Biomarkers of neurodegeneration and neural injury as potential predictors for delirium

Abstract

Objectives:

Determine if biomarkers of Alzheimer’s disease and neural injury may play a role in the prediction of delirium risk.

Methods:

In a cohort of older adults who underwent elective surgery, delirium case-no delirium control pairs (N=70, or 35 matched pairs) were matched by age, sex and vascular comorbidities. Biomarkers from CSF and plasma samples collected prior to surgery, including amyloid beta (Aβ)₄₂, Aβ₄₀, total (t)-Tau, phosphorylated (p)-Tau¹⁸¹, neurofilament-light (NfL), and glial fibrillary acid protein (GFAP) were measured in cerebrospinal fluid (CSF) and plasma using sandwich enzyme-linked immunosorbent assays (ELISAs) or ultrasensitive single molecule array (Simoa) immunoassays

Results:

Plasma GFAP correlated significantly with CSF GFAP and both plasma and CSF GFAP values were nearly two-fold higher in delirium cases. The median paired difference between delirium case and control without delirium for plasma GFAP was not significant (p=0.074) but higher levels were associated with a greater risk for delirium (odds ratio 1.52, 95% confidence interval 0.85, 2.72 per standard deviation increase in plasma GFAP concentration) in this small study. No matched pair differences or associations with delirium were observed for NfL, p-Tau 181, Aβ₄₀ and Aβ₄₂.

Conclusions:

These preliminary findings suggest that plasma GFAP, a marker of astroglial activation, may be worth further investigation as a predictive risk marker for delirium.

I. Introduction

Delirium, characterized by the acute onset of confusion, inattention, disorganized thought processes, and altered levels of consciousness, is a potentially serious condition among hospitalized older adults, frequently precipitated by illness, surgery, medications, and other factors. Delirium is associated with poor outcomes, including prolonged length of hospital stay, institutionalization, functional and cognitive decline, and death.^1,2 Persons over the age of 65 who develop delirium have a 31% increased risk over the subsequent 5 years of developing dementia³, while those with underlying dementia are 5 times more likely to develop delirium compared to those without underlying cognitive impairment¹. Delirium is associated with an acceleration in long-term cognitive decline (LTCD) in persons with^4–8 and without dementia^9,10. It remains unclear if delirium occurs as a consequence of underlying brain vulnerability, accelerates asymptomatic or “preclinical” AD/ADRD through direct neural injury, or both.^2,11 Fundamental understanding of the interface of delirium and Alzheimer’s Disease (AD) and AD-Related Dementias (AD/ADRD) may provide an important opportunity to advance our conceptualization of both disorders.

Well-established CSF biomarkers for AD - reduced levels of amyloid β (Aβ) ₄₂ protein and elevated levels of total (t)-Tau and phosphorylated (p)-Tau¹⁸¹¹² have been studied relative to delirium with variable findings^13–18. For example, some studies have associated low preoperative CSF Aβ₄₀/Tau and Aβ₄₂/Tau ratio with higher incidence and greater symptom severity of postoperative delirium¹⁶, and higher CSF t-Tau levels have been associated with delirium incidence¹⁵ and severity¹⁹. In contrast, other studies found preoperative CSF Aβ₄₂, t-Tau, and p-Tau¹⁸¹ levels did not differ between those with and without delirium¹⁷. Studies examining plasma AD biomarkers have found an association between p-Tau¹⁸¹ and postoperative delirium²⁰ and delirium severity²¹.

Biomarkers associated with neural injury are also of interest in dementia and delirium pathology. Neurofilament light (NfL) is a protein that provides structural support for myelinated axons. NfL levels increase in both CSF and plasma proportionally to the degree of axonal damage²² and have been shown to increase following stroke, traumatic brain injury and neurodegenerative conditions including AD. NfL has been reported to be increased after delirium ^23–25. Glial fibrillary acidic protein (GFAP), expressed by astrocytes and upregulated in reactive astrogliosis has also been shown to be elevated in neurodegenerative conditions including AD ²⁶. However, neither plasma ²¹. or CSF²⁷ GFAP has been demonstrated to be increased in delirium.

In this paper, we examine the relationship of CSF and blood biomarkers for AD and neural injury with delirium, specifically Aβ₄₂, Aβ₄₀, Aβ_42/40 ratio, p-Tau¹⁸¹, NfL, and GFAP, sampled preoperatively in a matched cohort of older adults undergoing elective orthopedic surgery. We hypothesized that these biomarkers reflect underlying brain vulnerability and, therefore, would be associated with an increased risk of delirium. Among these potential biomarkers, our aim was to identify the blood biomarker that was: (1) closely correlated with CSF levels; and (2) most strongly associated with post-operative delirium.

II. Materials and Methods

Study Participants

Eligible participants were age ≥65 years, English speaking, with an anticipated hospital length of stay of at least 24 hours, scheduled for total hip or knee replacement under spinal anesthesia, who were enrolled in one of two cohorts that following the same procedures: the Role of Inflammation after Surgery for Elders (RISE) cohort between April 26, 2017 and February 13, 2019 or in the Successful Aging after Elective Surgery (SAGES II) cohort between April 1, 2019 and June 13, 2022, both described in detail previously^28,29. For this project, a nested-case control design was chosen. Cases who experienced postoperative delirium (n=35) and controls without delirium (n=35), defined as patients with no full or subsyndromal delirium on any post-operative day (POD), were matched on 3 factors: age (within 5 years); sex; and presence of any vascular comorbidity (cerebrovascular, cardiovascular, peripheral vascular disease, or diabetes with end-organ damage). These matching factors were selected as potential confounders associated with both AD biomarkers and delirium risk, as used in our prior studies ^25,30,31 Written informed consent for study participation was obtained according to procedures approved by the institutional review boards of the Beth Israel Deaconess Medical Center and Brigham and Women’s Hospital, the 2 study hospitals, and Hebrew SeniorLife, the study coordinating center, which were all located in Boston, Massachusetts.

Timing of interviews and study variables

After an initial telephone screening interview and medical record review, an in home baseline cognitive and functional assessment was conducted. Variables were collected at pre-surgical baseline by patient interview or chart review, and included age, sex, race, education, marital status, and vascular comorbidity (myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, diabetes, and diabetes with end organ damage). Full details on these study variables have been provided previously^28,29

Assessment of Delirium and delirium-related outcomes

Delirium was assessed daily during hospitalization by trained research staff using the long-form Confusion Assessment Method (CAM)³² combined with a validated chart review method^33,34. The CAM was completed using information from patient interviews that included a brief cognitive screen (orientation, short-term recall, sustained attention)³⁵, the Delirium Symptom Interview³⁶ and information related to acute changes in mental status noted by nurses or family members. The CAM has high inter-rater reliability and sensitivity (94%) and specificity (89%) when compared to reference standard diagnoses³⁷ and is widely used as a standardized method for delirium identification. Patients were classified as delirious if either CAM or chart criteria were met on one or more hospital days.

Specimen Collection

Blood was collected in heparinized tubes, then centrifuged to separate plasma from cellular material. The plasma was aliquoted into at least 10 subsamples, labeled and color coded, split, and stored in three −80°C freezers at two different locations. Rigorous quality control standards have been followed to ensure the integrity of RISE and SAGES biospecimens. CSF sampling was performed by the assigned anesthesia team led by experienced anesthesiologists in the operating room under standard monitoring, immediately before the administration of spinal anesthesia for the planned surgical procedure. After confirming intrathecal placement and adequate CSF flow, 1.0–1.5 ml of CSF was collected by aspiration or dropwise collection directly into the collection tubes. To minimize potential contamination of CSF sample with blood, the collection tubes were centrifuged at 1000 rcf for 10 minutes then CSF was aliquoted into 200 µL subsamples, labeled, and stored in 0.5 mL polypropylene tubes at −80°C until analyzed.

Biomarker Assays

CSF p-Tau¹⁸¹,Aβ₄₀, and Aβ₄₂ were measured in singlicate using a TECAN Freedom EVO liquid handler (TECAN, Switzerland) using the EUROIMMUN p-Tau¹⁸¹ (EQ 6591-9601-L), Aβ₄₀ (EQ 6511-9601-L), and Aβ₄₂ (EQ 6521-9601-L) assays. CSF GFAP and NfL were measured in duplicate on a fully automated Quanterix HD-X analyzer (Quanterix, Billerica, MA) using the Neurology 2-Plex B (193520) assay. Plasma pTau¹⁸¹, Aβ₄₀, Aβ₄₂, GFAP, and NfL were measured in duplicate on the Quanterix HD-X analyzer using the pTau¹⁸¹ Advantage V2 (103714) and the Neurology 4-Plex E (103670) assay. All assays were performed according to manufacturer’s protocol. The EUROIMMUN p-Tau¹⁸¹, Aβ₄₀, and Aβ₄₂ assays have been assessed independently in a large, clinically diverse cohort study of AD (MassGeneral Institute for Neurodegenerative Disease [MIND] biobank, n=730). The laboratory-based analytic and clinical performance has been validated and normative/diagnostic cut-off values have been established. The EUROIMMUN assays are sandwich enzyme-linked immunosorbent assays (ELISAs) and the Quanterix assays are ultrasensitive single molecule array (Simoa) immunoassays.

Samples were thawed and clarified with brief centrifugation, diluted per kit protocol, and applied to the wells. EUROIMMUN assays were read on the TECAN Infinite F50 microplate reader and analyzed using the Magellan software (TECAN, Switzerland). Quanterix Simoa assays were run on the HD-X automated platform and analyzed using HD-X software. An 8-point standard curve was prepared using calibrator stock serially diluted in diluent buffer. Quality control samples (supplied in the kit) and reference pooled CSF and plasma samples (MIND biobank) were included on each plate to monitor accuracy and coefficients of variation (CV).

Statistical Analysis

We reported the mean and standard deviation for the distribution of the biomarkers. Spearman correlations³⁸ between plasma and CSF biomarkers were also reported. To account for the matched design, we first computed the difference of the biomarker concentration between the delirium and non-delirium subjects in each matched pair. We report the median estimate of these paired differences, along with their interquartile ranges. We also used the Wilcoxon Signed Rank test³⁹ to test the null hypothesis that the median estimate was equal to zero. We also computed the fold difference (ratio of the biomarker concentration in the delirium case to that of the control) for each pair, and reported the mean fold difference estimate. We modeled the effect of the biomarker on the probability of delirium incidence (Peak CAM-S) by using conditional logistic regression models^40,41 which took into account the matched pair design, and reported the estimates of the odds ratios and 95% confidence interval. In the analysis to assess the effects of the biomarkers on delirium severity, we first examined the distribution of the delirium severity. We use the general linear model⁴² to obtain the estimated mean change and 95% confidence interval of delirium severity per one standard deviation increase of the biomarker. Since these models correlate continuous biomarker levels of delirium severity (and did not use the matched pairs design), covariable adjustment was used for age, sex, and vascular comorbidity.

III. Results

Characteristics of the full sample (n=70) and the matched cases with delirium (n=35) and controls without delirium (n=35) are described in Table 1. The matched sample included older adults who were on average 74.7 (6.9) years old [mean(sd)]. Sixty-three percent underwent knee replacement surgery and 37% had hip replacement surgery. Patients were highly educated, with an average of 15.1(3.3) years [mean(sd)] of education, 80% were women, 47% were married and 20% identified as non-white. The matched sample was generally healthy with an average Charlson score of 1.5(2.0). Preoperative serum creatinine was 1.2(0.6) mg/dL [mean(sd)] in men and 1.2 (0.7) mg/dL in women.

Table 1.

Characteristics of the Study Sample

Study variables	Full sample (N=70)	Delirium (n=35)	No Delirium (n=35)
Age, years, mean, (SD)*	74.7 (6.9)	74.9 (6.9)	74.6 (6.9)
Female, n (%) *	56 (80)	28 (80)	28 (80)
Nonwhite, n (%)	14 (20)	13(37)	1(3)
Education, years, mean (SD)	15.1 (3.3)	14.3 (3.4)	15.9 (3.1)
Married, n (%)	33 (47)	13 (37)	20 (57)
Vascular comorbidity *	32 (46)	16 (46)	16 (46)
Surgical type
Knee replacement, n (%)	44 (63)	24 (69)	20 (57)
Hip replacement, n (%)	26 (37)	11 (31)	15 (43)

Abbreviations: SD=standard deviation

CSF assays for Aβ₄₀, Aβ₄₂, and p-Tau¹⁸¹ were previously described as having mean replicate CVs < 10% (Aβ₄₀: 1.86%, Aβ₄₂: 2.3%, p-Tau¹⁸¹ 4.2%), demonstrating excellent technical performance. For CSF GFAP and NfL, mean replicate CV across all samples was 4.43%. For plasma p-Tau¹⁸¹, all samples measured had replicate CVs < 20%, with a mean replicate CV of 3.56%. Plasma Aβ₄₀, Aβ₄₂, GFAP, NfL had replicate CVs < 20%, with a mean replicate CV of 2.38%. All CSF and plasma samples were run on a single plate, so plate-to-plate adjustment was not required.

Correlation of CSF and blood biomarkers is presented in Table 2. The correlation between plasma and CSF was 0.42 for GFAP, 0.60 for NfL and 0.43 for p-Tau (p<0.05), whereas the correlation was only 0.16 for Aβ₄₀ and 0.25 for Aβ₄₂. Plasma GFAP correlated with CSF NfL (0.35), p-Tau^{181 (}0.37), Aβ₄₀ (−0.33) and Aβ₄₂ (−0.46) while CSF GFAP correlated with plasma NfL (0.35), p-Tau^{181 (}0.41), Aβ₄₀ (0.26) and Aβ₄₂ (0.25). CSF GFAP also correlated with CSF NfL and p-Tau¹⁸¹, and plasma GFAP correlated with NfL and p-Tau¹⁸¹ (Data not shown).

Table 2.

Correlation of CSF and plasma biomarkers

		CEREBROSPINAL FLUID
		GFAP	NfL	p-Tau181	Aβ40	Aβ42
PLASMA	GFAP	0.42*	0.35*	0.37*	−0.33*	−0.46*
	NfL	0.35*	0.60*	0.37*	−0.05	−0.11
	p-tau181	0.41*	0.44*	0.43*	−0.14	−0.29*
	Aβ40	0.26*	0.10	0.09	0.16	0.18
	Aβ42	0.25*	0.16	0.004	0.17	0.25*

Spearman Correlation Coefficients, N=70 (35 matched delirium-no delirium pairs)

^*p<0.05

Aβ₄₀ = amyloid β40; Aβ₄₂ = amyloid β42; GFAP = glial fibrillary acid protein; NfL = neurofilament light

Concentrations of biomarkers and the median pair differences (MPD) of the biomarker level between each delirium case and no delirium control are presented in Table 3. In CSF, the MPD was not significant for GFAP, NfL, Aβ₄₀, Aβ₄₂, and p-Tau¹⁸¹. There was no significant difference in the MPD for these biomarkers in plasma; for GFAP the median difference between each delirium and no delirium pair was 47.4, with an interquartile range of −42.4 and 123.9 (p=0.074).

Table 3.

Distribution of Biomarkers overall and by Postoperative Delirium Status

	Overall Distribution (N=70)1	Delirium (n=35)1	No Delirium (n=35)1	Median Pair Difference (35 Pairs) (IQR)2	P-value
CSF biomarker (pg/ml)
GFAP	11130 (9079)	12891.6 (11779)	9368.7 (4707.6)	1119.6 (−2659.9,8151.2)	0.162
NfL	1025 (558.3)	1107.2 (621)	942.5 (482.6)	75.33 (−206.4,461.9)	0.206
p-Tau 181	61.7 (33.1)	62.5 (29.8)	60.9 (36.5)	3.19 (−20.9,26.4)	0.642
Aβ 40	7307 (2663)	6950.5 (2643.2)	7663.6 (2673.1)	−1382.6 (− 2956.8,1253.1)	0.138
Aβ 42	849.9 (385.5)	786.6 (370.9)	913.3 (394.6)	−260.3 (−617.2,330)	0.183
Plasma biomarker (pg/ml)
GFAP	171.6 (134.4)	198.7 (164.8)	144.4 (89.4)	47.4 (−42.4,123.9)	0.074
NfL	32.8 (21.2)	33.0 (20.5)	32.5 (22.2)	0.6 (−5.8,15.1)	0.630
p-Tau 181	2.8 (1.7)	3.0 (1.8)	2.6 (1.6)	0.3 (−0.5,1.2)	0.224
Aβ 40	130.5 (40.0)	134.3 (36.6)	126.7 (43.5)	10.7 (−16.1,34.3)	0.117
Aβ 42	8.8 (3.4)	8.5 (2.8)	9.2 (3.9)	−1.2 (−2.7,2.4)	0.430

¹Mean and Standard Deviation

²Wilcoxon Signed Rank test

Aβ₄₀ = amyloid β40; Aβ₄₂ = amyloid β42; GFAP = glial fibrillary acid protein; NfL = neurofilament light

Among all the biomarkers, GFAP had nearly two-fold higher values in delirium cases, with a 1.95− and 1.90-fold difference, in plasma and CSF respectively (Table 4). Other plasma and CSF biomarkers were also higher in delirium cases compared to control cases without delirium, but the magnitude was smaller than the changes in GFAP. Plasma GFAP had an odds ratio (OR) for postoperative delirium of 1.52 per each 1 standard deviation increase in GFAP concentration [95% confidence interval (CI) 0.85,2.72] (Table 4). OR for delirium with CSF GFAP was 1.72 (95% CI (0.85,3.49). Plasma p-Tau¹⁸¹ and CSF NfL had OR of 1.38 and 1.59, respectively.

Table 4.

Association between Biomarkers and Risk of Postoperative Delirium via Conditional Logistic Regression

Biomarker	Fold difference 1	Odds Ratio2	95% C.I.
Plasma GFAP	1.90	1.52	(0.85,2.72)
Plasma NfL	1.24	1.03	(0.63,1.67)
Plasma p-Tau181	1.43	1.38	(0.77,2.49)
Plasma Aβ40	1.12	1.34	(0.73,2.48)
Plasma Aβ42	1.07	0.78	(0.46,1.33)
CSF GFAP	1.95	1.72	(0.85,3.49)
CSF NfL	1.33	1.59	(0.83,3.04)
CSF p-Tau181	1.25	1.05	(0.65,1.71)
CSF Aβ40	1.03	0.70	(0.40,1.23)
CSF Aβ42	1.08	0.71	(0.43

¹The average of the ratios of pair-specific protein level of delirium to control without delirium

²Per 1 standard deviation change in the biomarker

Aβ₄₀ = amyloid β40; Aβ₄₂ = amyloid β42; GFAP = glial fibrillary acid protein; NfL = neurofilament light

No association between the biomarkers and delirium severity as measured by the increase in peak CAM-S score was found (Table 5). For plasma GFAP, for each 1 standard deviation increase in GFAP concentration, there was an increase of 0.60 in peak CAM-S (−0.02–1.22); for CSF GFAP the increase in peak CAM-S was 0.38 (95%CI (−0.24–1.01). Plasma p-Tau¹⁸¹ increased peak CAM-3 by 0.60 (−0.12–1.32).

Table 5 .

Association between Biomarkers and Postoperative Delirium Severity

Biomarker	Increase in Peak CAM-S1	95% C.I.
Plasma GFAP Plasma NFL	0.60 0.33	(−0.02, 1.22) (−0.31, 1.03)
Plasma p-Tau181	0.60	(−0.12, 1.32)
Plasma Aβ40	0.37	(−0.32,1.06)
Plasma Aβ42	−0.15	(−0.84,0.53)
CSF GFAP	0.38	(−0.24, 1.01)
CSF NFL	0.27	(−0.36, 0.91)
CSF p-Tau181	0.43	(−0.21, 1.08)
CSF Aβ40	−0.63	(−1.28,0.01)
CSF Aβ42	−0.72	(−1.35,−0.09)

per 1 SD increase in biomarker

General Linear Models adjusting for sex, age, vascular comorbidity.

Aβ₄₀ = amyloid β40; Aβ₄₂ = amyloid β42; GFAP = glial fibrillary acid protein; NfL = neurofilament light

IV. Discussion

We examined the AD biomarkers Aβ₄₂, Aβ₄₀, and p-Tau¹⁸¹, plus the neural injury marker NFL and astrocyte activation marker GFAP in CSF and plasma in a cohort of older adults undergoing elective orthopedic surgery. As has been shown previously⁴³, we found moderate correlations (range 0.42–0.60) between CSF and plasma values for p-Tau¹⁸¹, GFAP and NfL, suggesting that plasma levels of each of these biomarkers are closely associated with their CSF levels. Moreover, GFAP also correlated with all other biomarkers between CSF and plasma. In addition, plasma GFAP had an OR for delirium of 1.52 per each 1 standard deviation increase in GFAP level. While these increased risks are clinically relevant, they did not achieve statistical significance in this small preliminary study. Based on the significant correlation between CSF and plasma levels and the strong association with delirium risk, we identified plasma GFAP as the most promising candidate plasma biomarker among those we tested for prediction of delirium.

Our a priori hypothesis was that biomarkers of brain vulnerability would be associated with risk of delirium. In this study, plasma GFAP, a marker of reactive astrogliosis and neural injury, was identified to be the most promising candidate biomarker for delirium, more so than AD biomarkers and NfL. Elevations in CSF and plasma biomarkers occurs early in AD pathogenesis and continues from preclinical to dementia stage. During neurodegeneration, astrocytes are metabolically impaired as neural injury occurs, and this leads to impaired support of neuronal metabolism ⁴⁴. Astrocytosis also promotes further immune cell infiltration. Notably, plasma GFAP has been found to increase over the course of AD, distinguish AD from other dementias, predict risk of progression, and correlate with CSF and neuroimaging biomarkers⁴⁵, providing further evidence for plasma GFAP as a potential risk and disease biomarker of AD. However, plasma GFAP levels are reported increased, at least transiently, in a variety of other conditions, including stroke, concussion and traumatic brain injury, acute COVID-19, alcohol withdrawal, glioma, and multiple sclerosis ⁴⁶. Studies of GFAP as a potential biomarker in delirium have been limited to date. For example, in a small case-series of individuals undergoing cardiac surgery, elevated serum Tau predicted delirium, although NfL or GFAP did not²³. Another study of patients undergoing cardiac surgery found no association between serum GFAP levels and risk for delirium⁴⁷, nor did we find a significant association in our prior work²⁵. However a post-mortem study did find that in patients with delirium there was more astrocytosis compared to patients without delirium, yet coexisting dementia did not affect these findings ⁴⁸.

Prior studies examining AD biomarkers and delirium have found mixed results. A recent systematic review examining diagnostic biomarkers for AD/ADRD within the National Institute on Aging-Alzheimer’s Association (NIA-AA) research framework, identified an equivocal association between amyloid biomarkers, including plasma, CSF, and tissue markers, with delirium⁴⁹. Other biomarkers, specifically p-Tau¹⁸¹, NfL, and GFAP have shown more promise as plasma-based biomarkers in AD. For example, patients with mild cognitive impairment who were more likely to progress to AD had elevated baseline plasma levels of NfL, GFAP and p-Tau^{181 50}. In delirium, elevated t-Tau²¹ and p-Tau^{181 20} have been observed. Studies examining plasma NfL have found increased NfL to be associated with delirium^24,51. Reduced kidney function has been associated with increased levels of dementia-related blood biomarkers including GFAP but not increased dementia risk⁵².

A major strength of this study is the use of matched delirium cases and controls without delirium. This is a powerful study design, which maximizes efficiency for evaluating potential biomarkers⁵³. We were also able to look simultaneously at both CSF and plasma biomarkers. Another important strength is the rigorous, validated approach for identifying incident delirium.

A few limitations should be noted. First, the study was conducted in a relatively small, matched sample, which while suitable for biomarker discovery, lacks power and limited our ability to detect statistically significant differences or to control for multiple comparisons in this study. In post-hoc power calculations, our power was only 36% to detect a statistically significant association with the sample size of N=70. Second, the matching may inadvertently select a control sample that is non-representative of the full SAGES cohort or the general surgical population, limiting generalizability. This was the case where the matched cohort included a higher proportion of non-white participants by chance (20% in matched sample, vs. 8% in full SAGES cohort ⁵⁴), and a higher delirium rate (37% in matched non-white sample, vs. 10% in non-white sample in full SAGES cohort). While we plan to investigate racial differences in future studies, our many past analyses have not demonstrated any similar differences based on race. Third, It is important to note that identification of cutpoints for biomarkers to best identify delirium was beyond the scope of this study. This will be an important next step for future studies in larger and more representative samples. Fourth, there are inherent limitations when using blood-based biomarkers to accurately reflect what is occurring in the CNS in brain disorders, although the correlation of CSF and plasma in this study helps support the plasma findings. Fifth, in delirium, there is emerging evidence for endothelial cell dysfunction and increased blood brain barrier (BBB) permeability⁵⁵. Thus, plasma GFAP level may be affected by or dependent upon the degree of “leakiness” of the BBB, an important area for future investigation. Lastly, we make the assumption that the preoperative timepoint reflects underlying brain vulnerability, such as from pre-clinical neurodegeneration or neuroinflammation from trauma, contributes to an elevated GFAP level and indicates a “vulnerable brain”. For the current study, enrollment was limited to elective joint replacement procedures, thereby limiting the confounding effects of inflammation due to trauma or other acute medical conditions. Future studies looking at other markers of neuroinflammation, both prior to and during the course of delirium are needed to fully understand these relationships.

VI. Conclusions

In summary, we found that CSF and plasma levels of GFAP were highly correlated with each other, and also associated with higher delirium risk in a small matched case-control cohort study of older adults. While power was limited, these findings provide preliminary evidence that GFAP, a marker of neural injury and brain pathology, may have value as a predictive risk marker for delirium. This study lays the groundwork for future larger studies to build support for the role of preexisting brain disease impacting vulnerability to delirium, and providing an intriguing pathophysiologic link between delirium and dementia.

Key points.

• Delirium is a common complication of hospitalization associated with poor outcomes, but the underlying pathophsysiology of this condition is poorly understood

• There is a close interrelationship between delirium and dementia, therefore CSF and blood biomarkers for Alzheimer’s Disease and neural injury might be associated with delirium

• We found that plasma and CSF glial fibrillary acid protein (GFAP), a marker of reactive gliosis known to be elevated in neurodegeneration, appears to be related to delirium risk, with a nearly two-fold greater odds of delirium

• Further larger studies are needed to understand the potential role of GFAP as a predictive risk marker for delirium

Data availability statement:

The data that support the findings of this study are available from the corresponding author upon reasonable request.