Our study demonstrate that cerebral [¹⁸F]FDG PET provides a moderate to high diagnostic sensitivity and specificity for differentiation between subtypes of AP with significant sensitivity (> 80%) and specificity (> 90%). However, the sensitivity falls to approximately 62% for 4R-tauopathies, mirroring the known diagnostic challenges of these conditions. These findings support earlier studies that showed the efficacy of [¹⁸F]FDG PET in differentiating AP subtypes, which reported similar levels of sensitivity and specificity (> 85%) [10, 11].

Challenges arise with CBD due to its overlapping metabolic pattern with PSP, which was identified in a prior prospective study that reported a low PPV (67%) for CBD [12]. This is likely because CBD and PSP share certain neuropathological features, such as basophilic inclusions and distinct cytoskeletal abnormalities, making them difficult to differentiate [19]. Our study’s moderate sensitivity in diagnosing 4R-tauopathies, particularly CBD, aligns with the general difficulties encountered in their accurate clinical diagnosis, as reported in previous studies [20, 21].

The strengths of our study include the inclusion of a diverse patient population consecutively referred for cerebral [¹⁸F]FDG imaging due to suspected atypical parkinsonism, thereby reflecting real-world clinical practice. We also adhered to the standards for reporting diagnostic accuracy studies.

Limitations of our study include potential bias of the reference standard. The clinical diagnosis might have been influenced by neuroimaging results, challenging the independence of reference standards from the test results. A follow-up consultation would be necessary to assess the neurological status and clinical appearance without access to imaging data, which would significantly reduce the sample size of the most ill subjects, as 24 patients died in the study period, and a significant portion of the patient population, due to the severity of their illness, would be unable to participate. Another limitation involves the lack of post-mortem histopathological diagnosis, which raises questions about the follow-up diagnosis's robustness, especially in uncertain clinical cases. However, histopathological diagnosis is rarely performed, and selective inclusion of patients with a histopathological diagnosis would not be possible with the present material. In a post-mortem sample of 25 patients with MSA symptomatology included at our Neurology department before 2015, 22 showed MSA pathology post-mortem, 2 showed other pathology, and one was inconclusive (unpublished data). Further limitations include the access to previous imaging in the diagnostic reading of the [¹⁸F]FDG PET, which limits the assessment of the independent information by [¹⁸F]FDG PET. However, clinical [¹⁸F]FDG PET reading without the access to MRI or other structural information is not relevant according to practise guidelines [11], and DAT scans are shown to be unable to distinguish between parkinsonian syndromes [14] but is likely to have improved the discrimination between parkinsonian syndromes from healthy ageing. Thus, we chose to include previous imaging to reflect daily clinical practice. Additional readers would have strengthen the results and provided measures of interrater variation. A high specificity of [¹⁸F]FDG PET is believed to be of diagnostic, prognostic, and survival benefit, as is also seen in other patient groups [22]. There are currently ongoing clinical treatment trials for AP subtypes [23], for which an early diagnosis is essential for correct stratification of the patients, ensuring more reliable and robust outcomes. Despite these limitations, our findings reinforce the utility of cerebral [¹⁸F]FDG PET. Specific tracers for proteinopathies could potentially enhance sensitivity and specificity, as seen with amyloid imaging in Alzheimer’s disease. Promising tracers such as the tau tracer [¹⁸F]APN-1607 and the novel [¹⁸F]PI-2620 are under study and could greatly aid in early diagnosis [24, 25]. Unfortunately, an α-synuclein PET tracer is still missing [26]. The mitochondrial translocator protein (TSPO) binding [¹¹C]PBR28 shows potential as a neuroinflammation tracer. Recent studies suggest its use as an imaging biomarker for MSA with high sensitivity and specificity [27]. However, the necessity for genotyping for TSPO single-nucleotide polymorphism to exclude potential false negatives remains a hurdle [28]. Future research to examine the role of tau and neuroinflammation PET tracers in diagnosing early-stage AP is greatly encouraged.