Poor Olfaction Prior to Cardiac Surgery: Association with Cognition, Plasma Neurofilament Light, and Post-Operative Delirium

Abstract

Objectives:

Post-operative delirium (POD) affects up to 50% of cardiac surgery patients, with higher incidence in older adults. There is increasing need for screening tools that identify individuals most vulnerable to POD. Here, we examined the relationship between pre-operative olfaction and both incident POD and POD severity in patients undergoing cardiac surgery. We also examined cross-sectional relationships between baseline olfaction, cognition, and plasma neurofilament light (NfL).

Methods:

Individuals undergoing cardiac surgery (n=189; mean age=70 years; 75% men) were enrolled in a clinical trial of cerebral autoregulation monitoring. At baseline, odor identification performance (Brief Smell Identification Test), cognitive performance, and plasma concentrations of NfL levels (Simoa^™ NF-Light Assay) were measured. Delirium was assessed with the Confusion Assessment Method (CAM) or CAM-ICU, and delirium severity was assessed using the Delirium Rating Scale-Revised-98. The association of baseline olfaction, delirium incidence, and delirium severity was examined in regression models adjusting for age, duration of cardiopulmonary bypass, logistic European System for Cardiac Operative Risk Evaluation (EuroSCORE), and baseline cognition.

Results:

Olfactory dysfunction was present in 30% of patients, and POD incidence was 44%. Pre-operative olfactory dysfunction was associated with both incident POD (OR=3.17, p=0.001) and greater severity of POD after cardiac surgery (OR=3.94 p<0.001) in models adjusted for age, duration of bypass, and a surgical risk score. The addition of baseline cognition attenuated the strength of the association, but it remained significant for incident POD (OR=2.25, p=0.04) and POD severity (OR 2.10, p=0.04). Poor baseline olfaction was associated with greater baseline cognitive dysfunction (p<0.001) and increased baseline plasma NfL concentrations (p=0.04). Neither age, cognition, nor baseline NFL concentration modified the association of impaired olfaction and delirium outcomes.

Conclusions:

Olfactory assessment may be a useful pre-surgical screening tool for the identification of patients undergoing cardiac surgery at increased risk of POD. Identifying those at highest risk for severe delirium and poor cognitive outcomes following surgery would allow for earlier intervention and pre-operative rehabilitation strategies, which could ultimately impact the functional disability and morbidity associated with POD.

INTRODUCTION

Post-operative delirium (POD) is an acute fluctuating disturbance of consciousness and decompensated cognitive state. POD is associated with increased healthcare utilization and morbidity as well as with increased in-hospital and long-term mortality.¹ Multiple factors are implicated in the pathogenesis of POD including neuroinflammation, neurotransmitter imbalance, hypoxia, hypoglycemia, and impaired cerebral blood flow autoregulation, though baseline vulnerability is recognized as a core factor that predicts the development of POD.² As such, there is increasing need for screening tools that can help identify individuals most vulnerable to POD allowing for earlier intervention.

Impaired olfaction has gained increasing attention as a potential correlate of underlying brain vulnerability and risk of future cognitive decline in older adults. Olfactory loss at baseline is an independent predictor of transition to dementia in longitudinal studies of cognitively intact older adults.^3,4 There is also preliminary evidence that olfactory loss is associated with POD. Kim et al.⁵ examined olfactory performance preoperatively in Parkinson’s disease (PD) patients undergoing elective surgeries under general anesthesia. Poor olfaction and length of operation were significant predictors of POD, assessed using the Confusion Assessment Method. Prior work by Brown et al.⁶ was the first to examine the relationship between pre-operative olfactory performance and POD in 165 patients undergoing cardiac surgery. Using an established chart-review method, POD was observed in 33% of the patient sample. Patients with abnormal preoperative olfaction had greater baseline cognitive difficulties and significantly greater risk of POD.⁶ However, the chart-review delirium method is limited by reduced sensitivity compared to in-person psychiatric diagnoses.

In this study, we measured psychophysical olfaction, cognitive performance, and pre-operative plasma concentrations of neurofilament light (NfL), a well-known measure of neuronal injury and potential biomarker for delirium risk,⁷ in individuals undergoing cardiopulmonary bypass surgery. We also characterized the presence and severity of POD using a rigorous in-person assessment of delirium and a consensus expert diagnostic panel. Our primary aim was to examine the association between baseline olfaction and 1) incident POD and 2) severity of POD. As a secondary aim, we examined the relationship between olfaction and both baseline cognition and pre-operative plasma NfL levels. Based on prior work, we hypothesized a priori that poor baseline olfaction would be associated with 1) the presence and severity of delirium following surgery, 2) greater baseline cognitive dysfunction, and 3) increased plasma NfL.

MATERIALS & METHODS

Participants

The study was approved by the Johns Hopkins institutional review board (#00027003). All patients provided written informed consent. The current work represents a secondary analysis of data collected during a randomized controlled trial evaluating cerebral blood flow autoregulation and brain injury after cardiac surgery (NCT 00981474). In the main trial, participants were enrolled between 10/11/2012 and 5/10/2016 and were randomized one to one to blood pressure targets during cardiopulmonary bypass based on measures of cerebral autoregulation versus standard-of-care targets. Inclusion criteria were age 55+ years; undergoing primary or re-operative coronary artery bypass graft with or without valvular surgery, or ascending aorta surgery that required cardiopulmonary bypass; and high risk for neurologic complications (encephalopathy or stroke) as determined by a Johns Hopkins risk score.⁸ This score includes history of stroke, carotid artery bruit, hypertension, diabetes, and age and generally excluded patients in the lowest quartile of risk. Any patient meeting one of the following criteria were not enrolled: 1) non–English speaking, 2) inability to attend outpatient visits, 3) severe visual impairment, 4) liver dysfunction (alanine aminotransferase, aspartate aminotransferase, or alkaline phosphatase elevated to twice the upper limit of the reference range), 5) chronic renal failure including requiring dialysis, 6) contraindications to MRI, and 7) emergency surgery. The primary trial results were reported previously,^9,10 but the results of this investigation have not. During this nested cohort study, 2,764 patients were screened for eligibility criteria. In sum, 1,710 did not meet enrollment criteria, 273 declined participation, and 565 had other reasons, including staff availability. Of the 216 patients enrolled, 1 withdrew, 9 did not have baseline olfactory assessment, and 17 did not have POD assessments, leaving 189 individuals in the final analytic sample.

Perioperative Care

Standard institutional monitoring was implemented for all participants, which included radial arterial blood pressure. Midazolam (0.15 mg/kg), fentanyl (5–20 μg/kg) and isoflurane were used to induce and maintain general anesthesia with a skeletal muscle relaxant. Cardiopulmonary bypass was implemented using a nonocclusive roller pump and a membrane oxygenator. The circuit included a 40-μm or less arterial line filter. Nonpulsatile flow was maintained between 2.0 and 2.4 L/min/m2. For each participant, alpha-stat pH management was used with rewarming based on institutional standards. Nasopharyngeal temperature was maintained at less than 37°C. Propofol (20 to 75 μg/kg1 per minute) was employed for postoperative sedation until appropriate for tracheal extubation or for 24 hours after surgery. Patients requiring >24 hours of mechanical ventilation could receive fentanyl and midazolam.

Olfactory Testing

Odor identification accuracy was measured prior to surgery with the 12-item Brief Smell Identification Test^11,12 (BSIT, Sensonics Incorporated, Haddon Heights, NJ), a short form of the 40-item University of Pennsylvania Smell Identification Test (UPSIT¹³). The psychometric properties and test sensitivity are well-described.^11,12 Briefly, the BSIT is a booklet measure of odor identification that can be administered in <5 minutes. Microencapsulated odorants are embedded in 10- to 50-μm urea formaldehyde polymer microcapsules affixed to a strip on each page. Odors are released with a scratch of a vendor-supplied pencil tip. Participants identify the odorant from four possible responses in a forced-choice format. The total number of odors identified correctly were tallied. Thus, lower scores reflect poorer odor identification accuracy. The BSIT’s test–retest reliability over a 1-week interval was 0.71. Participants were classified as having impaired olfaction (BSIT ≤8 for men or ≤9 for women) using previously published normative data.¹¹

Delirium Assessment

The primary outcome was any incidence of delirium after surgery, using a consensus panel diagnosis with Diagnostic and Statistical Manual of Mental Disorders, 5^th Edition (DSM-5³) criteria. The secondary outcome was the maximum score on the 16-item clinician-rated Delirium Rating Scale-Revised-98 (DRS-R-98) severity scale.¹⁴ Delirium was assessed daily on three of the first four postoperative days using the Confusion Assessment Method (CAM) or the CAM-ICU,^15,16 followed by consensus panel adjudication using DSM-5 criteria. The CAM assessments were performed by trained research assistants and included a structured cognitive examination and open-ended queries of patients, nurses, families, and medical records for evidence of delirium. For intubated patients in the ICU, the CAM-ICU was used. On days which patients were not assessed in person (up to postoperative day 4, and generally because every single weekend day could not be staffed), a validated medical record review was used.¹⁷ Delirium severity was assessed with the DRS-R-98 with a maximum severity score of 39 points. Coma was assessed using the Richmond Agitation Sedation Scale (RASS¹⁸), with a score of −4 or −5 indicating coma. Patients who were comatose on all assessments were classified as having coma in this analysis and were censored.

Delirium assessors were formally trained by a delirium expert and psychiatrist (KN), with co-evaluation of patients every two weeks. During the parent study, κ statistics for agreement between assessors were 0.7 to 0.8, indicative of substantial agreement. During adjudication, all delirium assessments were presented by research assistants to a panel with substantial clinical and research expertise in delirium, consisting of four consultation psychiatrists and one geriatric psychiatrist. The DSM-5 criteria for delirium were rated separately by each panel member using a standardized approach.¹⁹

Cognitive Assessment

Cognitive testing was performed at baseline, typically two weeks prior to surgery. We assessed multiple cognitive functions of relevance to cardiac surgery, including attention, word generation, psychomotor processing speed, auditory-verbal and visuospatial memory. The test battery consisted of: the Rey Auditory Verbal Learning Test,²⁰ Rey Complex Figure Test,²¹ Controlled Oral Word Association Test,²² Symbol Digits Modalities Test,²³ Trail Making Test,²⁴ and Grooved Pegboard.²⁵ A composite cognitive z-score was generated from these scores, consistent with prior work.²⁶ To obtain this z-score, individual cognitive test z-scores at each timepoint were first calculated based on the mean and standard deviation of baseline tests of all patients in the parent study enrolled after delirium testing started. Second, timed tests were multiplied by −1 so that higher scores represented better performance. Third, the composite cognitive z-score was calculated by averaging the individual test z-scores.

Measurement of Neurofilament light

Baseline arterial blood was collected into glass tubes after anesthesia induction. Within two hours of collection, the samples were centrifuged at 1500g for 8 minutes and the serum was separated and stored at −80°C, with subsequent batch processing. Plasma NfL was examined by the Single Molecule Array (SIMOA)-based immunoassay technology on the HD-X platform. The Simoa^™ NF-Light Assay (Quanterix Corporation, Lexington, MA, USA) was used for NfL quantification. The assay uses a combination of monoclonal antibodies along with purified bovine NfL as a calibrator. All samples were measured in duplicate with an analytical sensitivity of <1.0 pg/mL. Samples were thawed for 30 minutes, vortexed for 10 seconds and centrifuged (10,000 G, 10 minutes, 5°C) before dilution. Samples were diluted four times per protocol. Intra-assay coefficient of variation was 6.1% and inter-assay coefficient of variation was 15% for quality control plasma sample.

Statistical Analysis

This analytic sample was based on the number of patients with available olfaction and delirium data in the parent study. Statistical analyses were conducted using SAS v9.4 (SAS Institute, Cary, NC). In our primary analyses, odor identification performance was modeled as a categorical variable (impaired vs. normal olfaction). In secondary analyses, olfaction was modeled as a continuous variable (total BSIT score). The primary outcome was any new diagnosis of delirium. Delirium severity was the maximum DRS-R-98 score across all assessments and was considered as categories (0–4, 5–8, 9–12, >12) since the DRS-R-98 score distribution was skewed. Variables included in regression models were considered a priori and included age, duration of cardiopulmonary bypass, logistic EuroSCORE (European System for Cardiac Operative Risk²⁷), and baseline cognitive status. Logistic regression was used for the primary delirium outcome, and ordinal logistic regression was used for the delirium severity outcome. Results were stratified by median age and composite cognitive z-scores, and an interaction p-value was used to determine the significance of interaction terms in the models examining associations of impaired olfaction with delirium and delirium severity. P-values < 0.05 were considered statistically significant in all analyses.

RESULTS

The final analytic sample was 189 patients. The mean age was 70 ± 7.6 years and 75% were men. Patient and perioperative characteristics for the overall sample and by olfactory status are described in Table 1. The median olfaction score at baseline was 10 (IQR 3). Based on population cutoffs, 30% (57/189) of our patient had impaired olfaction. The incidence of delirium in the cohort was 44% (83/189), and the median delirium severity score in the cohort was 8 (IQR 3).

Table 1.

Sample Characteristics

	All Patients	Normal Olfaction	Impaired Olfaction	p-value
Sample size, n (%)	189 (100.0)	132 (69.8)	57 (30.2)	--
Age (years), mean (SD)	70.1 (7.6)	69.2 (7.2)	72.3 (8.0)	0.009
Men, n (%)	141 (74.6)	100 (75.8)	41 (71.9)	0.579
Race, n (%)				0.116
White	152 (80.4)	111 (84.1)	41 (71.9)
Black	26 (13.8)	14 (10.6)	12 (21.1)
Other	11 (5.8)	7 (5.3)	4 (7.0)
Comorbidities, n (%)
Prior Stroke	23 (12.5)	17 (13.3)	6 (10.7)	0.628
Hypertension	176 (93.1)	122 (92.4)	54 (94.7)	0.564
Atrial Fibrillation	50 (26.5)	33 (25.0)	17 (29.8)	0.490
Myocardial Infarction	59 (31.2)	39 (29.6)	20 (35.1)	0.450
COPD	18 (9.6)	12 (9.2)	6 (10.5)	0.770
Tobacco (current)	16 (8.6)	10 (7.7)	6 (10.5)	0.524
Diabetes	92 (48.7)	60 (45.5)	32 (56.1)	0.177
Anemia	81 (43.1)	53 (40.5)	28 (49.1)	0.270
EuroSCORE, median (IQR)	4.6 (6.5)	4.12 (5.2)	5.74 (10.3)	0.011
Baseline Cognitive z-score, mean (SD)	-0.02 (0.7)	0.13 (0.6)	-0.37 (0.7)	<0.0001
Surgery, n (%)				0.715
CABG	96 (51.1)	64 (48.48)	32 (57.1)
CABG +Valve	33 (17.6)	24 (18.18)	9 (16.1)
Valve	56 (29.8)	42 (31.82)	14 (25.0)
Other	3 (1.6)	2 (1.52)	1 (1.8)
CPB duration (min), median (IQR)	115 (62.0)	116 (64.0)	105 (60.0)	0.511

Note: COPD = Chronic Obstructive Pulmonary Disease; EuroSCORE= Logistic European System for Cardiac Operative Risk Evaluation; CABG= Coronary Artery Bypass Graft; CPB= Cardiopulmonary bypass

Baseline Olfaction, Cognition, and Neurofilament Light

Compared to patients with normal olfaction, patients with impaired olfaction had lower composite cognitive z-scores (−0.37±0.71 vs. 0.13±0.59; p<0.0001). When considered as a continuous variable, worse baseline olfaction was linearly associated with worse baseline cognitive performance on a battery of neuropsychological tests (r=0.35, p<0.0001; Supplemental Figure 1A).

Compared to patients with normal olfaction, patients with impaired olfaction had higher baseline plasma NfL concentrations (median 23.0 pg/mL [IQR=15.0, 34.1] vs. 17.0 pg/mL [IQR=12.3, 24.6]; p=0.0038). When considered as a continuous variable, worse baseline olfaction was linearly associated with higher baseline blood NfL concentrations in the range of olfaction scores considered normal based on population cutoffs (r=−0.25, p=0.001; Supplemental Figure 1B).

Baseline Olfaction and Post-Operative Delirium

Patients with baseline impaired olfaction had a greater incidence of POD compared to patients with normal olfaction (61% [35/57] vs. 36% [48/132], p=0.0015) (Figure 1A). As shown in Table 2, the adjusted odds of delirium were higher in patient with impaired vs. normal olfaction after adjusting for age, duration of cardiopulmonary bypass, and logEuroSCORE (OR 3.17, 95% Confidence Interval (CI)=1.58, 6.35, p=0.001). The addition of baseline cognition to the model did attenuate the results, but statistical significance remained (OR 2.25, 95%CI 1.05, 4.85, p=0.038). When olfaction was considered as a continuous variable (score of 0–12), each unit decrease in baseline olfaction score was associated with increased delirium when adjusted for age, duration of cardiopulmonary bypass, and logEuroSCORE (OR 1.18, 95%CI 1.02, 1.36, p=0.03), but the model was no longer statistically significant when baseline cognition was added as a covariate (Table 2).

Figure 1.

Comparison of (a) delirium incidence and (b) delirium severity scores between patients with normal and impaired odor identification ability

Table 2.

Association of Baseline Olfaction with Incident Delirium

	Unadjusted model		Adjusted model 1a		Adjusted model 2b
	OR (95% CI)	P-value	OR (95% CI)	P-value	OR (95% CI)	P-value
Impaired vs. Normal Olfactionc	2.78 (1.47, 5.28)	0.002	3.17 (1.58, 6.35)	0.001	2.25 (1.05, 4.85)	0.04
Continuous Olfaction Scored	1.15 (1.01, 1.31)	0.03	1.18 (1.02, 1.36)	0.03	1.07 (0.91, 1.26)	0.43

^aModel 1: adjusted by age, EuroSCORE, and cardiopulmonary bypass duration

^bModel 2: adjusted by age, EuroSCORE, cardiopulmonary bypass duration, and baseline cognition z-score

^cImpaired olfaction is defined as a Brief Smell Inventory Test Score ≤8 for men; ≤9 for women

^dPer unit decrease in score

Baseline Olfaction and Post-Operative Delirium Severity

Patients with impaired olfaction at baseline had higher postoperative maximum delirium severity scores compared to patients with normal olfaction (median, 10 [IQR 3] vs. median, 7 [IQR 6], p<0.001; Figure 1B). As shown in Table 3, the odds of being in a greater quartile of delirium severity were greater for patients with impaired vs. normal olfaction in all models, including models adjusted for age, logEuroSCORE, duration of cardiopulmonary bypass, and baseline cognition (OR 2.10, 95%CI 1.04, 4.25, p=0.038). When olfaction was considered as a continuous variable (score of 0–12), each unit decrease in baseline olfaction score was associated with greater odds of being in a higher quartile of delirium severity in unadjusted models and models adjusted for age, duration of cardiopulmonary bypass, and logEuroSCORE. However, the addition of baseline cognition to the model attenuated these results, which were no longer significant.

Table 3.

Association of Baseline Olfaction with Increasing Category of Delirium Severity^a

	Unadjusted model		Adjusted model 1b		Adjusted model 2c
	OR (95%CI)	P-value	OR (95%CI)	P-value	OR (95%CI)	P-value
Impaired vs. Normal Olfactiond	4.00 (2.19, 7.30)	<0.001	3.94 (2.07, 7.51)	<0.001	2.10 (1.04, 4.25)	0.038
Continuous Olfaction Scoree	1.27 (1.12, 1.43)	<0.001	1.26 (1.11, 1.44)	<0.001	1.10 (0.96, 1.28)	0.18

^aCategories of delirium severity are defined by the DRS-R-98 score: 0–4, 5–8, 9–12, >12. Thus, the interpretation of each odds ratio is the odds of being in a higher category of delirium severity.

^bModel 1: adjusted by age, EuroSCORE, and cardiopulmonary bypass duration

^cModel 2: adjusted by age, EuroSCORE, cardiopulmonary bypass duration, and baseline cognition z-score

^dImpaired olfaction is defined as a Brief Smell Identification Test Score ≤8 for men; ≤9 for women

^ePer unit decrease in score

Stratified and Sensitivity Analyses

The associations of baseline olfaction and POD and delirium severity were examined in stratified analyses based on median values of age, baseline NfL concentrations, and baseline cognition. There were no differences in the association of baseline olfaction and either POD or delirium severity by any subgroup (all p-interaction values for median age, baseline NFL concentration, and baseline cognition>0.05). Additionally, we included sex as a covariate in all adjusted models described in Tables 2 and 3 as a post hoc sensitivity analysis, and all inferences were unchanged.

DISCUSSION

The current findings replicate and extend prior work from our group linking poor baseline olfaction to incident POD in patients undergoing cardiac surgery.⁶ In the present study, delirium diagnosis was strengthened with the use of a rigorous in-person assessment of both delirium and delirium severity, followed by a consensus diagnostic panel. We found that pre-operative olfactory impairment was significantly associated with both incident POD and greater severity of POD after cardiac surgery, although the strength of the associations were attenuated by the inclusion of baseline cognition in models where olfaction was assessed as a continuous variable. Furthermore, poor baseline olfaction was associated with other markers of brain vulnerability, including greater baseline cognitive dysfunction and increased pre-operative plasma NfL concentrations.

POD can affect up to 50% of patients undergoing cardiac surgery, with the highest incidence reported in older adults.¹ Although factors that trigger POD have not been fully elucidated, predisposing patient vulnerability has been put forth as a critical determinant in the development of POD. Our findings linking olfaction to POD may be explained by multiple factors. Numerous large-scale community-based studies have demonstrated that olfactory impairment is an independent predictor of future cognitive decline^3,4 and all-cause mortality.²⁸ Olfactory dysfunction can represent acute or more lasting sequela of conditions ranging from sinonasal disease and post-viral infection to head injury and neurodegenerative diseases like Alzheimer’s and Parkinson’s disease. Thus, olfactory loss may serve as a bellwether of underlying brain vulnerability. Olfactory impairment in older adults is also independently associated with other risk factors for delirium, including increased age, frailty, cardiovascular disease burden, depression and poor nutritional intake.^29–32 Inflammation is thought to play a role in the pathogenesis of delirium, and an observational study in a surgical cohort found that patients with poor baseline olfaction had significant elevations in IL-1β, IL-6, TNF-α pre- and post-anesthesia and in IL-1β and IL-6 levels from pre- to post-surgery.³³ Retrospective longitudinal studies have also demonstrated that subclinical atherosclerosis may be a risk factor for the development of olfactory dysfunction.³⁴

In our cross-sectional analysis of patients who were assessed prior to surgery, olfactory impairment was associated with greater cognitive dysfunction. There are several mechanisms that may underlie these findings. Accelerated olfactory loss is observed in the prodrome of Alzheimer’s disease (AD) and Lewy body disease,^35,36 and across a 15-year period, older adults with increased AD pathology showed higher within-person coupling of olfaction and memory decline.³⁷ In persons with mild cognitive impairment and AD, olfactory impairment is associated with reduced volumes of the olfactory bulb, olfactory cortex, amygdala, entorhinal cortex and hippocampus.^38–41 There is also evidence linking olfactory deficits in older adults to amyloid positivity, increased tau in temporoparietal regions and elevated plasma levels of IL-6.^42–44 Though a range of genetic and environmental factors contribute to olfaction, demographic factors and baseline cognitive abilities may explain part of the variance in odor identification ability. The ability to assign a verbal label to an odor engages semantic memory and executive functioning, whereas measures of odor detection threshold are less influenced by cognitive factors.⁴⁵ In a recent study, Lan et al.³³ assessed odor identification, odor detection threshold and cognition in older adults pre- and post-abdominal surgery. Odor detection thresholds were not associated with cognition, whereas odor identification and memory performance were associated. In sum, cardiac surgery patients with impaired olfaction may have increased atrophy of olfactory-eloquent brain regions which overlap with medial temporal lobe regions subserving memory and other key cognitive functions.

We found associations between impaired baseline olfaction and higher pre-operative plasma NfL concentrations. Plasma NfL has gained increasing attention as a blood-based biomarker of neuroaxonal injury, with increased NfL concentrations reported in POD.^7,46,47 In non-surgical patient populations, poor olfaction is associated with elevated plasma NfL concentrations in patients with Parkinson’s disease and multiple sclerosis.^48,49 In community-based cohort studies, plasma concentrations of NfL are associated with worse cognition in cross-sectional analyses and with worsening cognition over time. Our results of an association between impaired olfaction and blood levels of NFL further suggests that impaired olfaction provides insight into brain vulnerability prior to surgery.

In the current study, adjusting for baseline cognitive performance attenuated the association between baseline olfaction impairment and incident POD. Indeed, olfactory impairment is a demonstrated vulnerability marker for cognitive decline in older adults.⁵⁰ As noted previously, odor identification performance has overlapping yet distinct components with respect to cognitive performance on measures of attention, memory, language, and executive functioning. As such, the strength of the association observed between odor identification performance and cognition, though not reflective of multicollinearity, likely contributed to the attenuation of these associations when cognition was included in adjusted models. Interestingly, inclusion of baseline cognition did not diminish the relationship between poor olfaction and incident POD in our previous study. In that study, delirium was diagnosed through review of the medical record, which may be less sensitive for hypoactive delirium and preferentially identify patients with hyperactive delirium. In the current study, impaired olfaction was independently associated with greater delirium severity, a score which is weighted towards hyperactive delirium. Thus, in both studies, impaired olfaction appears to be associated with more severe and hyperactive rather than hypoactive forms of delirium, independent of cognition. On the other hand, the use of a rigorous assessment and consensus-based diagnosis of delirium in the current study resulted in greater sensitivity of overall delirium diagnosis and more inclusion of hypoactive forms of delirium. With the inclusion of more hypoactive delirium in the outcome definition, baseline cognition did attenuate the association of impaired olfaction with overall delirium. Taken together, these results imply that the association of impaired olfaction and delirium may differ by delirium subtype and/or severity, and this may provide insight to the pathophysiology of different delirium subtypes.

The novelty and strengths of this study include the formal assessment and consensus diagnosis of delirium, the inclusion of a robust battery for characterization of cognition, and the measurement of plasma NfL concentrations, a surrogate marker of neuroaxonal injury. There are several limitations to our results to consider. Though a short olfactory screening tool was chosen to minimize patient burden during the perioperative period, a broader assessment of olfactory function could inform what olfactory domain and test best captures risk for POD. Patients may also represent a selected population that are at high risk for delirium, given the screening criteria and high incidence of delirium in the study. Furthermore, less than 10% of our surgical population was enrolled which limits the generalizability of our findings. As with all observational studies, there is a potential for confounding variables (either measured or not measured) that might affect the results.

CONCLUSIONS

Collectively, our findings indicate that olfactory testing may provide insight into brain vulnerability, cognitive dysfunction, and delirium risk in individuals undergoing cardiac surgery. Given the associations of olfaction impairment with both incident delirium and risk factors for delirium, olfactory assessment may enhance current screening methods aimed at identifying and stratifying cardiac surgery patients at risk for severe POD. Identifying those at highest risk for severe delirium and poor cognitive outcomes following surgery would allow for earlier intervention and pre-operative rehabilitation strategies, which could ultimately impact the functional disability and morbidity associated with POD. Future studies evaluating olfactory dysfunction in delirium risk prediction models and with additional neurochemical and radiologic biomarkers of delirium are warranted.

Highlights:

• In this prospective nested cohort study of 189 cardiac surgery patients, impaired smell prior to surgery was associated with incident delirium and greater delirium severity after cardiac surgery.

• Poor baseline smell was associated with greater baseline cognitive dysfunction and increased baseline concentration of plasma neurofilament light, a measure of neuronal injury.

• Pre-operative smell testing may provide insight into brain vulnerability, cognitive dysfunction, and delirium risk in individuals undergoing cardiac surgery and could inform strategies for patient risk stratification and early identification of individuals at risk for post-operative delirium.

Data Availability:

Per the charter of the original grant submission, the database from this study will be made available by the senior author upon reasonable request with a clear statement of purpose and aims and with an institutional data use agreement.