Abstract
Background:
Olfactory deficits are early features of preclinical Alzheimer’s disease (AD). Whether olfaction is associated with PET biomarkers among community-dwelling older adults is less clear.
Objective:
Investigate cross-sectional and longitudinal associations of olfaction with mild cognitive impairment (MCI) and amyloid-β (Aβ) and tau deposition.
Methods:
We analyzed 364 initially cognitively normal participants (58% women, 24% black) who had baseline olfaction data and subsequent cognitive assessments during an average 2.4-year. A subset of 129 had PET-PiB (Aβ) (n = 72 repeated) and 105 had 18F-flortaucipir (FTP)-PET (tau) (n = 44 repeated). Olfaction was measured using a 16-item Sniffin’ Sticks Odor Identification Test. The association of olfaction with incident MCI was examined using Cox regression. Associations with PiB-distribution volume ratio (DVR) and FTP-standardized uptake value ratio (SUVR) were examined using partial correlation. We tested whether PiB+/–status modified these associations. Analyses were adjusted for demographics and olfactory test version.
Results:
17 (5%) participants developed MCI. Each unit lower odor identification score was associated with 22% higher risk of developing MCI (p = 0.04). In the PET subset, lower scores were associated with higher mean cortical DVR and DVR in orbitofrontal cortex (OFC), precuneus, and middle temporal gyrus (p≤0.04). The “olfaction*PiB+/–” interaction in OFC DVR was significant (p = 0.03), indicating the association was limited to PiB positive individuals. Greater decline in odor identification score was associated with greater increase in anterior OFC DVR and entorhinal tau SUVR (p≤0.03).
Conclusion:
Among community-dwelling older adults, poorer olfaction predicts incident MCI and is associated with overall and regional Aβ. Greater olfaction decline is associated with faster Aβ and tau accumulation in olfaction-related regions. Whether olfaction predicts AD-related neurodegenerative changes warrants further investigations.
INTRODUCTION
Olfactory function declines with aging, and the prevalence of anosmia, loss of smell, increases substantially with advancing age [1, 2]. Olfactory deficits are one of the earliest features of Alzheimer’s disease (AD) dementia, supported by both human studies and animal models [3–5]. Olfactory deficits also predict subsequent cognitive impairment and AD, and AD pathology in olfaction-related brain areas, including the orbitofrontal cortex, entorhinal cortex, and amygdala, may underlie these associations [6–11]. However, the relationships between in vivo biomarkers of AD, such as amyloid-β (Aβ) and tau pathology, are less clear among community-dwelling older adults.
Current knowledge on the relationship between olfaction and PET imaging of AD biomarkers is primarily focused on Aβ, and measures of Aβ are often limited to the overall Aβ burden. The specific distribution of brain Aβ burden with impaired olfaction is not well characterized. To date, only one study examined regional Aβ with olfaction in a small sample [12]. Information on the relationship between olfaction and tau pathology assessed with PET scanning is limited, especially among community-dwelling older adults [13]. To date, to the best of our knowledge, only three studies have reported on the relationship of olfactory function with tau PET, and all studies involved participants recruited from clinic-based samples [14–16]. These studies have yielded mixed results with two studies showing associations between poorer olfaction and higher tau burden, particularly in temporal and parietal regions, and one study showing no direct relationship with tau burden. These inconsistencies may be due to differences in the cognitive status of participants, olfactory assessment, and PET imaging analysis. It is also possible that patients from clinics are more likely to have other pathologies that affect olfaction.
Most, but not all, PET imaging studies of AD biomarkers indicate a relationship of olfactory function with Aβ and tau, but the relationship is often modest and driven by individuals who are cognitively impaired or Aβ positive [12, 15–18]. Thus, cognitive status or Aβ positivity appears to play a role in relationships with Aβ and tau. Understanding these relationships in community-dwelling older individuals may provide additional insights into the role of olfactory deficits in early AD pathology.
In this study, we investigate the associations between olfaction, the incidence of cognitive impairment, and AD biomarkers in a well-characterized community-dwelling sample from the Baltimore Longitudinal Study of Aging. First, we determine whether poorer olfaction is associated with incident MCI/AD among initially cognitively normal older adults. Next, in a subset of participants, we examine the relationships between olfaction and Aβ and tau burden measured by PET. Finally, we test whether amyloid positivity modifies the relation of olfaction with Aβ and tau burden.
METHODS
Study population
Participants were drawn from the Baltimore Longitudinal Study of Aging (BLSA), an ongoing longitudinal study with continuous enrollment that began in 1958 [19]. Starting in 2009, all eligible BLSA participants were continuously enrolled in the MRI study. Participants aged < 60 years were seen every four years, aged 60–79 years every two years, and aged≥80 years annually. A subset of the BLSA neuroimaging participants received PiB-PET scans beginning in 2005. The interval between olfactory testing and PET assessment was within 6 months, except for one participate (interval = 315 days). The median interval was 5 (interquartile range = 12) days. The olfactory testing, cognitive assessment, and brain MRI were completed at the same visit. The BLSA and PET protocols were approved by the Institutional Review Boards of the National Institutes of Health and the Johns Hopkins Medical Institutions, respectively. Participants provided written informed consent at each BLSA visit.
We identified 364 initially cognitively normal participants aged 60 and older who had their first olfaction assessment and subsequent assessments to determine cognitive status during a mean follow-up of 2.4 years, ranging between 1 and 5 years. The first olfaction assessment between 2015 and 2018 was considered “baseline” in this analysis. A subset of 129 participants with mixed cognitive status (120 cognitively normal, 8 MCI, 1 AD) had [11C]PiB PET Aβ imaging collected between 2016 and 2020, and 72 had repeated measures of Aβ imaging and olfaction over an average 2.2 years (average 2.4 visits per participant). Of the 129 with Aβ imaging, 105 had tau imaging with [18F]-Flortaucipir (FTP) PET (97 cognitively normal, 8 MCI), and 44 had repeated tau imaging and olfaction over an average 2.2 years (average 2.3 visits per participant).
Diagnoses of MCI and AD
Procedures for determination of cognitive status have been described previously [20]. Clinical and selected neuropsychological data from BLSA participants were reviewed at a consensus conference if participants screened positive on the Blessed Information-Memory-Concentration Test score (i.e., score≥4) [21], if their Clinical Dementia Rating (CDR) score was≥0.5 using subject or informant report [22], or if concerns were raised about their cognitive status. MCI was determined using the Petersen criteria and diagnosed when 1) cognitive impairment was evident for a single domain (typically memory) or 2) cognitive impairment in multiple domains occurred without significant functional loss in activities of daily living [23]. Both cross-sectional and longitudinal neuropsychological performance, as well as the CDR, are used to determine cognitive impairment. Neuropsychological tests used in the diagnosis include Mini-Mental State Exam (MMSE) [24], Trail Marking Test [25], Letter (F, A, S) and Category (animals, vegetables, fruits) Fluency [26, 27], Boston Naming Test [28], Clock Drawing Test [29], Constructions, Calculations, and a memory impairment screen with maximum score of 13 (MIS+). The MIS+was based on the 4-item short delay free and cued recall score from the MIS test (maximum score = 8) [30] plus the 5-item memory score from the Blessed Information-Memory-Concentration Test (maximum total score = 13). The use of this MIS+measure had been previously validated against use of the Picture Version Free and Cued Selective Reminding Test as a memory screen for diagnosis of dementia and MCI between 2008 and 2010. In BLSA, diagnoses of dementia and AD have continued to follow the Diagnostic and Statistical Manuel, the third edition, revised (DSM-III-R) and the National Institute of Neurological and Communication Disorders and Stroke-Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) criteria, respectively [31, 32]. Date of symptom onset was estimated for MCI/AD and was synonymous with the date of onset of MCI.
Olfaction
Olfactory function was measured as odor identification using the validated 16-item Sniffin’ Sticks Identification Test [33]. Participants were presented with a total of 16 common odors. For each odor, participants were given four choices (one correct odor). The range of the odor identification test score was between 0 and 16 with a higher score being higher olfactory function. Two versions of the 16-item Sniffin’ Sticks were introduced and randomized at the initial assessment in the BLSA to minimize potential learning effects over time.
MR imaging
Magnetization-prepared rapid gradient echo (repetition time = 6.8 ms, echo time = 3.2 ms, flip angle = 8°, image matrix = 256×256×170, voxel size = 1×1×1.2 mm3) scans were acquired on a 3 T Philips Achieva scanner. MPRAGE scans were used to compute anatomical labels and regional brain volumes with Multi-atlas region Segmentation using Ensembles of registration algorithms and parameters [34]. These anatomical labels were mapped onto PET scans to define regions of interest (ROIs) used in PET processing.
PET imaging
Dynamic [11C]PiB PET studies (33 time frames over 70 min immediately following bolus injection of approximately 555 MBq of radiotracer) were performed in 3D mode on a GE Advance (image matrix = 128×128×35 slices, voxel size = 2×2×4.25 mm3, 4.5 mm full-width at half-maximum [FWHM] at the center of the field of view) or Siemens High Resolution Research Tomograph (HRRT) (image matrix = 256×256, 207 slices, voxel size = 1.22×1.22×1.22 mm3, 2.5 mm FWHM at the center of the field of view) scanner at the Johns Hopkins University PET facility. Scans acquired on the Siemens HRRT were smoothed with a 3 mm isotropic Gaussian kernel and resampled to match their spatial resolution and voxel size to those of the GE Advance scans. The same image processing steps were then applied to both GE Advance and Siemens HRRT scans. Details of image acquisition and processing, including quantification of distribution volume ratio (DVR), are described elsewhere [35]. Regional DVRs were harmonized across the two scanners using an affine transformation whose parameters were calculated by leveraging longitudinal PiB PET data available on both scanners for 79 BLSA participants. Mean cortical DVR was calculated as the average of DVR values from the cingulate, frontal, parietal (including precuneus), lateral temporal, and lateral occipital cortices, excluding the sensorimotor strip. PiB+/–status was determined using a cutoff of 1.06, derived from a two-class Gaussian mixture model fitted to baseline mean cortical DVR data [36]. PiB positive individuals are more likely to accumulate greater Aβ burden over time compared to PiB negative individuals [37]. Regional Aβ burden was examined in the following ROIs: orbitofrontal cortex (anterior, posterior, medial, and lateral OFC), precuneus, amygdala, entorhinal cortex, inferior temporal gyrus, middle temporal gyrus, fusiform gyrus, and occipital fusiform gyrus.
18F-flortaucipir (FTP) PET scans (6 time frames over 30 min starting in post-injection of approximately 370 MBq of radiotracer) were acquired on a Siemens HRRT scanner (image matrix = 256×256×107, voxel size = 1.22×1.22×1.22 mm3, 2.5 mm FWHM at the center of the field of view). Scans were corrected for partial volume effects using the region-based voxel-wise method [38]. We computed standardized uptake value ratio (SUVR) images with inferior cerebellar gray matter as the reference region. Image processing steps are described in further detail in Ziontz et al. [39] Regional tau burden was examined in the following ROIs: entorhinal cortex, inferior temporal gyrus, and fusiform gyrus.
Statistical analysis
In the overall sample who were initially cognitively normal, we examined the association of baseline odor identification score with incident MCI using a Cox proportional hazards model, adjusted for baseline age, sex, race, test version, and additionally adjusted for Apolipoprotein E (APOE) ɛ4 carrier status.
In the PET subset, we examined cross-sectional associations of odor identification score with each neuroimaging biomarker of interest using Pearson correlation and partial correlation after controlling for age, sex, race, education, and test version. We also tested the strength of the association after further controlling for APOE ɛ4 carrier status. We further tested the “olfaction*PiB+/–status” interaction to determine whether amyloid positivity modified the association between olfaction and AD biomarkers using covariate-adjusted multivariable linear regression.
In a smaller subset with repeated PET imaging, we examined longitudinal associations of the annual rate of change in odor identification score with annual rates of change in biomarkers of interest using partial correlation, adjusted for age, sex, race, education, and test version. In a sensitivity analysis, we further adjusted for change in mental status measured by MMSE to ensure associations were not confounded by change in cognitive status [24]. Annual rates of change were computed as the difference from the first to last visit divided by follow-up time in years. We also tested the strength of the associations after further controlling for APOE ɛ4 carrier status. We explored whether amyloid positivity modified the association between rate of change in olfaction (denoted as Δolfaction) and rates of change in AD biomarkers by testing the “Δolfaction* PiB+/–status” interaction using covariate-adjusted multivariable linear regression.
We performed sensitivity analysis by excluding current smokers and those with depression determined by the Center for Epidemiologic Studies Depression Scale≥16 [40] and by adjusting for the presence of vascular disease. In this exploratory analysis, significance was set at p < 0.05. All statistical analyses were performed using SAS (version 9.4, Cary, NC).
RESULTS
Participant characteristics are presented in Table 1. During a mean 2.4 years of follow-up, 17 (5%) participants developed MCI, and none developed AD. Among 17 participants who developed MCI, based on clinical characterization, 11 were categorized as AD etiology, 4 vascular dementia, 1 frontotemporal dementia, and 2 unspecified. Baseline olfactory scores and scores for neuropsychological tests used in the diagnosis are presented in Supplementary Table 1 for the overall sample, those who subsequently developed MCI and those who remained cognitively normal. After adjustment for age, sex, race, education, and olfactory test version, each unit lower odor identification score was associated with 22% higher risk of developing MCI (HR = 1.222, 95% CI: 1.004–1.487, p = 0.04). This relationship remained similar after further adjustment for APOE ɛ4 carrier status (HR = 1.219, 95% CI: 1.000–1.487, p = 0.05). Figure 1 shows Kaplan–Meier plot for baseline odor identification score associated with MCI survival. Results remained similar after excluding current smokers and those with depression and adjusting for the presence of vascular disease (data not shown).
Participant characteristics
DVR, distribution volume ratio; SUVR, standardized uptake value ratio; CES-D, Center for Epidemiologic Studies Depression Scale. Vascular disease includes myocardial infarction, congestive heart failure, hypertension, peripheral artery disease, and coronary artery disease by self-report.
Kaplan–Meier plot for baseline odor identification score associated with MCI survival. Dashed line indicates anosmia defined as 10th percentile (score≤7) base on previous literature [2]; Solid line indicates free of anosmia (score > 7). Covariate-adjusted hazard ratio of anosmia = 4.18 (95% CI = 1.38–12.62, p = 0.011).
In the PET subset, lower odor identification score was cross-sectionally associated with higher Aβ burden, including mean cortical PiB DVR and DVR in anterior, posterior, lateral, and medial OFC, precuneus, and middle temporal gyrus after covariate adjustment (Table 2). Further adjustment for APOE ɛ4 carrier status slightly attenuated these associations but trends remained (all p≤0.10). The “olfaction*PiB+/–” interaction was significant in OFC DVR measures of interest (β= –0.015, p = 0.03), showing that the olfaction-OFC DVR association was limited to the PiB positive group (Fig. 2). Odor identification score was not significantly associated with DVR in other ROIs after covariate adjustment (Table 2), and associations with DVR in the inferior temporal gyrus and fusiform gyrus were at trend level (p = 0.060 and 0.057, respectively; Table 2). Lower odor identification score was bivariately correlated with higher entorhinal FTP-PET SUVR (p = 0.044) and marginally associated with higher fusiform gyrus FTP-PET SUVR (p = 0.050), but these associations were attenuated after adjustment for covariates (p = 0.119 and 0.142, respectively) (Table 2) and APOE ɛ4 carrier status (p = 0.212 and 0.325, respectively). The “olfaction*PiB+/–” interaction was not statistically significant with tau SUVR in the entorhinal cortex or fusiform gyrus as the outcome (both p > 0.05). Odor identification score was not significantly associated with tau SUVR in the inferior temporal gyrus with and without covariate adjustment (Table 2).
Cross-sectional associations between olfaction and PET measures of amyloid-β and tau
The bold number reflects significant associations at p < 0.05.
Scatter plots of odor identification score with regional orbitofrontal cortex (OFC) PiB DVR stratified by PiB+/–status. Filled circle indicates PiB positive. Empty circle indicates PiB negative.
In a smaller subset with repeated measures of PET imaging and olfaction over an average 2.2 years (SD = 0.8), greater decline in odor identification score was significantly associated with greater increase in Aβ PiB DVR in anterior OFC and entorhinal tau SUVR and there was a trend towards greater increase in PiB DVR in posterior OFC after covariate adjustment (p = 0.009, 0.035, and 0.074, respectively; Table 3; Fig. 3). Results remained similar after adjustment for concurrent change in MMSE (Table 3) and APOE ɛ4 carrier status (p = 0.009, 0.040, and 0.076, respectively). The “Δ olfaction * PiB+/–” interaction was not significant in changes of these biomarkers of interest (all p > 0.05). We did not observe significant associations between the rate of change in odor identification score and rates of change in the remaining ROIs (Table 3). After excluding current smokers and those with depression and adjusting for the presence of vascular disease, longitudinal associations with changes in PiB DVR in anterior and posterior OFC remained a trend, while cross-sectional associations were attenuated (data not shown).
Associations between the annual rate of change in odor identification score and annual rates of change in measures of amyloid-β DVR and tau SUVR
Due to a scoring lag, repeated measures of Mini-Mental State Exam (MMSE) were available for the majority (but not all) of the participants with repeated PET imaging.
Scatter plots of the annual rate of change in odor identification score with annual rates of change in amyloid-β (top) and tau (bottom) biomarkers identified in Table 3.
DISCUSSION
In this sample of community-dwelling older adults, we demonstrate that poorer olfactory function predicts incident MCI. Further, in a subset with PET imaging, poorer olfactory function is associated with higher levels of Aβ deposition, especially in OFC and temporal regions, and the olfaction-OFC Aβ association is limited to PiB positive individuals. Greater olfaction decline is associated with faster accumulation of Aβ and tau, localized to areas important for olfactory function. Our study extended prior knowledge by focusing on a diverse sample of community-dwelling older adults, examining longitudinal associations between change in olfactory function and change in PET biomarkers of AD, and including multiple regions of interest to determine the spatial distribution.
Consistent with prior studies, we found that poorer olfactory function predicted incidence of MCI during a mean 2.4 years of follow-up and this relationship remained similar after adjustment for APOE ɛ4 carrier status [6, 11]. Our study differs from previous studies in several aspects. First, previous study samples from the Rush Memory and Aging Project, Mayo Clinic Study of Aging, and Epidemiology of Hearing Loss Study are predominantly white, while the BLSA population is more diverse with 24% black. Notably, our findings were robust to the adjustment for race. Second, previous studies have slightly longer follow-up (up to 10 years and mean 3.5 years), while our study has a mean 2.4 years of follow-up due to the later introduction of olfactory testing. Third, previous studies had higher incidence of MCI or CI (30%, 17%, and 10%, respectively), while only 5% of BLSA participants developed MCI due to the healthier status of the BLSA population and shorter follow-up. Despite these differences, we observed a consistent relationship of poorer olfactory function with an increased risk of developing MCI.
Our study adds to the growing body of literature on olfaction in relation to Aβ deposition. We demonstrate novel findings on the spatial distribution of associations between Aβ deposition and olfaction. Specifically, we found associations in various parts of the orbitofrontal cortex, precuneus, and middle temporal lobe. Among five previous studies examining olfaction and Aβ PET [12, 42], only one examined regional Aβ burden and reported the relationship of poorer olfaction with higher Aβ burden in posterior cingulate, temporo-parietal, and lingual cortex [12]. Our findings on Aβ deposition in the precuneus and middle temporal gyrus are in line with this previous study [12]. Our findings on the relation between Aβ deposition in the OFC and olfaction are new and offer support that the regions involved in olfaction overlap with those showing early AD pathology. The observed olfaction-OFC Aβ level relationships are prominent in the PiB positive group, suggesting that the olfaction-OFC Aβ relationship is dependent on PiB+/–status. The lack of associations in the PiB negative group may reflect reduced variability and highlights potential issues in use of PiB as a continuous measure without accounting for the presence versus absence of detectable Aβ. Associations limited to the PiB positive group are also in line with previous findings that the relationship with Aβ deposition is primarily driven by those who are cognitively impaired [12].
Cross-sectionally, we observed the olfaction-tau relationship in the entorhinal cortex and fusiform gyrus at trend level, which was in line with previous studies [15, 16]. These modest cross-sectional relationships may be due to the healthier status of our study sample and relatively low tau burden. Our study focused on community-dwelling older adults, while the previous two studies included patients with MCI/AD seen in clinics. We did not find a significant “olfaction*PiB+/–“ interaction, though one recent study suggested that the olfaction-tau PET relation may also be driven by PiB positive individuals [15].
Our longitudinal findings demonstrate localized relationships between declining olfaction and accumulation of Aβ and tau over time. Specifically, greater olfaction decline is linked to faster accumulation of Aβ in anterior and posterior OFC and not in other brain areas of interest. Our neuroimaging findings are novel and consistent with prior postmortem findings, which suggest that olfactory impairment observed in early AD dementia is due to AD pathology in olfactory-related brain areas [3]. One previous cross-sectional study using voxel-based analysis did not find a relationship between olfaction and Aβ burden in the orbitofrontal cortex [12]. One possible explanation is the cross-sectional design which cannot assess within-individual change over time. Similarly, our longitudinal analyses of tau burden have shown that declining olfaction over time is associated with faster accumulation of entorhinal tau, but there was no cross-sectional association with entorhinal tau after covariate adjustment. We addressed limitations of prior cross-sectional studies by examining the rate of change in tau burden and olfactory function. Our findings for the entorhinal cortex are in line with previous reports that poorer olfaction is associated with higher tau burden in the temporal area [15, 16]. Notably, these longitudinal associations were robust to adjustment for APOE ɛ4 carrier status and further adjustment for change in mental status.
This study has limitations. The BLSA study population is healthier and better educated than the general older adult population. Our sample size of individuals with both olfactory testing and biomarkers is modest due to the later initiation of olfactory assessment. The cross-sectional and longitudinal associations with biomarkers do not address temporality, and we cannot determine causation. A limitation is that the diagnosis of verbal memory impairment in MCI was based on consensus case conferencing, using longitudinal change on an internal memory screening procedure (MIS+and verbal memory items from the BIMC) in combination with information from the CDR, rather than a specific cutoff on a published test of verbal memory. This study has several strengths. First, the study population is well characterized, which allowed us to examine the strength of the relationships with MCI and PET biomarkers after accounting for APOE ɛ4 carrier status. Our focus on community-dwelling older adults may provide additional insights into the role of olfaction in an early pathophysiologic stage of AD dementia. Second, rigorous adjudication of MCI and dementia allowed us to capture incident MCI over a relatively short follow-up. Third, we included PiB PET measures in multiple ROIs, which identified the spatial distribution of the relation of olfaction with Aβ. Fourth, we tested whether PiB positivity status modified the relationships with Aβ and tau burden. More importantly, this study included repeated measures of olfaction, Aβ and tau burden which allowed us to determine the longitudinal relationship between declining olfaction and accumulation of Aβ and tau burden.
In conclusion, among community-dwelling older adults, poorer olfactory function predicts incident MCI and is associated with overall and regional Aβ. Greater longitudinal decline in olfaction is associated with faster Aβ and tau accumulation in regions important for olfaction. Future studies are warranted to understand whether olfaction predicts AD-related neurodegenerative changes over time.
