Article Text

Download PDFPDF

Research paper
Cerebrospinal fluid Alzheimer biomarkers can be useful for discriminating dementia with Lewy bodies from Alzheimer’s disease at the prodromal stage
  1. Olivier Bousiges1,
  2. Stephanie Bombois2,
  3. Susanna Schraen3,
  4. David Wallon4,
  5. Muriel Muraine Quillard5,
  6. Audrey Gabelle6,
  7. Sylvain Lehmann7,
  8. Claire Paquet8,
  9. Elodie Amar-Bouaziz9,
  10. Eloi Magnin10,
  11. Carole Miguet-Alfonsi11,
  12. Xavier Delbeuck2,
  13. Thomas Lavaux12,
  14. Pierre Anthony13,
  15. Nathalie Philippi13,
  16. Frederic Blanc13
  17. For the ePLM network and collaborators
  1. 1 Laboratory of Biochemistry and Molecular Biology, and CNRS, Laboratoire de Neurosciences Cognitives et Adaptatives (LNCA), University Hospital of Strasbourg, Strasbourg, Alsace, France
  2. 2 Université Lille Nord de France, DISTALZ, Memory Center, Lille, France
  3. 3 UMR-S 1172 – JPArc-Centre de recherches Jean-Pierre Aubert Neurosciences et Cancer and CHU Lille, UF Neurobiologie, Université Lille, Lille, France
  4. 4 Department of Neurology, Rouen University Hospital, Rouen, France
  5. 5 Department of Biochemistry Laboratory, Rouen University Hospital, Rouen, France
  6. 6 CMRR (Memory Resources and Research Centre), Department of Neurology, CHU de Montpellier, Hôpital, Gui de Chauliac, Montpellier, France
  7. 7 Laboratoire de Biochimie et Protéomique Clinique, CHU de Montpellier and Université de Montpellier, IRMB, CRB, Montpellier, France
  8. 8 CMRR (Memory Resources and Research Centre) Paris Nord Ile de France and Histologie et Biologie du Vieillissement, Groupe Hospitalier Saint-Louis Lariboisière Fernand-Widal APHP, INSERM U942, Université Paris Diderot, Paris, France
  9. 9 Service de Biochimie et Biologie moléculaire, GH Saint-Louis-Lariboisière-Fernand Widal, APHP, Paris, France
  10. 10 Department of Neurology, Centre Mémoire Ressources Recherche Besançon Franche-Comté, CHU de Besançon, Besançon, France
  11. 11 Laboratoire de pharmacologie clinique, CHU de Besançon, Besancon, France
  12. 12 Laboratory of Biochemistry and Molecular Biology, University Hospital of Strasbourg, Strasbourg, France
  13. 13 Neuropsychology Unit, Neurology Service, and CNRS, ICube Laboratory UMR 7357 and FMTS, Team IMIS/Neurocrypto, University Hospital of Strasbourg, CMRR (Memory Resources and Research Centre), Geriatrics Day Hospital, Geriatrics Service, Strasbourg, France
  1. Correspondence to Dr Olivier Bousiges, Laboratory of Biochemistry and Molecular Biology, and CNRS, Laboratoire de Neurosciences Cognitives et Adaptatives (LNCA), University Hospital of Strasbourg, UMR7364, Strasbourg, France; olivier.bousiges{at}


Background Differential diagnosis between dementia with Lewy bodies (DLB) and Alzheimer’s disease (AD) is not straightforward, especially in the early stages of disease. We compared AD biomarkers (phospho-Tau181, total-Tau, Aβ42 and Aβ40) in cerebrospinal fluid (CSF) of patients with DLB and AD, focusing especially on the prodromal stage.

Methods A total of 1221 CSF were collected in different memory centres (ePLM network) in France and analysed retrospectively. Samples were obtained from patients with prodromal DLB (pro-DLB; n=57), DLB dementia (DLB-d; n=154), prodromal AD (pro-AD; n=132) and AD dementia (n=783), and control subjects (CS; n=95). These centres use the same diagnostic procedure and criteria to evaluate the patients.

Results In patients with pro-DLB, CSF Aβ42 levels appeared much less disrupted than in patients at the demented stage (DLB-d) (P<0.05 CS>pro-DLB; P<0.001 CS>DLB-d). On average, Aβ40 levels in patients with DLB (pro-DLB and DLB-d) were much below those in patients with pro-AD (P<0.001 DLB groups<pro-AD). The Aβ42/Aβ40 ratio in patients with pro-DLB remained close to that of CS. t-Tau and phospho-Tau181 levels were unaltered in patients with DLB (pro-DLB and DLB-d).

Conclusions Reduced levels of CSF Aβ42 were found in patients with DLB but rather at a later stage, reaching those of patients with AD, in whom Aβ42 levels were decreased even at the prodromal stage. At the prodromal stage of DLB, the majority of patients presented a normal CSF profile. CSF t-Tau and phospho-Tau181 were the best biomarkers to discriminate between AD and DLB, whatever the stage of disease.

Statistics from


Dementia with Lewy bodies (DLB) is the second cause of cognitive neurodegenerative disease after Alzheimer’s disease (AD). DLB was first described in 19611 and recognised as a common cause of dementia quite recently.2 Diagnostic criteria for DLB evolved over time and were subsequently revised in 2005.3 As in AD, the concept of prodromal stage has been described in DLB.4 Prodromal DLB (pro-DLB) can be named also as mild cognitive impairment (MCI) due to DLB or mild neurocognitive disorder of Lewy body disease as in Diagnostic and Statistical Manual IV (DSM 5).5 Prodromal DLB (pro-DLB) can be defined as the presence of the disease, but cognitive impairment is not sufficient to lead to functional deficits in activities of daily living and thus to dementia.6 Descriptions of prodromal DLB (proDLB) criteria have been proposed7: pro-DLB patients are those who have both the revised diagnostic criteria for DLB but instead of dementia, fit the criteria for MCI.6 The cognitive profile of prodromal DLB includes impairment on tests assessing visual memory, executive functions and visuoconstructive abilities with weaknesses for episodic memory, short time and working memory, verbal initiation, praxis, language, visuospatial abilities and social cognition.8 9 Brain MRI group studies of patients with pro-DLB have shown focal atrophy of the insula.10 Due to some clinical and neuropsychological similarities between AD and DLB, differential diagnosis is not straightforward. Furthermore, DLB and Parkinson’s disease (PD) may share parkinsonism (akinesia, rigidity and resting tremor), which may be absent or mild in DLB, especially at the beginning of the disease.2 7 Moreover, from a histopathological viewpoint, they share α-synuclein aggregates, commonly called Lewy bodies.11 However, patients with DLB show some additional, specific criteria compared with AD, such as illusions or visual hallucinations, fluctuating cognitive impairment and neuroleptic sensitivity.3 Despite the extremely high diagnostic specificity of these criteria (specificity for McKeith 1996 probable is 95.1% in early stages and 88% in late stages),12 their sensitivity remains low (32%) in the case of pure DLB.13 Their sensitivity was even lower (12%) when DLB was associated with AD.13 Consequently, two-thirds of patients with DLB are not diagnosed or are misdiagnosed.13 In addition, the MCI stage of DLB is not well characterised: the diagnosis relies on the clinical criteria of McKeith, and imaging measures (eg, MRI, brain perfusion SPECT, [123I]FP-CIT SPECT) are not modified early in the course of the disease. Although not specific to DLB, the preserved hippocampal volumes are associated with increased risk of probable DLB competing with AD dementia in patients with MCI, and this preservation may support prodromal DLB over AD, particularly in patients with MCI with non-amnestic features.14 The other potentially interesting examinations in the diagnosis of prodromal DLB are DAT-scan and iodine123-metaiodobenzylguanidine (MIBG) myocardial scintigraphy; however, it is difficult to advise their use, because very few studies have shown interest in the prodromal stage of patients with DLB. For demented forms, the utility of DAT imaging in distinguishing DLB from AD is well established, with sensitivity (78%) and specificity (90%).15 Normal DAT uptake may be reported in autopsy-confirmed DLB either because of minimal brainstem involvement and limited nigral neuron loss or a balanced loss of dopamine across the whole striatum, rather than predominantly in the putamen.16 MIBG scintigraphy quantifies postganglionic sympathetic cardiac innervation, which postmortem studies show as reduced in Lewy body disease. Early studies suggesting very high diagnostic accuracy were followed by lower, but still useful, sensitivity (69%) and specificity (87%) values for discriminating probable DLB from probable AD in multicentres settings, rising to 77% and 94% in milder cases with Mini-Mental State Examination (MMSE) >21.17 Nevertheless, it is essential to discover new biomarkers that are able to discriminate DLB from AD to increase differential diagnosis. At present, the only validated biological biomarkers are those analysed routinely in cerebrospinal fluid (CSF) for the diagnosis of AD: Aβ42, t-Tau, phospho-Tau181 and, to a lesser extent, Aβ40.

Therefore, we wanted to determine whether these CSF biomarkers could be useful to discriminate between DLB and AD diseases at an early stage. Overall, the literature suggests that there is no increase of CSF tau and phospho-Tau181 levels in DLB.18 19 Several studies have described a decrease of CSF Aβ42 in patients with DLB; however, this decrease appears only belatedly, at the stage of dementia, as we described previously.20 Moreover, CSF Aβ40 appears to be decreased in patients with DLB compared with AD,19 21 22 making the Aβ42/Aβ40 ratio very interesting for the differential diagnosis between these two diseases.20 21 Thus, based on our previous results, the aim of the present study was to assess the relevance of these biomarkers to discriminate between AD and DLB at an early stage using a large clinically based cohort of patients from the French ePLM centres group.


Study design and subjects

This retrospective multicentre study was based on the database of the ePLM network, which includes six French memory research centres performing both clinical and biological diagnosis of neurodegenerative disorders (Besançon, Montpellier, Lille, Paris-Lariboisière, Rouen and Strasbourg). Patients were selected in the database between January 2010 and December 2015. These centres used the same diagnostic procedures and diagnostic criteria. All the patients underwent an extensive examination, including physical, neurological and neuropsychological assessments (including the MMSE for general cognitive functions), laboratory tests and brain imaging. Each centre included patients with DLB according to McKeith’s and DSM-5 criteria (probable DLB, based on the existence of two core symptoms in addition to cognitive decline) for DLB demented patients.3 Patients withPro-DLB patients were defined as patients with MCI (Petersen criteria)23 and meeting the DSM-5 criteria (mild neurocognitive disorder of Lewy body disease). Patients with AD were selected according to Albert’s24 and Dubois’25 criteria for patients with pro-AD and McKhann’s criteria26 and Dubois’ criteria25 for demented AD patients, with the exclusion of the CSF criteria. Table 1 summarises the main demographics information of the patients, and table 2 described more accurately the clinical features of patients with DLB at the time of lumbar puncture. A total of 1221 patients were included in this study: 95 control subjects (CS), 57 patients with DLB at the prodromal stage (pro-DLB), 154 patients with DLB at the demented stage (DLB-d), 132 AD patients at the prodromal stage (pro-AD) and 783 AD patients at the demented stage (AD-d). The CS had various diagnoses (ie, frontotemporal lobe degeneration, psychiatric disorder (ie, depression and anxiety], traumatic brain injury, sleep apnoea syndrome, alcoholism, vascular MCI, cavernous angioma, prefragile X syndrome, Sneddon syndrome, encephalopathy, Sjögren’s syndrome) defined according to international criteria. For the purposes of this study, patients with dual disease (ie, AD/DLB or other mixed neurodegenerative disease) were excluded from the study. These patients were discriminated thanks to the criteria of AD,25 that is, storage impairment in verbal memory and hippocampal atrophy, and DLB,3 that is, probable DLB, without the results of the CSF. These patients with the two criteria were excluded.

Table 1

Population demographics and biomarker values

Table 2

Clinical characteristics of patients with DLB

CSF samples and analysis

CSF was obtained by lumbar puncture as part of the diagnostic work-up for patients with cognitive complaints. CSF was collected using standardised collection, centrifugation and storage conditions in the different centres. In each centre, CSF samples were collected in polypropylene tubes (Sarstedt, ref.: 62.610.201). Each CSF sample was transferred to the centre’s local laboratory less than 4 hours after collection; the sample was homogenised on receipt at the laboratory and was then centrifuged at 1000 g for 10 min at 4°C. Samples were then transferred in 0.5–1.5 mL polypropylene tubes (depending on the centre) and stored at −80°C until analysis. CSF Aβ1–42, tau and phospho-Tau181 were measured using standardised commercially available tests (Innotest sandwich ELISA) according to the manufacturer’s procedures (Fujirebio Europe, Ghent, Belgium). CSF Aβ1–40 was measured using different kits according to the centre: the Innotest kit for Strasbourg and Besançon, the IBL kit (Human Amyloid b (1–40) (N) Assay kit, IBL, Japan) for Paris, Montpellier, Lille and Rouen. An assay comparison was performed between the two tests by Fujirebio Laboratories. The correlation equation for Aβ1–40 between the Innotest and the IBL kit is: (Innotest Aβ1–40)=0.66 × (IBL Aβ1–40)+923. We used this equation to harmonise Aβ1–40 values between centres. Note that for Aβ1–40, we did not have the same number of patients as for other biomarkers, either because the dosage was not done systematically, or because there was insufficient CSF available to perform an additional Aβ40 assay. For this parameter, 442 patients had a dosage of Aβ40 and were distributed as follows: CS: n=52, pro-DLB: n=33, DLB-d: n=43, pro-AD: n=60, AD-d: n=281. All CSF assays were run as routine clinical neurochemical analyses. As the data were generated in the different centres through the routine activity of their laboratories, the lots of assay kits were variable within and between laboratories. The quality of the results was ensured by the use of validated standard operating procedures and, in each laboratory, the interassay variability was estimated using internal quality controls (QCs). For example, in Strasbourg, interassay coefficients of variations were: 5.4%–6.4% for Aβ42, 8.8%%–7.9% for t-Tau, 10.2%–16.4% for phospho-Tau181 and 3.7%–2.8% for Aβ40. The intra-assay variability observed in replicates was less than 10% in all four biomarkers. All the laboratories participated in an external QC assessment organised by the Alzheimer’s association’s QC programme for CSF biomarkers, ensuring also the quality of the results and the validity of the intersite data comparison. The cut-off values were determined according to the analysis of harmonisation of collection tubes on AD diagnosis performed by Lehmann and colleagues (ie, 700 ng/L for Aβ42, 400 ng/L for total t-Tau, 60 ng/L for phospho-Tau181).27

Statistical analysis

Statistical analyses were carried out using Graph-Pad PRISM, V.5 (GraphPad, San Diego, California, USA). Normally distributed data were analysed using one-way analysis of variance with Tukey’s post hoc analyses to determine between-group differences. In the case of non-Gaussian-distributed parameters, we used the Kruskal-Wallis test. In the case of two independent variables, a χ2 test was used. Receiver-operating characteristic (ROC) curve analysis was employed to evaluate the diagnostic value of CSF parameters. ROC curve comparisons were performed using MedCalc, V.12.7.0 (MedCalc Software, Ostend, Belgium).


Population demographic characteristics and mean CSF biomarkers values (Aβ42, Aβ40, t-Tau and phospho-Tau181) are presented in table 1 (and online supplementary table for results per centre). The results for CSF biomarkers are schematised in figure 1.

Supplementary file 1

Figure 1

Scatterplots of CSF analytes. CSF concentrations of t-Tau (A), phospho-Tau181 (B), Aβ42 (C), Aβ40 (for centres using the IBL kit, the Aβ40 values were calculated according to the equation: 0.66 × (IBL Aβ1–40)+923 (D), Aβ42/Aβ40 (E) in each patient group (for t-Tau, phospho-Tau181 and Aβ42 results, the number of patients per group was: CS n=95, pro-DLB n=57, DLB-d n=154, pro-AD n=132, AD-d n=783; for Aβ40 and Aβ42/Aβ40 results, the number of patients per group was: CS: n=52, pro-DLB: n=33, DLB-d: n=43, pro-AD: n=60, AD-d: n=281). ***P<0.001; *P<0.05 versus CS; §§§P<0.001 versus pro-AD and DLB-d. P values were calculated using ANOVA with Tukey’s post hoc analyses. Full lines represent the mean for each group; the dotted lines represent the cut-offs suggested by the manufacturer (Fujirebio). AD, Alzheimer’s disease; AD-d, AD-d, AD at demented stage; ANOVA, analysis of variance; DLB, dementia with Lewy bodies; DLB-d, DLB-d, DLB at demented stage; pro-AD, AD at prodromal stage; pro-DLB, pro-DLB, DLB at prodromal stage.

Amyloid biomarkers profile

As expected, CSF Aβ42 levels were significantly lower in the DLB-d group and AD groups (pro-AD and AD-d groups) compared with CS (P<0.001 for pro-AD, AD-d and DLB-d groups vs CS group) (figure 1C). No significant difference was observed between pro-AD, AD-d and DLB-d. Interestingly, Aβ42 values of these three groups were also significantly lower compared with those of the pro-DLB group (P<0.001 for pro-AD, AD-d and DLB-d groups vs pro-DLB group). The mean Aβ42 level of patients with pro-DLB was significantly lower than that of CS (P<0.05), indicating that Aβ42 levels began to be altered at the prodromal stage of DLB. For Aβ40, patients with pro-AD showed higher mean levels compared with CS, pro-DLB and DLB-d groups (P<0.001), whereas no difference was found between CS, pro-DLB and DLB-d groups (figure 1D). It is noteworthy that no significant difference in mean CSF-Aβ40 was found in the AD-d group compared with CS and the pro-DLB groups. However, the mean Aβ40 level in the AD-d group was significantly higher than in the DLB-d group (P<0.001); furthermore, the mean Aβ40 level diminished between the prodromal and the demented stage of AD (P<0.001 between pro-AD and AD-d). Compared with the CS group, Aβ42/Aβ40 ratio analyses showed no differences in the pro-DLB group but even slightly significant lower levels in the DLB-d group (P<0.01); however, no significant difference could be detected between the pro-DLB and DLB-d groups (figure 1E). Thus, the mean Aβ42/Aβ40 levels of the CS, pro-DLB and DLB-d groups were significantly higher compared with those of the AD groups (pro-AD and AD-d) (P<0.001 between CS, pro-DLB, DLB-d vs pro-AD and AD-d).

CSF Tau profile

The profile of the mean values between Tau and phospho-Tau181 for all the groups was quite similar. No significant variation was noted between the CS, pro-DLB and DLB-d groups (see figure 1A,B and table 1). As expected, the t-Tau and phospho-Tau181 values were significantly higher in the AD groups (pro-AD and AD-d) compared with the DLB and CS groups (P<0.001 for pro-AD and AD-d vs pro-DLB, DLB-d and CS groups for both t-Tau and phospho-Tau181 values).

Load of pathological biomarkers

After analysing the biomarker profile of each patient, we were able to classify patients according to the number of pathological biomarkers (figure 2). We found that the CSF profiles were different between patients with DLB and AD. In the pro-DLB group, half of the patients had no pathological biomarkers, and among those with just one pathological biomarker (36%), the majority presented a pathological level of Aβ42 (more than 70%). At the demented stage, only 28% of patients with DLB had no pathological biomarkers, and the proportion of patients with only Aβ42 as a pathological biomarker was 39% compared with 25% in the pro-DLB group.

Figure 2

Classification of patients according to the number of pathological biomarkers by diagnostic group. Pie chart segments in white: patients with no pathological biomarkers; segments in uniform grey: one pathological biomarker (t-Tau, phospho-Tau181 or Aβ42); segments in grey with white spots: two pathological biomarkers (t-Tau and phospho-Tau181, t-Tau and Aβ42 or phospho-tau181 and Aβ42); segments in black: three pathological biomarkers. Each pie chart segment indicates the pathological biomarker(s), the number and percentage of patients in the diagnostic group with this profile. AD, Alzheimer’s disease; AD-d, AD at demented stage; CS, control subjects; DLB, dementia with Lewy bodies; DLB-d, DLB at demented stage; pro-AD, AD at prodromal stage; pro-DLB, DLB at prodromal stage.

In the pro-AD group, the majority of patients presented three pathological biomarkers (61%), and only 4% of these patients had no pathological biomarker, supporting the notion that these biomarkers become abnormal very early in the disease, right from the prodromal stage. At the demented stage, 72% of patients with AD presented three pathological biomarkers and 2% had no pathological biomarkers.

Figure 3 shows the distribution of Aβ40, using cut-offs provided by the manufacturer (Fujirebio). In the pro-DLB group, the dispersion of values was very low: most patients (73%) had normal levels of Aβ40 in CSF, while 15% were above and 12% below the normal level. Compared with the pro-DLB group, a higher proportion (46%) of patients with DLB-d had a low level of Aβ40, whereas a lower proportion (35%) had a normal level. Among the AD groups, a majority of the patients at the prodromal stage (62%) had a high Aβ40 level, 33% presented a normal Aβ40 level and 5% a low Aβ40 level. Interestingly, patients with AD-d had CSF Aβ40 levels closer to those of the CS group than to those of the pro-AD group.

Figure 3

Aβ40 distribution within each of the diagnostic groups (for centres using the IBL kit, the Aβ40 values were calculated according to the equation: 0.66 × (IBL Aβ1–40)+923. Pie chart segments in white: patients with Aβ40 levels below 7000 ng/L; segments in uniform grey: patients with Aβ40 levels between 7000 ng/L and 12 000 ng/L; segments in black: patients with Aβ40 levels above 12 000 ng/L. The Aβ40 levels were considered normal when they were between 7000 ng/L and 12 000 ng/L according to the cut-off values provided by the manufacturer. Each pie chart segment indicates the number and percentage of patients in the diagnostic group with this profile. AD, Alzheimer’s disease; AD-d, AD at demented stage; CS, control subjects; DLB, dementia with Lewy bodies; DLB-d, DLB at demented stage; pro-AD, AD at prodromal stage; pro-DLB, DLB at prodromal stage.

Diagnostic relevance of CSF biomarkers

ROC curve analysis enabled us to determine the effectiveness of biomarkers at discriminating between AD and DLB (table 3). As expected, Tau and phospho-Tau181 were very good at discriminating between the two pathologies, whatever the stage (AUC=0.91 and 0.93 at the prodromal stage, for t-Tau and phospho-Tau181, respectively; AUC=0.87 at the demented stage for both t-Tau and phospho-Tau181; AUC=0.88 and 0.89 for all patients for t-Tau and phospho-Tau181, respectively). Interestingly, t-Tau and phospho-Tau181 each performed as well as all three biomarkers considered together (t-Tau+phospho-Tau181+Aβ42), whatever the stage (AUC=0.93 at the prodromal stage and AUC=0.87 at the demented stage). Aβ42 was less effective than t-Tau and phospho-Tau181 at discriminating between the two pathologies at the prodromal stage (AUC=0.76) and almost ineffective at discriminating between them at the demented stage (AUC=0.64). Aβ40 alone seemed to be quite a good discriminant (AUC=0.80) at the prodromal stage but was less good at the demented stage (AUC=0.72). Interestingly, at the prodromal stage, the Aβ42/Aβ40 ratio was the best tool to discriminate between AD and DLB (AUC=0.95) and a combination of t-Tau, phospho-Tau181 and the Aβ42/Aβ40 ratio achieved an AUC of 0.97. It should be noted that at the demented stage, the Aβ42/Aβ40 ratio was less effective, with an AUC of 0.72. The cut-offs used in our study were good enough at the prodromal stage but lost specificity at the demented stage (table 3, right columns).

Table 3

ROC analysis of CSF parameters

Correlations between CSF biomarkers and cognitive parameters

In both AD and DLB, Aβ42 and t-Tau were correlated to the MMSE score (Aβ42, rs=0.3521, P<0.0001 for DLB and rs=0.181, P<0.0001 for AD; t-Tau, rs=−0.1906, P=0.0131 for DLB and rs=−0.1762, P<0.0001 for AD) (figure 4). In contrast, Aβ40 was correlated with the MMSE score in AD cases but not in patients with DLB despite a slight trend (rs=0.1923, P=0.098 for DLB; rs=0.2276, P<0.0001 for AD). Conversely, phospho-Tau181 was correlated with the MMSE only in patients with DLB but not in AD (rs=−0.2085, P=0.0065 for DLB; rs=−0.07667, P=0.0597). However, Aβ42 and Aβ40 relative to each other seemed to be correlated for patients with DLB (rs=0.3614, P<0.0013) and were of course correlated for AD patients (rs=0.3309, P<0.0001). In the same way, Tau and phospho-Tau181 are correlated to each other, both in patients with DLB (rs=0.7926, P<0.0001) and AD patients (rs=0.8071, P<0.0001).

Figure 4

Correlation between MMSE and CSF biomarkers in patients with DLB and AD. For patients with DLB (pro-DLB+DLB-d), correlation between (A) Aβ42 and MMSE score, (B) Aβ40 and MMSE score, (C) Aβ40 and Aβ42, (D) t-Tau and MMSE score, (E) phospho-Tau181 and MMSE score, (F) phospho-Tau181 and t-Tau. For patients with AD (pro-DLB+DLB-d), correlation between (G) Aβ42 and MMSE score, (H) Aβ40 and MMSE, (I) Aβ40 and Aβ42, (J) t-Tau and MMSE score, (K) phospho-Tau181 and MMSE score, (L) phospho-Tau181 and t-Tau. Data were evaluated by Spearman’s rank correlation (rand P value). AD, Alzheimer’s disease; CSF, cerebrospinal fluid; DLB, dementia with Lewy bodies; DLB-d, DLB at demented stage; pro-DLB, DLB at prodromal stage; MMSE, Mini-Mental State Examination.


In clinical practice, DLB is sometimes difficult to diagnose, especially at the onset of the disease, when it is frequently misdiagnosed as AD. This diagnosis is essential because: (1) DLB represents the second cause of dementia after AD, (2) antipsychotic drugs can lead to clinical worsening and death in patients with DLB and thus it is important not to miss the diagnosis, (3) patients with DLB need specific care against falls, orthostatic hypotension, syncope and other autonomic impairment such as constipation; (4) misdiagnosis may lead to the failure of therapeutic trials in both AD and DLB. CSF AD biomarkers provide useful tools in this setting in clinical practice. Although these biomarkers are altered in some other neurodegenerative diseases, they can provide interesting information to discriminate between AD and DLB.

We have confirmed that the Aβ42 decrease only occurred late in DLB (ie, at the demented stage), half of the patients in the pro-DLB group having a normal CSF profile. In this study, the distribution of patients by number of pathological biomarkers differed from that of our previous work.20 This discrepancy may be partially explained by the different cut-offs used in the previous study and the multicentric design of the present study.

What is the explanation for the decrease in Aβ42 in patients with DLB? To date, several lines of evidence relate to interaction between the amyloid pathway and synucleinopathy. More than 80% of patients with DLB showed moderate or abundant cortical amyloid plaques,28 and α-synuclein pathology is also found in up to 50% of patients with AD (for a review, see ref 29), suggesting a close link between amyloidopathy and synucleinopathy.

An interesting study with autopsy verification showed that DLB patients with amyloid plaques presented significantly lower CSF Aβ42 concentrations than DLB patients without amyloid plaques.30 If we correlate these results to ours, it would suggest that the majority of our patients with pro-DLB did not have amyloid pathology, which occurs in the demented stage. In the study by Slaets et al,30 the majority of patients (73%) at the demented stage had amyloid plaques, which could explain why Aβ42 levels in patients with DLB-d were similar to those of the AD groups in our study. This delayed decrease in Aβ42 levels in CSF of DLB compared with patients with AD can be explained by the fact that amyloid deposits have been shown to occur late, after the α-synuclein aggregation in patients with DLB.28

Our results in patients with AD and DLB highlight a correlation between MMSE and Aβ42 peptides and a correlation between MMSE and Aβ40 only in patients with AD (figure 4). The correlation between CSF amyloid peptides and the severity of cognitive decline is a controversial point, especially for AD. Most of the studies, as described in the recent Blennow’s review, report in patients with AD no relation of Aβ42 levels according to the progress of the cognitive decline,31 supporting the idea that the change from normal to pathological levels most likely occurs during the preclinical asymptomatic phase of the disease. However, this observation is not shared by all.32–34 For example, Ganzer and colleagues found, although weak, a correlation between CSF-Aβ1–42 and MMSE score.35 Kaerst and colleagues, using a slightly different approach, showed in three groups of patients with DLB dementia classified according to disease severity (mild, medium and severe) that the greater the DLB progression, the lower the Aβ42 levels, which implies an evolution of the levels of Aβ42 throughout the pathology.33 More obviously, Mizoi and collaborators32 found, similarly to our results, a correlation of the MMSE as well with Aβ42 as with Aβ40. In the same way Gómez-Tortosa34 and colleagues indicate that β-amyloid level was the CSF marker that best reflected disease progression, decreasing with longer disease duration in DLB and with lower MMSE scores in AD. These studies suggest that Aβ42 could play a prognostic marker. This hypothesis is underpinned by a study in patients with PD (presenting an α-synuclein aggregation, as in the case of patients with DLB), in which the authors showed that patients with a low Aβ42 level in CSF had cognitive deficits (Mattis Dementia Rating Scale score) earlier than patients with a normal Aβ42 level.36 Likewise, Abdelnour et al reported a faster cognitive decline (MMSE) in patients with DLB having an Aβ42 decrease compared with DLB patients with a normal Aβ42 level.37 This study suggests that a pathological CSF Aβ42 is associated with a more rapid cognitive decline in synucleinopathies. Thus, at the individual level, a follow-up of Aβ42 level could have a potentially interesting role as a prognostic marker in these pathologies. It was established that t-Tau and phospho-Tau181 values are not correlated with the severity of the disease, even in patients with DLB presenting neurofibrillary tangles30; however, we find a correlation of t-Tau with MMSE in both DLB and AD, and a correlation only with patients with DLB for phospho-Tau181 and MMSE. These results suggest in patients with AD that phospho-Tau181 does not evolve over time and so is at the highest level from the MCI stage. In patients with DLB, our results suggest that both t-Tau and phospho-Tau181 are able to progress during the pathology.

CSF Aβ40 levels in the different groups showed the same variations as in our previous, monocentric study,20 that is, Aβ40 levels slightly though not significantly decreased in patients with DLB-d in comparison with patients with pro-DLB and CS. Of note, the CSF Aβ40 increase in patients with pro-AD coupled with the slight decrease in CSF Aβ40 in patients with DLB-d means that this biomarker was a little more discriminating than Aβ42, whatever the stage of AD and DLB (AUC=0.80 (Aβ40) and AUC=0.76 (Aβ42) for prodromal groups and AUC=0.72 (Aβ40) and AUC=0.64 (Aβ42) for demented groups). It is worth highlighting the fact that a lower level of Aβ40 has been clearly described in patients with DLB compared with patients with AD18 21 22; however, this reduction is not always demonstrated in relation to controls,18 even if others found a significant reduction in patients with DLB compared with controls.22

Finally, the Aβ42/Aβ40 ratio was a very good way of discriminating between AD and DLB, essentially at the prodromal stage (AUC=0.95), but this ratio lost its discriminatory power at the demented stage (AUC=0.72) (table 3). This result is in accordance with the literature, as most studies show that the Aβ42/Aβ40 ratio improves the ability to differentiate patients with AD from various neurodegenerative diseases (eg, vascular dementia and DLB), in comparison with Aβ42 alone,21 from the MCI stage.20 38 39

Concerning patient selection, our study has limitations, because the diagnosis of patients was based on clinical criteria, not on autopsy verification, and it was also a retrospective study. Another important limitation is the imbalance of patients per centre that could lead to a bias, for example, patients with pro-DLB mainly come from the Strasbourg team. This bias may interfere especially with the Aβ42 results, which are higher in the pro-DLB group than in the DLB-d, pro-AD and AD-d groups only in the Strasbourg team. However, we verified that no statistical difference exists between the centres for each biomarker (Aβ42, Tau and phospho-Tau) (data not shown). We conclude that the differences observed in the pro-DLB group are not due to an intercentre difference. Furthermore, the prodromal criteria for DLB have not been validated by follow-up and progression to DLB in our series. In addition, we tried to exclude patients with dual AD/DLB pathology by clinical tests. Our objective was to highlight differences between pathologies as much as possible pure. This may explain why Tau and phospho-Tau levels have such an important discriminating power. It would be interesting to include, in future studies, groups of patients with comorbidity in order to be closer to the ‘real life’ diagnosis conditions. Nevertheless, concerning the AD diagnosis, the CSF analysis in patients with AD at either the prodromal or demented stage was in line with our previous results,20 and the literature and so confirms the quality of the clinical diagnosis. Concerning DLB diagnosis, given the excellent specificity (96%–98%) of the McKeith criteria,3 it seems likely that the patients in our DLB group were indeed correctly diagnosed. Furthermore, biomarker differences between the AD and DLB groups, including t-Tau and phospho-Tau181 levels and the Aβ42/Aβ40 ratio, and the difference in profile between the two groups, may suggest that these groups really did not have the same disease and so reinforce the relevance of all of our findings.


To summarise, this study enabled us to determine the best biomarkers to discriminate between AD and DLB, in particular according to the stage of the pathology. T-Tau and phospho-Tau181 are the best biomarkers to differentiate AD and DLB, whatever the stage. However, at the prodromal stage, the Aβ42/Aβ40 ratio appears to be the best tool to differentiate between these two diseases. Aβ40 levels, though less discriminating than the Aβ42/Aβ40 ratio at the prodromal stage, have the same level of discrimination as the Aβ42/Aβ40 ratio at the demented stage. Of note, the Aβ40 increase in CSF of patients with pro-AD gives this biomarker a level of discrimination between AD and DLB that is at least equivalent to, or even slightly better than, that of Aβ42.

Our results highlight the interest of CSF AD biomarkers: in fact, in most cases, either none of the biomarkers are disturbed or only Aβ42 is decreased in CSF of patients with DLB. However, an isolated Aβ42 decrease is not sufficient to confirm DLB, as a similar decrease has been reported in many other types of dementia.40 Hence, new specific biomarkers of DLB still need to be found. α-Synuclein is a protein that specifically aggregates in DLB; therefore, it is likely that the latter is a potentially interesting biomarker. However, a few studies are rather contradictory when analysing the α-synuclein levels results in the CSF of DLB patients: decreased, unchanged and even elevated levels have all been reported.41 Furthermore, to date, no major study has evaluated correctly the interest of α-synuclein in addition to CSF biomarkers (other than conventional markers including Aβ42, the Tau protein and its phosphorylated form) in patients with MCI or mild dementia in order to distinguish AD and DLB. This point should be further investigated.

Until specific biomarkers of DLB can be found, AD biomarkers will be very useful for the differential diagnosis, even at early stages.



  • Contributors OB and FB: study concept, design and supervision; FB, NP and PA: selection of patients from Strasbourg; OB and TL: analysis and interpretation of all data. SB: selection of patients from Lille; SS and XD: analysis of CSF assays of patients from Lille; DW: selection of patients from Rouen; MMQ: analysis of CSF assays of patients from Rouen; AG: selection of patients from Montpellier; SL: analysis of CSF assays on patients from Montpellier; CP: selection of patients from Paris; EA-B: analysis of CSF assays of patients from Paris; EM: selection of patients from Besançon; CM-A: analysis of CSF assays of patients from Besançon.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.