Dopaminergic imaging is an established biomarker for dementia with Lewy bodies, but its diagnostic accuracy at the mild cognitive impairment (MCI) stage remains uncertain.

To provide robust prospective evidence of the diagnostic accuracy of dopaminergic imaging at the MCI stage to either support or refute its inclusion as a biomarker for the diagnosis of MCI with Lewy bodies.

We conducted a prospective diagnostic accuracy study of baseline dopaminergic imaging with [123I]N-ω-fluoropropyl-2β-carbomethoxy-3β-(4-iodophenyl)nortropane single-photon emission computerised tomography (123I-FP-CIT SPECT) in 144 patients with MCI. Images were rated as normal or abnormal by a panel of experts with access to striatal binding ratio results. Follow-up consensus diagnosis based on the presence of core features of Lewy body disease was used as the reference standard.

At latest assessment (mean 2 years) 61 patients had probable MCI with Lewy bodies, 26 possible MCI with Lewy bodies and 57 MCI due to Alzheimer's disease. The sensitivity of baseline FP-CIT visual rating for probable MCI with Lewy bodies was 66% (95% CI 52–77%), specificity 88% (76–95%) and accuracy 76% (68–84%), with positive likelihood ratio 5.3.

It is over five times as likely for an abnormal scan to be found in probable MCI with Lewy bodies than MCI due to Alzheimer's disease. Dopaminergic imaging appears to be useful at the MCI stage in cases where Lewy body disease is suspected clinically.

Recently published diagnostic criteria for mild cognitive impairment with Lewy bodies (MCI-LB) include five neuropsychiatric supportive features (non-visual hallucinations, systematised delusions, apathy, anxiety and depression). We have previously demonstrated that the presence of two or more of these symptoms differentiates MCI-LB from MCI due to Alzheimer's disease (MCI-AD) with a likelihood ratio >4. The aim of this study was to replicate the findings in an independent cohort.

Participants ⩾60 years old with MCI were recruited. Each participant had a detailed clinical, cognitive and imaging assessment including FP-CIT SPECT and cardiac MIBG. The presence of neuropsychiatric supportive symptoms was determined using the Neuropsychiatric Inventory (NPI). Participants were classified as MCI-AD, possible MCI-LB and probable MCI-LB based on current diagnostic criteria. Participants with possible MCI-LB were excluded from further analysis.

Probable MCI-LB (n = 28) had higher NPI total and distress scores than MCI-AD (n = 30). In total, 59% of MCI-LB had two or more neuropsychiatric supportive symptoms compared with 9% of MCI-AD (likelihood ratio 6.5, p < 0.001). MCI-LB participants also had a significantly greater delayed recall and a lower Trails A:Trails B ratio than MCI-AD.

MCI-LB is associated with significantly greater neuropsychiatric symptoms than MCI-AD. The presence of two or more neuropsychiatric supportive symptoms as defined by MCI-LB diagnostic criteria is highly specific and moderately sensitive for a diagnosis of MCI-LB. The cognitive profile of MCI-LB differs from MCI-AD, with greater executive and lesser memory impairment, but these differences are not sufficient to differentiate MCI-LB from MCI-AD.

Mild cognitive impairment (MCI) may gradually worsen to dementia, but often remains stable for extended periods of time. Little is known about the predictors of decline to help explain this variation. We aimed to explore whether this heterogeneous course of MCI may be predicted by the presence of Lewy body (LB) symptoms in a prospectively-recruited longitudinal cohort of MCI with Lewy bodies (MCI-LB) and Alzheimer's disease (MCI-AD).

A prospective cohort (n = 76) aged ⩾60 years underwent detailed assessment after recent MCI diagnosis, and were followed up annually with repeated neuropsychological testing and clinical review of cognitive status and LB symptoms. Latent class mixture modelling identified data-driven sub-groups with distinct trajectories of global cognitive function.

Three distinct trajectories were identified in the full cohort: slow/stable progression (46%), intermediate progressive decline (41%) and a small group with a much faster decline (13%). The presence of LB symptomology, and visual hallucinations in particular, predicted decline v. a stable cognitive trajectory. With time zeroed on study end (death, dementia or withdrawal) where available (n = 39), the same subgroups were identified. Adjustment for baseline functioning obscured the presence of any latent classes, suggesting that baseline function is an important parameter in prospective decline.

These results highlight some potential signals for impending decline in MCI; poorer baseline function and the presence of probable LB symptoms – particularly visual hallucinations. Identifying people with a rapid decline is important but our findings are preliminary given the modest cohort size.