FDG-PET

For predicting progression from MCI to clinically diagnosed AD, a high-quality systematic review from 2018 (studies published between 1999 and 2017) reported that diagnostic accuracy results for FDG-PET varied across studies (sensitivity 25% to 100% and specificity 15% to 100%).³ A systematic review from 2024, which included more up-to-date studies, reported that sensitivities in predicting progression from MCI to clinically or neuropathologically diagnosed dementia (mainly AD) ranged from 43% to 100%, and specificities from 63% to 94%.⁴ The wide range of values was due to heterogeneity between studies in how images were assessed, how studies were conducted and differences in study populations.

SPECT

For predicting the progression from MCI to AD, one lower-quality systematic review with meta-analyses evaluated the accuracy of SPECT and FDG-PET (studies published between 1998 and 2006).⁵ Five of the included studies were on FDG-PET and four were on SPECT. The authors reported that the sensitivity (81%) and specificity (74%) values for SPECT imaging were lower than the sensitivity (87%) and specificity (89%) values for FDG-PET. Limitations with the review mean that the results of the meta-analyses should be treated with caution.

FDG-PET

For differentiating subtypes of dementia, a high-quality systematic review from 2020 (three cohort studies published between 2007 and 2011) evaluated the accuracy of FDG-PET for distinguishing neuropathologically confirmed AD from non-AD.⁶ In two studies (n=182), the median sensitivity and specificity of FDG-PET for distinguishing between AD and non-AD was 89% (range 84% to 94%) and 74% (range 73% to 74%), respectively. For distinguishing between AD and frontotemporal lobar degeneration (FTLD), one study (n=45) reported median sensitivity and specificity for FDG-PET of 97% (range 96% to 98%) and 66% (range 59% to 73%), respectively. Based on these findings, the authors concluded that FDG-PET was highly sensitive and moderately specific for diagnoses of neuropathologically confirmed AD. A systematic review from 2024 (14 studies published between 2007 and 2021) using clinically (rather than neuropathologically) diagnosed dementia subtype as the reference standard reported similar sensitivity and specificity results to the 2020 review for distinguishing between AD and non-AD or FTLD.⁷

SPECT

For differentiating subtypes of dementia, a high-quality systematic review from 2020 (three studies published between 2007 and 2011, n=205), evaluated the use of SPECT to distinguish confirmed AD (based on postmortem examination) from various non-AD neuropathologic diagnoses.6 Median sensitivity was 63% (range 57% to 94%) and median specificity was 83% (range 76% to 92%).

We identified a 2020 retrospective observational study that directly compared FDG-PET with SPECT in diagnosing AD in people with MCI and dementia. The reference standard was an amyloid-PET scan.⁸ The sensitivity of FDG-PET and SPECT was 76% versus 43% (p<0.001), while specificity was 74% versus 83% (p=0.45). While these findings are broadly in concordance with the results from the systematic reviews, they must be treated with caution given the study limitations.

We did not identify any economic evidence comparing FDG-PET with SPECT imaging for the diagnosis of dementia.

We identified one study which included a survey of 98 people (68 with dementia and 30 controls) who had an FDG-PET and SPECT scan.⁹ Results suggested that both types of scan are generally acceptable to patients, and that perceived diagnostic accuracy tended to dictate preferences for one scan over the other.

The PET-CT service in NHSScotland is in high demand and expansion of the service to include dementia imaging will require infrastructure and resource investment to ensure that scanning capacity for other patients (for example, people with cancer) is not negatively affected. Existing barriers to use of PET-CT for dementia imaging include a lack of spare capacity for scans as well as the production and supply of radioisotopes.