The Long and Winding Road of Antipsychotics for Delirium: Straightening the Path Forward*

Delirium occurs in up to 50% of ICU adults despite the use of multidomain bundles known to reduce it (1,2). With the duration of delirium being associated with reduced post-ICU cognitive function (3), a high delirium symptom burden observed in many patients (1), and the challenges of rigorously delivering nonpharmacologic delirium treatment interventions (4), the discovery of effective pharmacologic delirium treatments remains a key ICU care priority. Haloperidol has been used in the ICU on an "off-label" basis to treat delirium and its symptoms for decades. Before the publication of the first two ICU antipsychotic randomized, placebo-controlled treatment studies in 2010 (5,6), the use of haloperidol to treat ICU delirium was driven not by rigorous evidence but by an ever-increasing awareness of delirium, results from uncontrolled case series, the publication of practice guidelines recommending use (7) and a wide acceptance the acetylcholine–dopamine pathway was mechanistically important (8). At the turn of this century, concerns about haloperidol safety gained prominence when the Food and Drug Administration added warnings to the haloperidol package insert about the risks of haloperidol-associated QTc interval prolongation and Torsade des Pointes in 2001 and use in older adults with dementia in 2004. Among the two 2010 treatment studies, the first study found treatment with haloperidol or ziprasidone, compared with placebo, not to be associated with any improved outcome (5); the second one found quetiapine, compared with placebo, to be associated with faster delirium resolution (6). However, only 36 patients were enrolled in this quetiapine study, and "as needed" haloperidol was frequently administered to both groups. When the 2013 Society of Critical Care Medicine (SCCM) Pain, Agitation/Sedation, Delirium, Immobility, and Sleep Disruption (PADID) guidelines subsequently addressed the narrow question of whether antipsychotic use reduced delirium duration, quetiapine was recommended but haloperidol was not (9). In 2016, the Centers for Medicare and Medicaid Services (CMS) proposed restricting the use of antipsychotics in hospitalized older adults to only patients at risk for causing harm to themselves or others. This draft policy was a response to concerns about the widespread use of "off-label" antipsychotics for delirium, antipsychotic safety issues, and the lack of rigorous evidence demonstrating antipsychotic benefit. Although never implemented, these proposed CMS restrictions, coupled with 2018 SCCM PADIS guidelines recommending delirium should not routinely be treated with antipsychotics, highlighted the need for large, prospective trials evaluating the safety and efficacy of antipsychotics for delirium treatment (4). In this issue of Critical Care Medicine, Carayannopoulos et al (10) publish a rigorous systematic review and meta-analysis evaluating whether the use of antipsychotics in critically ill adults with delirium impacts patient-important outcomes (10). The five placebo-controlled trials that met eligibility criteria, each deemed to have a low risk of bias, enrolled a total of 1750 patients. Haloperidol was evaluated in four trials, quetiapine in two, and ziprasidone in one. Across these five studies, antipsychotic use did not affect days spent without delirium or coma, mortality, the duration of mechanical ventilation, or the length of ICU or hospital stay. The use of antipsychotics did not result in an increased risk of adverse effects. Outcomes were not different between haloperidol and atypical antipsychotic groups. Meta-regression could not be performed due to an inadequate number of eligible studies and trial sequential analysis was inconclusive for the two outcomes evaluated (delirium-free days and mortality). Of note, statistical heterogeneity between studies was found to be moderate–high for each outcome evaluated. The study by Carayannopoulos et al (10) builds on the results of another recent systematic review and meta-analysis that focused solely on ICU delirium treatment studies comparing haloperidol to placebo, another antipsychotic, or benzodiazepines, opioids, or antiemetics (11). This second review, which included 11 randomized studies, found haloperidol did not affect mortality, days spent without delirium or coma, or serious adverse event occurrence. Although the results of these systematic reviews do not support a change in current 2018 PADIS antipsychotic delirium treatment recommendations (4), they help inform a research path moving forward regarding the role of antipsychotics as a treatment of delirium in critically ill adults. Antipsychotics are most frequently initiated when delirium is accompanied by symptoms like agitation or distress. However, in the meta-analysis performed by Carayannopoulos et al (10), 67% of the patients had hypoactive delirium at baseline, a motoric delirium subtype where agitation is generally absent (1). Compared with the MIND-USA and EuRIDICE trials (12,13), where haloperidol dose titration and continuation were based on daily delirium occurrence, the AID-ICU trial (14) administered a standard dose of haloperidol 2.5 mg IV q8h throughout the study but allowed additional "as needed" haloperidol doses for symptom management. The effect of antipsychotics on delirium symptoms, which often fluctuate daily (1), have a number of nondelirium causes, and may have phenotypical presentations (15), remains an important research gap. A number of additional challenges and evidence gaps should be considered when future antipsychotic treatment trials are designed. Nonpharmacologic delirium reduction interventions, like the ICU Liberation Bundle, should be optimized in all patients (2), yet only two trials in the review by Carayannopoulos et al (10) formally incorporated these approaches (12,13). A "one size fits all" approach to delirium treatment may be misguided given delirium is precipitated by multiple causes (1). Patient outcomes are known to vary by delirium phenotype (16); future antipsychotic trials should consider baseline phenotypical differences. The pharmacology of each antipsychotic is different; patient response may depend on the specific antipsychotic evaluated. The optimal timing for antipsychotic initiation remains unclear. Will delirium that occurs before the ICU respond the same way as delirium that first occurs after ICU admission? Delirium is challenging to detect in patients who are deeply sedated or neurologically injured but a valid delirium biomarker does not exist and electroencephalogram assessment approaches remain underdeveloped (1,4). Patients at greatest risk for experiencing the most severe outcomes of delirium, including post-ICU cognitive decline, may be important candidates for pharmacologic treatment. However, despite a published ICU delirium core outcome set that advocates long-term outcome evaluation (17), the effect of antipsychotics on post-ICU outcomes, particularly long-term cognition, remains poorly researched. Recently, antipsychotic use has been reported to be associated with a mortality benefit independent of a delirium effect (18–20). The AID-ICU trial recently reported 1-year mortality to be significantly lower in ICU patients treated with haloperidol compared with placebo (18). Administration of quetiapine to adults with neurotrauma has been reported to improve survival (19). Mechanisms for these potential mortality benefits, including the influence of the daily antipsychotic dose administered (20), need to be further explored. Although Carayannopoulos et al (10) do not report an increase in adverse events with antipsychotic use including QTc interval prolongation and extrapyramidal effects, individual trials generally administered low antipsychotic doses and patients at greater baseline risk for safety concerns were excluded (6,12–14). Until future antipsychotic trials are conducted, and benefit is demonstrated, the results from Carayannopoulos et al (10) support the currency of continuing to adopt 2018 PADIS guideline antipsychotic recommendations (4).