Introduction
Olfactory impairment is a common and early feature of many neurodegenerative diseases but is especially prominent in Lewy body (LB) diseases where pathologic involvement of all parts of the olfactory system is recognized (Attems et al., Reference Attems, Walker and Jellinger2014). Its high prevalence and early presence make it a key prodromal feature of Parkinson’s disease (PD), with a review indicating its onset decades before motor symptoms (Savica et al., Reference Savica, Bradley and Mielke2018). However, only a few studies have directly assessed olfaction in dementia with Lewy bodies (DLB), with studies using both clinical (Westervelt et al., Reference Westervelt, Stern and Tremont2003; Reference Westervelt, Bruce and Faust2016; Williams et al., Reference Williams, Williams, Combrinck, Christie, Smith and McShane2009; Yoon et al., Reference Yoon, Kim, Moon, Yong and Hong2015; Reference Yoo2018) and neuropathological diagnoses (Beach et al., Reference Beach2020; McShane et al., Reference McShane2001; Olichney et al., Reference Olichney2005) finding greater impairments in DLB than AD.
The early and accurate identification of DLB is recognized as important for optimizing patient management (Taylor et al., Reference Taylor2020). We have reported the diagnostic utility of both dopaminergic imaging with 123I-N-fluoropropyl-2β-carbomethoxy-3β-(4-iodophenyl) (FP-CIT) and cardiac metaiodobenzylguanidine (MIBG) imaging in mild cognitive impairment (MCI). Both had high specificity (88% for each; Roberts et al., Reference Roberts2021a; Reference Roberts2021b) but more modest sensitivity for identifying MCI with Lewy bodies (MCI-LB) compared with MCI-AD. There is therefore a need to find simple, inexpensive tests which can be applied in memory services, with the aim of identifying a large proportion of those with likely LB disease. This would enable these imaging investigations to be focused on people with a much higher probability of having LB disease. One approach is to test for olfactory impairment, especially odor identification.
Although odor identification tests discriminate well between PD and non-PD (Mahlknecht et al., Reference Mahlknecht2016), such studies have typically been undertaken in younger adults and olfactory function declines with age with hyposmia being present in about 25% of older adults (Murphy et al., Reference Murphy, Schubert, Cruickshanks, Klein, Klein and Nondahl2002). Olfactory deficits are also reported in AD, which might seem to further limit the potential value of olfaction tests in identifying DLB, but not only do these occur significantly more often in DLB than AD but also the early involvement of the olfactory organ by LB diseases and the associated prominence of hyposmia in prodromal PD (Savica et al., Reference Savica, Bradley and Mielke2018) suggests that olfactory impairments may discriminate between AD and DLB at the MCI stage. In addition, a large proportion of abnormal olfaction in clinically diagnosed AD is due to comorbid LB disease (Beach et al., Reference Beach2020). So olfactory testing may be a sensitive test for prodromal DLB. This is supported by a few small studies of olfaction. One found no differences between AD and DLB at the dementia stage, while such differences were present at the earlier MCI stage (Yoo et al., Reference Yoo2018); in a longitudinal study in which olfaction was tested in MCI patients, those who later progressed to DLB had greater impairments compared with those who developed AD dementia with a receiver operator characteristic area under the curve of 84% (Yoon et al., Reference Yoon, Kim, Moon, Yong and Hong2015); and a clinical assessment of mild DLB and AD (Mini-Mental State Examination, MMSE = 24) found odor identification again distinguished these two diseases with a high sensitivity for DLB of 81% (Williams et al., Reference Williams, Williams, Combrinck, Christie, Smith and McShane2009).
We therefore carried out the largest comparison, to date, of olfaction in MCI-LB and MCI-AD using the Sniffin’ Sticks-16 olfaction test. We hypothesized that not only would olfactory function be worse in probable MCI-LB than in MCI-AD, and in cognitively healthy older adults, but that the previously identified cutoff score of ≤ 10 on this test in PD (Mahlknecht et al., Reference Mahlknecht2016) would be too high in this population because of the impairments of olfaction with increasing age. So we also sought to assess whether a different cutoff for olfactory testing would be more appropriate for identifying MCI-LB than those previously used in PD.
Methods
Participants
As detailed previously (Donaghy et al., Reference Donaghy2020; Roberts et al., Reference Roberts2021a; Reference Roberts2021b), medically stable patients aged 60 years or older with a clinical diagnosis of MCI were recruited from local memory services in the North-East of England between April 2016 and September 2019. Potential study participants either reported the presence of any core clinical feature of DLB (complex visual hallucinations, rapid eye movement (REM) sleep behavior disorder, cognitive fluctuations, or parkinsonism not preceding cognitive impairment by more than 12 months), or any supportive clinical feature found in DLB, but not specific to this (e.g. mood change or sleep disturbance). Exclusion criteria were dementia at screening, no objective cognitive impairment, or possible vascular or frontotemporal etiology and parkinsonism present for more than a year before the onset of cognitive problems ("one year rule"). In addition, healthy comparison subjects (HC) with no evidence of cognitive impairment or parkinsonism or other brain diseases and a normal structural MRI brain scan were recruited through the Join Dementia Research platform, and from friends or families of the patients. All identified participants provided written informed consent prior to detailed screening and medical review before final inclusion.
Following consent participants underwent a research-level assessment involving a semistructured interview, clinical and neurocognitive assessment and neurological examination by a medical doctor (RD ans SL), and imaging with FP-CIT, MIBG, and MRI (Firbank et al., Reference Firbank2021; Roberts et al., Reference Roberts2021a; Reference Roberts2021b) at baseline, and then had longitudinal review at approximately annual follow-ups. Mean (SD) of maximum follow-up were 1.4 (0.98) years, with a maximum of 3.7 years from baseline.
Clinical assessment, imaging, and differential diagnosis
Assessment
The MDS Unified Parkinson’s Disease Rating Scale – Motor Examination (UPDRS-III), Epworth Sleepiness Scale, and Geriatric Depression Scale were administered to patients. The Instrumental Activities of Daily Living scale, North-East Visual Hallucinations Inventory, Neuropsychiatric Inventory, Mayo Sleep Questionnaire, Clinician Assessment of Fluctuation, and Dementia Cognitive Fluctuation Scale were administered to informants. Clinical Dementia Rating scale and Cumulative Illness Rating Scale for Geriatrics were completed on the basis of the clinical history and other research assessments. A detailed neuropsychological evaluation was also carried out as reported previously (Donaghy et al., Reference Donaghy2020) which included the ACE-R, a 100-point cognitive screening test from which MMSE score was derived. Dopaminergic 123I-N-fluoropropyl-2β-carbomethoxy-3β-(4-iodophenyl) single-photon emission computed tomography (FP-CIT) and cardiac MIBG imaging were offered to all participants as previously described (Roberts et al., Reference Roberts2021a; Reference Roberts2021b). FP-CIT images were visually rated as normal or abnormal by a five-person panel of experienced image analysts, blind to clinical information. MIBG images were classified as abnormal given a heart:mediastinum uptake ratio of < 1.86 based on data from locally recruited HC (Roberts et al., Reference Roberts2019).
Differential diagnosis
As detailed previously (Donaghy et al., Reference Donaghy2020; Roberts et al., Reference Roberts2021a; Reference Roberts2021b), diagnoses were made by a three-person expert clinical panel (AJT, PCD, and JPT) who independently reviewed research data and health service record and made MCI diagnoses according to NIA-AA criteria (Albert et al., Reference Albert2011). This consensus panel method has previously been validated against autopsy and is recognized by regulatory authorities as the clinical gold standard (McKeith et al., Reference McKeith2007).
To determine MCI etiology, the presence or absence of core LB symptoms was also rated by the panel, in accordance with the fourth consensus criteria for DLB (McKeith et al., Reference McKeith2017), and those with evidence (including on MRI) of vascular or frontotemporal etiologies, or parkinsonism predating cognitive impairment by more than 1 year, were also excluded. In accordance with the research diagnostic criteria for MCI-LB (McKeith et al., Reference McKeith2020), a diagnosis of probable MCI-LB was made if a patient had two or more core LB symptoms or one core symptom in addition to a positive FP-CIT or MIBG scan. Patients were diagnosed with possible MCI-LB when they had only one core symptom or one or more abnormal scans. MCI-AD was diagnosed following the criteria of Albert et al. (Reference Albert2011). Subjective and objective cognitive decline consistent with AD was established, along with generally maintained independence of function in everyday life, and the absence of dementia and other causes were then excluded as above. These diagnoses were updated at each annual follow-up, and a diagnosis of dementia was made when any participant was judged to no longer be functionally independent. Participants with dementia were not followed up further.
Sniffin’ Sticks assessment
Olfactory function was assessed using Sniffin’ Sticks-16 (SS-16) which was administered to each participant in accordance with the manufacturer’s instructions. The test consists of 16 pens impregnated with specific odors. These were in turn held about 2 cm below the nose of the participant who was then asked to smell and identify the odor from a forced choice of four written alternatives. A pause of about 30 s was allowed between each Sniffin’ stick.
Analysis
Analyses were conducted in R software with the epiR and pROC packages. Significance was considered as p < 0.05. Group differences in hyposmia measured by total Sniffin’ Sticks scores were assessed with the general linear model, adjusting for age (mean centered). Model diagnostics were checked by plotting residuals against fitted values, and with Q−Q plots.
Receiver operating characteristic (ROC) curves were plotted to assess the discriminatory utility of Sniffin’ Sticks total in identifying MCI-LB. Diagnostic cutoffs for discriminating MCI-LB from MCI-AD and HC were identified by Youden’s index.
Results
The SS-16 was completed by 122 participants (32 HC, 90 MCI) and baseline characteristics are in Table 1 and task performance in Figure 1. A Mann−Whitney U test found no significant sex differences in olfactory function (p = 0.70). The age-adjusted linear model (see Table 2) demonstrated that there were significant differences in olfactory function overall between groups (F(3,117) = 9.83, p < 0.001), with this being worse in probable MCI-LB than MCI-AD, and HC. We also included the possible MCI-LB group for information; this group were more similar to MCI-AD, with significantly better olfactory function than probable MCI-LB.
Figure 1. SS-16 task performance in each diagnostic group, 10- and 7-point cutoffs marked (dashed lines).
Table 1. Baseline characteristics of healthy comparison subjects and MCI subgroups who completed the Sniffin’ Sticks Smell Identification Test
Count (%), Mean (SD), or Median [Range].
Table 2. General linear model for estimated age-adjusted diagnostic group differences in overall performance on the 16-item Sniffin’ Sticks Smell Identification Subtest
Sniffin’ Sticks cutoff ≤ 10
Using the previously identified cutoff score of ≤ 10 (Mahlknecht et al., Reference Mahlknecht2016), the SS-16 had 84% sensitivity for diagnosis of probable MCI-LB (95% CI: 69–94%). Specificity was low in differentiating probable MCI-LB from MCI-AD at 30% (95% CI: 16–49%), though better in differentiating these from HC (66%, 95% CI: 47–81%).
ROC analysis
ROC curves were plotted, with best cutoffs identified with Youden’s index. A lower cutoff of ≤ 7 was found to best distinguish probable MCI-LB from both MCI-AD and HC in this cohort (see Figure 2). This cutoff had 63% sensitivity for probable MCI-LB, but better specificity versus MCI-AD (73%) and HC (97%).
Figure 2. ROC curve for Sniffin’ Sticks Smell Identification Test in distinguishing probable MCI-LB from HC (dashed line, AUC = 0.83) and MCI-AD (solid line, AUC = 0.67) with standard 10-point (circle) and 7-point (square) cutoffs marked.
Correlations with disease severity
Excluding HC, total score on the SS-16 was significantly associated with level of global cognitive function assessed with the Addenbrooke’s Cognitive Examination – Revised (Spearman’s r = 0.38, p < 0.001), but not with levels of motor impairment, as assessed by the Unified Parkinson’s Disease Rating Scale Part III (Spearman’s r = −0.13, p = 0.216).
Comparison with olfactory responses in questionnaire
All participants were asked if they had noticed a loss or reduction in their sense of smell when completing the Questionnaire for Symptoms Suggestive of Lewy Body Disease (QSSLBD). Of the 83 who did not recognize a reduction or loss of their sense of smell, 49 (59%) scored below the SS-16 cutoff. Of the 39 who did report a noticed loss of sense of smell, 31 (79%) scored below the cutoff and eight (21%) were above this threshold.
Discussion
Previous research has demonstrated that olfactory impairment is a highly prevalent and early feature of LB diseases, that in dementia it occurs more frequently in DLB than in AD and suggested such differences may be more prominent in MCI. We conducted a prospective analysis of olfaction in MCI and found as hypothesized that olfactory impairment is more frequent in MCI-LB than both in MCI-AD and in healthy older people and that a lower cutoff may be more appropriate in MCI than in PD (≤ 7 in our study vs ≤ 10), though this requires replication. We also found that questioning about a loss of sense of smell did not perform well in such patients and testing is required to identify their olfactory impairments.
Previously, we have reported that two imaging biomarkers recommended in diagnostic criteria for both DLB and MCI-LB have high specificity (both 88%) in patients with MCI (Roberts et al., Reference Roberts2021a; Reference Roberts2021b). This is similar to their specificities in dementia but, as expected in earlier disease, we found the sensitivities were lower (66% for FPCIT and 59% for MIBG) in MCI than in dementia (92%; OʼBrien et al., Reference OʼBrien2014). Even at the dementia stage, the diagnosis of DLB is delayed and frequently missed (Surendranathan et al., Reference Surendranathan2020), with a large study finding only 4.6% of dementia cases diagnosed with DLB in UK memory services (Kane et al., Reference Kane2018). This compares with a recent autopsy analysis of a large representative UK cohort of dementia in which 26.3% had LB disease sufficiently severe to cause dementia (McAleese et al., Reference McAleese2021). It is likely that even more cases are missed at the MCI stage than in dementia. Although it is unrealistic to expect every person with a given disease to be identified during life, the magnitude of the gap in DLB suggests that many more people with LB disease in memory services could be identified, a view supported by the wide variation in diagnostic rates in clinical services (Kane et al., Reference Kane2018). It is also not realistic to expect such services to utilize FPCIT or MIBG in all patients presenting with cognitive impairment and so identifying simple brief tests for early LB disease would enable such diagnostic imaging tests to be targeted on patients with a higher likelihood of having MCI-LB/DLB, thereby facilitating their early identification. This would in turn enable early optimization of treatment for this complex disease with multiple physical and neuropsychiatric symptoms (Taylor et al., Reference Taylor2020). Previously, we have reported that using DLB assessment toolkits was associated with a 35% increase in diagnosis of DLB in memory services (Surendranathan et al., Reference Surendranathan2021). We suggest that in addition such services could further improve their identification and diagnosis of DLB/MCI-LB by adding Sniffin’ Sticks to their assessment protocols. This test is simple, cheap, and popular with patients who enjoy the novelty of identifying the odors.
A major objection to this argument is that because most people in memory services have AD and some of these have abnormal olfaction (test positive), then testing olfaction for diagnostic scanning will still lead to most test positive patients having AD and so most positive tests will still be false with only a minority having LB disease. Using our identified SS-16 cutoff of ≤ 7 would mitigate this concern, but this objection assumes that those clinically diagnosed with AD do in fact have (only) AD and do not also have LB disease. We reported that many patients with an AD-like clinical presentation have high-grade LB disease (Thomas et al., Reference Thomas2018) and this is consistent with the high prevalence of LB disease in autopsy studies. In Alzheimer’s Disease Neuroimaging Initiative (ADNI) of those clinically diagnosed with AD, 45.5% had LB pathology (Toledo et al., Reference Toledo2013) and in brain bank studies in the US (Schneider et al., Reference Schneider, Aggarwal, Barnes, Boyle and Bennett2009), Finland (Oinas et al., Reference Oinas2009), and Japan (Wakisaka et al., Reference Wakisaka, Furuta, Tanizaki, Kiyohara, Iida and Iwaki2003) LB pathology was reported in 24.7%, 29% and 41.4% of those with dementia and the above UK study found LB pathology sufficient to cause dementia in 26.3% (McAleese et al., Reference McAleese2021). Thus, many of those in services diagnosed with AD have LB disease and abnormal olfaction in such "AD" is therefore likely due to LB disease with or without comorbid AD.
This point is not merely inferential. Other autopsy studies have consistently shown impairments in olfaction are strongly associated with LB disease rather than AD. LB density was significantly associated with olfactory impairment in a study comparing olfaction in AD and DLB (McShane et al., Reference McShane2001); anosmia was about three times more frequent in people who had LB disease together with AD compared with those with pure AD (Olichney et al., Reference Olichney2005); people with no clinical features of any LB disorders but who had LB disease at autopsy had an 11-fold increase in abnormal olfaction when tested during life (Ross et al., Reference Ross2006), suggesting such testing may be useful in identifying people without clinically manifest LB symptoms. Finally, a recent large study found that people with combined AD and LB pathologies were 17 times more likely to have olfactory impairment on testing with the University of Pennsylvania Smell Identification Test (UPSIT) olfaction test than those who had pure AD pathology (Beach et al., Reference Beach2020). Such evidence, from different brain banks around the world, strongly suggests that most of those with abnormal olfaction who have been diagnosed with clinical AD do in fact have LB disease either alone or along with AD pathology. This makes it likely that many of those with MCI-AD and abnormal olfaction in our study have LB disease and that apparent false positives in memory services would be highly likely to be true positives with abnormal olfaction correctly identifying the presence of occult LB disease. While such an argument needs direct investigation by future research, the evidence overall suggests olfactory testing is likely to be a useful means of identifying early LB disease.
Our exploratory analyses of SS-16 score with disease severity found that the previously reported association with severity of cognitive impairment (Yoo et al., Reference Yoo2018) appears to be already present through the MCI stage. This suggests that LB disease is present in the olfactory areas as well as neocortical areas during MCI, consistent with the evidence from pathology studies (Attems et al., Reference Attems, Walker and Jellinger2014). The absence of such a correlation with the UPDRS is perhaps to be expected in our patient group since by applying the "one year rule" to recruitment, we restricted this group to people with a recent onset of parkinsonism and so a large proportion of patients had low scores on the MDS UPDRS. However, other patients had higher scores even without parkinsonism due to the effects on aging and diseases such as arthritis, further complicating the use of the UPDRS to identify a relationship with SS-16.
We also chose to explore whether the patient report of hyposmia in the QSSLBD might perform as well as SS-16 and thus be an even simpler was of identifying LB patients. This was not the case because the majority of participants who reported normal olfaction in fact scored abnormally (≤10 on SS-16) and so would be missed if this question were used alone. We conclude that proper olfactory testing is necessary to help identify LB disease in this patient group.
Our study benefits from being a relatively large and well-characterized cohort of probable MCI-LB and MCI-AD with detailed clinical and cognitive assessments and both structural and radionuclide imaging biomarkers and from using an established objective test of olfaction. Although using autopsy diagnosis may be regarded as the gold standard, this is not realistic for MCI studies and our use of consensus clinical panel diagnosis is the standard recognized by regulatory authorities (McKeith et al., Reference McKeith2007) and has been validated against autopsy (McKeith et al., Reference McKeith2007). This is further strengthened by the prospective annual diagnostic reviews in this cohort. However, our study cohort was selected on the basis of the possible presence of symptoms characteristic of LB disease identified in memory services, such as core clinical diagnostic features or supportive features in the diagnostic criteria, such as depression, anxiety, postural hypotension, and falls. While this was necessary to ensure a high proportion of MCI-LB in the study sample, it does mean those diagnosed with MCI-AD may not be entirely representative of all AD in such services. Here though this would suggest that perhaps a higher proportion of those diagnosed with AD have LB disease than even the high frequency that previous autopsy data support, meaning an even larger proportion of those with AD and abnormal olfaction might be true positives for LB disease. Many participants, particularly those with MCI-LB, were receiving cholinesterase inhibitors or memantine. This reflects a willingness of clinicians to use these medications in the MCI phase where they are confident that a neurodegenerative process is present. Finally, although as expected (Kane et al., Reference Kane2018) there was a significant imbalance in sex between MCI-LB and MCI-AD groups, there was not any evidence for sex differences in olfactory function.
In conclusion, in this prospective analysis of olfaction in MCI, we found impairments were more frequent in MCI-LB than MCI-AD and testing for such abnormal olfaction may be useful for identifying such early LB disease. While direct investigation of this is needed our findings, and the wider research data on LB disease and olfaction, suggest olfactory testing might be a useful way of improving the identification of early LB disease in memory services. Furthermore, the high sensitivity for AD and DLB in MCI suggest it may also be useful in other settings for identifying early LB disease, such as for other recognized prodromal presentations of DLB (delirium onset and psychiatric onset) (McKeith et al., Reference McKeith2020).
Conflicts of interest
All authors declare no conflicts of interest.
Acknowledgements
This work was supported by Alzheimer’s Research UK (ARUK-PG2015-13) and by the NIHR Newcastle Biomedical Research Centre. GE Healthcare provided the FP-CIT ligand for this investigator-led study. The authors would like to acknowledge the support of Ms Helen Kain in the undertaking of this study, and the NIHR Clinical Research Network North East and Cumbria for their support in recruiting participants. JO is supported by the Cambridge NIHR Biomedical Research Centre and the Cambridge Centre for Parkinson’s Plus Disorders. LMA is supported by the National Institute for Health Research Applied Research Collaboration South West Peninsula. The views expressed in this publication are those of the author(s) and not necessarily those of the National Institute for Health Research or the Department of Health and Social Care.
Author contributions
AJT: drafting of manuscript, formulation of research question, design of the study, interpretation of data, and review and critique of manuscript; CAH: data collection, analysis and drafting of manuscript; SB: data collection and review and critique of manuscript; RD: data collection and review and critique of manuscript; SL: data collection and review and critique of manuscript; NB: study administration, data collection, and review and critique of manuscript; MF: data collection and review and critique of manuscript; GR: data collection and review and critique of manuscript; LMA: formulation of research question, design of the study, and review and critique of manuscript; JO: design of the study and review and critique of manuscript; JPT: design of the study, data collection, and review and critique of manuscript; PCD: design of the study, formulation of research question, data collection, and review and critique of manuscript.
Open access
