Task-Evoked Pupillary Response as a Potential Biomarker of Dementia and Mild Cognitive Impairment: A Scoping Review

Abstract

Pupil dilation functions as a proxy for cognitive effort and can be measured through automated pupillometry. The aim of this scoping review is to examine how individuals with cognitive impairment differ in task-evoked pupillary responses relative to cognitively healthy individuals. A systematic literature search across six databases was conducted to identify studies examining changes in pupillary responses evoked by cognitive tasks comparing patients with dementia to healthy controls. Eight articles met inclusion criteria and were included for review. Differences in task-evoked pupillary response between cognitively impaired and cognitively healthy participants were observed across studies. Pupil dilation is decreased in patients with Alzheimer’s Disease compared to controls, with no difference observed in patients with mild cognitive impairment. A mild, non-significant trend towards reduced pupil dilation in patients with either Parkinson’s Disease or Dementia with Lewy Bodies suggests a similar but less pronounced effect than in AD patients. Further research is required to examine the utility of task-evoked pupillary responses as a potential biomarker indexing cognitive decline in individuals transitioning to mild cognitive impairment and/or dementia.

Introduction

Dementia is defined as a decline in one or more cognitive domains leading to an impairment in activities of daily living, beyond the extent of normal aging. ¹ Alzheimer’s Disease (AD) is the most common type of dementia, while mild cognitive impairment (MCI) is often considered a prodromal syndrome to dementia, in which cognitive impairment is detectable but does not yet meet the criteria for dementia.

AD and other dementias are diagnosed on a primarily clinical basis, using a neuropsychological assessment of memory dysfunction and impairment of other cognitive domains, such as executive function, language, learning, attention, and visuospatial skills.^1,2 These evaluations commonly employ screening tests such as the Mini-Mental State Examination (MMSE) or the Montreal Cognitive Assessment (MoCA), which are clinically valuable but remain limited due to numerous potential confounds.^3,4 Further, they only assess functional cognitive impairment and do not reflect underlying pathophysiology, which may begin years prior to clinical presentation.^5,6 Imaging and advanced testing such as cerebrospinal fluid analysis are useful tools but are not diagnostic for dementia.^2,7 Definitive diagnosis can only be made upon histopathological examination, which is rarely performed in living patients due to the invasive nature of such procedures. Further investigation into alternative methods for the early assessment of MCI and dementia is therefore warranted.

Previous research has explored pupillary changes as a potential early biomarker of dementia. The seminal work by Hess & Polt (1964) first established that pupil dilation acts as a proxy for mental exertion during a mathematical task, such that pupil diameter positively correlates to cognitive effort in healthy adults. ⁸ This discovery was reinforced by Kahneman & Beatty (1966), who demonstrated incremental increases in cognitive load during a digit span task correlated to increases in pupillary diameter. ⁹ Further work by Granholm et al. (1996) found that these incremental changes in pupil dilation plateaued as the load approached a maximal cognitive capacity during a digit span task, and subsequently decreased as the cognitive load exceeded capacity. ¹⁰ This phenomenon has been consistently demonstrated across various cognitive tasks, such as visual search, attentional allocation, and language processing tasks.^11–13

The locus coeruleus (LC) within the pons modulates pupillary dilation via inhibition of the Edinger-Westphal Nucleus, stimulation of the interomedio-lateral column of the spinal cord, and stimulation of the basal forebrain.^14–16 Substantial LC degeneration is observed in both AD and Parkinson’s Disease (PD), and to a lesser extent in other forms such as frontotemporal dementia (FTD) and vascular dementia (VaD). ¹⁷ In the context of AD, LC degeneration is associated with build-up of neurofibrillary tangles early in disease progression, potentially preceding cognitive impairment^18,19. In PD, neuronal loss and alpha-synuclein pathology affects the LC earlier and to a greater extent than the nucleus basalis and substantia nigra.^20,21 In FTD, divergent proteinopathies confer differences in LC degeneration, though LC neuronal loss is reported in all FTD subtypes.^22,23

As LC degeneration is observed in multiple forms of dementia, often early in disease progression, pupillary response may serve as a proxy for LC function in assessing early pathology or incipient decline. Previous studies examining pupillometry in dementia using either anticholinergic eyedrops to induce pupil dilation or using a light reflex have reported variable effects.^24–27 Instead, pupillary response to a cognitive task may better index decline in cognitive function among people with various forms of dementia. The clinical utility of pupillometry in the evaluation of cognitive decline may be twofold: early detection of pathological processes outside the limits of normal cognitive ageing; and differentiation between dementia and non-degenerative causes of cognitive impairment.

The purpose of this brief scoping review is to present the current evidence and knowledge gaps relating to task-evoked pupillary response differences between patients with MCI or dementia and cognitively healthy adults. The research question driving this study was: Do individuals with dementia and/or mild cognitive impairment differ from cognitively healthy participants in task-evoked pupillary responses during cognitive tasks?

Methods

Search Strategy

For this review, the databases MEDLINE, Embase, APA PsycINFO, CINAHL, Scopus, and Web of Science were searched to find articles relevant to the topic, available as of September 2021. The key search terms were developed in conjunction with a health sciences librarian and included the following: (Alzheimer* or dementia or cognitive* impair* or cognitive* decline) and (pupillometry or eye track* or pupil* (dilat* or respon*)).

Study Criteria and Selection

Studies were included if (1) the patient population included MCI or any form of dementia (2) the subjects performed a cognitive task; (3) the study included at least one group of healthy controls or a reference group; (4) baseline pupil metrics were recorded or accounted for in pupillary data; (5) pupil metrics were taken during or after the cognitive task.

Studies were excluded if they (1) were non-experimental studies, reviews, or conference proceedings; (2) were not available as full-text articles in English; (3) included designs without baseline pupillary measurements; (4) included designs without pupillary measurements during or after a cognitive task. As dementia may develop later in Parkinson’s disease (PD) progression, studies including participants with PD were included if they featured a form of cognitive assessment and task.

Data Extraction

The Covidence platform was used to organize articles and assist in managing the screening, selection, and data extraction processes (Covidence Systematic Review Software, Veritas Health Innovation, Melbourne, Australia). A PRISMA flowchart of the data extraction process is presented in Figure 1. Six reviewers independently screened titles and abstracts of all articles to determine whether they were suitable for inclusion within the current study. Two reviewers independently screened full-text articles for inclusion and exclusion criteria described above. Any conflicts were resolved by a third reviewer.

Figure 1.

PRISMA flow diagram highlighting systematic review process.

The included full-text articles were divided such that two reviewers independently extracted data on the study type, control and patient population characteristics, methods, pupillometry metrics, and cognitive task outcomes using a standardised form. Conflicts in the extracted data were discussed between the two reviewers until consensus was reached. Tables containing data extraction information are available in Supplementary Table 1.

Quality Assessment

All eligible studies for review were case-control designs. A case-control specific critical appraisal tool from the Joanna Briggs Institute was used to assess the quality of the selected articles. ²⁸ Each article used in the review was assigned to two different reviewers who individually assigned scores, based on the checklist. Discrepancies between scores were discussed with the other assigned reviewer.

Results

Nine full-text articles met the inclusion criteria.^29–37 However, two studies drew populations of interest from the same sample (obtained from the Vietnam Era Twin Study of Ageing).^30,32 The Granholm et al (2017) results were chosen as they included a larger study size compared to Elman et al (2017) (128 MCI participants vs 48 MCI participants, respectively) in order to minimise the risk of patient overlap.^30,32 Altogether, eight studies were included. Figure 1 contains a PRISMA diagram outlining the search results and number of articles identified at each stage.

Quality assessment

Table 1 highlights the quality assessment scores for each study according to checklist items. Quality assessment using the JBI case-control appraisal tool demonstrated generally high scores, with no study failing to meet over half of the checklist items. Item 9 from the case-control studies checklist (“Was the exposure period of interest long enough to be meaningful?“) was excluded in our assessment. This item was determined to be not applicable to the selected studies, as the exposure period (of the cognitive task) was applied only once during the experimental session. Overall, there was an initial moderate level of agreement between raters’ scores, Fleiss’ κ = .577 (95% CI .565 - .590), P = .003. Sources of disagreement primarily stemmed from ambiguity surrounding whether confounding factors were suitably identified with strategies to address confounding factors (checklist items 6 and 7). Following discussion, a complete agreement was reached among reviewers.

Table 1.

Quality Assessment – JBI Case-Control Study Checklist.

First Author (Year)	1	2	3	4	5	6	7	8	9	10	Total Score
Dragan (2017) 29	Y	Y	N	Y	Y	N	N	U	-	Y	Y 5/9 N 3/9 U 1/9
Fletcher (2016) 31	U	Y	Y	U	Y	Y	Y	Y	-	Y	Y 7/9 N 0/9 U 2/9
Granholm (2017) 32	Y	Y	Y	Y	Y	Y	Y	Y	-	Y	Y 9/9 N 0/9 U 0/9
Podlasek (2019) 33	Y	Y	Y	Y	Y	Y	N	U	-	Y	Y 7/9 N 1/9 U 1/9
Porter (2010) 34	Y	Y	Y	Y	Y	Y	Y	Y	-	Y	Y 9/9 N 0/9 U 0/9
Ranchet (2017) 37	N	U	Y	Y	Y	U	Y	Y	-	Y	Y 6/9 N 1/9 U 2/9
Suzuki (2017) 35	Y	Y	Y	Y	Y	Y	Y	Y	-	Y	Y 9/9 N 0/9 U 0/9
Wang (2016) 36	Y	Y	Y	Y	Y	N	N	Y	-	Y	Y 7/9 N 2/9 U 0/9
Total item score	Y 6/8 N 1/8 U 1/8	Y 7/8 N 0/8 U 1/8	Y 7/8 N 1/8 U 0/8	Y 7/8 N 0/8 U 1/8	Y 8/8 N 0/8 U 0/8	Y 5/8 N 2/8 U 1/8	Y 5/8 N 3/8 U 0/8	Y 6/8 N 0/8 U 2/8	-	Y 8/8 N 0/8 U 0/8

Scoring of Items: Y = yes, N = no, U = unclear, – = N/A.

1 Were the groups comparable other than the presence of disease in cases or the absence of disease in controls? 2 Were cases and controls matched appropriately? 3 Were the same criteria used for the identification of cases and controls? 4 Was exposure measured in a standard, valid and reliable way? 5 Was exposure measured in the same way for cases and controls? 6 Were confounding factors identified? 7 Were strategies to deal with confounding factors stated? 8 Were outcomes assessed in a standard, valid and reliable way for cases and controls? 9 Was the exposure period of interest long enough to be meaningful? (Item 9 was not applicable to studies) 10 Was appropriate statistical analysis used?

Participants

The eight studies included 1198 participants, of which 917 were healthy controls. Of the 281 patient participants, 27 were classified as having Alzheimer’s disease, 43 having Parkinson’s disease, 11 having semantic dementia (SD), 13 having behavioural variant frontotemporal dementia (bvFTD), 8 having dementia with Lewy bodies (DLB), 6 as having progressive non-fluent aphasia (PNFA), 20 having dementia otherwise unspecified, and 153 having MCI. Participant characteristics for each study are described in Table 2.

Table 2.

Participant Characteristics.

First Author (year)	Participants	Mean Age (SD or Range)	Sex (M/F)	Neurocognitive Assessment	Inclusion Criteria
Dragan (2017) 29	AD 9/44 ROA 8/44 HOA 10/44 HYA 17/44	69.1 ± 7.8 69 ± 8.5 66.4 ± 8.5 22.8 ± 3.1	5/4 3/5 2/8 5/12	MoCA	MoCA ≤24: Early AD MoCA ≥26: Young, healthy
Fletcher (2016) 31	bvFTD 13/60 SD 11/60 PNFA 6/60 AD 10/60 Control 20/60	65.2 (52/76) 66.5 (53-78) 69.1 (61/77) 68.1 (54-78) 65.6 (57-71)	11/2 7/4 1/5 5/5 10/10	General neuro-psychological assessment	Fulfilling consensus criteria for dementia syndromes No neurological or psychiatric history
Granholm (2017) 32	aMCI 53/918 naMCI 47/918 mMCI 25/918 Control 793/918	62.17 ± 2.3 61.76 ± 2.5 61.6 ± 2.6 61.5 ± 2.4	918/0	18-Test neuro-psychological assessment	Wave 2 of vietnam era twin study of ageing 18 neuro-psychological tests across 6 cognitive domains, age 20 GCA, and jak-bondi actuarial-neuropsychological approach
Podlasek (2019) 33	Dementia 20/60 MCI 20/60 Control 20/60	76.4 ± 9.2 67.6 ± 14.5 62.4 ± 14.0	6/14 12/8 5/15	MoCA	MoCA ≤20 MoCA 21-25 MoCA >25
Porter (2010) 34	AD 8/21 Control 13/21	79.0 ± 3.85 78.7 ± 3.07	5/3 6/7	MMSE	Probable AD according to DSM IV, NINCDS-ADRDA Cognitively healthy
Ranchet (2017) 37	PD-CI 10/37 PD-NCI 11/37 Control 16/16	71 69 61	9/1 7/4 8/8	MoCA	MoCA 16-26 MoCA ≥26
Suzuki (2017) 35	DLB 8/17 Control 9/17	77.4 ± 6.9 71.4 ± 6.7	3/5 8/1	Clinical dementia rating; MMSE	Stage 1 dementia according to clinical dementia rating
Wang (2016) 36	PD 22/41 Control 19/41	67.4 68.6	14/8 9/10	MoCA	MoCA ≥24

ROA – at-risk older adults; HOA – healthy older adults; HYA – healthy younger adults; aMCI – amnestic single domain MCI; naMCI – non-amnestic single domain MCI; mMCI – multiple domain MCI; bvFTD – behavioural variant frontotemporal dementia; SD – semantic dementia; PNFA – progressive nonfluent aphasia; DSM IV – Diagnostic and Statistical Manual for Mental Disorders; NINCDS-ADRDA – National Institute of Neurological and Communicate Disorders and Stroke/Alzheimer’s Disease and Related Disorders Association; PD-CI – Parkinson’s disease with cognitive impairment; PD-NCI – Parkinson’s disease with no cognitive impairment; DLB – dementia with Lewy bodies.

There was considerable variability in the measures used to assess cognitive function. Four of the studies used the Montreal Cognitive Assessment (MoCA).^29,33,36,37 Two studies, those by Porter et al (2010) and Suzuki et al (2017), used the Mini-Mental State Examination (MMSE) to assess cognitive function.^34,35 Among those that utilized either MoCA or MMSE tools, cut-off scores for cognitive impairment classification varied widely (Table 2).

Types of Cognitive Tasks

Of the eight articles reviewed, four studies used cognitive tasks related to attention and working memory.^29,32–34 Four studies used cognitive tasks related to visual processing.^29,35–37 The study by Dragan et al. (2017) utilised a flicker change detection task that related to both working memory and visual processing, therefore it was included in the total for both categories. ²⁹ Studies by Ranchet et al (2017) and Wang et al (2016), which used tasks related to visual processing also examined voluntary executive function.^36,37 Fletcher et al. (2016) used a task examining auditory and semantic processing. ³¹ The specific cognitive task and corresponding categorization are described in Table 3.

Table 3.

Intervention Characteristics.

First Author (Year)	Cognitive Task	Cognitive Task Categorization	Pupillometry Equipment	Sampling Rate	Pupillometry Metrics
Dragan (2017) 29	Flicker change detection	Working memory, visual processing	iView X infrared eye-tracking system	60 Hz	Pupil diameter, peak pupil response velocity
Fletcher (2016) 31	Nonverbal auditory task	Auditory processing	EyeLink II head mounted eye-tracker	250 Hz	Change in pupil diameter relative to baseline, mean pupil maximum
Granholm (2017) 32	Digit span	Working memory	NeurOptics PLR-200 pupillometer	30 Hz	Change in pupil diameter relative to baseline
Podlasek (2019) 33	Sequence memorisation	Working memory	Gaze data explorer	120 Hz	Change in pupil diameter relative to baseline
Porter (2010) 34	Target-present/Absent search	Attention and inhibition	EyeLink II head mounted eye-tracker	N.R.	Change in pupil diameter relative to baseline
Ranchet (2017) 37	Prosaccade/Antisaccade	Visual processing, voluntary executive function	FOVIO remote eye tracker	60 Hz	N.R. – Data converted into cognitive workload index
Suzuki (2017) 35	Pareidolia	Visual processing, recognition	EMR-dFactory	60 Hz	Change in pupil diameter relative to baseline
Wang (2016) 36	Prosaccade/Antisaccade	Visual processing, voluntary executive function	EyeLink 1000	500 Hz	Pupil dilation, constriction size, time of maximum constriction

N.R. - Not Reported.

Performance on cognitive tasks was not consistently reported across studies or was not applicable according to the type of task employed. Among those directly reporting cognitive task performance, groups with cognitive impairment performed worse than controls, in that they had greater error rates or reduced accuracy.^34–36 Ranchet et al. (2017) reported no significant difference between PD and control groups. ³⁷ Any potential correlation between task performance and changes in pupillary metrics between those with impaired cognition compared to healthy older adults remains unclear.

Pupil Dilation

Reported Pupillary Metrics

All but one of the studies measured pupil dilation as a change in pupil diameter relative to baseline, while Dragan et al. (2017) primarily assessed pupil dilation velocity rather than a change in diameter. ²⁹ Secondary metrics taken included the mean maximum pupil dilation and constriction size. These metrics were not consistently measured; therefore, secondary metrics were not assessed in this review.

Differences in Pupillary Reactivity

Seven of the eight studies reported a significant difference in pupil reactivity among patient groups compared to healthy controls.^{29,31–33,35–37} Three studies reported an increase in pupil response in the patient groups,^31,35,37 while three studies reported a decreased pupil response among patient groups relative to healthy controls.^29,33,36 One study reported both increases and decreases in patient groups, ³² and one reported no differences between groups altogether. ³⁴

Pupillary Responses in Alzheimer’s Disease

Among studies involving patients with AD, Fletcher et al. (2016) reported increased pupil dilation among AD patients. ³¹ Dragan et al (2017) reported decreased pupil response velocity, while Podlasek et al (2019) reported reduced pupil dilation and Porter et al (2010) reported no significant difference when compared to healthy controls.^29,33,34 The population-weighted effect size of all studies showed a moderate significant effect among AD patients towards smaller pupil dilation (d = .485) among disease groups compared to healthy controls, as described in Figure 2a.

Figure 2.

Forest plot demonstrating effect size (Cohen’s d) weighted for each study according to patient type with corresponding overall effect sizes for each group. A positive integer indicates reduced pupil reactivity (dilation or velocity) among the patient group compared to control participants, whereas a negative integer indicates greater pupil reactivity among the patient group. The overall effect size is the mean Cohen’s d weighted by the sample sizes of each individual study. Patients were categorized into disease types according to how they were defined in their original studies. A. Alzheimer’s Disease, B. Parkinson’s Disease and Dementia with Lewy Bodies, C. Mild Cognitive Impairment.

Pupillary Responses in Parkinson’s Disease and Lewy-Body Dementia

Two studies included PD patients; Ranchet et al (2017) reported increased dilation and Wang et al (2016) reported decreased dilation.^36,37 Wang et al (2016) further reported an increased time to maximal constriction in the PD group compared to the control group. Only one study by Suzuki et al. (2017) reported on Dementia with Lewy Bodies (DLB), which found increased pupil dilation in the disease group compared to controls. ³⁵ Ranchet et al. (2017) reported data converted to an index of cognitive effort, and therefore could not be included in effect size calculations. ³⁷ Assessing PD and DLB studies together, the population-weighted effect size showed a mild, non-significant effect among PD and DLB patients towards smaller pupil dilation (d = .223) among disease groups compared to healthy controls, as described in.

Pupillary Responses in Mild Cognitive Impairment

Three studies included MCI groups. Podlasek et al (2019) reported no difference between control and MCI groups. ³³ Dragan et al. (2017) found that pupil response velocity was lower in MCI patients compared to younger adults, and non-significantly lower compared to older adults. ²⁹ Granholm et al. (2017) found a differential effect in which pupil dilation increased during a moderate-load cognitive task in single-domain (amnestic and non-amnestic) MCI but decreased in multiple-domain MCI. ³² In this same study, pupil dilation decreased under the high-load condition in all groups, but to a greater extent in single-domain MCI groups compared to controls and a lesser extent in the multiple-domain MCI group. The population-weighted effect size of all studies showed no difference between patient and control groups in pupil dilation (d = .074) among patient groups as compared to healthy controls, as described in Figure 2(c).

Discussion

The purpose of this review was to examine existing literature to determine whether individuals with dementia or MCI would exhibit different task-evoked pupillary responses compared to healthy controls. Results across the eight studies included in this review suggest that differences between these groups indeed exist. To our knowledge, this is the first review to assess the potential role of pupillometry in measuring task-evoked pupillary response for the assessment of dementia.

We report here that pupil dilation is decreased in AD patients compared to controls, with no difference observed in MCI patients. A mild, non-significant trend towards reduced pupil dilation in PD & DLB patients, possibly limited by fewer studies, suggests a similar but less pronounced effect than in AD patients. As pupil dilation is normally increased in response to greater cognitive effort, it is expected that patients with cognitive impairment would require greater effort relative to healthy controls completing the same task and therefore demonstrated greater pupil dilation in comparison. The paradoxical decrease in pupil dilation in patients with dementia is therefore likely due to an alternate process beyond changes in cognitive effort/demand.

Given that LC degeneration is present in multiple forms of dementia, impairment in pupil dilation may be indicative of disease status. LC changes typically occur early in disease progression - in both AD and PD, LC pathology precedes degeneration of other key centres associated with each disease and has been described to even occur within MCI states.^19,38,39 Therefore, alterations in pupil dilation seen in patients with AD is likely due to underlying LC degeneration and may hold diagnostic utility. In AD specifically, post-mortem histological examination of patients found that LC neuronal degeneration occurs in a stepwise manner from cognitively healthy, to MCI, and AD. ³⁸ Among MRI studies, LC contrast ratio, an estimate of LC neuronal integrity, is consistently reduced in AD but variable among MCI groups.^40,41 Although not included for analysis, the study by Kremen and colleagues, who also used a sample obtained from the Vietnam Era Twin Study of Ageing, found that pupil dilation was greater in individuals with increased risk for AD as assessed by polygenic risk scores. ⁴² Elman et al. (2017) drawing from the same population, further report that MCI status was associated with higher tonic activity within LC networks that was proposed to be indirectly due to damage to LC neurons. ³⁰ Interestingly, MCI groups in the current review demonstrated no significant changes in pupil dilation compared to healthy controls. Instead, MCI possibly reflects a transition state in which early LC degeneration is counterbalanced by increased cognitive effort required during the eliciting task.

Impairment in pupil dilation detected through task-evoked pupillary response may function as a potential biomarker of early dementia, with specific utility in detecting AD. Further, as pupillary dysfunction is likely secondary to degenerative changes within the LC, pupillometry may also hold clinical utility in the differentiation of dementia from other clinical presentations, such as delirium or depression. As a preliminary review, we report decreased pupillary dilation during a cognitive task in individuals with AD compared to healthy controls, with no differences found in MCI populations.

Limitations

Our study is limited by the vast heterogeneity of studies present in our review, specifically the range of cognitive tasks employed to evoke pupillary responses. Another limitation was the variability in the classification of MCI vs dementia according to neurocognitive assessment tool cut-off scores (MoCA/MMSE), in which MCI in one study may have been considered dementia in another. Due to the limited number of studies featuring standard cognitive assessment tools, correlation between cognitive assessment scores and pupillary metrics were not assessed within this scoping review.

Future Directions

Given the potential clinical utility in detecting early dementia, further investigation into the sensitivity and specificity of pupil dilation in AD and other dementias is crucial. Further studies exploring the ability for pupil dilation to differentiate between various forms of dementia, as well as between dementia and dementia-mimics may also yield considerable insight into the utility of task-evoked pupillary responses as a clinical biomarker of dementia. It is also unclear whether the specific cognitive task affected pupillary responses - further research should assess whether tasks specific to affected cognitive domains (i.e., working memory in AD) produces different pupillary responses compared to unaffected domains.

ORCID iDs

Michael Zeeman, BHSc https://orcid.org/0000-0001-7111-5831

Tolani Brimmo https://orcid.org/0000-0003-0969-4451

Esther S Kim, PhD https://orcid.org/0000-0003-3748-7310