The Clinical Use of Alzheimer’s Biomarkers in Patients With Mild Cognitive Impairment: A European Alzheimer’s Disease Consortium Survey

Background. This study aims to investigate the clinical use of the main Alzheimer’s disease (AD) biomarkers in patients with mild cognitive impairment (MCI) by examining the beliefs and preferences of clinicians and biomarker experts of the European Alzheimer’s Disease Consortium (EADC). Methods. Out of 306 contacted EADC professionals, 150 (101 clinicians, 43 biomarker experts, and 6 falling into other categories) lled in an online survey from May to September 2020. The investigated biomarkers were: medial temporal lobe atrophy score (MTA) on structural MRI, typical AD (i.e. temporoparietal and posterior cingulate) hypometabolism on FDG-PET, CSF (Aβ 42 , p-tau, t-tau), amyloid-PET and tau-PET. show


Background
The clinical understanding of Alzheimer's disease (AD) and its diagnostic implications have radically changed over time. Currently, AD is de ned as a biological disease characterized by the abnormal accumulation of amyloid plaques and neuro brillary tangles in the brain [1,2], and is clinically considered a continuum ranging from a long preclinical phase (abnormal biomarkers in asymptomatic individuals) to mild cognitive impairment (MCI) and nally dementia. Even if this mainly represents a research framework, it points to the clinical relevance of AD biomarkers, and supports their increasing use in clinical practice. Consistently, several studies demonstrated that the use of biomarkers has different clinical implications [3][4][5][6] with MCI patients being the ones who might bene t the most from biomarker testing [7]. Indeed, an early etiological diagnosis of AD in MCI patients should initiate speci c counseling [8] and allow interventions to potentially delay the progression to dementia [9]. For an etiological diagnosis of early symptomatic patients, this concept has already been picked up by most academic memory clinics and is implemented in clinical practice. However, to date, no clinical guidelines recommend a "biomarker-based diagnosis of AD" as the state-of-the-art for the etiological diagnosis of MCI (or even dementia) patients.
The main biomarkers able to capture AD pathophysiological changes can be classi ed into two categories: (i) diagnostic / pathophysiological biomarkers are measures of amyloid (i.e. positive amyloid-PET, low CSF Aβ 42 ) and tau (i.e. positive tau-PET, high CSF p-tau) deposition, and have the necessary speci city for a diagnosis of AD [10]; (ii) progression / topographical biomarkers include measures of neurodegeneration (e.g. medio-temporal lobe atrophy (MTA) on structural MRI, hypometabolism on FDG-PET, high CSF t-tau), are not speci c to AD but can be used to track the disease progression [10]. Among all these biomarkers, a major progress in the last years is the advent of amyloid-PET and tau-PET: this novel molecular imaging technique makes possible to visualize the topography of amyloid and tau deposits in the brain, previously measurable only indirectly via CSF. Speci cally, given the strong association between tau pathology and symptoms, already known from neuropathological studies, and given the relative speci city of tau-PET tracers for tau deposits in AD, tau-PET could represent the most e cient imaging modality in AD: a positive tau-PET might be an indicator of both amyloid and tau pathology and thus provide a de nite diagnostic answer with only one test [11,12]. However, this interpretation might depend on the relative role attributed by the interpreting physician and clinician to amyloid and tau pathology in the physiopathology of AD [13].
A previous European survey conducted in 2012 investigated the use of biomarkers in MCI patients, and showed that physicians felt comfortable delivering a diagnosis of MCI due to AD when both amyloid and neurodegeneration biomarkers were abnormal [14]. The most frequently used biomarkers were found to be MTA (75% of responders reported to use it at least frequently), Aβ 42 , p-tau, t-tau levels in CSF (22%), FDG-PET (16%), and amyloid-PET (3%). In light of the notable advances recently occurred in the eld of AD biomarkers, namely clinical studies on real-word patients and introduction of newly developed techniques, an up-to-date depiction of their clinical use is needed. Thus, the aim of this study is to investigate the clinical use of the main AD biomarkers in MCI patients by examining the beliefs and preferences of clinicians (i.e. neurologists, geriatricians, psychiatrists) and biomarker experts (i.e. nuclear medicine, radiology, and laboratory physicians) within the framework of the European Alzheimer's Disease Consortium (EADC, http://www.eadc.info), a functional network comprising 70 European memory clinics of excellence working in the eld of AD (from 20 countries), providing a setting in which to increase the scienti c understanding of AD.

The survey
A survey consisting of an online questionnaire assessing the clinical use of biomarkers in MCI patients (see the Supplementary Material for the full questionnaire) was made accessible from May 14th to September 30th, 2020. The number of items/questions depended on the respondent specialty and answers: 26 for clinicians and 14 for biomarker experts, i.e. nuclear medicine, radiology, and laboratory physicians; the difference being due to 12 questions speci c to the treating clinicians.
The questionnaire was subdivided in ten sections assessing respectively (i) respondent's details, i.e. specialty, center, city, country, role in memory clinic, years of experience in the eld of neurodegenerative disorders, competence to answer to questions regarding neurodegenerative disorders; (ii) for clinicians only, the request to provide names and contacts of nuclear medicine, radiology, and laboratory physicians associated with the respondent's memory clinic; (iii) beliefs about the pathogenic role of amyloid and tau in Alzheimer's disease; (iv) respondent's clinical work, i.e. consultation with MCI, number of new MCI patients per month; (v) frequency of use of biomarkers (MRI, FDG-PET, CSF, amyloid-PET, tau-PET); (vi) use of imaging biomarkers in clinical reports (vii); use of biochemical biomarkers in clinical reports; (viii) diagnostic additional value of biomarkers; (ix) clinical vignette where respondents were asked to rate the diagnostic con dence on the basis of abnormality of the biomarkers (alone or combined); and (x) perceived clinical utility of amyloid-PET and tau-PET.
Concerning tau-PET, the following disclaimer was added to the survey: "Please note that some of the following questions concern tau-PET. We are aware that while amyloid-PET tracers are well established and show similar performances, tau-PET tracers have been more recently developed and are less established. e.g., Flortaucipir is the most used tracer and has been validated against neuropathology and second-generation tracers are promising for an increased diagnostic accuracy. For this reason, we chose not to specify further which tau-PET tracer, and we would ask you to consider in answering a "theoretical" tau-PET tracer with a diagnostic accuracy deemed adequate for clinical use in AD." The survey took place in two steps: rst, we contacted the clinicians working in the EADC memory clinics asking them to ll in the online questionnaire and to provide names and contacts of nuclear medicine, radiology, and laboratory physicians associated with their memory clinics; second, we contacted the nuclear medicine, radiology, and laboratory physicians recommended by the clinicians during the rst step.

Outcomes of interest
In the present study we assessed the i) responders' belief about the pathogenic role of amyloid and tau in Alzheimer's disease pathology and symptoms (Sect. 2.2.1), ii) frequency of use of AD biomarkers in MCI patients (Sect. 2.2.2), iii) use of AD biomarkers to support etiological diagnosis in MCI (Sect. 2.2.3), iv) additional value of FDG-PET, CSF, amyloid-PET and tau-PET over neuropsychological testing and structural MRI in an MCI patient (Sect. 2.2.4), v) responders' con dence in making a diagnosis of MCI due to AD in a typical MCI patient based on different AD biomarkers (Sect. 2.2.5), vi) responders' opinion on the perceived clinical utility of amyloid-PET vs tau-PET in MCI and mild dementia patients (Sect. 2.2.6).

Belief on the pathogenic role of amyloid and tau in AD
The following question was posed to all responders: "What is your belief/opinion about the pathogenic role of amyloid and tau in Alzheimer's disease pathology and symptoms?". Possible answers went from 0 to 10 (11-point Likert scale): 0, the abnormal accumulation of amyloid is the initial cause of AD; 5, amyloid and tau have the same relevance in causing AD, or neither amyloid nor tau are the initial cause of AD; 10, the abnormal accumulation of tau is the initial cause of AD. Answers were grouped into three categories: 0-4, favoring amyloid; 5, neutral; 6-10, favoring tau.

Frequency of use of AD biomarkers in MCI patients
The following question was posed: "In MCI, in your clinical practice, please state frequency of use for medial temporal lobe atrophy (MRI), FDG-PET, CSF (e.g. Aβ 42 , p-tau, t-tau), amyloid-PET, tau-PET".
Possible answers were: not used (0%), rarely (< 10%), regularly (20-60%), frequently (60-80%), always (> 80%). Answers were grouped into three categories: no use, rare-to-regular use (rarely or regularly), frequent-to-constant use (frequently or always). This item was applicable only to clinicians. Since the frequency of use of biomarkers is strongly dependent on the center expertise, policies, guidelines and facilities, we included only the answers of the head of the memory clinic or, if not possible, the answers of the most experienced (years of expertise in the eld of neurodegenerative disorders) clinician of that memory clinic.

Use of AD biomarkers to support etiological diagnosis in MCI
The following question was posed: "Do you use imaging biomarkers / CSF collection (e.g. Aβ 42 , p-tau, ttau) / APOE genotyping to support your etiological diagnosis in MCI?". Possible answers were yes or no. Concerning imaging biomarkers, an additional question was posed: "Do you use any quantitative reading tool (e.g. SPM) or scale (e.g. MTA scale; Scheltens et al., 1992) for your clinical reports?". Possible answers were yes or no. In case of a rmative answer, a further question was posed: "In MCI please state what kind of quantitative reading tool (e.g. SPM) or scale (e.g. MTA scale; Scheltens et al., 1992) you use for your clinical reports for the answers indicated in previous question". These items were applicable only to clinicians.

Additional value over neuropsychological testing and structural MRI in MCI
The following question was posed to all responders: "Assuming that clinical examination with neuropsychological testing and brain structural MRI are the most feasible procedures in most memory clinics, please rate the additional diagnostic value (i.e. the ability to provide diagnostic information in excess of that already provided by neuropsychological testing and brain structural MRI) in an MCI patient of FDG-PET, CSF markers (e.g. Aβ 42 , p-tau, t-tau), amyloid-PET, tau-PET". Possible answers were: none, little, moderately signi cant, greatly signi cant, decisive. Answers were grouped into three categories: none-to-little (none or little), moderate (moderately signi cant), great-to-decisive (greatly signi cant or decisive).

Con dence in an etiological diagnosis of AD in MCI
The following case vignette was proposed to all responders: "A 75 years old person comes into your o ce complaining of memory deterioration in the past 6-12 months, he/she is in good physical health, has no problems in his/her daily chores, but is clearly worried. Routine labs are normal, but he/she performs 1.5 SD below the age-and education-adjusted mean on a test of verbal or non-verbal recall. How con dent would you be with a diagnosis of MCI due to AD (or prodromal AD) on the basis of i) evidence of clear-cut medial temporal lobe atrophy alone, ii) clear-cut temporoparietal and posterior cingulate hypometabolism on FDG-PET alone, iii) clearly abnormal CSF levels of Aβ and tau alone, iv) clearly positive amyloid-PET, v) clearly positive tau-PET, vi) at least one clearly positive amyloid marker and at least one clearly positive neuronal injury marker.". Possible answers were: not at all comfortable, moderately comfortable, comfortable, very comfortable, extremely comfortable. Answers were grouped into three categories: not comfortable, su ciently comfortable (moderately comfortable or comfortable), very-to-extremely comfortable (very comfortable or extremely comfortable).

Perceived clinical utility of amyloid-PET vs tau-PET in MCI and mild dementia
The following question was posed to all responders: "Independent of any speci c patient's feature and based on your clinical experience with patients usually seen in your memory clinic, what is, in your opinion, the most clinically useful exam for etiological diagnosis of MCI and mild dementia?". Possible answers went from 0 to 10 (11-point Likert scale): 0, amyloid-PET is in general the most useful exam; 5, amyloid-PET and tau-PET are equally useful, or neither amyloid-PET nor tau-PET are the most useful exam; 10, tau-PET is in general the most useful exam. Answers were grouped into three categories: 0-4, favoring amyloid-PET; 5, neutral; 6-10, favoring tau-PET.

Statistical methods
The outcomes of interest were assessed using proportion test (χ 2 ). Signi cance was set at p < 0.05 and post-hoc pairwise comparisons were adjusted using Bonferroni correction.
All statistical analyses were performed with R, version 3.4.2 (R Foundation for statistical computing, https://www.r-project.org/).

Responders' features
First, 168 clinicians working in the EADC memory clinics were contacted, and 51% of them (86/168) lled in the survey questionnaire recommending 138 additional professionals (20 clinicians, 42 biomarker experts, 1 neuroscientist and 1 psychologist) related to their memory clinics, 46% (64/138) of whom lled in the survey. In total, 150 professionals lled in the survey questionnaire, with an overall answer rate of 49% (150/306). Among them, 6 did not fall into the pre-de ned categories of interest (i.e. physicians and biomarker experts), being research staff (n = 4) or general practitioner (n = 2), and were therefore excluded from the analyses. Among the remaining responders, 5% (7/144) declared that they were not competent enough in the eld of neurodegenerative disorders to ful ll the questionnaires, and therefore did not proceed with the questionnaire. The nal sample consisted of 100 clinicians and 37 biomarker experts (9 radiologists, 18 nuclear medicine physicians, and 10 laboratory physicians). Among clinicians, 45% (45/100) were head of their memory clinics, while the remaining ones were memory clinic staff or collaborators. Clinicians had more years of experience in the eld of neurodegenerative disorder than biomarker experts (clinicians: 19 ± 9, biomarker experts: 13 ± 8; p < 0.001). On average, clinicians consulted 17 ± 13 new patients with MCI in a typical month. Detailed respondent's characteristics are reported in Table 1, while Fig. 1 shows the geographic distribution of responders.  Figure 2A shows the beliefs on the pathogenic role of amyloid and tau in AD reported by clinicians and biomarker experts, while gure S1A shows the beliefs on the pathogenic role of amyloid and tau in AD reported by different biomarker experts groups, i.e. radiologists, nuclear medicine physicians and laboratory physicians. Clinicians deemed the abnormal accumulation of amyloid (answers from 0 to 4, 42% of cases) rather than tau (answers from 6 to 10, 20% of cases, p < 0.005; Fig. 2A) as the initial cause of AD. A similar, although not signi cant possibly due to the smaller sample size, pattern was observed among biomarker experts (amyloid: 43%, tau: 27%, p = 0.223; Fig. 2A). When comparing the proportion of clinicians and biomarker experts favoring amyloid, no signi cant difference emerges. The same result was observed for tau. Figure 3 shows the detailed frequency of use of each biomarker. Among clinicians, 8% (8/100) did not provide clinical consultation for patients with MCI, and were therefore excluded from the analyses on the frequency of use of AD biomarkers in MCI. Moreover, due to the above-explained reasons, we included only the answers of the head of the memory clinic or, if not possible, the answers of the most experienced (as measured by years of expertise in the eld of neurodegenerative disorders) clinician of that memory clinic.

Frequency of use of AD biomarkers in MCI patients
The frequency of clinicians reporting a frequent-to-constant use of MTA (77%, 41/53) is higher than that of those reporting a frequent-to-constant use of CSF (45%, 24/53; p = 0.014), FDG-PET (32%, 17/53, p < 0.001), amyloid-PET (8%, 4/53; p < 0.001) and tau-PET (2%, 1/53; p < 0.001). Moreover, the frequency of clinicians reporting a frequent-to-constant use of CSF and FDG-PET is higher than that of those reporting a frequent-to-constant use of amyloid-PET (p < 0.001 and p = 0.035) and tau-PET (p < 0.001 and p = 0.001). Figure 4 shows the use of AD biomarkers to support etiological diagnosis in MCI. Among clinicians, 8% (8/100) did not provide clinical consultation for patients with MCI, and were therefore excluded from the analyses on use of AD biomarkers to support etiological diagnosis in MCI. Clinicians providing clinical consultation for patients with MCI used imaging (90%, 83/92) and CSF (87%, 80/92) biomarkers to support etiological diagnosis in MCI more frequently than APOE genotyping (27%, 25/92; p < 0.001).

Use of AD biomarkers to support etiological diagnosis in MCI
As for imaging biomarker, 67% of clinicians (62/92) used quantitative reading tools and scales for reporting imaging biomarkers data in clinical reports. Speci c results on quantitative tools or scales are presented in Table 2. For each imaging biomarker, the number of clinicians using any quantitative reading tool or scale for that biomarker in clinical reports is reported in the rst row. Number of clinicians using each speci c quantitative tool are reported in subsequent rows.
3.5. Additional value over neuropsychological testing and structural MRI in MCI Figure 5 shows the detailed reported additional value over neuropsychological testing and structural MRI in MCI, while gure S2 shows the detailed reported additional value over neuropsychological testing and structural MRI in MCI by different biomarker experts groups, i.e. radiologists, nuclear medicine physicians and laboratory physicians. Clinicians report more frequently a great-to-decisive additional value of Aβ 42 , p-tau, t-tau levels in CSF (85%, 85/100) with respect to amyloid-PET (72%, 72/100; p < 0.05), tau-PET (54%, 54/100; p < 0.001) and FDG-PET (35%, 35/100; p < 0.001). Amyloid-PET, instead, has been reported as having a great-to-decisive additional value more frequently than FDG-PET only (p < 0.001). No signi cant differences emerged between amyloid-PET and tau-PET (p = 0.077), and between FDG-PET and tau-PET (p = 0.063).
No signi cant differences were observed when comparing the frequency of clinicians and biomarker experts reporting a great-to-decisive additional value for each technique.
3.6. Con dence in an etiological diagnosis of AD in MCI Figure 6 shows the detailed level of con dence in an etiological diagnosis of AD in MCI, while gure S3 shows the detailed level of con dence in an etiological diagnosis of AD in MCI by different biomarker experts groups, i.e. radiologists, nuclear medicine physicians and laboratory physicians. Clinicians reported to be more frequently very-to-extremely comfortable with an etiological diagnosis of AD on the basis of at least one clearly positive amyloid marker and at least one clearly positive neuronal injury marker (86%, 86/100) with respect to abnormal CSF levels of Aβ 42 , p-tau and t-tau (64%, 64/100; p = 0.009), positive amyloid-PET (52%, 52/100; p < 0.001), positive tau-PET (41%, 41/100; p < 0.001), typical AD (i.e. temporoparietal and posterior cingulate) hypometabolism on FDG-PET (22% 22/100; p < 0.001), and evidence of MTA on structural MRI (14%, 14/100; p < 0.001). Moreover, clinicians reported to be more frequently very-to-extremely comfortable with an etiological diagnosis of AD on the basis of abnormal CSF levels of Aβ and tau or of a positive amyloid-PET than typical AD hypometabolism on FDG-PET (p < 0.001 and p < 0.001 respectively) and evidence of MTA (p < 0.001 and p < 0.001 respectively). Additionally, they reported to be more frequently very-to-extremely comfortable with an etiological diagnosis of AD on the basis of abnormal CSF levels of Aβ and tau rather than of positive tau-PET (p = 0.028). Finally, clinicians reported to be more frequently very-to-extremely comfortable with an etiological diagnosis of AD on the basis of positive tau-PET than evidence of MTA (p < 0.001).
Biomarker experts stated to be more frequently very-to-extremely comfortable with an etiological diagnosis of AD on the basis of at least one clearly positive amyloid marker and at least one clearly positive neuronal injury marker (68%, 25/37) rather than on the evidence of MTA (14%, 5/37; p < 0.001). Among biomarker experts we observed no other signi cant differences, possibly due to the smaller sample size, with the frequency of those stating to be very-to-extremely comfortable with an etiological diagnosis of AD being 46% (17/37) for positive amyloid-PET, 43% (16/37) for typical AD hypometabolism on FDG-PET, 41% (15/37) for positive tau-PET and 38% (14/37) for abnormal CSF levels of Aβ and tau.
No signi cant differences were observed when comparing the frequency of clinicians and biomarker experts reporting to be very-to-extremely comfortable with an etiological diagnosis of AD on the basis of MTA, amyloid-PET, and tau-PET. Biomarker experts reported to be very-to-extremely comfortable with an etiological diagnosis of AD on the basis of typical AD hypometabolism on FDG-PET more frequently than clinicians (p = 0.024). Conversely, clinicians reported to be very-to-extremely comfortable with an etiological diagnosis of AD on the basis abnormal CSF levels (p = 0.010), or of at least one clearly positive amyloid marker and at least one clearly positive neuronal injury marker (p = 0.028) more frequently than biomarker experts. Figure 2B shows the perceived clinical utility of amyloid-PET vs tau-PET in MCI and mild dementia reported by clinicians and biomarker experts, while gure S1B shows the perceived clinical utility of amyloid-PET vs tau-PET in MCI and mild dementia reported by different biomarker experts groups, i.e.

Perceived clinical utility of amyloid-PET vs tau-PET in MCI and mild dementia
radiologists, nuclear medicine physicians and laboratory physicians. Clinicians considered amyloid-PET and tau-PET equally useful to support an etiological diagnosis in MCI and mild dementia patients (amyloid-PET: 35% vs tau-PET: 24%, p = 0.121; Fig. 2B). A non-signi cant trend towards an amyloid-PET preference was observed among biomarker experts (amyloid-PET: 46% vs tau-PET: 24%, p = 0.088). When comparing the proportion of clinicians and biomarker experts favoring amyloid-PET or tau-PET, no signi cant difference emerged.

Discussion
This study reports the results of the largest multidisciplinary survey on the use and the perceived utility of Alzheimer's biomarkers in clinical practice in MCI patients.
The investigation on the professionals' beliefs on the initial cause of AD revealed that a prevalent pathogenic role was attributed to amyloid (42% clinicians, 43% biomarker experts) rather than tau (20% clinicians, 27% biomarker experts). However, when the focus switched from the theoretical belief to the perceived clinical utility of PET scans assessing amyloid and tau, responders did not show a clear preference for amyloid-PET. Consistently, a recent work assessing the diagnostic value of amyloid-PET and tau-PET in a memory clinic population, showed that the two exams signi cantly impacted diagnosis and diagnostic con dence in a similar way [4]. Notably, clinicians and biomarker experts proved to be perfectly aligned both with regard to beliefs on the pathogenic role of amyloid and tau and to the perceived clinical utility of amyloid-PET and tau-PET.
In the clinical assessment of MCI patients, the most widely used biomarker is MTA on structural MRI (78% frequent-to-constant use), followed by Aβ 42 , p-tau, t-tau levels in CSF (45%), typical AD hypometabolism on FDG-PET (32%), amyloid-PET (8%) and tau-PET (2%). These results highlight an overall increase in the use of all biomarkers, as compared with a previous survey launched in 2012 [14], and with another one launched in 2014 among members of European Academy of Neurology (EAN) and EADC [15]. In a period of less than 10 years, amyloid-PET went from being frequently or constantly used in 3% of cases to 8% (and from being used at least regularly in 16% of cases to 31%), thus highlighting a remarkable spread of this biomarker in the clinical practice. Moreover, tau-PET, which has been recently introduced [16] and was not even included in the previous survey, now appears to be used frequently or constantly used in 2% of cases and at least regularly in 6%. The use of these advanced techniques might be limited by the absence of disease modifying therapies. Their use will probably increase as new effective therapies will become available. It must be remarked that the frequency of use of each biomarker is clearly in uenced by its local availability, which varies signi cantly from center to center with MRI being already widely available and tau-PET being available only in research contexts, despite the promising preliminary results exhibited since its recent introduction [4,17]. Of note, the limited frequency of use detected for amyloid-PET and tau-PET might be in uenced by the lack of reimbursement for this examination. As for CSF markers, their frequency of use might be further in uenced not only by their availability but also by the attitude toward lumbar puncture among clinicians and patients. Performing lumbar puncture, indeed, requires speci c training and evaluation of potential contraindications; however, the overall risk of complications is relatively low [18].
Imaging and CSF biomarkers were found to be widely used to support the etiological diagnostic process in MCI (i.e. ≃90%), while APOE genotyping was used only in 27% of cases, denoting a slight decrease from previous survey (38% in 2012). This result seems at odds with non-invasivity and affordability of APOE genotyping and with the crucial information that this exam might reveal. Indeed, even though the estimated risk conferred by an ε4 allele varies between studies, odd ratios ranging between 1.8 and 9.9 have been reported [19]. Moreover, it has been recently shown that APOE ε4 carriers have an increased rate of progression from preclinical AD to MCI, and from MCI to mild AD dementia [20]. APOE ε4 allele, however, is neither necessary nor su cient to cause AD. Thus, APOE genotyping should not be used alone for a diagnosis of MCI due to AD in a single case [21]. Indeed, current guidelines do not do not recommend APOE genotyping in the workup of dementia and MCI either in isolation or as part of the diagnostic process [22,23].
Clinicians deemed Aβ 42 , p-tau, t-tau levels in CSF as the most valuable biomarker in terms of additional diagnostic value over neuropsychological testing and structural MRI (85% great-to-decisive additional value), followed by amyloid-PET (72%), tau-PET (54%) and lastly typical AD hypometabolism on FDG-PET (35%). Notably, clinicians perceived the diagnostic value of CSF and amyloid-PET dramatically increased from previous survey (60% and 43% in 2012), whilst that of FDG-PET (46% in 2012) decreased. Similarly, biomarker experts attributed more value to CSF (76%), amyloid-PET (70%), tau-PET (63%) and lastly FDG-PET (49%). The remarkable level of perceived utility attributed to Aβ 42 , p-tau, t-tau levels in CSF can be easily explained in light of the different pathophysiological process which is able to capture (i.e. amyloid load, brain tauopathy, neurodegeneration). However, Ramusino and colleagues recently compared the relative incremental diagnostic value of amyloid-PET and CSF (Aβ42, p-tau, t-tau), showing that amyloid-PET induces greater changes in the diagnosis of AD patients as compared to CSF [3]. A recent survey further con rmed the value of amyloid pathology markers, highlighting that these biomarkers are perceived by clinicians as the most valuable to predict progression and rate of progression in MCI patient [8].
As for tau-PET, both clinicians and biomarker experts appear to already rely on this exam in clinical practice. This might be due to its ability to differentiate between amyloid-positive and negative neurodegenerative diseases with high accuracy [11] and to the evidence that the accumulation of pathologic tau is closely related to functional and structural deterioration in the AD spectrum [24]. The relatively low additional value attributed to FDG-PET might be due its nature of nonspeci c measure of neurodegeneration, detecting damage that may derive not only from AD but from a variety of etiologies, for example cerebrovascular injury [2].
Concerning diagnostic con dence in an etiological diagnosis of AD in MCI, both clinician and biomarker experts quite predictably deemed the combination of amyloidosis and neuronal injury biomarkers as the most convincing in vivo signature of AD, while MTA alone was perceived as the less reliable biomarker. As for other biomarkers, different considerations emerged between clinicians and biomarker experts. Typical AD hypometabolism on FDG-PET, indeed, was perceived as of great-to-decisive additional value by 44% of biomarker expert but only 22% of clinicians. Conversely, CSF collection was perceived as more useful by clinicians (64%) than biomarker experts (38%). A more balanced pattern emerged for amyloid-PET (52% vs 46%) and tau-PET (41% vs 40%). Notably, among experts in the eld of neurodegenerative disorders, newly developed biomarkers mainly used in research and still not clinically validated, i.e. amyloid-PET and tau-PET, have already reached (and in some cases exceeded) the level of reliability of well-established and validated biomarkers, in agreement with their level of clinical validity, namely for Flortaucipir as tau-PET tracer, as recently assessed [25,26].

Limitations
The here-reported results allow an up-to-date depiction of the clinical use of Alzheimer's biomarkers in patients with MCI in European countries and of their perceived utility among a multidisciplinary group of experts in the eld of neurodegenerative disorders. Nevertheless, this study has some limitations. Firstly, the participation of EADC centers was only partial (51% response rate). Secondly, the nature of the responding centers, i.e. memory clinics with a clinical research background, might prevent to generalize our results to ordinary memory clinics. Thirdly, since tau-PET tracers have been developed only recently and are less established, we had to ask responders to consider a "theoretical" tau-PET tracer with a diagnostic accuracy deemed adequate for clinical use in AD, maybe affecting their answers concerning this technique. Fourthly, plasma biomarkers were not included in the present survey. However, since they are emerging as potentially scalable and valuable biomarkers [27], their inclusion in future survey is needed. Lastly, the results of the survey might be in uenced by the local indications for prescription, reimbursement policies, and costs of biomarkers, which are not always consistent across Europe [7].
Indeed, the same biomarker might be indicated without restrictions and reimbursed in some countries (e.g. CSF in France), and indicated only in speci c cases (e.g. CSF in UK) or not reimbursed (e.g. CSF in Spain) in others [7]. As a consequence, in some cases, wealthy patients might pay out of pocket to have access to advanced diagnostic exams that would otherwise be inaccessible. In other words, the European heterogeneity in indications for prescription and reimbursement policies and other nancial aspects might have in uenced the results of the survey.

Conclusions
Altogether, the results of this ample and multidisciplinary survey suggest that both traditional and newlydeveloped (e.g. amyloid-PET and tau-PET) biomarkers are widely adopted across European memory clinics in patients with MCI, not only for research purposes but also in clinical practice. Overall, we observed that CSF is currently considered as the most useful biomarker by clinicians and biomarker experts, followed by amyloid-PET. Moreover, the use of molecular imaging in the diagnostic work-up of MCI patients is increasing over time.

List Of Abbreviations
Alzheimer's disease (AD) Mild cognitive impairment (  The question posed to responders was: "Independent of any speci c patient's feature and based on your clinical experience with patients usually seen in your memory clinic, what is, in your opinion, the most clinically useful exam for etiological diagnosis of MCI and mild dementia?". Answers were grouped into three categories: 0-4, favoring amyloid-PET; 5, neutral; 6-10, favoring tau-PET.

Figure 4
Use of AD biomarkers to support etiological diagnosis in MCI. The question asked to clinicians was: "Do you use imaging biomarkers / CSF collection (e.g. Aβ42, p-tau, t-tau) / APOE genotyping to support your etiological diagnosis in MCI?". Possible answers were yes or no. Clinicians reported to use imaging to support their etiological diagnosis in MCI in 90% of cases, CSF in 87% of cases and APOE in 27% of cases.

Figure 5
Additional value over neuropsychological testing and structural MRI in MCI. The question asked to responders was: "Assuming that clinical examination with neuropsychological testing and brain structural MRI are the most feasible procedures in most memory clinics, please rate the additional diagnostic value (i.e. the ability to provide diagnostic information in excess of that already provided by neuropsychological testing and brain structural MRI) in an MCI patient of FDG-PET, CSF markers (e.g. Aβ42, p-tau, t-tau), amyloid-PET, tau-PET". Possible answers were: none, little, moderately signi cant, greatly signi cant, decisive. Answers were grouped into three categories: none-to-little (none or little), moderate (moderately signi cant), great-to-decisive (greatly signi cant or decisive).

Figure 6
Con dence in an etiological diagnosis of AD in MCI. The following case vignette was proposed to responders: "A 75 years old person comes into your o ce complaining of memory deterioration in the past 6-12 months, he/she is in good physical health, has no problems in his/her daily chores, but is clearly worried. Routine labs are normal, but he/she performs 1.5 SD below the age-and educationadjusted mean on a test of verbal or non-verbal recall. How con dent would you be with a diagnosis of MCI due to AD (or prodromal AD) on the basis of i) evidence of clear-cut medial temporal lobe atrophy alone, ii) clear-cut temporoparietal and posterior cingulate hypometabolism on FDG-PET alone, iii) clearly abnormal CSF levels of Aβ and tau alone, iv) clearly positive amyloid-PET, v) clearly positive tau-PET, vi) at least one clearly positive amyloid marker and at least one clearly positive neuronal injury marker.".
Possible answers were: not at all comfortable, moderately comfortable, comfortable, very comfortable, extremely comfortable. Answers were grouped into three categories: not comfortable, su ciently comfortable (moderately comfortable or comfortable), very-to-extremely comfortable (very comfortable or extremely comfortable).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. ARTUseBiomarkersCaprioglio07SupplementaryMaterial.docx