Background
Studies based on clinical and neuropathological diagnoses have shown that Alzheimer’s disease (AD) is the most common cause of dementia [
1‐
4]. However, there are several neurodegenerative [
5‐
8] and non-neurodegenerative pathologies [
1,
9‐
12] that are known to contribute to cognitive impairment and a dementia diagnosis. Different clinical dementia syndromes have degrees of clinico-pathological correlation, therefore if clinical diagnosis is used to estimate the accuracy of biomarkers and their cutoffs, inaccurate results may occur [
13‐
15]. In addition, coincident neurodegenerative diseases (NDDs) and vascular pathology are common findings in subjects with AD in autopsy series [
1,
9,
14,
16‐
18]. However, clinical studies of dementia of the Alzheimer type (DAT) and other NDDs assign a single primary clinical diagnosis to patients. Accordingly, most biomarker studies are based on clinical diagnoses and report results on subjects using a single NDD diagnosis as the outcome. While the use of clinical diagnoses is helpful for the screening and the evaluation of new biomarkers, substantial follow up studies are needed to establish the performance of these biomarkers. Such studies should include in the analysis consecutive series of patients with NDDs and non-NDDs together with multimodal biomarkers assessing their performance in complex settings with several coincident diseases. Previously, retrospective studies have analyzed the correlation between neuropathological findings and CSF [
14,
19,
20], magnetic resonance imaging (MRI) [
21‐
24] and positron emission tomography (PET) [
21,
25‐
27]. Most of these studies tried to categorize patients into a single diagnostic category. Here, we instead tried to assess how different combinations of biomarkers can detect different coincident pathologies and therefore predict the different combinations of neuropathological substrates of the cognitive impairment in the studied subjects to help identify homogeneous cohorts of patients for clinical studies and clinical trials. This is especially true of clinical trials for DAT in which one pathology is targeted for study such as therapies that target Aβ or tau mediated mechanisms of neurodegeneration. Thus, here we specifically tested cerebrospinal fluid (CSF) biomarkers for the diagnosis of AD (Aβ
1-42, total tau (t-tau) and phosphorylated tau (p-tau
181) and dementia with Lewy bodies (DLB) (α-synuclein), MRI hippocampal and occipital pathology for the diagnosis of coincident DLB and medial temporal lobe (MTL) pathologies and occipital hypometabolism for the diagnosis of DLB. In addition, we tested the neuropathological association of hallucinations, memory and executive dysfunction. For this study, we examined the first 22 patients in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) who were longitudinally followed to death and underwent postmortem examination.
Discussion
We performed a multimodal biomarker analysis of consecutive ADNI subjects who came to autopsy after longitudinal follow-up to death. Most of the subjects were late amnestic MCI and probable DAT at baseline visit, without any atypical clinical presentation. In addition, subjects with vascular disease or a Hachinski Ischemic Score >4 were excluded from the ADNI study [
29]. Despite this, only four out of 22 subjects had a neuropathological diagnosis of pure AD (13.6%), while DLB, MTL-TDP and infarcts were present in 45.5%, 40.9% and 22.7% of the patients, respectively. A predominantly executive dysfunction was associated with the presence of DLB and hallucinations, as recorded in the NPI-Q. CSF fluid Aβ
1-42, t-tau and p-tau
181 were associated with the phosphorylated tau (neurofibrillary tangle) Braak stage and the NIA-AA criteria A score, but not with other pathologies. Baseline occipital hypometabolism accurately predicted the presence of DLB pathology, whereas MRI GM occipital atrophy did not. MTL-TDP and AGD were associated with greater hippocampal atrophy.
Despite the neuropathological heterogeneity of the patients, all the demented subjects had a DAT probable clinical diagnosis and the MCI patients had and AD-like profile; three subjects had parkinsonian signs, of those two had DLB pathology and one did not. Therefore, an overall classical DAT presentation does not rule out the presence of coincident vascular disease or other neurodegenerative disease even in the presence of CSF and PET amyloid imaging findings compatible with AD pathology. This is not surprising because memory impairment is the most common presenting clinical symptom of clinically diagnosed DLB (cDLB) patients (with confirmed abnormal dopamine transporter imaging) [
49] and most of the DLB/AD + DLB cases were diagnosed as DAT, showing a low sensitivity of the clinical criteria, at least in this small series [
50]. One difference between our study and other previous studies is that we did not include any patients with a cDLB diagnosis and patients with coincident DLB in our study had a typical DAT profile. Nevertheless, we found two clinical markers that were highly predictive of coincident pathologies: a predominant dysexecutive syndrome and the presence of hallucinations. A previous neuropathological study found a high specificity of visual hallucinations in DLB (with a prevalence of 1% early in the course of AD), although prevalence of visual hallucinations was low and therefore not was not a sensitive biomarker [
51]. In our study, we found that during the progression of disease a prominence of dysexecutive impairment in the presence of an amnestic profile is a marker for coincident DLB. This is consistent with a previous study that described a worse executive function in subjects with DLB and worse memory in patients with AD, although no classification performance was reported [
52]. However, a predominantly disexecutive syndrome might not be a specific biomarker and other NDDs like frontotemporal lobar degeneration could have a similar profile.
Currently, Aβ amyloid PET imaging and CSF Aβ
1-42 are the most widely accepted research biomarkers for AD which have shown an important correlation with brain Aβ amyloid deposition [
20,
26,
53] and with each other [
54]. Confirming the results of our previous study in which one fourth of the patients had coincident pathologies [
14], mainly DLB, and studies for other groups [
19], CSF Aβ, t-tau and p-tau
181 levels can reliably predict AD pathology even in the presence of other coincident pathologies and subjects with Aβ levels above the published cutoff [
34] had a low burden of AD.
Three studies with FDG-PET and several neuropathologically confirmed cases have previously reported occipital FDG-PET hypometabolism independent of coincident AD: Albin et al. included three DLB and three with AD + DLB [
25], Kantarci et al. included 2 AD and 3 DLB (and a larger number of clinically diagnosed cases) [
21] and Minoshima et al. included 7 AD + DLB, 4 DLB and 10 AD (and a larger number of clinically diagnosed cases) [
27]. The last study reported a sensitivity of 90% and a specificity of 80% for the diagnosis of DLB with or without AD based on hypometabolism in the occipital cortex [
27]. Kantarci et al. carried out a study mostly based on clinically diagnosed patients, and they described an area under the curve (AUC) in the receiver operating characteristic (ROC) of 0.84 for the FDG-PET with cDLB patients showing an occipital hypometabolism independently of Aβ deposition measured by PiB PET [
21]. In our study, we found an 80% sensitivity and a 100% specificity based on occipital FDG-PET hypometabolism. All of our DLB subjects had coincident AD and this did not affect the accuracy of the classification. In addition, the only DLB case that was classified as non DLB by the occipital FDG-PET cutoff was the only one that did not have diffuse neocortical LBs. The percentage of predicted DLB pathology in the ADNI-1 DAT subjects was similar to the one observed in the autopsied subjects. Interestingly, it has been suggested that occipital hypometabolism might be an preclinical biomarker of DLB [
55]. On the other hand, functional neuroimaging approaches that measure striatal dopaminergic innervation and myocardial sympathetic nerve integrity might be more specific to changes associated with LB pathology, specially the latter which captures postganglionic denervation which is present in PD and DLB patients, but not in patients with other atypical parkinsonisms [
56]. However, doing these additional tests would increase the, cost, time and inconvenience for patients, whereas FDG PET is also helpful for the differential diagnosis in non-parkinsonian syndromes. Therefore a FDG PET measure that specifically predicts coincident DLB pathology would be preferable. The voxels that contribute to HCI are located in temporal, occipital and parietal cortices and it might be possible that the association with HCI with the presence of DLB and not Braak stage might reflect that this measure is also capturing the posterior cortical metabolism characteristic of DLB. Nevertheless, the AD group without DLB was less impaired at baseline and we did not analyze the FDG PET scans that were matched for clinical severity.
Neither our study nor others have found occipital lobe atrophy in cDLB or DLB [
21,
23]. In addition, we found no differences in hippocampal volume based on the presence of coincident DLB. Conversely, it has been reported that cDLB patients have similar hippocampal volume as CN subjects but lower hippocampal volume than DAT patients with an high diagnostic accuracy [
21] and that a semiquantitative visual rating of MTL MRI atrophy had a high accuracy to classify AD against DLB and pathologically diagnosed vascular cognitive impairment patients [
24]. The definition of the DLB group might explain these differences. For example, the study by Burton et al. included cDLB diagnosis patients [
24] and the multimodal study by Kantarci et al. was mostly comprised of cDLB patients [
21], therefore it can be expected that the pattern of atrophy is different in DLB with a cDLB diagnosis compared to those AD + DLB with a DAT diagnosis. In addition, the study by Kantarci et al. described a large sample of neuropathologically diagnosed subjects in whom only DLB subjects with high DLB probability as defined by McKeith criteria [
46] had similar hippocampal volume as CN subjects, whereas intermediate and low probability DLB had similar hippocampal atrophy as AD subjects [
22]. Therefore, hippocampal volume might not be a good marker of coincident DLB in a cohort of DAT subjects.
Late MCI ADNI 1 patients and DAT patients recruited in ADNI represent amnestic, cognitively impaired, subjects without any cognitive signs or symptoms suggestive of non-AD pathologies and a low vascular risk profile [
29]. Therefore, these patients with multiple coincident pathologies represent the typical patients recruited in AD clinical trials. These multiple coincident pathologies have important implications for clinical trials and the approach for treating patients. Studies of the brains of patients treated with Aβ immunotherapy have shown a decrease of total Aβ burden and a decrease of neurite curvature ratio and a [
57‐
59] and an increase of amyloid deposition in neocortical blood vessels that might decrease over time [
59], without any strong effect on tau [
57] or α-synuclein [
60] clearance. Therefore, multimodal biomarker approaches that aim to detect the different coincident pathologies (instead to categorizing patients into a single diagnostic category), will be needed to select homogenous populations for protein-specific targeted clinical trials and to tailor the treatment for each patient.
Competing interests
M.W.W.: stock options, Elan, Synarc; travel expenses, Novartis, Tohoku University, Fundacio Ace, Travel eDreams, MCI Group, NSAS, Danone Trading, ANT Congress, NeuroVigil, CHRU-Hopital Roger Salengro, Siemens, AstraZeneca, Geneva University Hospitals, Lilly, University of California, San Diego–ADNI, Paris University, Institut Catala de Neurociencies Aplicades, University of New Mexico School of Medicine, Ipsen, Clinical Trials on Alzheimer’s Disease, Pfizer, AD PD meeting, Paul Sabatier University; board membership, Lilly, Araclon, Institut Catala de Neurociencies Aplicades, Gulf War Veterans Illnesses Advisory Committee, VACO, Biogen Idec, Pfizer; consultancy, AstraZeneca, Araclon, Medivation = Pfizer, Ipsen, TauRx Therapeutics, Bayer Healthcare, Biogen Idec, ExonHit Therapeutics, Servier, Synarc, Pfizer, Janssen; honoraria, NeuroVigil, Insitut Catala de Neurociencies Aplicades, PMDA = Japanese Ministry of Health, Labour, and Welfare, Tohoku University; commercial research support, Merck, Avid; government research support, DOD, VA. Other authors report no conflicts of interest. JM has consulted for Eisai, Glaxo-SmithKline, Novartis, Estavo Jansen Alzheimer Immunotherapy Program, Pfizer and Eli Lilly/Avid Radiopharmaceuticals. D. Aisen serves on a scientific advisory board for NeuroPhage and as a consultant to Elan Corporation, Wyeth, Eisai Inc., Bristol-Myers Squibb, Eli Lilly and Company, NeuroPhage, Merck & Co., Roche, Amgen, Abbott, Pfizer Inc, Novartis, Bayer, Astellas, Dainippon, Biomarin, Solvay, Otsuka, Daiichi, AstraZeneca, Janssen, Medivation, Inc., Theravance, Cardeus, and Anavex and receives research support from Pfizer Inc., Baxter International Inc.
Authors’ contributions
All authors read and approved the final manuscript, contributed to interpretation of the data and critical review of the manuscript and study concept. NJC and JCM established the neuropathological diagnoses and gradings. DC and EH collected and performed immunohistochemistry. XD and CD processed and analyzed the MRI data. KC, AF, NA, AR, RJB and EMR processed and analyzed the FDG-PET data. JBT drafted the manuscript and performed the statistical analyses. JQT drafted the manuscript.