Charlson Comorbidity Index	2 [1–3]	1 [0–3]	2 [1–3]
Received antipsychotic treatment
Before admission	4 (4)	5 (5)	9 (8)
Between admission and randomization	8 (8)	11 (11)	10 (9)
Admitting diagnosis
Acute respiratory distress syndrome	22 (21)	16 (17)	22 (20)
Sepsis	18 (17)	20 (21)	16 (14)
Airway protection	35 (33)	22 (23)	31 (28)
COPD, asthma, or other pulmonary disorder	11 (10)	13 (13)	14 (13)
Surgery	8 (8)	11 (11)	13 (12)
CHF, myocardial infarction, or arrhythmia	3 (3)	4 (4)	4 (4)
Cirrhosis or liver failure	3 (3)	2 (2)	0 (0)
Seizures or neurological disease	1 (1)	2 (2)	0 (0)
Other	6 (6)	7 (7)	12 (11)
Admitted to the surgical ICU	33 (31)	32 (33)	33 (30)
APACHE II at ICU admission	29 [23–32]	27 [22–33]	28 [23–34]
SOFA at randomization	11 [8–14]	11 [8–13]	10 [7–13]
Received assisted ventilation before randomization
Invasive	100 (94)	94 (97)	110 (98)
Noninvasive	3 (3)	2 (2)	2 (2)
Shock before randomization	32 (30)	22 (23)	28 (25)
Days from ICU admission to randomization	2.0 [1.5–3.1]	2.1 [1.6–3.2]	2.6 [1.7–3.4]

Data are median [IQR] or n (%). APACHE=Acute Physiologic Assessment and Chronic Health Evaluation. ARDS=acute respiratory distress syndrome. CHF=congestive heart failure. COPD=chronic obstructive pulmonary disease. ICU=intensive care unit. IQCODE=Informant Questionnaire on Cognitive Decline in the Elderly. SOFA=Sequential Organ Failure Assessment.

At both 3-month and 12-month follow-up, cognitive, functional, psychological, quality-of-life, and readmission endpoints were similar in the three treatment groups (table 2). Cognitive impairment and inability to return to work were common in all three treatment groups (table 2). Adjusted analyses found no evidence that long-term cognition was affected by treatment of delirium with haloperidol or ziprasidone (figure 2 and appendix). Adjusted analyses did indicate an overall effect on 3-month Katz ADL scores (global p=.003; see appendix), but the effects of haloperidol (OR 0.66, 95% CI: 0.36 to 1.19) and ziprasidone (OR 1.05, 95% CI: 0.60 to 1.82) were not significant when compared with placebo, and adjusted analyses found no significant effect of haloperidol or ziprasidone on 12-month Katz ADL scores (global p=.07; see appendix). Adjusted analyses of effects on instrumental activities of daily living (FAQ score), quality of life (EQ-5D-3L visual analog index score), post-traumatic stress (PCL score), employment status, and post-discharge inpatient care utilization also found no evidence of treatment effects. Sensitivity analyses using complete cases yielded similar results (appendix).

Table 2:

Long-term outcomes by treatment group

	Placebo	Haloperidol	Ziprasidone
Cognitive Endpoints
3-mo Telephone Interview for Cognitive Status T-score*	39 [30–51]	42 [30–51]	39 [33–51]
3-mo cognitive impairment	33 (37%) of 90	25 (30%) of 82	29 (33%) of 87
12-mo Telephone Interview for Cognitive Status T-score*	42 [30–51]	46 [31–51]	42 [33–51]
12-mo cognitive impairment	23 (29%) of 78	22 (33%) of 66	26 (33%) of 78
Functional, Psychological, and Quality of Life Endpoints
3-mo Katz activities of daily living	1 [0–4]	0 [0–2]	1 [0–4]
12-mo Katz activities of daily living	0 [0–2]	0[0–1]	0 [0–2]
3-mo Functional Activities Questionnaire	5 [1–12]	5 [2–12]	6 [1–14]
12-mo Functional Activities Questionnaire	5 [1–9]	2 [0–8]	4 [0–12]
3-mo employment status†
Unable to work due to cognitive or physical limitations	27 (55%) of 49	28 (64%) of 44	24 (53%) of 45
Unable to work for another reason	15 (31%) of 49	9 (20%) of 44	16 (36%) of 45
Working but slowed by cognitive or physical limitations	4 (8%) of 49	3 (7%) of 44	0 (0%) of 45
Working at same or better pace and rate	2 (4%) of 49	3 (7%) of 44	3 (7%) of 45
12-mo employment status†
Unable to work due to cognitive or physical limitations	31 (55%) of 56	25 (61%) of 41	31 (57%) of 54
Unable to work for another reason	14 (25%) of 56	6 (15%) of 41	14 (26%) of 54
Working but slowed by cognitive or physical limitations	4 (7%) of 56	5 (12%) of 41	4 (7%) of 54
Working at same or better pace and rate	6 (11%) of 56	4 (10%) of 41	3 (6%) of 54
3-mo Post-Traumatic Stress Disorder Checklist score	27 [22–38]	26 [23–38]	27 [20–36]
12-mo Post-Traumatic Stress Disorder Checklist score	26 [20–38]	25 [22–37]	28 [22–36]
3-mo EQ-5D quality of life index	0.7 [0.4–0.8]	0.6 [0.5–0.8]	0.7 [0.5–0.8]
12-mo EQ-5D quality of life index	0.8 [0.5–0.8]	0.8 [0.5–0.8]	0.8 [0.5–0.8]
Healthcare Utilization Endpoints
Inpatient care 0–3 months after hospital discharge	35 (34%) of 103	30 (32%) of 94	40 (38%) of 106
Inpatient care 3–12 months after hospital discharge	45 (52%) of 87	32 (44%) of 72	40 (44%) of 91

Data are median [IQR] or n (%).

^*Scores are age-adjusted.

^†The employment questionnaire was added as an additional endpoint 2.5 years after recruitment began, so the total number of patients eligible for this outcome is smaller.

Scoring Interpretation: Telephone Interview for Cognitive Status T-Scores have a normalized mean of 50 with a standard deviation of 10 points. Lower scores indicate worse cognition. Scores of 35 or lower are consistent with cognitive impairment. Katz Activities of Daily Living scores range from 0 to 12 with higher scores indicating worse disability in basic activities of daily living. Functional Activities Questionnaire scores range from 0 to 30 with higher scores indicating worse disability in instrumental activities of daily living. Post-Traumatic Stress Disorder Checklist-Civilian (PCL-C) scores range from 17 to 85 with higher scores indicating worse PTSD symptoms. EQ-5D Quality of Life Index scores range from 0.0 to 1.0 with higher scores indicating better quality of life.

Figure 2: Long-term cognition by treatment group.

In survivors of critical illness, 3-month (upper panel) and 12-month (lower panel) global cognition, measured with the Telephone Interview for Cognitive Status (TICS), was not significantly different after treatment of delirium with haloperidol or ziprasidone as compared with placebo. Treatment effects were estimated after adjusting for prerandomisation predictors of cognition, including age, sex, race, baseline frailty, level of education, and baseline short IQCODE.

Analyses for heterogeneity of treatment effect found no evidence that age, education level, or baseline cognition (measured with the short IQCODE) consistently modified the effects of haloperidol and ziprasidone on long-term outcomes (appendix).

Discussion

In this long-term follow-up study of a large, randomized trial comparing haloperidol versus ziprasidone versus placebo for delirium during critical illness, we found no evidence that treatment with either typical or atypical antipsychotics alters long-term outcomes. As advancements in care have improved in-hospital survival for critically ill patients, understanding long-term functional outcomes has become increasingly important.¹⁹ Many ICU survivors value maintenance of cognitive and physical function at least as highly as survival,²⁰ necessitating that long-term outcomes be measured comprehensively and accurately in critical care trials. Observational studies linking delirium during critical illness with subsequent long-term cognitive impairment ^3,4 and disability ^6,7 suggest that delirium treatments might influence long-term outcomes. Though one-third of survivors in this trial were cognitively impaired and more than half of those assessed for employment status were unable to return to work, we found no consistent evidence that either haloperidol or ziprasidone affected long-term cognitive, psychological, functional, and quality-of-life outcomes.

Multiple prospective studies of critically ill patients have demonstrated that prolonged periods of delirium predict cognitive impairment^3,4 and disability ²⁰ up to a year later, but clinical trial evidence is needed to determine whether interventions targeting delirium in the acute setting also have long-term benefits. Though antipsychotics have been widely used to treat delirium in the ICU, our findings²⁰ and those from other placebo-controlled trials²¹ call this approach to management into question. The current findings suggest that these commonly administered drugs do not alter long-term outcomes associated with delirium and point to the need for future studies to identify effective treatments for both delirium and related long-term outcomes. Notably, the Agents Intervening against Delirium in the Intensive Care Unit (AID-ICU) trial²¹ reported that 90-day and one-year mortality²² were lower for ICU patients treated with haloperidol than for those who received placebo, but neither the primary endpoint (hospital-free days) nor the delirium-related secondary endpoint (days alive without delirium or coma) were significantly affected by haloperidol. In contrast, long-term follow-up in MIND-USA—which was completed before AID-ICU but is being reported after a delay due to pandemic-related delays and investigator institution changes—found no differences in 3- or 12-month mortality with use of haloperidol or ziprasidone versus placebo. While additional research is needed to determine whether the mortality results in AID-ICU have a biological underpinning or are explained by type I error, our trial indicates that differences in patient-oriented outcomes after critical illness are likely not driven by a yet undiscovered biological mechanism influenced by antipsychotic use during hospitalization.

Our findings also expand upon prior work evaluating long-term outcomes after receipt of antipsychotics during critical illness. Rood and colleagues assessed the impact of prophylactic haloperidol on post-critical illness quality of life in the pRophylactic haloperidol usE for DeliriUm in iCu patients at high risk for dElirium (REDUCE) study²³ and found no evidence that physical or mental quality of life at 6 months differed between the haloperidol and placebo groups. We expanded these findings by utilizing a larger battery of patient-oriented outcomes, including long-term cognitive impairment, functional disability, PTSD, unemployment, and post-discharge healthcare utilization, and similarly found no evidence of long-term effects. The recently published EuRIDICE trial also measured long-term outcomes, including cognitive, mental health, and quality of life outcomes,²⁴ and found no differences except that participants in the haloperidol group were less likely to remember their ICU admission and more likely to rate general health better than those in the placebo group. These results should be interpreted with caution since the trial was terminated early due to futility, and sample sizes at long-term follow-up were very small. In contrast, MIND-USA had high rates of follow-up (89% or greater) and a larger follow-up sample size and found no difference in outcomes with the use of antipsychotics. Collectively, these results—along with insufficient evidence of short-term benefit²⁵ and studies showing inappropriate continuation of antipsychotics at hospital discharge²⁶—indicate that antipsychotics should not be used routinely to prevent or treat delirium in critically ill adults and likely do not improve long-term outcomes.

Current Society of Critical Care Medicine (SCCM) clinical practice guidelines recommend against routinely using haloperidol or an atypical antipsychotic to treat delirium during critical illness.²⁷ The guidelines do, alternatively, conditionally recommend “a multicomponent, nonpharmacologic intervention that is focused on (but not limited to) reducing modifiable risk factors for delirium, improving cognition, and optimizing sleep, mobility, hearing, and vision in critically ill adults” based on low-quality evidence. Quasi-experimental studies have found that adherence to multicomponent, nonpharmacologic, best-practice bundles (such as the ABCDEF bundle) is associated with reductions in delirium and other adverse outcomes, but none of these studies examined long-term outcomes.²⁸ We observed that one-third of patients had cognitive impairment at 12-month follow-up and over half of patients assessed for employment status were unable to return to work due to related physical or cognitive limitations. Given that new or worsening deficits in cognitive, psychological, and physical function after critical illness (known collectively as post-intensive care syndrome)²⁹ are common, our study highlights the importance of critical care trials measuring patient-centered long-term outcomes whenever possible.

Though nearly always important, the assessment and analyses of long-term outcomes in critical care trials is often not straightforward. Follow-up is difficult because many survivors of critical illness remain at high risk for events—such as recurrence of acute illness, rehospitalization, and death—that prevent the assessment of outcomes. In addition, a percentage of survivors fail to remain connected to the health care system. If any factor that prevents follow-up occurs at a different rate in one treatment group than in another, bias can be introduced due to differential follow-up. Several statistical methods can be used to account for differential follow-up if needed.²⁸ These methods are uncommon³⁰ and present interpretation difficulties, however, and survivor-only analysis can provide unbiased estimates of treatment effect when mortality doesn’t differ between groups.³⁰ Since there was no difference in mortality between treatment groups in our trial, we conducted survivor-only analyses.¹⁶ Nevertheless, we cannot completely rule out the possibility of confounding by differential follow-up with a survivors-only analysis.

Strengths of our trial included use of a placebo with study drugs delivered in a double-blind fashion, a large study population recruited in 16 centers throughout the United States, and robust measurement of multiple, patient-centered long-term outcomes using validated assessments deployed by trained research personnel. One limitation was the administration of cognitive assessments via telephone rather than in person, an approach that may have reduced the sensitivity of cognitive assessments but enhanced feasibility of long-term outcomes assessments in a large, multicenter trial. Additionally, while we excluded patients with severe cognitive impairment prior to ICU admission using the IQCODE, this tool can only estimate baseline cognitive function and is not a direct measure. A small proportion of participants in each treatment group were not assessed during follow-up. As previously noted, if missingness occurred for different reasons in each treatment group, results may have been biased, but follow-up rates were similar in all three groups and were higher than those typically achieved in studies of critical illness survivors. Due to limited numbers of participants in certain subgroups (e.g., the very old and very young), further studies of targeted subgroups are needed to identify heterogeneity of treatment effect. Lastly, because our analyses were hypothesis-driven and prespecified, we did not adjust results for multiple comparisons.

In conclusion, we found no clear evidence in the MIND-USA Study that treatment of delirium during critical illness with haloperidol or ziprasidone improves long-term cognitive, psychological, or functional outcomes for patients who survive critical illness. These findings, along with those of other placebo-controlled trials of antipsychotics for delirium during critical illness, suggest that the routine use of antipsychotics for delirium likely does not improve long-term patient-centered outcomes and is not warranted. Further research is needed to identify optimal pharmacological and non-pharmacological therapies for the prevention and management of ICU delirium and its long-term sequelae.

Research in context.

Evidence before this study

We searched PubMed for relevant original research articles assessing long-term outcomes after treatment of delirium during critical illness with antipsychotic medications. Using combinations of the terms “delirium,” “antipsychotic,” “critical illness,” “cognition,” “disability,” and “quality of life,” we filtered search results by the article type Clinical Trial and limited results to English-language articles published from database inception to August 1, 2022. We identified only one trial that investigated the impact of haloperidol on long-term quality of life after delirium during critical illness and no trials that evaluated long-term cognitive outcomes

Added value of this study

This large, multicentre, randomised, double-blind, placebo-controlled trial is the first, to our knowledge, to show that neither haloperidol nor ziprasidone, when used to treat delirium during critical illness, significantly improve long-term cognitive function, functional disability, mental health, or quality of life in survivors. As found in numerous observational cohorts, a high percentage of survivors in this trial had adverse cognitive, functional, psychological, and quality of life outcomes months after critical illness, highlighting that long-term outcomes should be included, when possible, in future critical care trials.

Implications of all the available evidence

Our findings, along with insufficient evidence of short-term benefit and frequent inappropriate continuation of antipsychotics at hospital discharge, indicate that antipsychotics should not be used routinely to treat delirium in critically ill adults

Data sharing

Deidentified data and the data dictionary will be shared with approval from the MIND-USA Steering Committee and a signed data access agreement. All requests should be sent to timothy.girard@pitt.edu.