Abstract
Background:
Total Tau concentration in cerebrospinal fluid (CSF) is widely used as a biomarker in the diagnosis of neurodegenerative process primarily in Alzheimer’s disease (AD). A particularly high Tau level may indicate AD but may also be associated with Creutzfeldt-Jakob disease (CJD). In such situations little is known about the distribution of differential diagnoses.
Objective:
Our study aimed to describe the different diagnoses encountered in clinical practice for patients with dementia and CSF Tau levels over 1000 pg/ml. We studied the p-Tau/Tau ratio to specify its ability to distinguish AD from CJD.
Methods:
Patients (n = 202) with CSF Tau levels over 1000 pg/ml were recruited in three memory clinics in France. All diagnoses were made using the same diagnostic procedure and criteria.
Results:
Patients were diagnosed with AD (n = 148, 73.2%), mixed dementia (n = 38, 18.8%), CJD, vascular dementia (n = 4, 2.0% for each), Lewy body dementia, and frontotemporal dementia (n = 3, 1.5% for each). Dispersion of CSF Tau levels clearly showed an overlap between all diagnoses. Using the p-Tau/Tau ratio suggestive of CJD (<0.075), all CJD patients were correctly categorized and only two AD patients were miscategorized. This ratio was highly associated with CJD compared to AD (p < 0.0001).
Conclusion:
Our study showed that in clinical practice, extremely high CSF Tau levels are mainly related to diagnosis of AD. CJD patients represent a minority. Our results support a sequential interpretation algorithm for CSF biomarkers in dementia. High CSF Tau levels should alert clinicians to check the p-Tau/Tau ratio to consider a probable diagnosis of CJD.
INTRODUCTION
Diagnosis of dementia may be difficult due to many cognitive symptoms shared by distinct neurodegenerative processes, especially in atypical forms of dementia. For decades, diagnosis of Alzheimer’s disease (AD) was based solely on clinical criteria [1] and postmortem neuropathological examination was the only means of providing a definite diagnosis. Low specificity levels for clinical diagnosis of AD reflect overlapping clinical profiles between AD and other dementias, especially at the earliest stages [2]. In this context, cerebrospinal fluid (CSF) biomarkers represent a new diagnostic tool [3 –5]. These biomarkers are well validated, achieving sensitivity, specificity, and diagnostic accuracy levels consistently exceeding 80% [6, 7]. Furthermore, several studies have shown that the use of CSF biomarkers improves the validity of clinical criteria [8] and the utility of CSF biomarkers is widely demonstrated in clinical research and practice [9]. In France, the prescription of CSF biomarkers for diagnosis of AD is guided by the recommendations of the French Health Authority (HAS).
The CSF biomarker profile of AD consists of decreased levels of amyloid-β 1-42 (Aβ42) and elevated levels of total Tau protein (Tau) and its phosphorylated form at threonine 181 (p-Tau) [10, 11]. Tau levels are believed to correlate with the rate of axonal degeneration across several neurologic diseases, while p-Tau may be a more specific marker for the formation of neurofibrillary tangles [12]. Overlap of the CSF levels of Tau and Aβ42 between AD and frontotemporal dementia (FTD), Lewy body dementia (LBD), and vascular dementia (VaD), is known to limit the diagnostic accuracy of these biomarkers [13]. Such overlap may not always be explained by the co-occurrence of AD pathology and other neurodegenerative processes. For instance, AD lesions were found in 23–31% of a pathological series of LBD patients [14].
The situation also arises between AD and Creutzfeldt Jakob disease (CJD). On the one hand, Tau levels may be very high in AD and for some, may predict rapid cognitive decline [15], but on the other hand, many studies have described extremely high Tau levels in CJD [16, 17]. Blennow et al. proposed a threshold of Tau >1400 pg/ml as suggestive of CJD [11]. As a consequence, in current practice clinicians should question high Tau levels in CSF and should consider alternative diagnoses, especially CJD. To the best of our knowledge, there is no study exploring the distribution of diagnoses in patients with high Tau levels. Moreover, p-Tau, which is generally not elevated in other forms of dementia, seems to be a more specific marker for AD [12, 18]. Since CSF total Tau concentrations may reflect the extent of the neurodegenerative process and hyperphosphorylation of Tau (associated with neurofibrillary pathology) is not usually present in CJD, several authors have suggested that the p-Tau/Tau ratio might improve diagnostic accuracy between CJD and AD [19]. A p-Tau/Tau ratio <0.075 was considered as highly suggestive of CJD [20].
In clinical practice, high CSF Tau levels are commonly found when biomarkers are used to explore cognitive decline. However, little is known about the distribution of possible diagnoses when confronted with particularly high CSF Tau levels, often leading the clinician to consider CJD. Such information is thus of real importance prior to engaging other complementary investigations. To the best of our knowledge, all previous studies that assessed the diagnostic value of CSF biomarkers recruited patients based on final diagnosis. The aim of our study was to detail the distribution of all diagnoses encountered after measuring CSF biomarkers in a cohort of patients fulfilling dementia criteria and presenting a high level of CSF Tau, in accordance with real clinical practice. The secondary objective was to confirm whether p-Tau/Tau ratio was able to distinguish AD from CJD in our sample.
MATERIAL AND METHODS
Study design and subjects
Of 1,654 patients with dementia who underwent lumbar puncture (LP) between September 2005 and September 2014 in three distinct memory clinics in France (Rouen, Paris Lariboisière, and Lille), 208 patients were included in this study according to our inclusion criteria. All three centers are specialized in the care management of patients with cognitive disorders, and used the same diagnostic procedure and criteria.
Patients underwent a comprehensive clinical examination including personal medical and family history, neurological examination, neuropsychological assessment, LP with CSF biomarker analysis, structural brain imaging (by 1.5T or 3T MRI) and functional imaging by (ECD-SPECT or FDG-PET).
As recommended by the French Health Authority, CSF biomarkers may be used in clinical practice in cases of atypical clinical presentation and/or rapidly progressive cognitive decline and/or in cases of diagnostic uncertainty. Neuropathological examination, if available, was performed using standard procedures and all diagnostic data were collected [21]. When diagnosis of CJD was possible, an electroencephalography and CSF 14-3-3 protein dosage was systematically recorded as part of the diagnostic criteria.
In each center, two neurologists and a biologist expert in CSF biomarkers provided a consensus diagnosis. For each patient, diagnosis was made considering CSF results and according to validated clinical diagnostic criteria for probable AD [4], probable FTD [22], cortico-basal degeneration [23], probable CJD [24], and probable LBD [25]. Patients with clinical LBD features and biochemical AD signature were included in the AD group. A diagnosis of mixed dementia (MD) was retained in cases of a combination of clinical, neuroimaging, and biochemical criteria of AD and vascular dementia [26, 27]. All patient diagnoses were validated in a second step by an independent neurologist before collecting the results of this study. The diagnostic assessment involved a review of full medical history, clinical symptoms, Mini-Mental Status Evaluation (MMSE) and neuropsychological assessment, analysis of brain imaging, and CSF biomarker results. If there was any disagreement between neurologists regarding the final diagnosis, patients were excluded from the analysis. Patients with recent history of stroke, head trauma, or seizure, 3 months before LP were excluded given that these conditions are well known to temporarily increase the Tau level of CSF.
CSF analyses
CSF was obtained by LP between the L3/L4 or L4/L5 intervertebral space, using a 25-gauge needle. In an effort to reduce intersite variability in CSF readings, all three centers used a common 10-ml polypropylene tube to collect the CSF (catalog number 62.610.201; Sarstedt, Nümbrecht, Germany) [28]. All samples were aliquoted after centrifugation into polypropylene Eppendorf tubes, then frozen at –80°C within 1 h. Aβ42, Tau, and p-Tau protein measurements were taken using an enzyme-linked immunosorbent assay (ELISA) (Fujirebio-Europe, Ghent, Belgium) according to the manufacturer’s instructions. The analysis was performed in duplicate and a coefficient of variation (CV) less than 15% was considered as acceptable. In this case, the mean of the two measured values was taken as final result. When the CV was above 15%, samples were reanalyzed in duplicate. For Tau dosage, if the concentration obtained was over 1200 pg/ml, the sample was then diluted if possible and a new dosage was performed according to the industrial guidance for Tau measurement in CSF. The quality of the results was ensured by the use of validated standard operating procedures and internal quality controls (QC) as the three centers were already involved in nationwide studies [28, 29]. The range of the QC coefficient of variation for Aβ42 across the different sets for the three laboratories was 5% –11%. For Tau and p-Tau the range was 8% to 14% and 6% to 14%, respectively. The use of external QC also ensured both the quality of the results and the validity of the intersite data comparison [30]. Finally, all three sites belong to the same national ePLM network, which was created to enhance harmonization of procedures regarding CSF biomarkers in AD.
To determine the optimal cut-off for each biomarker, regarding an AD biochemical signature, the area under the ROC curve (AUC) values were computed in each center using a separate group of AD patients (not involved in this study) and normal controls according to a previously published method [28]. The optimal cut-offs for each biomarker were defined using the highest Youden index, which was selected to maximize sensitivity and specificity on the basis of ROC curve analyses. For Rouen center, the respective cut-offs compatible with AD were Aβ42: <550 pg/ml, Tau: >350 pg/ml, and p-Tau: >60 pg/ml; for Lille center, Aβ42: <700 pg/ml, Tau: >400 pg/ml, and p-Tau: >60 pg/ml; and for Paris Lariboisière, Aβ42: <700 pg/ml, Tau: >400 pg/ml, and p-Tau: >60 pg/ml.
To be included in this study, each patient had to present a CSF Tau level over 1000 pg/mL. Extremely high level of Tau was arbitrarily defined by 2000 pg/ml.
Statistical analyses
CSF biomarker levels were compared between patients, divided in 7 groups: AD, CJD, FTD, LBD, MD, VaD, and other dementia group (OD) with unclassified patients within the previous groups. Patient groups were characterized by mean with standard deviation (SD) (mean ± SD) or by median with minimum and maximum (median [min-max]).
For statistical analysis, NCSS v6.0 was used with a Fisher’s exact test for categorical data, with nonparametric analyses (Mann-Whitney U Test) to compare groups and Spearman correlation coefficient to identify possible correlations between continuous variables. The level of statistical significance was set at 0.05.
RESULTS
Distribution of diagnoses
Our study included 208 patients with an elevated CSF Tau level (>1000 pg/ml). Six of the 208 were excluded because of diagnostic uncertainty or disagreement between neurologists (Fig. 1). Baseline demographic and clinical characteristics and final diagnoses are described in Table 1.
Of the remaining 202 patients, a total of 148 were diagnosed with AD. Distribution of the other forms of dementia is described in Table 1. Regarding the two remaining patients one had alcohol-related dementia and one had corticobasal degeneration.
Clinical data
The median age of the patients was 65 years [39–86], 64.3% were women, and 67 (33.1%) were less than 60 years old. One hundred and thirty-five patients (66.8%) presented predominant memory disorder. Over 31% of the patients had behavioral disorders. Patients were followed for a mean period of 3.6 years (±2.7). Concerning the seven AD patients with extremely high Tau level (i.e., >2000 pg/ml), clinical presentation was mainly amnesic for 5 of 7 patients, with no behavioral disorder and no rapid progression. Mean annual rate of change in MMSE score was 2.38 (±2.1), which was not significantly different from that of the other AD patients.
Concerning the 38 MD patients, they all had leukoaraiosis on MRI, 6 (15.8%) had microbleeds, and 17 (44.7%) showed at least one ischemic sequelae. Ten patients had sudden onset of symptoms. Demographic characteristics (gender, age of onset, delay between first complaint and LP, and annual rate of change in MMSE) and clinical features (rate of behavioral disorder, memory disorder, and aphasia) were not significantly different from the AD group.
Concerning the 4 CJD patients, 2 of 4 patients showed behavioral disorders, myoclonus, and aphasia. One initially had a dysexecutive syndrome. All CJD patients presented with rapidly progressive evolution and 2 of them died within 3 months. No cognitive tests were available due to the severity of the disease when these patients were explored by LP. All MRI were characterized by hyper intense signals on diffusion-weighted sequences and hyper intense signals on FLAIR in basal ganglia and only 1 patient showed EEG with periodic sharp wave complexes. No pathological examination was carried out.
CSF analyses
The median delay for CSF biomarker investigation was 2.9 years [0.2–20] after onset of the first symptoms. There was no correlation between CSF Tau and delay between the first symptoms and LP. Seventy-one CSF Tau dosages were out-of-range concentrations without dilution (>1200 pg/ml) and could not be dosed after dilution. Fifty-four (76%) were from AD patients and 17 (24%) from MD patients. They were excluded from our statistical analyses of CSF, in particular ratio analyses (Fig. 1).
The CSF levels of the three biomarkers for each group are shown in Table 2. Dispersion of CSF Tau levels showed an overlap between AD and other causes of dementia (Fig. 2). However, the median level of Tau was highest in the CJD group (1739.5 pg/ml) and LBD group (1579.0 pg/ml), followed by AD (1164.0 pg/ml) and VaD (1162.0 pg/ml). Seven of 94 AD, 1 of 21 MD, and only 1 of the 4 CJD had extremely high levels of Tau (i.e., >2000 pg/ml). CSF 14-3-3 protein dosage was performed for all CJD patients and 5 other patients (4 AD and 1 OD). Of the 4 CJD patients, only one 14-3-3 was clearly positive, 2 were doubtful and 1 was negative.
P-Tau/Tau ratio
Using p-Tau/Tau ratio and its cut-off suggestive of CJD (<0.075), all of the CJD patients were correctly categorized. Only 2 of 94 (2.1%) AD patients had a ratio <0.075, which might have led to differential diagnosis of CJD (Fig. 3). Furthermore, 6 of 7 patients with an extremely high value of Tau (i.e., >2000 pg/ml) showed a non-CJD ratio, i.e., over 0.075. Finally, a p-Tau/Tau ratio under 0.075 was significantly associated with diagnosis of CJD compared to AD using a Fisher’s exact test (p < 0.0001). No such significant association was found between CJD and the other groups of dementia (MD, VaD, FTD, or LBD).
DISCUSSION
The aim of this retrospective multicentric study was to describe the distribution of differential diagnoses of dementia for 202 patients with a CSF Tau level above the threshold of 1000 pg/ml defined in the literature [15].
First, we showed that AD and MD represented the final diagnosis for more than 92% of all recruited patients. Second, despite the fact that high CSF Tau level was frequently associated with CJD pathology [8], the number of CJD patients amounted to only 1.98% of our total cohort. As our patients were recruited in three memory clinics, our cohort was representative of a wider population and allowed us to address the question of potential differential diagnoses between CJD and other forms of dementia when Tau level is high. Such a situation is common in clinical practice; for example, in this study, elevated Tau represented 12.5% of all CSF biomarker dosages during the same period. Our results support the notion that an elevated level of Tau does not necessarily indicate a diagnosis of CJD. Indeed, other diagnoses were LBD and FTD, to a relatively similar degree. In line with the literature, we showed that CSF Tau does not significantly contribute to the differentiation of AD from non-AD, or CJD. A previous study reported a substantial overlap between AD and non-AD dementias regarding Aβ42 and Tau levels [11]. Another study demonstrated no significant differences in Aβ42 and Tau concentrations between AD and non-AD patients, whereas p-Tau concentrations were significantly higher in AD compared to non-AD patients [31]. The elevated concentration of CSF p-Tau is believed to be closely related to AD pathology and to be relatively specific for AD [12]. However, p-Tau level can also be high in CJD [32] due to massive neuronal death and following Tau elevation. In our study, three of four CJD patients presented an abnormal p-Tau dosage.
Our secondary objective was to see if the p-Tau/Tau ratio might distinguish AD from CJD. For some authors, distinction between CJD and other diagnoses was obtained using the association of Tau >1400 pg/ml and p-Tau/Tau ratio <0.04 [11]. For others [19, 33], a cut-off of p-Tau/Tau ratio at 0.05 had a 99% specificity. Dorey et al. demonstrated in a recent, large study that a p-Tau/Tau ratio value lower than 0.075 achieved 94.2% sensitivity and 98.8% specificity [20]. In our study, we showed that this ratio had correctly categorized 90 of 92 AD patients. In accordance with Blennow et al. [11], we suggest that this ratio is even more relevant in challenging situations of extremely high Tau levels. In France, the CSF 14-3-3-protein dosage is usually performed when a CJD diagnosis is suspected, but this marker may lack some sensitivity [34], as it may initially be negative or doubtable in some CJD cases.
The major limitation of the present study is the small number of non-AD dementia cases, precluding group comparisons. However, the distribution in this study reflects the real use of CSF biomarkers in clinical practice following French Health Authority guidelines. CSF biomarkers are mainly intended to reveal AD pathology and then to support AD diagnosis criteria or to exclude FTLD [22]. The fact that we found only four patients with CJD means that clinicians applied recommendations based on clinical, imaging, as well as CSF and electrophysiological criteria. In our study, three of four CJD patients presented atypical clinical characteristics as aphasia and dysexecutive syndrome explaining why AD diagnosis was also suspected. Another limitation of our study is the lack of neuropathological data to ascertain the final diagnosis. However a homogeneous diagnosis procedure in two steps for all centers and strict respect of validated criteria were applied to restrain this weakness.
Many patients in our study presented FTD-like clinical symptoms. However, only three were finally diagnosed with FTD. This low number is in accordance with the normal or reduced level of Tau in FTD [8, 35]. Our inclusion criteria regarding Tau threshold probably led to exclusion of most FTD patients, as well as LBD patients for whom CSF Tau dosage was mainly normal [36]. The last limitation is the attrition induced by 71 patient dosages which were out of range for the ELISA test and not available for dilution. This point highlights the importance of dilution, following the industrial dosage procedure, permitting precise Tau dosage in order to calculate p-Tau/Tau ratio.
CJD represents a minority of all diagnoses with extremely high CSF Tau levels. Our results support a sequential interpretation algorithm for CSF biomarkers in dementia. High CSF Tau levels should alert clinicians to check the p-Tau/Tau ratio. Under the threshold of 0.075, a diagnosis of CJD is highly probable. Our results pave the way for further, prospective studies, prior to fully implementing CSF P-Tau/Tau ratio as a biomarker for the clinical diagnosis of CJD.
