Introduction

Postoperative and intensive care unit (ICU) delirium remains a challenge for patients, families, and caregivers. Identified more than half a century ago in cardiac surgery patients, delirium today is characterized by criteria of the Diagnostic and Statistical Manual of Mental Disorders (DSM)-V, which can be summarized as a fluctuating disturbance of consciousness evolving over a short period of time, a change in cognition, and evidence from the current history, physical examination, or laboratory findings that the disturbance is caused by the direct physiological consequences of a general medical condition.

The suffering of delirious patients is severe as they may be restless, hallucinate, and be filled with fear. Unfortunately, problems continue even after the resolution of delirium. This syndrome may be associated with prolonged ICU and hospital stay [1, 2], more hospital readmissions [3], reduced quality of life, loss of independence, and increased mortality [1, 4,5,6,7]. Furthermore, the duration of delirium is associated with worse long-term cognitive function [8, 9]. The increased socioeconomic burden should also not be underestimated [10]. However, a recently updated delirium guidance paper on prevention and management of pain, agitation/sedation, delirium, immobility, and sleep disruption in adult ICU patients summarizes that delirium in critically ill adults has not been consistently shown to be associated with ICU length of stay, discharge disposition to a place other than home, depression, functionality/dependence, or mortality [11].

Ranging from 10 to 80% [12,13,14] or even up to 90% depending on the type of surgery [15], the overall incidence of delirium is high during hospital stay, especially in elderly patients. Delirium usually develops within 72 h after surgery and/or ICU admission. However, its impact is likely to be underestimated due to the predominance of hypoactive delirium.

Although the pathophysiology of delirium remains poorly understood, we know that the pathogenesis of the cognitive impairments associated with delirium is multifactorial. Certain entities such as drug overdose [16] but also drug withdrawal [17] bear delirium risk. As with so many disparate etiologies, it is highly unlikely that a single mechanism is solely responsible [18]. Therefore, research focuses on the assessment of modifiable pre-, intra-, and postoperative risk factors (e.g., dehydration, fluid balance, immobilization, analgesia, and sleep deprivation) associated with delirium, [18] as well as prediction, prevention, early detection, and treatment of this common psychiatric syndrome. One promising approach is the detection of elevated or lowered biomarkers as predictors or indicators of delirium [19, 20]. Furthermore, serum biomarkers may aid in risk stratification, diagnosis, and monitoring of delirium [19] and, finally, may help to find an effective treatment. This review aims to summarize the current state of knowledge on serum biomarkers of delirium.

Methods

We performed an updated review on biomarkers of delirium based on previous publications [19]. As age is one of the most consistently reported risk factors for developing delirium, we restricted our search to publications including patients aged 60 years and older [18, 21]. Study selection and quality assessment were performed by two independent authors (AH and KT). The results were compared, and disagreements were reviewed (MS).

Literature search

An electronic search of PubMed, Cochrane, Embase, and MEDLINE databases was performed. The detailed search strategy is available in “Appendix” (Additional files). Search terms used for each biomarker are listed in Additional file 1: Table S1. Every biomarker term according to Additional file 1: Table S1 was searched with “delirium”, “acute brain dysfunction”, “stroke”, “hemorrhagic stroke”, “ischemic stroke”, “traumatic brain injury”, and “septic encephalopathy”. The date of the last search was June 1, 2019.

Inclusion and exclusion criteria

Only studies that met the following criteria were included: patients aged 60 or older, sample size of 10 or higher, use of standardized approach to diagnose delirium [e.g., Intensive Care Delirium Screening Checklist (ICDSC), Confusion Assessment Method (CAM or CAM-ICU); Nursing Delirium Screening Scale (NuDESC); Delirium Observation Screening Scale (DOS); Diagnostic and Statistical Manual of Mental Disorders (DSM)-III/IV; Delirium Symptom Interview (DSI); Delirium Rating Scale Revised-98-T (DRS-R98-T); Memorial Delirium Assessment Scale (MDAS)], ICU/hospital cohort (e.g., excluding studies performed in nursing homes), and English language. Reviews from all cohorts (i.e., medical, surgical, mixed, ICU) were included. Age limitation was chosen due to significantly higher reported incidence of delirium in patients 60 to 65 and older, and to help clarify results within the flood of information on delirium biomarkers available to date for the age category at highest risk. Studies reporting data that included patients with cognitive dysfunction due to preexisting psychiatric disorders, known dementia, or alcohol-related delirium (delirium tremens) were excluded. Reviews/meta-analyses, and studies reporting animal data or cerebrospinal fluid (CSF) biomarkers were also excluded.

Data extraction

Study-relevant information was extracted by two independent investigators (AH and KT) for each included study. Any conflict of opinion was resolved by consensus with a third party (MS). Study location and date of study conduct, patient characteristics, past medical history including drug therapy prior to hospitalization, risk assessment scores (e.g., Charlson comorbidity index), outcome data (i.e., ICU and hospital length of stay, mortality), total number of patients, and study-specific procedures—including drug therapy during ICU and/or hospital stay—were considered relevant for data extraction. Observational and interventional study design was distinguished.

Results

Trial identification

In June 2019, a search of the PubMed, Cochrane, Embase, and MEDLINE databases using the search terms “delirium”, “acute brain dysfunction”, “stroke”, “hemorrhagic stroke”, “ischemic stroke”, “traumatic brain injury”, and “septic encephalopathy” [22] AND “biomarker” (term “biomarker” and each biomarker suggested by the authors, Additional file 1: Table S1) for papers published until June 1, 2019, retrieved 1839 publications (Fig. 1). After removal of 57 duplicates, a critical review of the titles and abstracts was performed, and another 616 articles without study-specific data were excluded. Thirty-two studies (29 observational and three interventional) remained after further exclusion of 1134 articles based on title, abstract, or full text as indicated in Fig. 1. The full texts of these remaining studies (Table 1) were reviewed for data extraction by two independent investigators (AH and KT). Due to age limitation a total of 73 publications on mostly mixed populations (sample size ranging from 10 to 1183) and 34 publications on the pediatric cohort have been excluded from our analysis (Fig. 1).

Fig. 1
figure 1

Flow diagram according to literature search

Full size image
Table 1 Summary of current evidence on biomarkers in elderly delirious patients determined by literature search
Full size table

Study characteristics

All 32 studies were published before May 2019 and included information on 7610 patients. Twenty-four studies reported data from surgical patients [23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46], of which two studies analyzed the same patient cohort [29, 30]. Of these 24 studies, 12 reported data collected from delirium high-risk surgical cohorts: Two studies reported data from cardiac surgery [39, 42], and ten reported data from hip surgery patients [23, 24, 28, 32, 33, 35,36,37, 41, 45]. Five studies reported data from medical patients (1026 patients) [47,48,49,50,51], and three studies reported data from a mixed cohort (i.e., surgical and medical patients or not defined; 1086 patients) [52,53,54]. Twenty-nine were observational studies, and three were interventional studies (outlined in more detail below).

Biomarkers

The authors initially screened for biomarkers already known as possible markers of delirium (Additional file 2: Table S2). A second comprehensive screening of the literature was for biomarkers mentioned particularly in the context of other neurological diseases, but bearing a possible association with delirium, including dementia, delirium tremens, hypoxic brain injury, and Parkinson’s disease (Additional file 2: Table S2). These searches resulted in 11 additional biomarkers that were also investigated in the context of delirium in the elderly (Additional file 2: Table S2).

Biomarkers were grouped according to their biochemical function (i.e., cytokine, enzyme, growth factor, hormone, metabolic product, neuronotrophic factor, neurotransmitter, transcription factor, transport protein, or other; Table 2). Overall, 20 biomarkers were reported to detect or to be associated with delirium (Table 3). Of these, higher levels of 14 biomarkers [i.e., IL-6, cortisol, prolactin, amyloid, creatinine, C-reactive protein (CRP), neopterin, metalloproteinase-9 (MMP-9), neutrophil–lymphocyte ratio (NLR), phenylalanine–tyrosine ratio, procalcitonin, thioredoxin, serpin family A member 3 (SERPINA3 (alpha 1-antichymotrypsin)), and 8-iso-prostaglandin F2α] and lower levels of 6 biomarkers [i.e., brain-derived neurotrophic factor (BDNF), leptin, acetylcholine, albumin, insulin-like growth factor-1 (ILGF-1), and alpha-2 glycoprotein (AZGP1)] were reported in delirious patients. However, apart from CRP clinical relevance of the presented biomarkers was either questioned or denied by the authors (Table 3). With the exception of IL-1β and IL-12 all inflammatory biomarkers could be linked to delirium (Additional file 3: Table S3; Additional file 4: Table S4). Moreover, a connection to delirium was found for all four biomarkers of metabolism (Additional file 3: Table S3; Additional file 4: Table S4).

Table 2 Grouping of suggested biomarkers of delirium according to their main function
Full size table
Table 3 Role of suggested biomarkers of delirium according to literature reports
Full size table

Surgical cohort

Overall, 19 studies reported on the incidence of delirium. Event rate of postoperative delirium ranged from 13.4 to 43.8% (Table 1). Four studies reported data from ICU patients (delirium incidence range 15.8 to 43.8%; Table 1) [39]. Within the studies reporting data from surgical patients, three were interventional and as such did not group patients into delirious and non-delirious prior to investigation. Each of these three studies showed a lower incidence of delirium in the experimental arm (Table 1): One compared olanzapine to placebo (delirium incidence 14.3% vs. 40.2%) [23], one compared fast-track surgery to the standard procedure (delirium incidence 3.4% vs. 12.9%) [34], and one compared hypertonic to normal saline (delirium incidence 11.7% vs. 38.3%) [33].

Biomarker assessment of postoperative delirium

Biomarkers reported from investigations of surgical cohorts are shown in Table 1. Two studies from the surgical population and one from the medical population evaluated the role of the acetylcholine pathway in delirium. Whereas anticholinergic treatment was suggested as a promising strategy to reduce the incidence of delirium [23], reports on cholinesterase were not as clear [28]. Albumin was the main focus of one study [43], but also evaluated in four other studies resulting from our literature search [23,24,25,26]. It was almost consistently reported to be lower in delirious patients. Amyloid β1-40, a protein associated with dementia, was the main focus of Sun and colleagues [27] who reported a possible role in the detection of delirium. No difference in ASAT was found among postoperatively delirious and non-delirious patients. Like albumin, ASAT was not the focus of the study cited here [24]. Among inflammatory markers, increased procalcitonin [27], CRP [24, 27, 28, 30,31,32, 41, 45, 46], and IL-6 [25, 27, 28, 34, 35, 44, 46] were consistently reported to be associated with delirium with the exception of one study that found no increase in IL-6 levels [33]. Of note, in this study, IL-6 was collected early (i.e., venous blood was drawn at 06:00 on the first day after surgery). Cortisol [27] and leptin were the two reported hormones with a possible link to postoperative delirium. Leptin levels were significantly lower in patients with delirium compared to those without [36]. Finally, AZGP1 [29], MMP-9 [44], neopterin [42], phenylalanine–tyrosine ratio [42], SERPINA3 [29], thioredoxin [37], and 8-iso-prostaglandin F2α [38] were linked to postoperative delirium.

Medical cohort

Within the studies reporting data from medical patients, three studies reported an incidence of delirium ranging from 12.3 to 29% (Table 1).

Biomarker assessment of delirium in elderly patients admitted for medical reasons

Biomarkers reported from investigations on patients admitted for medical reasons are outlined in Table 1. The study reporting results on the assessment of cholinesterases (i.e., acetylcholinesterase and butyrylcholinesterase) in medical patients found no association with delirium [47]. As opposed to the surgical cohort, ASAT was found to be lower in delirious patients admitted for medical reasons [47]. Lower BDNF levels were clearly linked to delirium in oncology patients [48]. Findings on CRP [47, 49, 50] and IL-6 [51] concurred with those reported from the surgical cohorts. NLR was additionally linked to delirium [49].

Mixed cohort

The three studies to report data from both, medical and surgical patients, reported a combined incidence of delirium of 23.5% (Sanchez [54]) and 35.7% (Van Munster [53]) in non-ICU patients, and of 78% (Nguyen [52]) in ICU patients with sepsis (Table 1).

Biomarker assessment of delirium in mixed cohorts of elderly patients

Biomarkers reported from investigations of mixed cohorts are also outlined in Table 1. The two markers reported to have a possible association with delirium in mixed patient cohorts were leptin [54] and prolactin (ICU cohort; [52]).

Discussion

Our updated findings are consistent with a review published 7 years ago, in 2011, by Khan and colleagues [19] who had reported a lack of evidence for the clinical use of biomarkers to aid in earlier detection, prevention, or treatment of delirium. As long as there are no adequate means of intervention, altered levels of biomarkers are unlikely to initiate any change in clinical practice. So far, no biomarker has been identified that would enable development of treatment strategies to lower the incidence, severity, or duration of delirium. A useful biomarker needs to be easily identifiable and reliable to enable targeted therapy while being cost-effective. These criteria were not applicable for any biomarker found in this literature review. Of note, we only reported data from patients aged 60 and older, as opposed to the review by Khan and colleagues [19].

Our main conclusion that none of the biomarkers described help to lower the incidence of delirium is based on several considerations. Pathophysiological explanations such as those addressed by the neurotransmitter hypothesis have already been described nearly 40 years ago [55]. Despite the knowledge of the role of neurotransmitters in the pathophysiology of delirium, with central cholinergic deficiency remaining the leading hypothesis [56], we have not yet been able to find a way to decrease the incidence of delirium. Dopamine excess and inflammation are important assumptions competing with or contributing to the hypothesis of cholinergic deficiency [57], but the focus should lie on modifiable factors causing delirium [18]. This includes stress reduction due to minimization of light and noise disturbances at night and adequate pain management among other things, but also preventive drug therapy in high-risk cohorts [58, 59]. Most importantly, the pathophysiology of delirium remains poorly understood [60] despite being first described by Hippocrates more than 2500 years ago [61]. So on the one hand, our findings are consistent with established theories such as the role of neurotransmitters, inflammation, and stress response in delirium (i.e., reported association of acetylcholine, IL-6, CRP, procalcitonin, and cortisol). On the other hand, there are new approaches that, however, lack sufficient evidence or validation for implementation into everyday clinical practice (i.e., BDNF, leptin, MMP-9, neopterin, phenylalanine–tyrosine ratio, prolactin, NLR, thioredoxin, 8-iso-prostaglandin F2α, AZGP1, and SERPINA3). As such, the latter are not readily available. Delirium biomarkers that are included in standard blood examinations (i.e., albumin and creatinine) may be useful to attribute higher delirium risk but are not targets specific to delirium prevention since they are routinely addressed with respect to other clinical questions. It is nevertheless interesting that an elevated creatinine level was reported to be an independent risk factor for delirium in a cardiac surgery cohort; clinicians should keep this in mind [39].

An almost exclusive association of the numerous investigated inflammatory biomarkers with delirium offers room for and discussion on a novel approach: The implementation of a widely used inflammatory marker (i.e., CRP or procalcitonin) as a diagnostic tool might help facilitate diagnosis of hypoactive delirium. The same can be said for biomarkers of metabolism. Moreover, these biomarkers could be implemented into a tool to assess delirium risk. Finally, the accepted role of amyloid β1-40 for delirium risk has also been confirmed. But clinical relevance here can be reduced to knowledge of its existence, since patients suffering from dementia are known to bear higher risk of delirium [62].

We acknowledge the following limitations of the presented review. First, we did not include any reports on biomarkers with a sample size lower than 10. Important biomarkers to be further explored in larger cohorts may thus have escaped our attention. Second, we focused on patients aged 60 and older, which led to the exclusion of numerous publications reporting other potentially important biomarkers for the elderly population including key articles on delirium biomarkers [63,64,65,66,67,68]. Interestingly, putative delirium biomarkers such as NSE and S-100β were reported/found to be irrelevant [59]. Third, there is an imbalance between medical and surgical patients investigated. In addition, high-risk surgical groups (i.e., hip and cardiac surgery) are overrepresented, whereas another important high-risk group (i.e., ICU patients) is underrepresented in the reported literature. Last, all observational studies have outlined biomarker levels comparing delirious to non-delirious patients. Overall, delirium is diagnosed clinically by established delirium assessment methods and biomarker groups and could be helpful to diagnose patients where a delirium is not so obvious (i.e., hypoactive delirium). In addition, it is not easy to influence levels of the biomarkers reported here other than by following guidelines (e.g., hygiene directives, pain control, prevention and management of infections, avoidance of unnecessary surgery, and adequate nutrition).

Conclusion

The concluding observations offer no ground-breaking recommendation for the implementation of a specific biomarker of delirium. Inflammatory biomarkers and biomarkers of metabolism could assist in diagnosing delirium and in assessing delirium risk. Expert opinions state that especially the hypoactive form is frequently undiagnosed even when using established tools to diagnose delirium. The implementation of these biomarkers in delirium assessment tools could represent a new approach. However, authors found inflammatory biomarkers not consistently reported as delirium risk factors. Their level of evidence should first be investigated in a meta-analysis.