Dopamine transporter imaging for the diagnosis of dementia with Lewy bodies

Target condition being diagnosed

Dementia with Lewy bodies (DLB) is a common cause of neurodegenerative dementia of old age, with an estimated prevalence of 15% to 25% of cases based on autopsy studies (Perry 1990; Heidebrink 2002). Community prevalence estimates vary more widely, with rates reported of 0% to 5% in the elderly population and 0% to 30.5% among elderly people with dementia (Zaccai 2005).

Diagnosis during life rests on a set of consensus clinical criteria that were adopted in 1996 and revised in 2005 (McKeith 1996; McKeith 2005). We show both the 1996 and 2005 versions in Appendix 1. Probable or possible DLB diagnoses may be made based on the number of core features or suggestive features, or both. The core features of the illness are recurrent visual hallucinations, fluctuating cognition, and Parkinsonism. The consensus criteria state that these features should occur early in the course of the illness.

As for other forms of dementia, the diagnostic gold standard is the neuropathological findings at postmortem. The neuropathological criteria for DLB have also evolved over time and remain a subject of debate. The standards currently considered the most robust combine assessments of Lewy body (LB) pathology and Alzheimer's disease (AD) pathology in order to estimate the probability of an association between the LB pathology and the dementia syndrome. These are either the standards established by McKeith 2005 or a combination of LB pathology assessed according to the method described by Braak 2003 with an accepted AD pathological standard (including as a minimum Braak neurofibrillary tangle staging). Comparisons of pathological and clinical findings suggest that the typical DLB clinical syndrome is seen most clearly in people with 'pure' DLB pathology. In contrast, people who also have significant AD pathology are found to have an attenuated clinical syndrome that is harder to recognise in life.

Changing neuropathological standards for the diagnosis of DLB clearly impact on assessment of the accuracy of the consensus clinical criteria. Several groups have reported sensitivity and specificity using the earlier clinical and neuropathological criteria (McKeith 1996). Overall, a clinical diagnosis of probable DLB in these studies had high specificity but often poor sensitivity (McKeith 2004), although it is likely that overdiagnosis of DLB using older neuropathological standards contributed to poor sensitivity estimates. There are few published data on the accuracy of a diagnosis of possible DLB. Although the possible DLB diagnosis appears to be more sensitive than the probable DLB diagnosis, this is at the cost of a loss of specificity, with the only reported autopsy series finding a specificity of just 28% (Verghese 1999). Nelson, et al. have reported on a very large clinicopathological data set from the National Alzheimer's Disease Coordinating Center (NACC) from 2000 onwards that used current neuropathological methods (Nelson 2010). In line with earlier reports, they found a high specificity of an antemortem diagnosis of DLB (95%), but again low sensitivity (32.1% for pure DLB and 12.1% for DLB plus AD). These authors found that the accuracy of DLB diagnoses in the specialist dementia centres contributing to the NACC database is actually declining with time.

An important but poorly understood issue is how the accuracy of the clinical criteria changes with the stage of the illness. Core features, such as visual hallucinations and Parkinsonian motor symptoms, occur more commonly in other forms of dementia by the moderate to severe stages. Thus, the ability of the criteria to predict the presence of Lewy body pathology declines as dementia progresses (Nelson 2010).

Index test(s)

Brains of people with DLB show a variety of reductions in indices of dopaminergic function both in vitro and in vivo that distinguish them from healthy elderly subjects and from people with AD (Tatsch 2008). Among the distinguishing features is a reduction of pre-synaptic dopamine transporter (DAT) in the corpus striatum, reflecting degeneration of dopaminergic nigrostriatal neurons. It is possible to image striatal dopamine transporter sites in vivo with single-photon emission computed tomography (SPECT) or positron emission tomography (PET) using a variety of radioligands. PET offers better image resolution, but SPECT has the advantages of being cheaper and more widely available. Most of the radioligands have been used only for research purposes, but the SPECT ligand [123I]-N-(3-fluoropropyl)-2β-carbomethoxy-3β-(4-iodophenyl) nortropane (123I-FP-CIT) is commercially available (under the brand name DaTscan^TM). We show approved indications for DaTscan^TM in Europe and the United States in Appendix 2.

In healthy subjects, ¹²³I-FP-CIT SPECT images show the corpus striatum (caudate nucleus and putamen) lit up as a comma-shaped structure bilaterally, reflecting specific uptake of the DAT ligand. Most experience of abnormal DAT images derives from people with Parkinson's disease. The typical pattern in Parkinson's disease is for the 'comma' to disappear, starting with the tail (the putamen) unilaterally. A similar pattern is seen in the majority of cases of DLB, although the change is usually more symmetrical. In a minority of DLB cases there may be more global loss of specific uptake throughout the corpus striatum from the outset. Images can be rated for the degree of abnormality either visually or using semiquantitative or quantitative methods. Visual ratings may be reported either in a binary fashion as normal or abnormal, or degrees of abnormality (typically two or three) may be reported. Excellent agreement can be reached between experienced visual raters (McKeith 2007). Semiquantitative or quantitative methods use a region of interest technique and derive ratios of ligand uptake in the caudate and in the putamen to nonspecific background uptake, usually in the occipital cortex.

A number of technical and clinical factors are important in image interpretation. The method of image processing can influence the results significantly, so uniformity of method within a study is essential. Particularly in multicentre studies, there is potential for systematic variation in results between centres due to use of different equipment and procedures. A pan-European initiative recently proposed a framework to address this potential variation (Dickson 2012). This framework has been used in the production of a normative database for ¹²³I-FP-CIT SPECT images, addressing the effects of age, gender, and image-processing method (Varrone 2013), that may be used as a reference resource in future studies. Vascular lesions of the corpus striatum or nigrostriatal pathway can lead to abnormal DAT images, so in order to reduce the risk of false positive results, structural brain images may be necessary to aid interpretation. A small number of prescribed medications or drugs of abuse may affect DAT images. In any large series, a small proportion of images will be hard to classify and should be rated as 'indeterminate'. Guidelines for brain SPECT using ¹²³I-labelled DAT ligands, which cover patient procedures, image acquisition and processing, image interpretation and reporting, were published in 2010 by the European Association of Nuclear Medicine and in 2012 by the Society of Nuclear Medicine (Darcourt 2010; Djang 2012).

Several studies have investigated the relationship between a clinical diagnosis of DLB and the result of FP-CIT SPECT imaging. Cross-sectional case-control studies have consistently found striatal binding of FP-CIT to be lower in groups of people with a diagnosis of DLB than in groups of healthy controls or people with a diagnosis of AD (Walker 2002; Ceravolo 2003; O'Brien 2004). The largest study was a multicentre European study of 326 people with dementia, including 151 with probable (two or more core clinical features) or possible DLB (McKeith 2007). An abnormal FP-CIT SPECT scan had a sensitivity of 77.7% (95% confidence interval (CI) 64.1% to 88.3%) and a specificity of 90.4% (95% CI 82.1% to 95.5%) for probable DLB diagnosed by a consensus panel, indicating a high degree of concordance between clinical and imaging data. Considering the performance of the clinical criteria against neuropathological findings, this result still allows for a wide range of possible performances of the imaging test against neuropathology. For example, there is no way of knowing if the nearly one-quarter of clinically probable DLB cases whose FP-CIT SPECT scan was rated as normal were all clinical false positives (although specificity of the clinical criteria in expert hands is generally high) or if SPECT scanning was missing even more true positives than the clinical criteria. Nor is there any way of knowing how many cases of DLB were missed by both the clinical criteria and the SPECT scan. Although AD is generally regarded as the main differential diagnosis for DLB, other diagnoses may also need to be considered, including frontotemporal dementia (FTD). FP-CIT SPECT imaging may be less accurate at discriminating between DLB and FTD than between DLB and AD (Morgan 2012).

On the basis of studies demonstrating a high degree of concordance between an abnormal FP-CIT SPECT result and clinical diagnosis, low dopamine transporter uptake on SPECT or PET was added to the consensus clinical criteria at their revision in 2005 as a suggestive feature, that is a diagnosis of probable DLB could be made on the basis of only one core clinical feature if there was also an abnormal dopamine transporter scan. The test consequently became more widely used, initially in Europe. DaTscan^TM was not approved for use in the United States until 2011.

Clinical pathway

In current clinical practice, an FP-CIT SPECT scan is used to confirm or refute a clinical suspicion of DLB, so subjects undergoing the test are likely to be those who already meet other criteria for possible or probable DLB. In most cases an FP-CIT SPECT scan would be ordered for patients being seen by specialist dementia services after a standard work-up for dementia, usually including some form of structural imaging and after a specialist physician had determined that a patient met some or all of the core or suggestive clinical features of DLB.

Performing DAT imaging only after applying the clinical diagnostic criteria for DLB means that the estimated accuracy of DAT imaging is limited by the sensitivity of the clinical criteria. This trade-off of sensitivity against feasibility and cost may be inevitable in clinical practice, but in some research contexts a higher priority may be placed on identifying 'pure' diagnostic groups. Hence, it would also be of value to know the accuracy of DAT imaging for identifying DLB in secondary care patients with dementia who are not selected for any other DLB clinical features.

Currently, DAT imaging is the only special investigation included in the consensus diagnostic criteria (although polysomnography may be used to confirm the presence of rapid eye movement (REM) sleep behaviour disorder, another suggestive feature). Several other biomarker tests, including brain perfusion SPECT, fluorodeoxyglucose PET, I¹²³-metaiodobenzylguanidine cardiac scintigraphy, and various cerebrospinal fluid and blood biomarkers, have been shown to discriminate between DLB and AD and have been suggested as having diagnostic utility (Aarsland 2008). However, at present none of these have been approved for use in the diagnostic pathway for DLB, and none is yet in regular clinical use.

Rationale

Increasing importance is attached to making precise diagnoses of the subtypes of dementia. Although patients and relatives attach value to accurate diagnosis, its real importance comes when it leads to differences of clinical management that translate into better outcomes for patients. Currently, the most important implication of a DLB diagnosis for patient management is the recognition by clinicians of the high risk of severe adverse reactions to antipsychotic medication. Although antipsychotic use is regarded as a risk factor for mortality among all people with dementia (Schneider 2005), there is evidence that the most severe effects occur largely in people with Lewy body disease, with Aarsland finding severe sensitivity reactions among 8 out of 15 people with DLB and 0 out of 17 with AD (Aarsland 2005). Given the high rate of neuropsychiatric symptoms in DLB, including psychosis, these patients may be at increased risk for antipsychotic exposure, and it is important for clinicians to recognise this risk. The other key reason for being interested in more accurate diagnosis of DLB is the prospect of future treatment. If, as is widely hoped, increasing understanding of the pathophysiology of the dementing illnesses is translated into disease-specific therapies capable of modifying the course of the disease, then accurate diagnosis at the earliest possible stage of illness will be essential to direct research efforts and subsequently treatments to the correct patients.

In the case of DLB, current diagnostic criteria are particularly limited by relatively poor sensitivity even in specialist research centres. A biologically derived diagnostic test of sufficient accuracy could be a great help. At present, FP-CIT SPECT is the most highly developed supplementary test for DLB, but there is uncertainty about its accuracy and its place in clinical practice. As the test is expensive, it is in practice likely to be used in a two-stage diagnostic process, that is after a clinical suspicion of DLB has been established.

This review aimed to provide a rigorous assessment of the accuracy of DAT imaging for DLB diagnosis in order to identify whether and how it can be usefully employed in research and clinical practice.

Secondary objectives

The secondary objective was to use subgroup analyses to investigate the following potential sources of heterogeneity.

(a) Age of participants.

(b) Severity of dementia. The performance of the index test itself may vary with dementia severity. For objective B, since the performance of the clinical criteria may also depend on the severity of dementia, the 'added value' to be obtained from the index test may vary significantly.

(c) For objective B, diagnostic status at inclusion (possible or probable DLB, or both).

(d) DAT imaging method (SPECT or PET ligand).

Criteria for considering studies for this review

Search methods for identification of studies

We used a variety of information sources, aiming to retrieve as many relevant studies as possible. We devised terms for electronic database searching in conjunction with the team at the Cochrane Dementia and Cognitive Improvement Group.

Data collection and analysis

Results of the search

The 'raw' search returned 9125 references. After de-duplication and first assessment by Anna Noel-Storr, Cochrane Dementia and Cognitive Improvement Group Trials Search Co-ordinator, 199 references were passed to the review authors for further consideration. One or both review authors considered 51 references to need closer scrutiny, and the full text of the paper or conference abstract was obtained. We identified one paper from a reference list, and also examined the full text of this. We could rapidly exclude the majority as review papers or on the basis of the index test (not dopamine transporter imaging) or the reference standard (not neuropathology). We identified five studies as possibly including data that fulfilled our inclusion criteria. Of these, we included one and excluded one on the basis of published data. We sought further information from the authors of the other three; we subsequently excluded two and classified one as ongoing. We illustrate the process of study selection in Figure 1.

Methodological quality of included studies

We rated the included study (Walker 2009) using the QUADAS-2 tool plus additional items tailored to this review. The study had a two-gate design with alternative diagnosis controls (Rutjes 2005), including secondary care patients on the basis of their meeting consensus clinical criteria for DLB or National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) criteria for Alzheimer's disease, or both. This design has an intrinsically high risk of spectrum bias, hence we rated the study as being at high risk of bias on the patient selection domain. We also considered that there was a high risk of the interpretation of the index test introducing bias because SPECT scans were not interpreted in conjunction with structural images (one of the additional items). This risks false positive classifications on the index test and can therefore lead to underestimates of test specificity. We judged the study to be at low risk of bias in the reference standard domain. We labelled the risk of bias in the flow and timing domain 'unclear'. In studies of this nature, the application of the reference standard is inevitably delayed until patient death. In this case the mean interval was 42 months (range 6 to 106 months), during which additional pathology could have developed. There is some risk of bias associated with this, but it is unavoidable and hard to quantify.

Findings

The most recent publication in 2009 reported data on 22 eligible participants with dementia and clinical diagnoses of DLB or AD. In this group, using the autopsy findings as the reference standard, the consensus clinical criteria for DLB had a sensitivity of 0.89 and a specificity of 0.46 for the discrimination of DLB from non-DLB (data taken from table 2, Walker 2009). Of 15 participants who were given an initial clinical diagnosis of DLB, 7 were false positives.

We considered first the complete group of 22 participants. Using semiquantitative analysis of the scans, with an abnormal scan defined as worst-affected posterior putamen to occipital cortex binding ratio greater than 2 SDs below controls, ¹²³I-FP-CIT SPECT had a sensitivity of 1.00 (95% CI 0.66 to 1.00) and a specificity of 0.92 (95% CI 0.64 to 1.00) for the diagnosis of DLB (Data table 1, Figure 2). Only 1 of 22 cases had discordant imaging and neuropathology. This was a false positive―an abnormal index scan in a patient who at autopsy was found to have AD and cerebrovascular disease that affected the substantia nigra and putamen. The 2009 paper did not report visual ratings, but the 2007 paper reported visual ratings for 19 eligible participants. In this case, ¹²³I-FP-CIT SPECT had a sensitivity of 0.86 (95% CI 0.42 to 1.00) and a specificity of 0.83 (95% CI 0.52 to 0.98) (Data table 2, Figure 3).

The intention of the study was to address the discrimination of DLB from AD (our objective A) rather than primarily to estimate the value of DAT imaging in clinical practice (objective B), but we considered that it did provide the best evidence available to assess the accuracy of DAT imaging among people identified clinically as having possible or probable DLB. We therefore also considered only the 15 participants who met consensus clinical criteria for DLB at baseline. The 2007 publication described 13 of these in more detail. Twelve of these 13 had 2 or 3 core features of DLB and could be classified as having 'probable DLB'; the remaining participant had only 1 core feature of DLB and may have met criteria for 'possible DLB' (suggestive features not reported). The clinical features of the two additional participants in the 2009 paper were not described.

In this group of 15 participants, ¹²³I-FP-CIT SPECT scanning, analysed semiquantitatively with an abnormal scan defined as a worst-affected posterior putamen to occipital cortex binding ratio greater than 2 SDs below controls, had a sensitivity of 1.00 (95% CI 0.63 to 1.00) and a specificity of 1.00 (95% CI 0.59 to 1.00) for the diagnosis of DLB (Data table 3, Figure 4). Visual scan ratings were also available for the 13 participants in the 2007 paper. Rated visually, accuracy in this group was lower with a sensitivity of 0.83 (95% CI 0.36 to 1.00) and a specificity of 0.71 (95% CI 0.29 to 0.96) (Data table 4, Figure 5). One participant had a scan rated visually as normal but had DLB diagnosed at autopsy; two participants had scans rated visually as abnormal but had diagnoses of AD plus cerebrovascular disease and frontotemporal lobar degeneration at autopsy.

Summary of findings

Summary of main results

Objective A. We included only one study in the review. This study addressed the accuracy of ¹²³I-FP-CIT SPECT scanning for the discrimination of DLB from AD in secondary care. We found ¹²³I-FP-CIT SPECT, analysed semiquantitatively, to have a sensitivity of 100% and a specificity of 92% for the diagnosis of DLB at autopsy among secondary care patients with moderately severe dementia selected on the basis of a clinical diagnosis of either DLB or AD. Due to the small size of the study, the confidence intervals for both estimates were wide. No participant who had a normal SPECT scan at presentation was found to have DLB at autopsy. One participant with a SPECT scan classified as abnormal did not have neuropathologically confirmed DLB, the reduced ligand uptake being due to focal cerebrovascular disease. The study's authors acknowledged that this error would have been avoided if the SPECT images had been interpreted in conjunction with high-quality structural images. The specificity is therefore likely to be an underestimate in comparison with standard practice. Visual rating of SPECT scans was slightly less accurate, with a sensitivity of 86% and a specificity of 83%, again with wide confidence intervals.

Objective B. Using data from the same study, we found that among 15 participants with initial clinical diagnoses of DLB (mostly probable) made by expert clinicians, ¹²³I-FP-CIT SPECT, analysed semiquantitatively, had a sensitivity of 100% and a specificity of 100% for confirmed DLB at autopsy. Again this result was imprecise; the data are also compatible with much lower sensitivity and specificity.

Strengths and weaknesses of the review

Despite an extensive literature search, we were able to include only one study with 22 eligible participants, hence our accuracy estimates were relatively imprecise. This paucity of evidence reflects the very significant challenges inherent in conducting long-term prospective studies of well-characterised participants, followed up to the point of autopsy. However, we consider that neuropathological diagnosis is the only reference standard that will lead to accurate estimates of the diagnostic accuracy of DAT imaging. The high level of accuracy in the one included study strongly reinforces the possibility that DAT imaging is more accurate than clinical diagnosis of DLB and, therefore, that clinical diagnosis cannot serve as a reference standard for estimating the diagnostic accuracy of DAT imaging.

The one included study did have significant methodological limitations that weaken confidence in the findings of the review. The study had a case-control-type design (more strictly, a two-gate design with alternative diagnosis controls (Rutjes 2005)). This study design tends to exclude patients who may present the most diagnostic challenge and is therefore at high risk of spectrum bias. The interpretation of the index test, which was done without reference to structural imaging, also risked introducing bias (underestimating specificity) and, in fact, this risk was realised in the study. The delayed verification design is associated with an unavoidable risk of the development of additional pathology between the index test and the reference standard, which may also lead to underestimates of specificity.

A strength of the study, and hence of the review, is that the neuropathological diagnoses were made according to rigorous standards.

Applicability of findings to the review question

The included patients were recruited from standard secondary care settings in the United Kingdom and met clinical criteria for DLB or for AD, the major differential diagnosis for DLB. The index test used employed the only DAT imaging ligand licensed for clinical use, and the reference standard used was likely to have classified patients as well as possible. The study was applicable to our first review objective (objective A).

We intended our second review objective (objective B) to be more applicable to clinical practice. It asked how accurately DAT imaging identifies DLB in people suspected of having DLB after a clinical work-up. We found no studies specifically designed to answer this question. However, data from the included study do suggest that DAT imaging is highly accurate, at least among people with moderately severe dementia and clinically probable DLB. It is important to note that the study did not include participants who might present the greatest diagnostic challenge, such as those with very mild dementia or those with only one core or suggestive DLB feature. DAT imaging in this context cannot add to the sensitivity of the clinical assessment (false negatives, or missed DLB diagnoses, on clinical assessment cannot be 'corrected'), but these data suggest that it can enhance specificity, that is a normal scan in this context can accurately exclude DLB. The specificity of the initial clinical diagnosis of DLB in the included study was only 46%, which is unusually low in comparison with most published reports of the accuracy of clinical criteria (McKeith 2004; Nelson 2010). This might exaggerate the apparent value added by SPECT imaging; on the other hand, it may be that this level of specificity is common in routine clinical settings.

Appendix 1. Consensus criteria for the clinical diagnosis of possible and probable dementia with Lewy bodies

Appendix 2. Marketing approvals for DaTscan^TM

European Union

First marketing authorisation in 2000.

Currently approved for the following therapeutic indications:

This medicinal product is for diagnostic use only.

DaTscan is indicated for detecting loss of functional dopaminergic neuron terminals in the striatum:

• In patients with clinically uncertain Parkinsonian Syndromes, in order to help differentiate Essential Tremor from Parkinsonian Syndromes related to idiopathic Parkinson's Disease, Multiple System Atrophy and Progressive Supranuclear Palsy. DaTscan is unable to discriminate between Parkinson's Disease, Multiple System Atrophy and Progressive Supranuclear Palsy.

• To help differentiate probable dementia with Lewy bodies from Alzheimer's disease. DaTscan is unable to discriminate between dementia with Lewy bodies and Parkinson's disease dementia.

United States

FDA approval in 2011 for the following indications and usage:

DaTscan (Ioflupane I 123 Injection) is a radiopharmaceutical indicated for
striatal dopamine transporter visualization using single photon emission
computed tomography (SPECT) brain imaging to assist in the evaluation of
adult patients with suspected Parkinsonian syndromes (PS). In these patients,
DaTscan may be used to help differentiate essential tremor from tremor due to
PS (idiopathic Parkinson's disease, multiple system atrophy and progressive
supranuclear palsy). DaTscan is an adjunct to other diagnostic evaluations.

Appendix 3. Sources searched and search strategies

Appendix 4. Additional items for assessing methodological quality in index test domain

Question 1: Were uninterpretable or intermediate test results reported?
Question 2: Were structural brain images available for comparison?
Question 3: Was the method of image reconstruction consistent throughout the study?
Question 4: Had test operators had appropriate training?
Question 5: Were data on observer variation in DAT image interpretation reported and within an acceptable range?