Study design

We conducted a single-centre prospective cohort study into the accuracy of ¹²³I-FP-CIT SPECT imaging in the diagnosis of probable MCI-LB in patients with one or more clinical symptoms at baseline that could indicate Lewy body disease. All patients were diagnosed with MCI on entry to the study; some developed dementia during follow-up.

Our index test was the dichotomised baseline FP-CIT image consensus panel rating result (see Image acquisition and processing). Our reference standard was consensus clinical diagnosis at most recent assessment of either probable Lewy body disease (comprising probable MCI-LB or probable DLB) or Alzheimer's disease (comprising MCI-AD or Alzheimer's disease dementia). Consensus clinical diagnosis at most recent assessment incorporated core features and cardiac metaiodobenzylguanidine (mIBG) imaging result where available (see Clinical diagnosis). The presence of core clinical features was assessed masked to imaging biomarker results. Patients with uncertain diagnoses of possible MCI-LB or possible DLB were included in the study, but not in the main diagnostic accuracy calculation, because of the greater diagnostic uncertainty in this group.

Our primary research question was as follows: what is the sensitivity, specificity and overall accuracy of ¹²³I-FP-CIT SPECT for the diagnosis of probable MCI-LB?

Patient recruitment

Patients aged 60 or older with an existing clinical diagnosis of MCI were recruited from local memory services in the North-East of England between April 2013 and September 2019. The medical records of all patients meeting the above criteria were reviewed to assess eligibility. In addition to the diagnosis of MCI, records had to include one or more clinical symptoms supportive of Lewy body disease (for example mood changes, sleep disturbance or autonomic symptoms) and/or the presence of core DLB features. Written informed consent was obtained from all patients. Following consent, participants underwent interview, clinical assessment and neurological examination by a medical doctor (R.D., S.L.). Determination of parkinsonism for diagnostic purposes was based on the neurological examination.

The MDS Unified Parkinson's Disease Rating Scale – Motor Examination (UPDRS-III),^{Reference Goetz, Tilley, Shaftman, Stebbins, Fahn and Martinez-Martin6} Epworth Sleepiness Scale^{Reference Johns7} and Geriatric Depression Scale^{Reference D'Ath, Katona, Mullan, Evans and Katona8} were administered to patients. The Instrumental Activities of Daily Living (IADL) scale,^{Reference Lawton and Brody9} North-East Visual Hallucinations Inventory,^{Reference Holiday, Pirogovsky-Turk, Malcarne, Filoteo, Litvan and Lessig10} Neuropsychiatric Inventory,^{Reference Cummings, Mega, Gray, Rosenberg-Thompson, Carusi and Gornbein11} Mayo Sleep Questionnaire,^{Reference Boeve, Molano, Ferman, Smith, Lin and Bieniek12} Clinician Assessment of Fluctuation^{Reference Walker, Ayre, Cummings, Wesnes, Mckeith and O'brien13} and Dementia Cognitive Fluctuation Scale^{Reference Lee, Mckeith, Mosimann, Ghosh-Nodial, Grayson and Wilson14} were administered to spouses or close family members acting as informants. The Clinical Dementia Rating scale (CDR)^{Reference Hughes, Berg, Danziger, Coben and Martin15} was completed using clinical history and research assessments. A detailed neuropsychological evaluation was also carried out as reported in our recent publication^{Reference Donaghy, Ciafone, Durcan, Hamilton, Barker and Lloyd16} which included the Addenbrooke's Cognitive Examination – Revised,^{Reference Mioshi, Dawson, Mitchell, Arnold and Hodges17} a 100-point cognitive screening test from which a Mini-Mental State Examination score was derived. Other tests included FAS Verbal Fluency,^{Reference Borkowski, Benton and Spreen18} the Trail-making Test Parts A and B,^{Reference Reitan19} the Graded Naming Test,^{Reference McKenna and Warrington20} the Rey Auditory Verbal Learning Test,^{Reference Rey21} simple and choice reaction times^{Reference Ballard, O'Brien, Gray, Cormack, Ayre and Rowan22} and line angle discrimination.^{Reference Donaghy, Ciafone, Durcan, Hamilton, Barker and Lloyd16,Reference Firbank, Mosimann, Watson, Barber and Blamire23}

Patients recruited from April 2016 onwards were offered cardiac sympathetic innervation imaging with mIBG, the results of which were incorporated into diagnoses. Cardiac mIBG uptake was quantified using the heart-to-mediastinum count ratio as a diagnostic indicator, as described previously.^{Reference Roberts, Kane, Lloyd, Petrides, Howe and O'Brien24,Reference Roberts, Lloyd, Kane, Durcan, Lawley and Howe25}

The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. All procedures involving human patients were approved by the National Research Ethics Service Committee North East - Newcastle & North Tyneside 2 (Research Ethics Committee Identification Number 15/NE/0420).

Clinical diagnosis

A three-person consensus clinical panel of experienced consultant old age psychiatrists (A.J.T., P.C.D., J.-P.T.) independently reviewed the research assessment and clinical notes and confirmed diagnoses of MCI according to National Institute on Aging—Alzheimer's Association (NIA-AA) criteria.^{Reference Albert, DeKosky, Dickson, Dubois, Feldman and Fox26} This consensus panel method has previously been validated against autopsy and is recognised by regulatory authorities as the clinical gold standard for living patients.^{Reference McKeith, O'Brien, Walker, Tatsch, Booij and Darcourt27,Reference McKeith, Ballard, O'Brien, Barber, Ince and Jaros28} This was based on evidence of minimal functional impairment and a CDR of 0 or 0.5, and a history of subjective and objective cognitive decline on assessment. Neuropsychological test results were not used to confirm MCI. Anyone with dementia or only subjective impairment was excluded. To determine the aetiology, the presence or absence of core Lewy body features were also rated by the panel, in accordance with the fourth consensus criteria for DLB^{Reference McKeith, Boeve, Dickson, Halliday, Taylor and Weintraub3} and the recently published consensus research criteria for MCI-LB.^{Reference McKeith, Ferman, Thomas, Blanc, Boeve and Fujishiro4} The panel reviewed the notes from the clinical and neurological examination done during the research assessment as well as the health service records for this.

Determination of parkinsonism for diagnostic purposes was based on the presence of bradykinesia (defined as slowness of movement and decrement in amplitude or speed), rest tremor or rigidity. Participants all had baseline research assessments and most had annual review data available by the time of data locking. Annual review data (up to 7 years) was used for the consensus panel diagnosis where available. Cardiac mIBG results were later incorporated into diagnoses, but the panel decisions on symptom presence were made initially masked to these findings. FP-CIT results were not included in the diagnosis, and the panel had no access to these.

Participants received a diagnosis MCI-AD when they had no core Lewy body features, a normal mIBG scan and evidence of decline that was characteristic of Alzheimer's disease, i.e., they met the additional NIA-AA criterion of ‘etiology of MCI consistent with Alzheimer's disease pathophysiologic process’.^{Reference Albert, DeKosky, Dickson, Dubois, Feldman and Fox26} Biomarker tests for Alzheimer's pathology were not conducted, in line with research practice when the study was developed.

The study is concerned with the detection of Lewy body disease and the presence of Alzheimer's disease or other aetiologies does not exclude Lewy body disease. Probable MCI-LB was diagnosed in patients with either two or more core Lewy body features, or one or more core feature and abnormal mIBG, in accordance with consensus criteria.^{Reference McKeith, Ferman, Thomas, Blanc, Boeve and Fujishiro4} Patients were assigned the diagnosis of possible MCI-LB if they presented with only one core feature and their mIBG scan was normal, or if their mIBG scan result was abnormal but they had no core features.

Image acquisition and processing

Patients were scanned within 1 month of baseline clinical assessment, unless an ¹²³I-FP-CIT scan had been acquired for clinical reasons within the previous 6 months, in which case it was not repeated, in accordance with our ethical approvals. This was the case for four patients, whose images were obtained for rating. These clinical images were acquired using a very similar protocol to the study scan protocol below, but not all were acquired on the same gamma camera.

Patients were scanned 3–6 h following a bolus intravenous injection of 185 MBq of ¹²³I-FP-CIT (Ioflupane (DaTSCAN) GE Healthcare, UK) (scan duration, 25 min) using a double-headed gamma camera (Siemens Symbia S or Siemens Intevo) fitted with a low-energy high-resolution parallel hole collimator. A total of 120 (60 per detector) 25 s views over a 360° orbit were acquired on a 128 × 128 matrix with a zoom of 1.23× giving a pixel size 3.9 mm × 3.9 mm. Image processing and display was then performed on a Hermes workstation (Hermes Medical Solutions, Stockholm, Sweden).

Images used in our previous publication were reconstructed without attenuation correction using filtered back projection and a Butterworth filter (order 10, cut-off 1.3 cycles/cm). New FP-CIT images were reconstructed using iterative reconstruction with resolution recovery, uniform attenuation correction and Monte Carlo scatter correction. For all images, transverse data was manually re-oriented to correct for any head tilt and to provide a consistent display.

Visual rating of FP-CIT images

Visual assessment of all scans was undertaken masked to clinical diagnosis and information. Briefly, scans were rated independently by each panel member using an established FP-CIT visual rating procedure^{Reference Benamer, Patterson, Grosset, Booij, de Bruin and van Royen30} that has also shown diagnostic value in the differential diagnosis of DLB and Alzheimer's disease.^{Reference O'Brien, Colloby, Fenwick, Williams, Firbank and Burn31} Raters were provided with age-corrected specific binding ratio results generated using DaTQUANT v1.0 (GE Healthcare, Chalfont St Giles, UK) prior to April 2016 and BRASS v2.5 (Hermes Medical Ltd, Stockholm, Sweden) for more recent scans. The consensus panel consisted of a group of four or five raters experienced at reviewing FP-CIT images: A.J.T., J.L., P.C.D. and G.P.; G.R. from 2016. The panel members were sent sets of anonymised images to review in a randomised order by an independent member of the team (S.J.C.). Panel members used their professional judgement in cases where visual assessment and semi-quantification did not agree.

Each rater independently dichotomised the scans as normal (non-Lewy body appearance) or abnormal. Mild balanced loss of dopaminergic uptake throughout both striata was designated as within normal limits, as this pattern was seen in controls in our paper using autopsy-confirmed diagnoses.^{Reference Lloyd, Petrides, Donaghy, Colloby, Attems and O'Brien32} Moderate-to-severe balanced loss was rated abnormal. After rating all scans, any scan where there was not agreement between at least four raters was then subsequently reviewed at a panel meeting, where a full consensus rating of normal or abnormal was agreed. If an infarct along the nigro-striatal pathway was suspected to be affecting uptake, images from magnetic resonance imaging (MRI) were reviewed retrospectively and the participant excluded if confirmed. Example images categorised as normal and abnormal are shown in Supplementary Figure 1 available at https://doi.org/10.1192/bjp.2020.234.

Statistical analysis

The Statistical Package for Social Sciences software (SPSS version 25) was used to produce summary statistics. Continuous variables were analysed for differences between the MCI-AD and probable MCI-LB groups using Student's t-test or Mann-Whitney U-test for independent samples. The χ^2-test was used for determining whether there was a difference in the proportions of binary variables.

BRASS quantification was used to calculate FP-CIT whole striatum and putaminal specific binding ratios (SBRs) for all participants. We checked for difference in mean SBR between the probable MCI-LB and MCI-AD population using an independent samples t-test, as data was normally distributed. We tested for a difference in the proportion of abnormal scans in the probable MCI-LB group and the MCI-AD group using a χ²-test. The accuracy of semi-quantification alone was calculated from the proportion of scans in each diagnostic group with Z-scores below –2, i.e. more than 2 s.d. below the mean of age-matched controls in the BRASS database.

The diagnostic accuracy of FP-CIT visual rating as a biomarker for probable MCI-LB (sensitivity, specificity and overall accuracy values) was calculated from a 2 × 2 frequency table. Likelihood ratios were calculated to estimate the added value of dopaminergic imaging in the diagnosis of probable MCI-LB. As a secondary analysis we assessed whether sensitivity appeared greater in those patients with parkinsonism at baseline, compared with those without, recognising that the study would not necessarily be powered to detect a significant difference.

To assess the potential impact of a positive FP-CIT result on diagnosis in clinical practice, we reviewed our probable MCI-LB group, identifying those with fewer than two core features at baseline. From this subset we calculated the proportion with a positive FP-CIT scan.